Introduction history overrides social factors in explaining genetic structure of females in Mediterranean mouflon

Abstract Fine‐scale spatial genetic structure of populations results from social and spatial behaviors of individuals such as sex‐biased dispersal and philopatry. However, the demographic history of a given population can override such socio‐spatial factors in shaping genetic variability when bottlenecks or founder events occurred in the population. Here, we investigated whether socio‐spatial organization determines the fine‐scale genetic structure for both sexes in a Mediterranean mouflon (Ovis gmelini musimon × Ovis sp.) population in southern France 60 years after its introduction. Based on multilocus genotypes at 16 loci of microsatellite DNA (n = 230 individuals), we identified three genetic groups for females and two for males, and concurrently defined the same number of socio‐spatial units using both GPS‐collared individuals (n = 121) and visual resightings of marked individuals (n = 378). The socio‐spatial and genetic structures did not match, indicating that the former was not the main driver of the latter for both sexes. Beyond this structural mismatch, we found significant, yet low, genetic differentiation among female socio‐spatial groups, and no genetic differentiation in males, with this suggesting female philopatry and male‐biased gene flow, respectively. Despite spatial disconnection, females from the north of the study area were genetically closer to females from the south, as indicated by the spatial analysis of the genetic variability, and this pattern was in accordance with the common genetic origin of their founders. To conclude, more than 14 generations later, genetic signatures of first introduction are not only still detectable among females, but they also represent the main factor shaping their present‐time genetic structure.

portant role in determining gene flow, and consequently the genetic structure of a population (Slatkin, 1987). Among other benefits, gene flow often contribute to maintain genetic diversity and heterozygosity in populations (Garant, Forde, & Hendry, 2007;Říčanová et al., 2011;Segelbacher et al., 2010), which in turn limit inbreeding depression (Keller & Waller, 2002) and preserve both immunocompetence (Kloch et al., 2013;MacDougall-Shackleton et al., 2005) and adaptability of populations to changing environments (Frankham, Ballou, & Briscoe, 2004). Investigating gene flow is therefore of considerable interest for conservation and management perspectives, especially in the current context of climate change and habitat fragmentation (Wasserman et al., 2012).
Studying the genetic structure is particularly suitable in (re)introduced populations as (re)introductions and translocations can have long-term impacts on the genetic makeup of populations. They generally involve only a limited number of individuals thus retaining only a subset of the total genetic diversity of the source population, the occurrence of a strong genetic bottleneck being a likely event (Biebach & Keller, 2009;Hedrick, Gutierrez, & Lee, 2001).
Nevertheless, a different origin of the founder individuals may genetically impact descendants for many generations (e.g., Biebach & Keller, 2009;Latch & Rhodes, 2005). Many studies have focused on the genetic consequences of introduction history (e.g., Barbanera et al., 2015;Biebach & Keller, 2009;Stephen et al., 2005), but the genetic consequences of (re)introductions and translocations remain difficult to study at the intrapopulation level due to several existing confounding factors (gene flow between translocated and other populations, unknown translocation history and founder genotypes, Mock, Latch, & Rhodes, 2004). However, as these processes represent the main tools within current conservation and management strategies (see Armstrong & Seddon, 2008;Batson et al., 2015;Latch & Rhodes, 2005), there is an essential need to assess the long-term impacts of these strategies on the genetic structure of populations.
The fine-scale spatial genetic structure of gregarious species is also strongly influenced by their social structures (Coltman, Pilkington, & Pemberton, 2003;Hazlitt et al., 2004;Storz, 1999). When individuals reproduce within social groups, the genetic differentiation among groups increases due to genetic drift (Storz, 1999). Concurrently, relatedness and inbreeding increase within groups and a socio-spatial genetic substructuring of the population consequently appear. In mammals in which males are more prone to disperse than philopatric females (Greenwood, 1980), female behavior is thus expected to determine the spatial genetic structure of the population. There is often a stronger genetic structure in females than in males because of increased relatedness among spatially close individuals (e.g., for wild boars Sus scrofa, Podgórski, Scandura, & Jedrzejewska, 2014). Although socially mediated fine-scale spatial genetic structure has been well characterized in various mammalian societies that exhibit stable social bonds (e.g., Hazlitt et al., 2004 in brush-tailed rock-wallabies Petrogale penicillata, Städele et al., 2015 in hamadryas baboons Papio hamadryas), studies are still scarce on species where group structure can be quite loose and characterized by fission-fusion dynamics, such as the large herbivores (but see Coltman et al., 2003 for Soay sheep Ovis aries, Archie et al., 2008 for African elephants Loxodonta africana).
To date, only a few studies have investigated the influence of the location of historic release sites on the current spatial genetic structure at the intrapopulation level (but see Simpson et al., 2013) and, to F I G U R E 1 Map of the traps and sites of release in the Caroux-Espinouse massif, southern France. The black line indicates the boundaries of the national fauna reserve, yellow points represent the traps of capture, red points represent the sites of release of founder individuals (see also our knowledge, almost none concurrently assessed the relative contribution of the social-spatial structure and the translocation history of the population. We tried to achieve this challenging task by studying an isolated Mediterranean mouflon (Ovis gmelini musimon × Ovis sp.) population in the Caroux-Espinouse massif (southern France) 60 years after 19 founder individuals of diverse origins were introduced. Ungulates are especially well-suited for such a task as most of their populations were (re)introduced worldwide during the last century (e.g., Ovis canadensis mexicana, Hedrick et al., 2001; Alpine ibex Capra ibex ibex, Biebach & Keller, 2009). Furthermore, these include social species with philopatric females composing matrilineal groups (e.g., wild boars Sus scrofa, Poteaux et al., 2009;Podgórski et al., 2014). Based on genetic data from 16 microsatellite loci combined with intensive visual and GPS monitoring of marked individuals, we thus aimed at disentangling the relative impacts of past introduction and socio-spatial structure on the fine-scale genetic structure of the population.
In the studied population, females are philopatric and establish their home range on their birth range (Dubois, Gerard, & Maublanc, 1992;Dubois et al., 1994;Dupuis et al., 2002). Regarding males, several forms of philopatry were previously observed. In relation to natal dispersal, dispersers and sedentary males (faithful to their birth range) have been observed (see Dubois, Quenette, & Bideau, 1993;Dubois et al., 1996;King & Brooks, 2003). Concerning reproductive dispersal, philopatric and unfaithful males have also been observed (Dubois et al., 1993. Additionally, individual movements and home ranges appeared to be strongly influenced by various natural and anthropogenic linear landscape features acting as behavioral barriers (Marchand et al., 2017). We expected these behavioral characteristics to lead to a strong socio-spatial structure (see Garel et al., 2007;Martins et al., 2002), determining in turn the genetic structure of the population. In addition, we expected both structures to be sex-specific due to the different use of space between males and females (Marchand et al., 2015(Marchand et al., , 2017 and the reproductive dispersal by males during the rutting period. Finally, the combination of marked socio-spatial structure and geographic isolation of the population may magnify the historical genetic signature of introductions and be favorable to the maintenance of a genetic footprint (e.g., in red squirrels Sciurus vulgaris, Simpson et al., 2013).

| Study population, species, and data collection
Data were collected in a National Hunting and Wildlife Reserve (1,658 ha, 532-1,124 m above sea level) in the Caroux-Espinouse massif (43°38′N, 2°58′E, 17,000 ha, 130-1,124 m a.s.l., southern France; Figure 1). This low mountain area is characterized by deep valleys indenting plateaux and creating a mosaic of ridges and thalwegs (Marchand et al., 2017). Vegetation within the wildlife reserve is an irregular mosaic of beech, chestnut, and coniferous forests with open areas dominated by rocky slopes and broom and heather moorlands (Marchand et al., 2015). Human activities are strictly regulated in the wildlife reserve: Hunting is forbidden, and recreational activities are restricted to hiking on a few main trails (Marchand et al., 2014a).
Ewes of Mediterranean mouflon are commonly viewed as monotocous (twinning rate <3%) and sexually mature from 1.5 years of age (Garel et al., 2005). Births occur from late March to early June (Bon, Dardaillon, & Estevez, 1993). Although probably sexually mature at 2 years (see Festa-Bianchet, 2012 in bighorn sheep), only rams three or more years old have been observed in rutting activities (Bon et al., 1992(Bon et al., , 1995. Mating system is expected to be polygynous with a few large males achieving most of the mating (see Geist, 1971 for Dall's sheep Ovis dalli; see Jarman, 1983;Hogg, 1987 for Bighorn sheep Ovis canadensis). Females seem to be philopatric (Dubois et al., 1992Dupuis et al., 2002;Martins et al., 2002;Maublanc et al., 1994) while males might tend to disperse as they get older (King & Brooks, 2003), although sedentary males have also been observed in this population (see Dubois et al., 1996). Mature males and females segregate most of the year but less during the rut as males join females to reproduce (Bon & Campan, 1989;Bon et al., 1992;Cransac et al., 1998;Dubois et al., 1993;Le Pendu et al., 1996). Faithfulness of males to their rutting site was also observed in this population (Dupuis et al., 2002;Martins et al., 2002).
The population of mouflon in the Caroux-Espinouse massif originally grew from 19 Mediterranean mouflon released in the wildlife reserve between 1956 and 1960 (Table 1; Figure 1; Cugnasse, 1990;Garel et al., 2005). Since 1973, there has been hunting in surrounding unprotected areas from the beginning of September to the end of February (Garel et al., 2007)  highway is separating the two areas. Accordingly with dispersal abilities of Mediterranean mouflon (see e.g., Dubois et al., 1996) and the impacts of anthropogenic linear landscape features on individual movements (Marchand et al., 2017), the Caroux-Espinouse population is isolated and no natural gene flow could arise from other mouflon populations.

| Microsatellite genotyping
We used 262 hair samples from individuals trapped within a time period (2010)(2011)(2012)(2013)(2014), that is short in comparison with the mouflon generation period (4.21 years, Hamel et al., 2016), preventing us from the existence of any temporal genetic structure. Analyzes were performed on individuals of both sexes having a fixed home range (in theory no longer expected to disperse), that is females two or more years old and males four or more years old (e.g., Dubois et al., 1992Dubois et al., , 1993Dupuis et al., 2002).
Genotyping was performed by the Antagene laboratory sults were considered as missing data. Using such reading process, genotyping error rates were estimated to be <2% in our dataset (Queney et al., unpublished data). Individuals were genotyped between one and four times to obtain at least 13 markers with no missing data.

| Genotyping errors
MICROCHECKER v.2.2.3 (Van Oosterhout et al., 2004) was used to detect null alleles. A Correspondence analysis (CA) was performed to identify and exclude outlier individuals. Data were checked for the presence of twin genotypes using the matching option in GenAlex v.6.501 (Peakall & Smouse, 2006, 2012. When detected, one of the two twin genotypes was randomly deleted to prevent bias in subsequent analyzes.

| Genetic structure of the population
We performed sex-specific analyzes (genetic and spatial), as different genetic and socio-spatial structures between the sexes were expected due to (i) their different utilization of space (Marchand et al., 2015(Marchand et al., , 2017, (ii) the segregation of males and females most of the year (Cransac et al., 1998), and (iii) the philopatric behavior of females and much more mobile behavior of males (Dubois et al., 1992(Dubois et al., , 1993Dupuis et al., 2002).
We In complement to the DAPC, we used a sPCA (Spatial Principal Component Analysis, library adegenet, Jombart et al., 2008), to compare the spatially explicit patterns of genetic variability in our population with the socio-spatial structure and spatial locations of founder events. The sPCA takes into account the spatial autocorrelation of allelic frequencies between neighbors (measured by Moran's index, I, Moran, 1948Moran, , 1950

| Socio-spatial structure of the population
We defined the socio-spatial units (groups of individuals living together) present in the population (Garel et al., 2007;Martins et al., 2002) using spatial data and arithmetic centers of traps (see above) to create groups of traps capturing individuals with a shared home range.

| Basic population genetics
Allele frequencies, number of alleles per locus (N a ), allelic richness (A r , calculated using the rarefaction method (El Mousadik & Petit, 1996) to avoid bias due to variable sample sizes among different sociospatial units, and expected heterozygosity (He sensu Nei's gene diversity, Nei, 1973) were determined within each socio-spatial unit for both sexes using FSTAT v.2.9.3.2 software (Goudet, , 2001.

| Measurement of history footprint on the current population
We tested for the occurrence of recent bottlenecks (less than 15 generations ago, founder events) using the approach implemented in Bottleneck v.1.2.02 software (Cornuet & Luikart, 1996;Piry, Luikart, & Cornuet,1999) (within each socio-spatial unit when relevant). When a population has faced a bottleneck, H e is expected to be larger than expected heterozygosity under migration-drift equilibrium (H eq ) and a heterozygosity excess is detectable (Piry et al., 1999). Following the recommendations of Piry et al. (1999) for microsatellite datasets of less than 20 loci, we used the onestep stepwise mutation (SMM) and the two-phase (TPM) models of evolution. The TPM was parameterized with a variance among multiple steps of 12, 95% single step mutations and 5% multistep mutations. We performed 1,000 iterations and used Wilcoxon's signed rank test to assess the occurrence of heterozygote excess.
Assumptions of sampling in a random mating and isolated population were respected (see Results).
All analyzes involving R packages were conducted with R 3.2.1 (R core team, 2016).

| Genetic data
Among the genotyped samples, 237 individuals with fixed home range and known socio-spatial unit had at least 13 successfully genotyped markers. Two outliers and five pairs of twins were identified among these individuals. Data on the 230 remaining individuals (69 males and 161 females) were considered in subsequent analyzes and were assigned to a socio-spatial unit (see below). Comparison of the observed genotypes with the distribution of randomized genotypes generated with the program MICROCHECKER revealed that there were no null alleles in the dataset.

| Genetic clustering
The DAPC revealed three or four genetic clusters for females ( Figure 2; Supplementary material C, Figure S1). In the following, we kept K = 3 because the break in BIC values occurred between two and three clusters and K = 3 was the only clustering solution with nonoverlapping clusters (Figure 2). Cross-validation procedures led us to keep the 10 first components of the PCA in subsequent DAPC. We then retained the two-first discriminant functions of the DA which explained 50.7% of the total variance in the genetic data. Assignation successes within clusters were very high (98%, 96% and 100% for clusters 1, 2, and 3, respectively) and proportions of females assigned to a genetic cluster with membership probability inferior to 0.80 were relatively low (15%, 12%, and 10% for clusters 1, 2, and 3, respectively) meaning that the occurrence of three genetic clusters was well supported by the data.
For males, the optimal number of genetic clusters was K = 2 ( Figure 2; Supplementary material C, Figure S2). We kept the first 10 PCs in the male DAPC and the first discriminant component of the DA which explained 58.3% of the total variance. Assignation success in each cluster reached 1 and membership probabilities were higher than 0.80 for all the males. Therefore, three genetic groups for females and two for males were identified in the population.

| Spatial genetic structure
The sPCA results corroborated those obtained with the DAPC. There

| Socio-spatial structure of the population
Based on the DAPC analyzes (see above), we set the number of sociospatial units in the hierarchical clustering of spatial data to three for females and to two for males. For females, the three socio-spatial units corresponded to groups of traps located in the northern, central and southern parts of the study area (hereafter called "Nf," "Cf," and "Sf," respectively; Figure 4). For males, one socio-spatial unit grouped traps located in the northern ("Nm") and the other in the southern part ("Sm") of the study area ( Figure 4). Traps in the central part of the reserve were principally grouped with traps from the north.
Socio-spatial units of females and males in the population were at HWE as indicated by randomization tests (p = .99 and .19 for females and males, respectively) and confirmed by F is values which were not significantly different from zero (  (Table 3).
Due to significant genetic differentiation between female sociospatial units, genetic bottleneck detection tests were performed within each socio-spatial unit for females, but not for males. Except for the southern socio-spatial unit of females (Sf), we found evidence supporting a recent bottleneck (<15 generations ago, founder events) for both sexes under both models of evolution (Table 4).

| Correspondence between socio-spatial and genetic structures
Although no direct correspondence could be seen between sociospatial units and genetic clusters (Table 5), females from the north were genetically closer to females from the south of the wildlife reserve as evidenced by the sPCA (Figure 3; see above). Indeed, the largest proportion of females from socio-spatial units Nf (56%) and Sf (64%) were grouped in genetic cluster 3 of the DAPC while females from socio-spatial unit Cf were assigned to genetic cluster 1 (41%) or 2 (48%, Table 5

| DISCUSSION
In males Mediterranean mouflon, the less philopatric sex, genetic structure was low compared to females for which the major axis of genetic structuring appeared to be a north-south grouping (DAPC and sPCA).
We also found a low but significant genetic differentiation between female socio-spatial units (pairwise F st ). However, this socio-spatial organization effect was not detected by DAPC and sPCA suggesting that other processes, such as a residual effect of introduction history (e.g., Biebach & Keller, 2009), would be more in play than social factors in determining the genetic structure of females in this population.

| Genetic diversity
As the isolated study population was introduced 60 years ago from 19 individuals, we could expect to observe low genetic diversity.
Surprisingly, heterozygosity was not as low as we could expect due to the founder events and is comparable to the one reported  (Biebach & Keller, 2012;Maudet et al., 2002) and with propagule pressure (repeated introduction events prevent inbreeding) (Hufbauer et al., 2013). As the number and the genotypes of founder individuals determine the number of alleles (N a ) present in the founded population, they have more detrimental effects on the number of alleles than on heterozygosity (Biebach & Keller, 2009, 2012Kekkonen & Brommer, 2015). While N a is mainly dependent on population size, heterozygosity is mainly impacted by the postbottleneck growth rate (Hedrick, 2011). In closed populations, N a can only be restored by mutations; heterozygosity is thus expected to be restored faster and to be easier to save than N a when conditions are favorable. In our population, founders from three different origins (Table 1) were released on four occasions, and growth rate after introduction is thought to have been as high as in other introduced mouflon populations (e.g., Kaeuffer et al., 2007) and most herbivore populations (Forsyth & Caley, 2006). These conditions could thus have favored heterozygosity maintenance at a relatively high level, comparable to unbottlenecked populations, while allelic richness still bears the scars of the founder events. Our results therefore support the theoretical work of Hufbauer et al. (2013) and Biebach and Keller (2012), and confirm that populations founded by multiple introductions of genetically diverse individuals avoid inbreeding effects.

| Influence of socio-spatial structure on genetic structure
We confirmed the previously highlighted low genetic differentiation between the different socio-spatial units of the studied population T A B L E 2 Sample size (n), number of allele (N a ), allelic richness (A r ), observed and expected heterozygosity (H o and H e , respectively) averaged overall loci ± SE, and F is values for the different socio-spatial units of the Mediterranean mouflon population of Caroux-Espinouse massif  (Petit et al., 1997). Global and all pairwise F st values were nonetheless significantly different from zero for females, indicating genetic differentiation between socio-spatial units, while a high degree of admixture between socio-spatial units was observed for males. DAPC results showed a clear separation of genetic clusters which did not match with the socio-spatial units, indicating that the socio-spatial structure was not the main driver of the genetic structure in this population. Therefore, socio-spatial structure contributed to the overall genetic structure of the population but other population processes contributed more (see below).
Differences in philopatric behavior between the sexes may lead to a more pronounced genetic structure in the philopatric sex (Goudet, Perrin, & Waser, 2002;Podgórski et al., 2014;Storz, 1999). As females showed a higher level of genetic structure than males (significant F st , more genetic clusters, and significant spatial genetic structure), our results confirm the philopatric behavior of females (Dubois et al., 1992Martins et al., 2002) and suggest that gene flow is essentially male-biased in this population, as evidenced also in Bighorn sheep (Boyce et al., 1999). Nevertheless, as in other mountain sheep species (Festa-Bianchet, 1991), permanent natal dispersal of Mediterranean mouflon rams seems to be rare in our population . Instead, rams have been shown to make excursions (temporary movement outside an established home range) during the rutting period (Marchand et al., unpublished data), and mating opportunities during these excursions may insure gene flow among the different socio-spatial units. In the same vein, it was shown that while females show a stable movement pattern year round (moving preferentially toward familiar areas), males exhibit a more labile pattern and are more prone to move toward unfamiliar areas during the rutting period (Marchand et al., 2017;Martins et al., 2002). These sex-specific behaviors lead to a higher spatial stability for females than for males. An interesting way to go further in the study of this sex-biased reproductive dispersal would be to investigate mtDNA and Y-chromosome, that is nonrecombining sex-specific markers (Handley & Perrin, 2007).

| Influence of introduction history on genetic structure
Surprisingly, given the spatial fidelity observed in female Mediterranean mouflon (Dubois et al., 1992Martins et al., 2002) and a northsouth spatial disconnection (see Supplementary G), females from the north and the south were genetically closer to each other than to females from the central part of the reserve (see Table 5; Figure  intrapopulation and small geographic scale like in the present study, Simpson et al. (2013) showed that the site of release of founder individuals introduced at the end of the 19th century still has an impact on the current spatial genetic structure of the population of red squirrels.
These results highlight that even after many generations, the genetic signature of source populations can still be present at intrapopulation scale. Moreover, female philopatry can be a factor explaining the persistence of this genetic signature in our population because global genetic structure (males and females considered in same analyzes) corresponds to female structure (results not shown).

| CONCLUSION
Our findings highlighted important elements for wildlife management. First, despite the existence of male-mediated gene flow, we showed that introduction history had long-term consequences on T A B L E 5 Proportion of individuals assigned with a membership probability higher than 0.80 from each socio-spatial unit in each genetic cluster determined in DAPC intrapopulation spatial genetic structure, probably due to the high female philopatry which has favored the persistence of the historical genetic signature of introductions. These results confirm the importance for wildlife managers to account for sex-specific spatial behaviors and gene flow when determining introduction/reinforcement strategies to ensure genetic mixing among introduced individuals from different sources. Indeed, as introductions often involve only a limited number of founders (Kekkonen & Brommer, 2015), genetic diversity may be low and a lack of genetic mixing between released individuals may result in introduction failure due to inbreeding depression and declining immunocompetence (see Armstrong & Seddon, 2008). Second, albeit allelic richness was relatively low, this study did not provide evidence that the Caroux-Espinouse Mediterranean mouflon population shows depauperate genetic diversity. Such a result stresses the importance of choosing the right number, level of admixture (e.g., individuals from different origins) and place of release of individuals translocated when planning a (re)introduction program, to ensure that introduced genetic diversity expands to the entire population. Finally, the present study provided evidence that different levels of genetic structuring can co-occur at fine spatial scale (1,700 ha) as compared to a mouflon home range (201 and 301 ha for females and males, respectively, Marchand, 2013). Such results would not have been obtained if sampling had been performed at a higher spatial resolution, highlighting the importance of designing sampling schemes in accordance with species biology and ecology. Additionally, such small-scale spatial genetic structure suggests that small-scale spatial population dynamic processes can occur in this population (see Coulson et al., 1999). All the results can be helpful for other populations, including endangered Corsican populations (Ovis gmelini musimon var. Corsicana) as they reveal important factors shaping the genetic structure of mouflon populations.

ACKNOWLEDGMENTS
We warmly thank all the professionals from the Office National de

CONFLICT OF INTEREST
All the authors disclose any potential sources of conflict of interest.

DATA ACCESSIBILITY
Microsatellite and spatial data sets will be available from the Dryad Digital Repository.