Genetic and demographic recovery of an isolated population of brown bear Ursus arctos L., 1758

Introduction

In recent centuries, large carnivore populations have been declining worldwide due to human intervention and habitat destruction (Treves & Karanth, 2003), but in the past 40 years, species resilience, species protection, land sharing programmes, and ongoing conservation of wilderness zones has supported partial recovery in areas of Europe and America (Chapron et al., 2014; Gilroy, Ordiz & Bischof, 2015; Gompper, Belant & Kays, 2015). The brown bear Ursus arctos may be a good model for study of the impact of population distribution on the genetic diversity of large mammals (Davison et al., 2011; Karamanlidis et al., 2012; Straka et al., 2012; Taberlet & Bouvet, 1994). Once widespread throughout Europe, most brown bear populations have undergone a reduction in numbers and geographic distribution over the past millennium, particularly since the 15th century, as a result of anthropogenic factors (Taberlet & Bouvet, 1994; Valdiosera et al., 2008).

The recent increase, expansion, and secondary contact processes occurring in some fragmented bear populations may have helped to improve their demographic status. An example of this is the recovery of the brown bear in Finland (Hagen et al., 2015), where the range contraction a century ago produced genetic structuring and led to at least two separate populations. Conservation during the second half of the 20th century, accompanied by immigration from Russia (Kopatz et al., 2014; Kopatz et al., 2012), has resulted in increasing numbers of bears, which dispersed further north and west over time. As a result, the Finnish population increased, and genetic screening has provided evidence of range expansion and gradual disappearance of population substructure along with increasing genetic diversity and admixture. Assignment probabilities of individuals suggested expansion from the southern subpopulation of Finland, which was supported by gradually increasing heterozygosity, allelic richness, and average numbers of alleles in the southern subpopulation (Hagen et al., 2015).

Nevertheless, some populations are so small and fragmented that natural recovery has failed in spite of the costly conservation programmes implemented by governments and NGOs (Woodroffe, 2001). The threshold under which a population is unrecoverable is difficult to assess, depending on a complex mixture of demographic, genetic, ecological, and socio-economic factors that are difficult to quantify and not always well known (Allendorf & Luikart, 2007).

Most brown bear populations assumed to have more than 100 individuals in 1950–1970 are currently recovering (Chapron et al., 2014), but smaller populations that have been isolated and cannot be rescued by large neighbouring populations have faced challenges to recovery or have become extinct. For some of these populations their genetic variability is still unknown. That has been the case with four isolated brown bear populations in Western Europe that survived at least until the 1980s in the Apennines (Italy), the eastern Alps (Italy), the Pyrenees (France and Eastern Spain), and the Cantabrian Mountains (Western Spain) (Chapron et al., 2014). In the Pyrenees and the Alps, bears were in decline during the last decades of the 20th century. When there were single or few bears left, populations were restored by introduction of animals from Slovenia (Clark, Huber & Servheen, 2002; Tosi et al., 2015). In the Apennines, after many decades of protection and conservation programs, 51 bears remain, but the population does not seem to be increasing (Ciucci et al., 2015). Of these four populations, the brown bear of the Cantabrian Mountains is the only isolated population in Western Europe showing a clear trend to natural recovery. This population has been isolated from that geographically nearest, the Pyrenean population, for at least 400 years (Nores & Naves, 1993).

During the first decades of the 20th century, the Cantabrian population split into western and the eastern subpopulations separated by a strip of land of 50–100 km wide with poor habitat quality and an accumulation of structures and roads (García et al., 2007; Nores & Naves, 1993). Bears in the two Cantabrian nuclei declined in number until the mid-1990s. In 1982–1995, the western population comprised 50–60 bears and showed an annual decrease of 4–5% (Wiegand et al., 1998). The eastern subpopulation comprised 20–25 bears in 1990 (Palomero, Fernández & Naves, 1993).

Surveys conducted in the late 1990s and early 2000s found significant genetic differentiation between the western and eastern subpopulations, likely increased by the evolutionary processes of genetic drift and selection since the population split (García-Garitagoitia, Rey & Doadrio, 2006; Pérez et al., 2009; Rey et al., 2000). These works indicated that the eastern Cantabrian subpopulation showed some of the lowest genetic variation among brown bear populations in Europe (Swenson, Taberlet & Bellemain, 2011). As consequence, both subpopulations are considered critically endangered in the Red Book of Spanish mammals (Palomero, 2007). Over the past 50 years, the Cantabrian bears seemed on the path to extinction, similar to the Pyrenean and the Alpine populations. However, this trend has been recently changing. Semi-annual monitoring of the Cantabrian bears, based on the number of females and cubs of the year (COY) (Palomero, 2007), and genetic surveys indicate that both the western and eastern subpopulations have increased since the mid-1990s. Despite some controversy about the reliability of the count of females with COY to determine population trends, the reported annual increase from 1990–2000 was 7.5% and 3.0% for the western and eastern subpopulations, respectively (Palomero, 2007). A genetic census conducted in 2006 estimated Nc is 203 bears in the western subpopulation (CI 95% = 168–260) and Nc is 19 (CI 95% = 12–40) bears in the eastern subpopulation (Pérez et al., 2014).

Connectivity was previously detected between Cantabrian brown bear populations; three males belonging genetically to the western subpopulation were found in the eastern subpopulation, and one male from the eastern subpopulation was found in the western subpopulation (Pérez et al., 2009; Rey et al., 2000). However, only in 2008 were two admixed individuals at the western limits of the eastern subpopulation range identified, indicating genetic flow between subpopulations (Pérez et al., 2010). In the most recent studies of Cantabrian bears (Pérez et al., 2014; Pérez et al., 2010), the majority of genetic samples were collected in 2006, with a few from 2008. The majority of the samples providing reliable genotypes were from the western subpopulation, and little information is available on the eastern subpopulation.

The goal of this study was to assess the demographic and genetic effects of reconnection on the eastern Cantabrian brown bear subpopulation. Our hypothesis was that the eastern subpopulation had experienced population growth and altered genetic composition through movements of individuals and effective genetic transfer of alleles from the western subpopulation. We assessed the eastern subpopulation genetic variation and gene flow, investigated possible movements of individuals from the western to the eastern subpopulation, and evaluated the impact of this migration process on genetic diversity. We employed several methods of determining the level of relatedness among individuals and estimated the effective population size (Ne) of the eastern subpopulation.

Complementary to the genetic data, data concerning females with COY from twenty-six years of field-based monitoring in both subpopulations were used to evaluate demographic changes.

Materials and Methods

Sampling collection and DNA extraction

Non-invasively sampled material from the eastern subpopulation was collected from June 2013–August 2014 in the Cantabrian Mountains (Fig. 1). Similar samples were collected from the western population to compare genetic factors and determine the direction of migration. In total, 152 non-invasive samples including hair (n = 122) and scat (n = 30) were collected. Samples were captured following either systematic (part of the monitoring campaign) or opportunistic (bear-watching and sign surveys or reports of beehive damage from regional rangers) methods, under permission of authorities of the autonomous region of Castile and Leon. The geographic distribution included 116 samples from the eastern subpopulation and 36 from the western (Table S1). Scat samples were dehydrated with silica gel and stored at constant temperature, and hair samples were stored in non-bleached paper envelopes again at constant temperature, until DNA extraction. The research did not involve animal experimentation and complied with international guidelines on ethical behaviour.

DNA was extracted from the hair root using the QIAamp DNA Investigator kit (Qiagen) and from faeces using the QIAamp DNA Stool kit (Qiagen) following manufacturer’s instructions and eluting the DNA in 40 μl of water for hair and 100 μl of water for faeces. The DNA extraction was performed in a room designed for ancient DNA extraction at the Museo Nacional de Ciencias Naturales of Madrid, Spain (MNCN-CSIC), using a tube with no DNA as a negative control for the extraction. No more than 16 samples, including respective negative controls, were processed in one set.

Results

Population monitoring

Monitoring indicated that the Cantabrian bear population has increased steadily from the mid-1990s (Fig. 2). Since the breeding interval of females of this population is normally two years (Palomero et al., 2006), the biennia with minimum and maximum numbers of females with COY found were 1993–1994 and 2013–2014, respectively. In 1993–1994, the number of females with COY was seven in the western and three in the eastern subpopulations. In 2013–2014, 54 and 10 females with COY were recorded in the western and eastern subpopulation, respectively (Fig. 2). Using Poisson regression, the estimated rate of exponential growth from 1994 (when both subpopulations were at the lowest numbers observed in the survey) to 2014 was 10.1% (CI 95%: 7.8–12.4; p < 0.0001) for the western subpopulation and 10.4% (CI 95%: 5.0–16.4; p = 0.0002) for the eastern subpopulation (Fig. 3).

Population structure and assignment of individuals

The number of populations and the assignment of individuals to each population was estimated using Bayesian inference (Pritchard, Stephens & Donnelly, 2000). The analysis showed best population structure, with the highest change in LnP(D), for the model with K = 2, corresponding to the subpopulations analysed. Accordingly, the modal value of ΔK (Evanno, Regnaut & Goudet, 2005) was shown at K = 2 (Fig. 4A). Based on the average proportion of membership (Q) obtained, for the eastern subpopulation, 46% of individuals were assigned to one of the two clusters found (five to the eastern and seven to the western clusters) with a high probability value (Q > 90%), whereas the remaining samples could not be assigned exclusively to a cluster of origin, since Q values ranged from 30–50%. The western subpopulation was clearly defined by a single cluster, with the majority of the samples assigned with high probability (Q > 90%) (Fig. 4B).

Similar levels of genetic differentiation between the two regions were suggested by the significant F_ST estimate (overall F_ST = 0.055) and the principal coordinate analysis (PCoA). Results of the PCoA separated samples into the two main groups. Some individuals from the eastern subpopulation overlapped with the western samples, indicating a degree of continuity between the two regions (Fig. 5). However, the three first principal coordinates of the PCoA explained only 34.5% of the molecular variation of the microsatellite loci used, hence this result must be interpreted with caution. Finally, the results of SAMOVA revealed two high F_CT values for clusters. The highest F_CT value, 78.77%, corresponded to an arrangement of populations in K = 2 clusters. The division into K = 3 clusters showed the second highest variance among groups. Both clearly differentiated the western subpopulation as a separate group.

Discussion

The results reveal a steady increase in the number of females with COY for both brown bear subpopulations on the Cantabrian Mountains. Reported migration of some individuals (Pérez et al., 2010) was also confirmed, with the movement of bears mainly from the larger and more densely populated western region to the eastern one. This has had a direct effect on the genetic composition of Cantabrian bears, which has shifted from two separate and clearly structured populations to subpopulations with some admixture. Although showing less genetic diversity and population growth, the Cantabrian bear populations exhibit some similarities to the Finnish brown bears (Hagen et al., 2015) in terms of expansion, connectivity, and homogeneity. In Finland this process resulted in part from immigration of Russian bears (Kopatz et al., 2014; Kopatz et al., 2012), which comprise a large and diverse population. In contrast, the Cantabrian bears have been isolated for several centuries, and nearly a century ago the decreasing population split into two subpopulations. Conservation measures, along with bear resilience, has contributed to increased populations in the two nuclei, and their connection has been partially re-established, although the bear range has seen only limited expansion.

We have attempted to overcome some of the limitations and challenges of non-invasive sampling, including low quality or degraded DNA, by minimizing the genotyping errors and applying robust statistical analyses (Bellemain et al., 2005; Taberlet & Luikart, 1999). To this end, we increased the minimum number of genotyped loci per individual to 16, a higher threshold than the eight (Rey et al., 2000); (Kruckenhauser et al., 2009), 13 (Ciucci et al., 2015), and 14 (Pérez et al., 2014; Pérez et al., 2009) used in previous studies of brown bear genetics, of which we are aware (Table 4). As a consequence, the P_ID and P_ID-Sib values obtained were low, indicating high discriminatory power. We consider that this difference is the result of an increase in subpopulation size rather than difference in loci used and number of samples analyzed, since the obtained percentage of samples giving a unique genotype was similar among studies.

For the first time, after 26 years of monitoring, we detected changes in the genetic composition in the eastern subpopulation. Previous studies supported the existence of two genetically differentiated subpopulations of brown bear in the Cantabrian Mountains (García-Garitagoitia, Rey & Doadrio, 2006; Pérez et al., 2010; Pérez et al., 2009; Rey et al., 2000), as was found in our study. Hence, the east-west division of the Cantabrian brown bears found previously was confirmed here with the Bayesian assignment analyses of population structure, which defined two clusters corresponding to the subpopulations analysed (García-Garitagoitia, Rey & Doadrio, 2006; Pérez et al., 2010; Pérez et al., 2009; Rey et al., 2000). However, we found larger differences in the genetic composition of the eastern population from that of earlier years, with the presence of individuals showing western population genotypes. In addition we detected a substantial degree of overlap between the two clusters, with a relatively high number of individuals from the eastern subpopulation that could not be unambiguously assigned to their cluster of origin. These individuals were sampled in the eastern region but showed genotypes with ambiguous values of membership coefficient (Q) assignment. Those with low values (Q = 30–50%) were likely the result of admixture, whereas those assigned (Q > 90%) to a different cluster than that of the geographical region in which they were sampled indicate migration between subpopulations, in this case, from the western to the eastern region. Results obtained with F_ST estimates, PCoA and SAMOVA correspond to the genetic pattern identified by STRUCTURE.

Migration between subpopulations was first detected in 1992 based on genotype composition (Rey et al., 2000), when a male with genetic profile of the western subpopulation was identified on the eastern side. Subsequent migration activity was detected in 2004–2006 with west-east movement of three males (Pérez et al., 2010; Pérez et al., 2009). Gene flow between subpopulations was detected in 2008 based on two genetically admixed individuals sampled in the eastern subpopulation (Pérez et al., 2010). Using Bayesian cluster analysis and sex determination, we observed an increased trend in brown bear dispersion and gene flow between subpopulations. Of the 26 unique genotypes detected in the eastern subpopulation, 14 (54%) presented admixture composition (Q = 30–50%) and seven (27%) were determined to be migrants (Q > 90%) from the western subpopulation. The two migrants successfully sexed were male, an insufficient number to determine whether the migration was a sex-mediated process. However, since dispersal in brown bears has been reported to be sex biased, with males leaving the natal area, while young females establish home ranges close to their mothers (McLellan & Hovey, 2001; Støen et al., 2005), it is likely that this is the case.

The effective population size (Ne) obtained in our study represents the number of individuals that effectively contribute to the population (Frankham, 1995). Our results indicate that Ne varied with the linkage disequilibrium (LD) and Bayesian computation methods, from 22.4 for LDNe to 12.1–13.7 for ONeSAMP. Discrepancies between the two methods are not unexpected, because they rely on different assumptions. The LD method is more restrictive because it assumes selective neutrality, unlinked markers, and a single, closed population (Palstra & Ruzzante, 2008). If migration is taking place among subpopulations, a closed population cannot be assumed, and the results from LDNe could be biased. The estimate obtained with ONeSAMP seems to fit better with our demographic data (COY value in 2014 of 10), since in the eastern subpopulation there are at least ten mature females (females with COY from 2013 plus those from 2014), an unknown number of breeding males, and perhaps some mature females that did not reproduce during the most recent two years. Despite these differences, both methods detected an increased Ne compared to previous estimates (2006, Ne = 9, CI 95% = 8–12; (Pérez et al., 2014)). This can be due to immigrant males in the eastern subpopulation and a consistent increase in the number of reproducing females during recent years. However, our results need further confirmation with larger sample sizes and additional years of sampling. In any case, these numbers are far short of the Ne = 50 adults required to avoid the adverse effects of inbreeding, and the Ne = 500 to avoid extinction due to the inability to cope with environmental change (Frankham, Bradshaw & Brook, 2014). Therefore, management of the eastern brown bear subpopulation should concentrate efforts on enhancing population growth.

The apparent increase of the Cantabrian bear population could be due to the reduction in mortality when effective conservation programs were implemented. For example, the 15-year period 1980–1994 saw 36 cases of illegal killing of bears in the western and 18 cases in the eastern region (Naves et al., 1999), while in 1995–2009, only seven and nine cases, respectively, were reported (Palomero et al., 2011). These nine cases included only one adult female; the others were adult males, young bears, or undetermined. Nevertheless, further cryptic mortality cannot be ruled out, but it is clear that mortality has decreased in recent years due to public awareness and law enforcement. When poaching dropped drastically around the mid-1990s, bear numbers began to increase, filling the gaps in the areas of their subpopulations. Apparently, the progressive saturation of the bear range triggered the dispersal of males. Contrary to the pattern described in Scandinavia (Swenson, Sandegren & So-Derberg, 1998), no dispersal of females has been detected in our study area (Pérez et al., 2010; Pérez et al., 2009; Rey et al., 2000). The migration was more common from west to east, from the subpopulation showing a higher rate of increase to the smaller eastern subpopulation, which entered recovery later. In addition to the increased public tolerance of bears, the rural exodus has led to reversion of the land to a state providing greater coverage and development of processes associated with recovery of natural forest stages (Navarro & Pereira, 2012; Ordiz et al., 2013). The transition from grasslands to shrub and to early-growth oak Quercus pyrenaica forests is improving the habitat of the potential corridor between the subpopulations (Naves et al., 2003).

Another factor that may have influenced the direction of the dispersion is conspecific attraction (Stamps, 1988). As rivers run north-south in the Cantabrian range, typical bear movements in past decades were detected in this direction (Clevenger & Purroy, 1991), with bears taking advantage of river corridors to move outside their home area. The connection between the studied subpopulations is hampered by a major highway (AP-66) as well as roads, railways and dams on rivers with a north-south orientation. In recent decades, the permeability of these barriers has not improved; nevertheless, bear presence in the corridor between subpopulations has noticeably increased. In spite of the increased number of bears crossing these barriers, none have been reported killed by traffic, so apparently some have learned to find their way across the passages and tunnels of the highway and through other barriers.

Improved connectivity and increasing population size are presumed to increase the genetic diversity and the long-term viability of populations (Frankham, 1996; Reed et al., 2003). When gene flow is re-established among subpopulations that have been isolated for a long time, spatial population structure decreases, followed by an increase in genetic diversity within subpopulations (Hagen et al., 2015; Ramakrishnan, Musial & Cruzan, 2010). Our results confirmed an increase in genetic flow accompanied by increased genetic diversity. Mean H_O for the eastern subpopulation (0.54) was similar to that reported for other small brown bear populations, such as the H_O of 0.50 in Italy in 2011 (n = 45) (Ciucci et al., 2015). Although direct comparison of results among studies presents limitations due to differences in the number of loci and samples used, etc., when compared with the genetic variation of the same subpopulation over the years, we observed an increasing trend in genetic diversity (Table 4). However, interpretation of the results should be made with caution, and a larger number of individuals from the eastern subpopulations should be included in further analyses.

As shown in other studies (Hagen et al., 2015), the migration of bears from the western to the eastern subpopulation has effected a rapid reduction of population substructure and increasing genetic diversity and admixture. Nevertheless, the isolation of the Cantabrian population as a whole prevents the long distance immigration that usually preserves genetic diversity by reshuffling alleles across the landscape (Bialozyt, Ziegenhagen & Petit, 2006; Frankham, 1996). The Cantabrian population has suffered a gradual contraction during the past five centuries (Nores & Naves, 1993), and slow contraction processes have a more pronounced effect on genetic diversity than do rapid contraction processes and also are less likely to preserve the initial genetic diversity; hence leaving the isolated populations with lower genetic difference (Arenas et al., 2012). Severe inbreeding has produced distinct morphological and physical characteristics in brown bears bred in captivity in zoological gardens (Laikre, 1999) as well as in other large carnivores, for example kinked tail, cowlicks, cryptorchidism, and heart defects in the Florida panther Puma concolor (Culver, 2010). In spite of the long isolation and the small size of the Cantabrian bear population, especially the eastern subpopulation, no morphological characteristics typical of severe inbreeding have been detected, but further studies are needed.

The demographic monitoring carried out for more than 25 years in the Cantabrian population of Ursus arctos has led to increased understanding of changes in a fraction of the bear population. The genetic monitoring programs represent a step forward and could detect demographic and genetic trends and other factors to aid the recovery of this isolated, but seemingly increasing, population.

Supplemental Information