INTRODUCTION

The use of molecular genetics to detect population structure is a well-established, practical alternative to direct methods such as mark-and-recapture or biotelemetry, and it has the added advantage of giving a temporal perspective to gene flow (Slatkin, 1985; Neigel, 1997). Wright's F statistics are used extensively for estimating the levels of gene flow and genetic dissimilarity among predefined patches or populations. However, there is not always much point in using such predetermined definitions of groups, because the genetic structure of a population is not necessarily reflected in the geographical proximity of individuals. Populations that are distributed in a spatially continuous way can be genetically structured along environmental gradients, such as climate or vegetation gradients, that function as cryptic barriers to gene flow (e.g., Zannèse et al., 2006).

Recent advances in the development of landscape genetic tools have enabled us to detect dissimilar individuals within spatially continuous populations. These tools were developed under the framework of Bayesian statistics, especially for approximations of the posterior probability of assigning individuals to their groups of origin; the updating of plural parameters is estimated in a deterministic order (Pritchard et al., 2000; Falush et al., 2003; Guillot et al., 2005). In addition to being used to find the number of genetically distinctive clusters, these approaches are particularly useful for mapping the borders of clusters and detecting migrants; they could therefore be powerful tools for monitoring the illegal translocation of nonnative individuals (Frantz et al., 2006) and for inferring the level of admixture in a genome (Corander and Marttinen, 2006).

Studies of genetic structure based on mitochondrial D-loop variation have demonstrated that the Boso population of the sika deer (Cervus nippon) consists of two major haplotypes (Nagata et al., 1999; Yoshio et al., 2008). Marked heterogeneity has been detected in the distribution of these haplotypes, and spatial discontinuity in the constitution of the haplotypes has been detected at the edge of a local motorway, Route 81; for details see Yoshio et al. (2008). Because Route 81 on the Boso peninsula passes along a steep-walled ravine, the natural landscape here may have functioned as a cryptic barrier to gene flow (Yoshio et al., 2008). Because the resolution of mtDNA polymorphism analysis is generally lower than that of other polymorphic marker analyses, such as AFLP or microsatellite analyses, more details of the genetic structure can be detected by using such other markers.

Our primary purpose was to dissect the genetic discontinuity by using microsatellites, which are codominant nuclear markers with higher resolution than mtDNA. If potential configurations or other environmental factors suggested by the analysis of mitochondrial D-loop variation were to impede migration and promote the genetic differentiation of underlying groups, then we expected that a similar pattern of genetic discontinuity would be detected by using other molecular markers. We conducted a spatially explicit Bayesian analysis using nine microsatellite markers to delineate the heterogeneity of the genetic structure of the Boso population.

MATERIALS AND METHODS

Amplification and genotyping of microsatellite DNA

Nine microsatellite loci (BOVIRBP, OarFCB193, BL42, BM203, BM888, BMC1009, Cervid14, ETH225, and BM4107) were used for genotyping (Table 1). These loci, apart from Cervid14, which is derived from the white-tailed deer (Odocoileus virginianus; DeWoody et al., 1995), are derived from cattle or sheep (Bishop et al., 1994; Crawford et al., 1995) and are mutually unlinked on a genetic map of red deer (Cervus elaphus), which is closely related to the sika deer (Slate et al., 2002). All of these loci have been revealed as polymorphic markers in the Kinkazan Island population of the sika deer (Okada and Tamate, 2000; Tamate et al., 2000).PCR was carried out in reaction volumes of 25 µl containing 2.5 µl of 10×NH₄ buffer (Bioline, London, UK), 2.5 mM MgCl₂, 0.2 mM each dNTP, 0.4 µM fluorescently labeled forward primer, 0.4 µM reverse primer (see Table 1), and 1.25 units of BIOTAQ DNA polymerase (Bioline). PCR cycling conditions consisted of an initial denaturation at 94°C for 1 min, followed by 30 cycles of denaturation at 94°C for 30 s, annealing at an appropriate temperature (see Table 1) for 30 s, and extension at 72°C for 30 s. After PCR, 0.1 µl of the product for each locus was mixed with 10 µl of deionized formamide and run on an ABI 3730 capillary sequencer (Applied Biosystems, Foster, CA, USA) with 0.2 µl of GeneScan-500 LIZ size standard (Applied Biosystems). Genotyping was performed with GeneMapper software, version 3.5 (Applied Biosystems).

Table 1.

The nine microsatellite loci used; annealing temperatures for amplification; allele sizes (in base pairs); numbers of alleles observed; observed and expected heterozygosities (H _o and H_e, respectively); F_IS (coefficient of inbreeding) values, and deviation from Hardy-Weinberg equilibrium (P).

^a Significance levels for P value after Bonferroni correction: ^*, α<0.05; ^**, α<0.01; ^***, α<0.001.

RESULTS

Spatial genetic structure of the Boso population

Five replicated runs using STRUCTURE showed that the mean posterior probability [LnP(D)] was highest when K=2 (Fig. 1). We found that the second-order polynomial fitted better (AIC=195 for the linear regression and AIC=169 for the second-order polynomial) and that the posterior probability was ∼1 when K=2. Along these lines of evidence, we concluded that K=2 was the best K. We then estimated the genetic population structure by using GENELAND (Fig. 2). The boundary estimated between identified clusters was situated at 140.11°E±0.0221° (median±SE) longitude; no distinctive latitudinal boundary was detected. The position of the boundary detected between the two clusters was fairly consistent, irrespective of the different priors used in the D-model and the F-model (data not shown). The distribution boundary of the two clusters is surrounded by Route 81 and two dirt forest roads (the Nishimine and Matsumori routes). Six of the nine loci contained private alleles, and most of these were found in individuals in the eastern area (see Fig. 2). Weir and Cockerham's θ, an estimator of genetic differentiation between the western and eastern individuals that was identified from the discontinuity line described above, was 0.017, deviated significantly from zero (P<0.001, 5×10⁴ permutations), implying that a distinguishing genetic differentiation existed in the population.

Fig. 1.

Plot of estimated logarithms of probability of data [LnP(D)] against number of clusters tested (K).

DISCUSSION

All nine microsatellite loci tested were polymorphic in this isolated sika deer population on the Boso Peninsula, central Japan. Significant heterozygote excess and deficit were found at a single locus (OarFCB193) and at two loci (BM203 and BM888), respectively (Table 1). The deficits in heterozygosity might have been due to a bottleneck effect and subsequent inbreeding resulting from a severe reduction in population size in the past (Koganezawa et al., 1976; Chiba Prefecture and Deer Research Group on Boso, 1994).

There is serious concern worldwide about the ecological impact of an overabundance of deer in relation to direct economic losses in agriculture and forestry and the transmission of several animal and human diseases (Côté et al., 2004). From an ecological viewpoint, deer overabundance has indirect negative effects on the whole community structure and nutrient cycle (Rooney and Waller, 2003). On the Boso Peninsula, intensive studies of the impacts of overabundance of the sika deer have revealed a marked increase in damage to the richness of plant and arthropod communities with increasing population density of the deer (Miyashita et al., 2004; Suzuki et al., 2008). In the 19th century, sika deer were distributed throughout the Boso Peninsula (Chiba Prefecture and Deer Research Group on Boso, 2004). A severe population reduction was observed around the 1940s, probably as a result of human settlement of the foothills and indiscriminate hunting (Chiba Prefecture and Deer Research Group on Boso, 2004). In 1973–74, the lineage population became very small and was restricted to one area in the southern part of the peninsula (Koganezawa et al., 1976). In addition to the relaxation of hunting pressure when protection of the deer was instigated, extensive deforestation for conifer planting, which provided new feeding sites, was probably the impetus for the population resurgence that began in the late 1980s. This range expansion of the deer population has continued up to the present. The estimated number of deer had increased to around 3000 by 2001, and the distribution area changed from an estimated 40 km² in 1973–74 to about 440 km² in 2001 (Chiba Prefecture and Deer Research Group on Boso, 2004). Our principal purpose in this study was to examine whether a genetic population structure was detectable in this spatially continuous sika deer population on the Boso Peninsula, and, if so, to compare the boundary positions of clusters as inferred by both mtDNA and microsatellite analyses.

Fig. 2.

GENELAND assignment of individuals. Individuals assigned to the western cluster are represented by orange triangles and those assigned to the eastern cluster are represented by yellow circles. Golf courses are shown by green round-edged squares, arrows point at Route 81 and dirt roads (see text), and roads broader than 5 m are shown by red lines. Genetic discontinuity was estimated from local polynomial fitting (LOESS) indicated by the blue dotted line (the thin blue dotted lines indicate±SE). The vertical axis on the graph indicates dummy variables (“0” was assigned to individuals with yellow circles and “1” was assigned to individuals with orange triangles).

We found two meaningful clusters, in the western and eastern areas flanking Route 81. In a study of mitochondrial D-loop variation, spatial genetic discontinuity between two main haplotypes was detected in the area surrounding Route 81 (Yoshio et al., 2008); the results of this mtDNA analysis were in agreement with those of our present study, although the mtDNA analysis also suggested that a network of artificial features (two golf courses and two dams) might have contributed to the restriction of deer dispersal. Although further details of the genetic structure were not revealed by using the microsatellites, such incongruence in the results obtained by using the two different types of markers is not extraordinary, because mtDNA is generally inherited from the dam while microsatellite polymorphism arises from both the sire and the dam (e.g., Elmer et al., 2007; Ronka et al., 2008). The finding of an approximate concordance in the spatial patterns of genetic discontinuity, despite the use of disparate genetic markers, is intriguing in terms of dispersal patterns of males and females.

The sika deer might have been fragmentarily distributed in the southern area in the 1970s, because at this time field signs were found only on the margins of this area (Koganezawa et al., 1976). One conceivable explanation is that the present sika deer population on the Boso Peninsula consists of genetically different lineages, and that spatially discontinuous clusters have come into contact in the vicinity of Route 81, where a steep-walled ravine exists. Not Route 81 itself (because of its low volume of traffic), but two rivers and three ridges encompassing Route 81, are also potential candidate geographic barriers to gene flow. Recent studies of red deer have revealed that some landscape features, such as roads and slopes, can impede migration on a local spatial scale, supporting the above view (Pérez-Espona et al., 2008).

Because the mountains on the Boso Peninsula are relatively low and thus probably do not hamper migration, other potential factors, such as the distribution of suitable foods (Takada et al., 2002), are worth considering as facilitators of population structuring. In terms of plant communities, the landscape of the Boso Peninsula is a mixture of fine-scale elements, including broad-leaved forests, Japanese cedar forests, and agricultural fields (Miyashita et al., 2007). In fact, the forest edges distributed sporadically in the study area are generally known to act as large sources of food for the deer (e.g., Sinclair, 1997); furthermore, a wide area of agricultural fields in the southern part of the distribution, to the west of Route 81, supplies the deer with high-quality food (Takada et al., 2002; Miyashita et al., 2007). These spatially biased distributions of food tend to affect migration distance (Kareiva and Odell, 1987; Dwyer and Morris, 2006); this variation in migration distance would, in turn, promote the formation of a cryptic genetic structure in the population.

It is also worth mentioning the effect of social behavior in the sika deer. In general, sika deer are potentially capable of dispersing great distances. According to knowledge of the white-tailed deer and roe deer, for example, dominant females occupy a home range that is at the center of the social group, with younger individuals occupying home ranges that overlap and extend the periphery (Porter et al., 1991). Populations expand as an array of family units composed of females and showing strong philopatry (“the petals of a rose” metaphor; see Porter et al., 1991), while males tend to disperse further than females (Purdue et al., 2000; Wang and Schreiber, 2001). It remains an open question whether such social behavior also facilitates the formation of a global spatial genetic structure; such a characteristic of dispersal could be a factor in the acceleration of genetic structuring on a restricted spatial scale (e.g., Mathews and Porter, 1993; Purdue et al., 2000). In a future study we will need to examine the extent to which the factors mentioned above have contributed to the organization of spatial genetic structure in the Boso population.