Male-biased dispersal causes intersexual differences in the subpopulation structure of the gray-sided vole.

The genetic structure of gray-sided voles was investigated at a spatial scale of 2 km using mtDNA sequences. The control region (674bp) of 162 voles was sequenced and 18 haplotypes were identified. Within 0.5-ha trapping plots (n = 8), the number of haplotypes and gene diversity was significantly greater in males than in females. The fixation index among plots for females (F GP = 0.241) was 3 times as large as that for males (0.075), implying male-biased dispersal. A simulation analysis showed that the observed genetic structure in males could be generated by modifying the observed haplotype distribution of females by adding the effects of local male dispersal. Half of the pairwise F GP (15/28) showed significant differentiation in females, whereas almost none (1/28) were significant in males. Isolation by distance was observed in females, whereas no clear spatial pattern was observed in males. Most pairwise F GP for females were not significant in the short- and intermediate-distance classes (≤1.0 km) as with those for males, whereas all showed significant differentiation in the long-distance class (>1.0 km) for females, but not for males. These findings indicate that the extent of subpopulations within which individuals interact differs between sexes.

Dispersal is a key to identifying a population or subpopulation (Waples and Gaggiotti 2006). When considering subpopulations linked by individual dispersal, 2 extreme cases are assumed. When the subpopulations are isolated completely, they are considered different independent populations. Conversely, when mating occurs randomly within the entire population with frequent dispersal among subpopulations, the population is panmictic and the subpopulations are arbitrary. In reality, most situations are intermediate between these 2 extremes.
Sex-biased dispersal might result in different gene frequency distributions between sexes within and among populations or subpopulations. Recently, many studies have reported intersexual differences in spatial genetic structure at fine scales when animals show sex-biased dispersal (e.g., Banks and Peakall 2012;Busch et al. 2009;Peakall et al. 2003;Temple et al. 2006). Because a "fine scale" can be defined as the spatial scale within which individuals have an opportunity to interact with each other like an ecological population (Waples and Gaggiotti 2006), studies at fine scales may reveal more details about the biological mechanisms underlying the observed pattern of spatial structure. However, generalizations to larger scales might not be simple. Larger spatial scales will usually incorporate greater abiotic and biotic heterogeneity than small scales, including greater genetic or phenotypic variation among individuals. Consequently, simple extrapolations that assume linear scaling relationships will not enable good predictions (Underwood et al. 2005) and results obtained at a fine scale are not easily transferred to larger scales (Holderegger et al. 2006). Therefore, intersexual differences in genetic structure observed at a fine scale might not always be true at larger scales. It is essential to know the limitations of the underlying mechanisms that give rise to spatial genetic structure at a fine scale in order to understand population or subpopulation structure at larger scales.
Because mitochondrial DNA (mtDNA) molecules are generally inherited maternally in animals (Avise 2004), the spatial genetic structure revealed by mtDNA sequence variation is a combined product of maternal lineage structure accumulated for multiple generations (i.e., a historical timescale) and individual dispersal for a single generation (i.e., an ecological time-scale). Intersexual differences in the spatial genetic structure of mtDNA would be generated by sexual Ishibashi et al. • Different Subpopulation Structure between the Sexes differences in dispersal behavior in one generation (Escorza-Treviño and Dizon 2000;O'Corry-Crowe et al. 1997). Recently, Cooper et al. (2010) developed an analytical framework based on intra-class genetic correlations to estimate dispersal based on mtDNA and showed that it is possible to use mtDNA alone to infer the relative dispersal of both sexes.
Extending the approaches of Cooper et al. (2010), we examined the effect of male-biased dispersal on the spatial genetic structure of mtDNA at various spatial scales (0.1-1.9 km) in the gray-sided vole, Myodes rufocanus. The gray-sided vole is suitable because 1) it clearly shows male-biased dispersal and strong female philopatry (Ishibashi and Saitoh 2008a;Saitoh 1995), providing clusters of close relatives in females, but not in males at the fine scale (Ishibashi et al. 1997(Ishibashi et al. , 1998, 2) because its dispersal ability is limited to a few hundred meters at most, the effect of individual dispersal can be covered on a relatively small spatial scale, and 3) considerable information on dispersal behavior has already been obtained (see "Study animal" below). Therefore, we predict that the genetic diversity of mtDNA would be lower for females than for males at a fine scale, because female relatives are usually located in close proximity, whereas males do not show such kin clusters. This sexual difference in dispersal patterns might also give rise to some genetic features at larger scales. Isolation by distance (IBD) is determined by the balance between gene flow (dispersal) and genetic drift (Hutchison and Templeton 1999). Because the effects of dispersal in males might be larger than those in females, the IBD patterns would differ between sexes.

Study Animal
The gray-sided vole is a small rodent with a body weight of ~50 g (Ohdachi et al. 2009). Its mating system is promiscuous (Ishibashi and Saitoh 2008b). Females produce several litters during a breeding season (Ishibashi and Saitoh 2008b;Saitoh 1990). The mean home-range lengths (i.e., the farthest trapping points in an individual's home range) of breeding females and males are 16.3 and 32.7 m, respectively (Ishibashi and Saitoh 2008a). Males disperse further than females from their birthplace. In a 3-ha outdoor enclosure, approximately 70% of the females began to breed within 2 home-range lengths of their natal sites (<33 m), whereas most breeding males (>80%) established home ranges farther than 33 m from their natal sites (Ishibashi and Saitoh 2008a). The average distances of natal dispersal were 64.9 ± 51.1 m (mean ± standard deviation[SD]) and 35.3 ± 45.6 m for males and females, respectively, in a 2-ha enclosure, whereas long-distance dispersal (≥200 m) was observed for 3 males (2.7%) and for 1 female (1.2%) (Saitoh 1995). Males do not move much once they have reproduced successfully (Ishibashi and Saitoh 2008a).

Study Site and Sampling Scheme
We collected vole tissue samples (toe tips) without sacrificing them in 8 trapping plots (I-VIII) in a deciduous broad-leaved forest consisting mainly of Japanese emperor oak (Quercus dentate) at Ishikari, Hokkaido, Japan following published guidelines (Committee on Scientific Names and Specimens 2009) with permission from the Hokkaido Prefectural Government (Shiribeshi #46). If samples are collected before natal dispersal occurs, no genetic differentiation is found between sexes (Prugnolle and de Meeus 2002;Vitalis 2002). Consequently, animals were trapped during the late breeding season (30 September-2 October 2009 for plots I-IV and 6-8 October 2009 for plots V-VIII). These trapping plots were set in a straight line separated from each other by distances of 0.1-1.9 km. Landscape features such as elevation, forest cover, and roads can influence animal movement and gene flow (e.g., Gauffre et al. 2008;Short Bull et al. 2011). However, the focal forest was relatively flat. A dwarf bamboo species (Sasa senanensis) grew homogeneously as the groundcover; this is the typical habitat of the gray-sided vole in Japan. Because there were no roads or waterways inside the study forest, the movement of voles among the survey plots did not appear limited. Therefore, our study scale might not have extended beyond a single landscape (Holderegger et al. 2006) and only geographic distance was a limiting factor in vole movement. In each trapping plot, 50 live-traps were set for 3 days in a 5 × 10 array at around 10-m intervals (~0.5 ha). Upon a first capture, each individual was sexed, weighed, and the toe tips clipped for identification (2 toes per individual). The clipped toes were immersed immediately in 99.5% ethanol in the field.
Each study plot covered the home ranges of 15-32 individuals (see below) that might have interacted with each other. Consequently, the genetic structure within plots may be defined as being of a fine scale. Here, we used intraclass correlations for pairs of genes among groups (F GP ) as the fixation index (Cooper et al. 2010;Rousset 2007). Because positive F GP estimates mean that the identity probability of mtDNA haplotypes within groups is larger than that among groups (by definition; Cooper et al. 2010;Rousset 2007), significant differentiation among groups indicates that individuals with the same mtDNA haplotype are observed more frequently within groups. Then, the extent of subpopulations would be shown by the border between nonsignificant and significant differentiations, when pairwise F GP estimates between plots are plotted with the geographic distance. Cooper et al. (2010) originally applied this analytical framework to the collared peccary, which forms social groups. Group members can be defined easily in the collard peccary. In contrast, the gray-sided vole does not form apparent groups and so some transitory individuals might be included in a sample from a trapping plot. In previous studies (Ishibashi and Saitoh 2008a;Ishibashi et al. 1998), we observed that gray-sided voles, especially adults, rarely dispersed at the end of the breeding season. Here, we captured voles in the trapping plots late in the breeding season and evaluated mainly adults (see below). Therefore, the examined voles likely consisted of those with a fixed home range and a few transitory individuals.

DNA Analyses
DNA was extracted from one of the toes using a DNeasy Blood and Tissue Kit (QIAGEN). Partial nucleotide sequences (674 bp) of the mtDNA control region were determined as described previously (de Guia et al. 2007), with some modification. Gene diversity (haplotype diversity), nucleotide diversity, and Tajima's D values (Tajima 1989) were calculated using Arlequin 3.5.1.2 (Excoffier and Lischer 2010). For each sex, pairwise F ST estimates (Weir and Cockerham 1984) and their P values (10 000 permutations) were also calculated between study plots using Arlequin.

F-Statistics as Functions of Identity Probabilities
After Cooper et al. (2010), the intra-class correlations for pairs of genes among trapping plots, F GP , were estimated using identity probabilities (Rousset 2007); for each sex, we obtained the intra-class correlations, F GP , and calculated the 95% confidence intervals (CI) and P values for the observed F GP estimates using a bootstrapping procedure and a random re-assigning technique, respectively (25 000 times each) using Mathematica ver. 8.0.4.0 (Wolfram Research).
As described in Cooper et al. (2010), with the resampling scheme, we tested whether the estimated fixation indices (F GP ) for each sex departed significantly from the null hypothesis that dispersal is independent of the sex of individuals. For each sex, 25 000 randomized datasets were generated by re-assigning the sex of each haplotype randomly within each plot. By doing so, the number of individuals from each sex was kept constant within each plot. The probabilities of identities between pairs of genes for each resampled dataset were calculated, and the distribution of sex-specific F GP estimates was obtained under the null hypothesis that dispersal ability is independent of sex. The resampling tests were also performed with Mathematica.

Simulation
Furthermore, we assessed the instantaneous effects of male dispersal on the spatial genetic structure by comparing the observed F GP estimate for males with a simulated distribution of F GP values obtained by modifying the mtDNA haplotype distribution of females by adding the effects of male dispersal. For the simulation, we estimated the numbers of emigrated and immigrated males per study plot under 3 assumptions based on the ecology of gray-sided voles: (1) equal numbers of young of both sexes are born per litter; (2) the survival rate of the sexes is equal at the natal sites; and (3) all females stay at their natal plots. Under these assumptions, we compared the observed numbers of males and females for each mtDNA haplotype within plots. For a haplotype, when there were fewer males than females in a plot, the number obtained by subtracting the observed number of males from that of females was thought to have emigrated from the focal plot. If there were more males for a shared haplotype, the excess males were considered immigrants into the plot. This is because the same number of voles should be found for both sexes if males did not disperse. When males had mtDNA haplotypes that were not observed in females in a plot, those males were also considered immigrants into that plot. Based on the observed frequencies of mtDNA haplotypes, we estimated that per plot on average 5 males emigrated and 4 males immigrated.
In the simulation, mtDNA haplotypes of males were generated for every plot from the observed set of female mtDNA haplotypes by randomly eliminating 5 haplotypes (equivalent to the estimated number of emigrants) from the observed set and by adding 4 haplotypes (equivalent to the estimated number of immigrants) to the observed set; the added haplotypes were selected randomly from a list of those allowing repeated selection. The list of haplotypes constituted the mtDNA haplotypes observed in the focal plot and the neighboring 2 plots, except for the plots at both ends of the transect: plot I used the haplotypes observed in plots I and II and plot VIII used those in VII and VIII. For each trial, a fixation index among plots (F GP ) was calculated. After 25 000 iterations, the observed F GP estimate of males was compared with the distribution of F GP values obtained by the simulation. The simulation was performed with Mathematica.

Detection of Isolation by Distance
As an extension of the approach presented by Cooper et al. (2010), pairwise F GP estimates between plots were calculated using Mathematica for both sexes to reveal the effects of male-biased dispersal on genetic structure. A P value for each estimate was obtained by re-assigning haplotypes randomly among all individuals of the focal plots (25 000 randomizations). Owing to multiple comparisons, the significance of the obtained P values for all pairwise combinations was judged using the sequential Bonferroni technique (Rice 1989). Then, to search for the correlation between pairwise F GP and geographic distance between plots, Mantel tests (Mantel 1967) were performed using GenAlEx 6.5 Smouse 2006, 2012). The significance value for the Mantel tests was determined based on 9 999 permutations. Other statistical analyses were performed using JMP ver. 9.0.1 (SAS Institute). The author (YI) will provide the Mathematica programs used herein on request.
In the 162 sequences, there were 29 variable sites and no indels. The nucleotide diversity across all sequences was 0.00746 ± 0.00403 (SD). Based on the variable sites, 18 haplotypes were defined (ISK01-18). Five to 10 haplotypes were observed in each plot (Table 1). Tajima's D was positive or negative at each trapping plot and no D value differed from zero (P > 0.20 for all plots). For the pooled data, D was also not different from zero (D = −0.05, P = 0.533). Hence, there was no evidence for either population fusion or any selective pressures, suggesting that the sequences evolved neutrally.
The number of haplotypes observed within plots was correlated with the number of samples in males (Spearman's rank correlation coefficient ρ = 0.752, P = 0.031), but not in females (ρ = −0.344, P = 0.404). Within plots, the number of haplotypes observed in females was significantly lower than that in males (Wilcoxon matched-pairs signed-rank test, S = 11.50, P = 0.0391, one-tailed). The gene diversity varied among plots, ranging from 0.63 to 0.90 (Table 1). The gene diversity in females was also significantly lower than that in males in each plot (S = 18.00, P = 0.0039, one-tailed). These data support our prediction that the genetic diversity of mtDNA would be lower for females than for males at a fine scale.
When males and females were analyzed separately, a contrasting genetic structure was revealed. The F GP estimate for females was significantly higher than that for males, and the value for females was 3 times as large as that for males (females, F GP = 0.241, 95%CI = 0.194-0.266, P < 0.0001; males, 0.075, 0.057-0.088, P = 0.002). Under the null hypothesis that dispersal is not sex biased, we expect the observed F GP of males and females not to depart significantly from the null distribution. For males, however, there was a significantly larger proportion of randomized data with a larger F GP than observed (P = 0.992; Figure 1). In contrast, for females, only a small proportion of randomized datasets gave a F GP larger than observed (P = 0.021). These results support male-biased dispersal from natal sites and strong female philopatry in the gray-sided vole.
The simulation analysis, in which the mtDNA haplotype frequencies of males were generated from the observed set of female mtDNA haplotypes by adding the effects of local male dispersal, demonstrated that the observed F GP for males was not different from the simulated values (Figure 2). The observed F GP value for males was within the 95% CI of the simulated values. Consequently, sex-biased dispersal behavior of individuals plays an important role in forming spatial genetic structure at the present spatial scale.
In females, the pairwise F GP estimates (range 0-0.369) showed significant differentiation in half of the plot combinations (15/28). The F GP estimate between plots was significantly correlated with geographic distance, showing isolation by distance (Mantel test, P = 0.027; Figure 3). In males, however, F GP between plots was small (range 0-0.211), almost all (27/28) were not significant, and no correlation was observed between geographic distance and F GP (P = 0.245). Becasue the dispersal ability of the gray-sided vole is limited to several hundred meters at most (Ishibashi and Saitoh 2008a;Saitoh 1995), the 2-km spatial scale of this study would include several demes or subpopulations beyond a fine-scale structure Table 1 Distribution of the 18 mtDNA haplotypes (ISK01-18) in the 8 trapping plots (I-VIII; ~0.5 ha each) at Ishikari, Hokkaido, Japan. a Values are multiplied by 1 000. The haplotype sequences were deposited in DDBJ under accession nos. AB720952-AB720968, except for haplotype ISK07, which had the same sequence as those deposited previously (AB259887 and AB259888) Haplotype based on the home ranges of several individuals. Genetic differentiation should be significant among subpopulations and those subpopulations might be located remotely. These features could be detected as an isolation-by-distance (IBD) pattern. Becasue IBD is determined by the balance between gene flow and genetic drift (Hutchison and Templeton 1999), the IBD patterns are predicted to differ between sexes in the gray-sided vole. Supporting this prediction, the genetic differentiation between plots was positively correlated with geographic distance in females, whereas no such a correlation was observed in males. When the geographic distances among plots were categorized into 3 classes; that is, short (<0.5 km, n = 10), intermediate (0.5-1.0 km, n = 7), and long (>1.0 km, n = 11), the F GP estimates for the long-distance class showed significant differentiation with large values for all combinations for females (11/11; range 0.188-0.346), whereas no significant values were observed for males (0/11; range 0-0.211; see Figure 3). These proportions of genetic differentiation differed significantly between the sexes (Fisher's Exact test, P < 0.0001). Conversely, for the short-and intermediate-distance classes, the results were not contrasting between the sexes: only 2 of 10 combinations (range 0-0.369) showed significant differentiation for females and 1 of 10 (range 0-0.199) for males for the short-distance class (P = 1.00). Two of 7 (range   Relationship between the natural logarithm of geographic distance (GGD) and pairwise F GP estimate between study plots for each sex. A significant correlation was observed for females (Mantel test, r = 0.378, P = 0.027), but not for males (r = 0.094, P = 0.245). Significant and nonsignificant F GP estimates are indicated by closed and open circles, respectively, with the significance of P values judged using the sequential Bonferroni technique (Rice 1989). Squares, hexagons, and circles indicate short-, intermediate-, and long-distance classes, respectively (see text). Negative estimates of F GP are shown here as zero. 0.097-0.350) differed significantly for females and none (range 0-0.156) for males for the intermediate-distance class (P = 0.462). Similar patterns were observed in the pairwise F ST estimates (data not shown).
These results provide a clue to the extent of subpopulations, indicating that the spatial scale of subpopulations within which individuals can interact with each other differs between sexes and was up to 1 km for females at a maximum, while it reached 2 km or more for males. Becasue the spatial genetic structure of mtDNA includes effects of matrilineal structure accumulated over multiple generations, the spatial scale within which females actually interact with each other might be smaller than the present estimation. Our 2-km study range might have included a few subpopulations for females.
The spatial genetic structure of males was considerably different from that of females (Figure 3). No significant differentiation was observed among the 8 plots, except for one combination of plots. Therefore, for males, a subpopulation extended at least 1.9 km. These results seem to contradict the tenet that the dispersal ability of individual voles is limited to a few hundred meters. The individual dispersal distance was so short that individual males cannot directly homogenize the genetic content among the 8 plots, which were separated by a maximum of 1.9 km. Our simulation analysis, however, indicated that the observed genetic structure for males could be generated from the observed mtDNA haplotype distribution of females by adding the effects of local male dispersal from neighboring plots (Figure 2). The spatial genetic structure of mtDNA is a combined product of the matrilineal structure accumulated for multiple generations and individual dispersal for a single generation. The scale of the female subpopulations was 1 km. Mitochondrial DNA haplotypes in a subpopulation might be exchanged through male dispersal to neighboring subpopulations and novel haplotypes might be provided to each subpopulation. This mechanism indicates that the mtDNA structure is easily homogenized through the local dispersal of males (assuming 500 m) for up to 2 km beyond the subpopulations of females. If this mechanism occurs among all subpopulations, homogenization extending to more than 2 km could be achieved through the local dispersal of males.
The gray-sided vole does not form apparent groups, although closely related females are usually located in proximity to each other (Ishibashi and Saitoh 2008a;Ishibashi et al. 1997Ishibashi et al. , 1998. Cooper et al. (2010) originally applied this analytical framework to the collared peccary, which forms clear social groups. The present study demonstrates that their approaches are applicable not only to group-forming species, but also to solitary species. It is easier to analyze genetic variation in mtDNA sequences than in nuclear DNA sequences, such as microsatellite DNA, because "universal primers" are available to amplify and determine the sequence of the mtDNA control region in animals (Kocher et al. 1989). This analytical framework using mtDNA can be applied to quantify sex-biased dispersal, which is known to occur widely in animal populations.
This study investigated the spatial genetic structure of the gray-sided vole in a homogenous habitat at a scale of 2 km. In an area greater than this spatial scale, individuals must move across several heterogeneous habitats; that is, landscapes. In the gray-sided vole, such movement is likely rare for both sexes. Furthermore, other factors might be more influential than the sex-biased dispersal of individuals, such as landscape geometry, habitat quality, predation risks, and social conditions (Le Galliard et al. 2012). Consequently, for mtDNA, at larger spatial scales than studied here, the sexual difference in genetic structure would become unclear and significant genetic differentiation among sites would be detected for both sexes. It will be interesting to explore the extent of the influence of dispersal behavior on genetic structure while considering habitat heterogeneity and other ecological factors.

Funding
Japan Society for the Promotion of Science Grant-in-Aid for Scientific Research (22370006).