Meinke.indd 75Meinke et al.: Polar Research 20(1), 75–83 The Arctic fox (Alopex lagopus) is a small mammal weighing 2-5 kg, with circumpolar dis- tribution. It has a home range ranging from 4 - 100 km2 (Frafjord & Prestrud 1992; Anthony 1997) but several ear-tagging and radio-tracking studies have revealed that the Arctic fox is capable of undertaking long-distance movements (Eber- hardt & Hanson 1978; Eberhardt et al. 1983; Fraf- jord & Prestrud 1992; Russian papers referenced by Fay & Rausch 1992). The longest movement recorded was 2300 km (Eberhardt, unpubl. data referenced by Garrott & Eberhardt 1987). The Arctic fox inhabits the entire coastal zone of Greenland and is found in two interfertile colour morphs: a blue and a white. The colour of a fox is determined by the alleles at one locus, where “the blue allele” is dominant over the reces- sive white one (Adalsteinsson et al. 1987). Based on observations and hunting statistics, Hersteins- son (1989) found a positive correlation between the length of time the ground was snow-covered and the proportion of white foxes in Icelandic populations. This situation had previously been noticed in Greenland by Braestrup (1941), who described the blue and the white fox as two sub- species: the white one subsisting on lemmings Genetic differentiation of populations of Greenlandic Arctic fox Pia G. Meinke, Christian M. O. Kapel & Peter Arctander Most microsatellites are very polymorphic. This makes them powerful markers for observing genetic differentiation between closely related pop- ulations. The population structure of the Greenlandic Arctic fox (Alopex lagopus) was studied genetically by analysing six polymorphic microsat- ellite loci of 75 foxes from four populations in different parts of Green- land. Genotypes were determined at the six loci for most of the indi- viduals. Population differentiation was quantified in three different ways both within the total population and pairwise between all populations. The tests were Fisher’s exact test, Rho estimates and Fst estimates, all of which supported a highly significant subdivision of the total population, and they showed significant differentiation in allele frequencies between all pairs of localities. It is concluded that the known long-distance migra- tion of the Greenlandic Arctic fox has not resulted in complete genetic mixing of the populations. Fisher’s exact test was also used to estimate levels of genetic differentiation between the two colour morphs: white and blue. No difference was found between allele frequencies of the two color morphs in any of the locations, and it was concluded that the white and blue morphs of the Greenlandic Arctic fox share the same habitat, at least during the mating season. P. G. Meinke, Dept. of Evolutionary Biology, Copenhagen University, Universitetsparken 15, DK-2100 København Ø., Denmark; C. M. O. Kapel, Danish Center for Experimental Parasitology, The Royal Veteri- nary and Agricultural University, Bulowsvej 17, DK-1870 Frederiksberg C, Denmark; P. Arctander, Dept. of Evolutionary Biology, Copenhagen University, Universitetsparken 15, DK-2100 København Ø., Denmark. 76 Genetic differentiation of populations of Greenlandic Arctic fox and living in the mostly snow-covered inland and the blue one depending to a large extent on products of the sea and living near the coast. Braestrup ś studies, based on fur trade statistics, were continued by Vibe (1967), who included field studies and climate statistics. He stated that unstable climatic periods during the last centu- ries, with the climate favouring the blue morph in some periods and the white one in other periods, have resulted in the “lively bastardization and cre- ation of mixed populations which prevail nearly everywhere in coastal regions today”. Braestrup (1941) found large fluctuations in the proportion of white foxes in the hunting bags from west Greenland and attributed this to a massive influx of white foxes to west Greenland from Canada and east Greenland in years with low lemming populations. In a morphological study of the foxes of Greenland, the blue foxes from different dis- tricts showed significant differences in (metric) bone measures while the white ones did not (Berg 1993). Berg suggested that the blue colour morph is more stationary than the white, due to the food supply in coastal areas being less variable between years than in inland habitats. Russian observations (reviewed by Wrigley & Hatch 1976) showed that foxes migrate from November to Jan- uary and return in February and March to breed. If this is the case, samples collected in winter could include migrants and could therefore be misleading. The aim of the present study was to delineate the genetic relationships of four Arctic fox popu- lations and the two colour morphs of Arctic fox in Greenland. The following questions were put for- ward: 1) Are geographically distant populations significantly differentiated? 2) Are the two colour morphs genetically differentiated? 3) Are there genetic differences between foxes sampled at the same locality during different seasons? To answer these questions we used six poly- morphic microsatellite loci. Due to the generally high mutation rate (10-2 - 10-5 mut./ locus/ gen- eration) (Weber & Wong 1993), microsatellites tend to be very polymorphic, which make them powerful markers for observing the differentia- tion between closely related populations. How- ever, their mutational modalities also result in a risk of homoplazy (Valdes et al. 1993; Weber & Wong 1993) and some microsatellites might have alleles that are not detectable (null alleles) (Callen et al. 1993). Also, microsatellite evolution is not yet properly understood and it could be problem- atic to consider them as strictly neutral markers: constraints on allele size have been implicated (Garza et al. 1995) and numerous indications of microsatellites involved in gene expression and function have been found (reviewed by Kashi et al. 1997). Furthermore, the possible effect of a selected locus on a closely linked microsatellite has been discussed (Slatkin 1995a). Several muta- tion models have been suggested to work for mic- rosatellites. Two are most used: the infinite allele model (IAM) (Estoup et al. 1995) and the step- wise mutation model (SMM) (Shriver et al. 1993; Valdes et al. 1993). Materials and methods Samples Seventy-five Greenlandic Arctic foxes from four widely spaced localities are included in this study; see Fig. 1 for sample details. All the foxes were 80 70 60 N Greenland Canada Søndre Strømfjord (Kangerlussuaq) 20 individuals 6 blue; 14 white Sampled: 19 in May-Aug. 1992 1 in Feb. 1993 Thule Air Base (Pituffik) 16 individuals 15 blue; 1 white Sampled: 8 in Feb. 1992 8 in July 1992 Scoresbysund (Ittoqqortoormiit) 19 individuals 8 blue; 11 white All sampled Oct. 1993-Apr. 1994 Julianehåb (Qaqortoq) 20 individuals 12 blue; 8 white Sampled: 7 in Mar.-Apr. 1992 13 in Mar. 1993 400 800 0 Km Fig. 1. Map of Greenland showing sites, sizes and dates of Arctic fox sampling. 77Meinke et al.: Polar Research 20(1), 75–83 caught in traps, killed and kept frozen at –20 °C in a period of two to three months prior to sampling of muscle tissue. DNA was extracted using Pro- teinase K and phenol/chloroform-solutions (e.g. Sambrook et al. 1989). The foxes analysed here have been included in studies on parasitology, diet and population composition (Kapel et al. 1996; Kapel & Nansen 1996; Kapel 1999). Analyses of microsatellite loci Several microsatellite loci originally described in studies of wolf and dog were screened. The primer sets designed for dog microsatellite loci turned out to give the best results, and six of them were chosen and optimized for the Arctic fox: cph3, cph6, cph9, cph15, cph16, and cph18 (Fred- holm & Winterø 1995). All six microsatellite loci used are dinucleotide repeats, and all but cph3 have perfect repeats (see Fredholm & Winterø 1995). The final PCR conditions were as follows (Table 1): denaturation at 94 °C for 2 min., x cycles of denaturation at 94 °C for y sec., anneal- ing at z °C for y sec. and extention at 72 °C for y sec., x, y and z are specified for each locus in Table 1. The amplification quality in different PCR machines differed from locus to locus. The final choice of machines is also shown in Table 1. DNA was amplified in a 10 µl reaction volume (10 mM Tris-HCL, 1.5 mM MgCl 2 , 50 mM KCl, pH 8.3, 200 µM of each dNTP, 200 pM of each primer, and 0.025 units of Boehringer-Mannheim Taq DNA polymerase). The products were run on 4.24 % acrylamidgels on an ABI 377 sequencing machine (Perkin Elmer) using dye-labeled prim- ers (one primer in each primer set) and ROX 500 (Perkin Elmer) as internal standard. Genotypes were determined at the six loci for most of the individuals (see Table 2). Statistical tests Deviations from Hardy-Weinberg equilibrium (HWE) was tested using the “exact HW test” Locus Annealing Sec. at each Cycles PCR machine temp. (z) step (y) (x) cph3 50 °C 15a 37 Unitec2042 cph6 49 °C 50a 37 Robocycler Gradient cph9 55 °C 15 35c Unitec2042 cph15 54 °C 15 34 Hybaid cph16 54 °C 15a 53 Unitec2042 cph18 66 °C 30a,b 31 Robocycler Gradient a Last cycle with extension time: 5 min. b All but last cycle with extension time: 40 sec. c Touch down PCR: two cycles with annealing temp. at each Table 1. PCR conditions. degree from 45 °C to 54 °C, followed by 25 cycles at annealing temp. 55 °C. Locus/ Thule Scores- Søndre Juliane- allele Air Base bysund Strømfjord håb cph3/151 0.026 153 0.536 0.132 0.316 155 0.107 0.026 0.079 157 0.036 0.184 0.211 0.100 161 0.250 0.474 0.105 0.350 163 0.071 0.053 0.132 0.350 165 0.079 0.026 0.175 167 0.053 169 0.026 0.025 171 0.026 173 0.026 179 0.026 n 28 38 38 40 cph6/109 0.222 0.158 0.400 119 0.667 0.028 121 0.053 0.250 123 0.028 0.105 0.125 125 0.278 0.316 0.075 127 0.222 0.211 0.150 129 0.222 0.158 133 0.333 n 30 36 38 40 cph9/152 0.344 0.028 0.368 0.675 154 0.031 0.250 0.237 0.075 156 0.438 0.639 0.342 0.175 158 0.188 0.053 0.050 162 0.025 164 0.083 n 32 36 38 40 cph15/153 0.167 0.132 0.050 0.050 155 0.033 0.079 0.075 0.200 157 0.667 0.658 0.675 0.550 159 0.053 0.125 0.150 161 0.133 0.026 0.075 0.050 163 0.026 165 0.026 n 30 38 40 40 cph16/155 0.075 159 0.156 0.105 0.175 0.025 161 0.656 0.632 0.250 0.100 163 0.188 0.263 0.525 0.725 165 0.025 0.075 167 0.025 n 32 38 40 40 cph18/253 0.028 0.184 255 0.059 0.053 257 0.125 0.056 0.441 0.289 259 0.167 0.029 0.026 263 0.031 0.028 0.206 0.053 265 0.781 0.083 0.118 0.026 267 0.063 0.056 0.088 0.105 269 0.500 0.158 271 0.083 0.059 0.105 n 32 36 34 38 Table 2. Observed allele frequency distributions by locus and population. Private alleles are shown in italics. Boldface indi- cates the most frequent allele/locus/population. N indicates number of chromosomes. 78 Genetic differentiation of populations of Greenlandic Arctic fox in GENEPOP (Raymond & Rousset 1995). The sequential Bonferroni technique (Rice 1989) was applied to test significant deviation from HWE at a “table-wide” α-level of 0.05. Test for genotypic disequilibrium was per- formed in GENEPOP using Fisher ś exact test (Raymond & Rousset 1995). Fisher ś exact test was also used to estimate levels of genic differen- tiation between colour morphs and sampling sea- sons within locations. The sequential Bonferroni technique (Rice 1989) was applied to test for sig- nificance at the “table-wide” α-level of 0.05. Population differentiation was quantified in three different ways both within the total popu- lation and between all populations pairwise: 1) by Fisher ś exact test in GENEPOP; 2) by esti- mating Rho using Goodmaǹ s (1997) analogue to Slatkin ś Rst (calculated in Rst-calc 2.2; Rst esti- mates are based on the SMM); and 3) by estimat- ing the Weir & Cockerham (1984) analogues to Wright ś Fst calculated in Arlequin. Fst esti- mates are based on the IAM. All three methods were applied to the total population and to each pair of populations. GENEPOP and Rst-calc 2.2 provided probability values and Rho estimates, respectively, for individual loci and overall loci, while Arlequin only provided Fst estimates of overall loci. Furthermore, Rst-calc 2.2 provided bootstrap values as well as significance values for all estimates, making it possible to see if two Rho values differ significantly from each other. Estimates of the number of migrants per gen- eration (Nm) between all pairwise populations was computed from the approximations: Nm= 1/4 (1/Fst -1) (Slatkin 1995b) and Nm=1/4 (1/Rho -1) (pers. comm., Bo Simonsen). Nm was also esti- mated using Slatkin ś rare alleles method which makes use of the fact that the logarithm of Nm is approximately linearly related to the logarithm of the average frequency of private alleles in a sample of alleles from the population (Slatkin 1985). This method has shown to be relatively insensitive to changes in other parameters than Nm and the number of individuals sampled per population (Slatkin 1985). Results Genetic variation and deviation from Hardy- Weinberg equilibrium The six loci—cph3, cph6, cph9, cph15, cph16 and cph18—were polymorphic with 12, 8, 6, 7, 6 and 9 alleles, respectively (Table 2). The aver- age number of alleles per locality ranged from 4 in cph16 to 7 in cph3 and cph18, and average expected heterozygosity values for all loci ranged from 0.54 in Thule Air Base to 0.73 in Søndre Strømfjord (see Table 3). There was highly significant deviation from HWE in the total population (p ≤ 0.002). All loci showed deficiency of heterozygotes (Table 3). No significant deviation from the HWE at the “table- wide” 0.05 level was found in any of the four sep- arate populations (Table 3). Looking at each locus across localities, two loci , one in each locality, showed significant deviation from HWE due to Ho He P(HWE) Thule Air Base cph3 0.57 0.66 0.077 NS cph6 0.27 0.46 0.232 NS cph9 0.69 0.68 0.394 NS cph15 0.33 0.53 0.012 NS cph16 0.50 0.53 0.827 NS cph18 0.44 0.38 1.000 NS Scoresbysund cph3 0.68 0.73 0.169 NS cph6 0.72 0.80 0.927 NS cph9 0.61 0.54 0.863 NS cph15 0.47 0.55 0.308 NS cph16 0.53 0.53 0.831 NS cph18 0.61 0.72 0.108 NS Søndre Strømfjord cph3 1.00 0.84 0.035 NS cph6 0.89 0.86 0.251 NS cph9 0.68 0.71 0.229 NS cph15 0.50 0.53 0.377 NS cph16 0.60 0.65 0.184 NS cph18 0.59 0.76 0.011 NS Julianehåb cph3 0.65 0.73 0.196 NS cph6 0.65 0.75 0.899 NS cph9 0.35 0.52 0.079 NS cph15 0.50 0.65 0.041 NS cph16 0.35 0.46 0.070 NS cph18 0.79 0.85 0.395 NS Total pop. cph3 0.74 0.81 0.001 S cph6 0.64 0.86 0.000 S cph9 0.58 0.69 0.001 S cph15 0.46 0.57 0.002 S cph16 0.49 0.64 0.001 S cph18 0.61 0.85 0.000 S Table 3. Expected (He) and observed (Ho) heterozygocity and probability values for “the exact HW test” calculated for each locus in every population. S and NS indicate significance and nonsignificance, respectively, at the “table-wide” 0.05 level. Nonsignificant p-values across populations are shown in italics. 79Meinke et al.: Polar Research 20(1), 75–83 foxes, while the other five loci showed nonsignifi- cant results (Table 4). To compare population dif- ferentiation of the white foxes with the blue foxes, Rho estimates were calculated for each colour. These Rho estimates were calculated for all pair- wise localities in which the colours were sampled, and for the total population (Table 5). The results for the blue foxes showed highly significant (p < 0.001) subdivision of the total sample; there was significant population differentiation (p < 0.05) in all pairwise comparisons except between Thule Air Base and Scoresbysund and between Søndre Strømfjord and Julianehåb. White foxes also showed significant subdivision of the total sample (p < 0.01) and in all pairwise population comparisons (p < 0.05). None of the Rho esti- mates calculated for blue and white foxes were significantly different from each other or from the Rho estimates of the populations of both col- ours (overlap in 95 % confidence intervals, see Table 5). In the test for difference in allele frequencies between the seasons, foxes at Thule Air Base showed a significant difference in allele frequency between two sampling occasions (winter and summer) in two out of six loci (Table 6). No sig- nificant difference was found between the allele frequencies of two sub-samples from Scoresby- sund (winter and spring), nor beween two sub- samples from Julianehåb (two consecutive spring seasons) (Table 6). Population structure The results for all loci from the three different tests for population subdivision are presented in Table 7. The first of the results to be noted is that Fish- er ś exact test showed highly significant subdivi- sion (p < 0.00001) of the total population in all but one locus (cph15). In five of the six pairs of localities, significant differentiation in allele frequencies was supported Locus Scoresby- Søndre Julianehåb Scoresbysund sund Strømfjord + S. Strømfjord + Julianehåb cph3 0.5390 NS 0.5928 NS 1.0000 NS 0.502 NS cph6 0.7613 NS 0.1179 NS 0.1356 NS 0.082 NS cph9 0.2312 NS 0.1627 NS 0.0240 NS 0.160 NS cph15 0.2731 NS 0.4658 NS 0.0641 NS 0.307 NS cph16 1.0000 NS 0.0103 NS 0.4752 NS 0.294 NS cph18 0.7009 NS 0.0195 NS 0.0092 NS 0.007 S Table 4. Results (p-values) from Fisher ś exact test for dif- ferentiation between blue and white foxes. S and NS indicate significance and nonsignificance, respectively, at the “table- wide” 0.05 level. Rho est., Thule Air Base Thule Air Base Thule Air Base Scoresbysund Scoresbyund S. Strømfjord Total pop. all loci −Scoresbysund −S. Strømfjord −Julianehåb −S. Strømfjord −Julianehåb −Julianehåb Blue 0.065 NS 0.152* 0.222*** 0.224** 0.191** 0.050 N 0.159*** foxes (0.020; 0.222) (0.152; 0.374) (0.174; 0.364) (0.152; 0.430) (0.121; 0.352) (0.007; 0.303) (0.141; 0.278) White 0.080* 0.180** 0.144** 0.128** foxes (0.022; 0.246) (0.101; 0.390) (0.080; 0.367) (0.086; 0.290) Table 5. Population differentiation with populations separated according to fur colour. Significance levels: ∗ = p < 0.05; ∗∗ = p < 0.01; ∗∗∗ = p < 0.001. NS indicates a nonsignificant result. 95 % confidence interval is provided in parentheses. heterozygote deficiency: cph15 in Thule Air Base and cph18 in Søndre Strømfjord (shown in italics in Table 3). All other loci across localities showed no deviation from HWE. Linkage disequilibrium was found in no pair of loci in the global test across all localities (prob- ability values ranging from 0.07 to 0.96; data not shown). Within the populations, the same test demonstrated two cases of linkage disequilib- rium (data not shown): the locus pair cph9−cph18 showed significant linkage in Thule Air Base (p = 0.007) and the pair cph9−cph16 showed linkage in Scoresbysund (p = 0.03). Dividing populations according to colour and time of sampling To test for differentiation in allele frequencies between the two colour morphs we divided each sample according to fur colour. Thule Air Base was omitted from this test as white foxes there were represented by only one animal. The remain- ing three localities contained sub-samples of at least six individuals each (Fig. 1). No significant differentiation was found between allele frequen- cies of the two colour morphs in any of the locali- ties (Table 4). By using the same test on the three localities pooled, cph18 showed a significant dif- ference in allele frequencies of blue and white 80 Genetic differentiation of populations of Greenlandic Arctic fox by from four to all of six loci. It was only between Thule Air Base and Søndre Strømfjord that half of the loci showed no significant differentiation in allele frequencies. The second of the results to be noted is that Rho estimates showed a highly significant population subdivision within the total population (p < 0.001) and a significant popula- tion differentiation between all pairwise combi- nations of populations (p < 0.05). Furthermore, bootstrap results showed that none of the Rho results from the pairwise locality studies were significantly different from another (overlap between all the 95 % confidence intervals), although they ranged from 0.065 (Thule Air Base−Søndre Strømfjord) to 0.226 (Thule Air Base−Julianehåb). Third, Fst estimates also showed a highly sig- nificant population subdivision (p < 0.001) within the total population. Furthermore, this estimate showed highly significant population differentia- tion (p < 0.001) between all combinations of pop- ulations. To find possibly deviant loci we looked at the population differentiation results locus-wise and found the following facts notable: 1) When cor- recting the significance values locus-wise (hori- zontally in Table 7) according to the sequential Bonferroni technique (results not shown), cph15 showed nonsignificant population differentiation in all pairwise population comparisons and non- significant subdivision of the total population. Cph9 showed nonsignificant result in two of the six pairwise population comparisons, cph3,6 and 16 in only one, and cph18 showed no nonsignifi- cant probability values. 2) There was a large span in the Rho estimates of individual loci within all pairwise population comparisons, ranging from a negative value (around -0.03) to 0.246 (Table 7). Migration The estimated number across the total popula- tion of migrating individuals calculated from the Rho and the Fst estimates was 1.71 and 1.32 individuals/generation respectively (Table 8) and 2.05 individuals/generation when estimated by using private alleles according to Slatkin (1985) (data not shown). In the pairwise population com- Locus Thule Air Base Scoresbysund Julianehåb (winter 1992 (winter 1993 (spring 1992 −summer 1992) −spring 1994) −spring 1993) cph3 0.0015 S 0.8920 NS 0.0828 NS cph6 0.7060 NS 0.4773 NS 0.4178 NS cph9 0.0746 NS 0.4729 NS 0.0563 NS cph15 0.4734 NS 0.3427 NS 0.3745 NS cph16 0.1435 NS 0.8863 NS 0.8987 NS cph18 0.0060 S 0.6906 NS 0.5028 NS Table 6. Results (probability values) from Fisher ś exact test for difference in allele frequency between foxes sampled indif- ferent seasons. S and NS indicate significance and nonsignifi- cance, respectively, at the “table-wide” 0.05 level. Table 7. Population differentiation. Probability values from Fisher’s exact test. S and NS refer to significance and nonsignifi- cance, respectively, at the “table-wide” 0.05 level. Rho estimates and Fst estimates are shown. Significance levels: *= p < 0.05; ** = p < 0.01; *** = p < 0.001. 95% confidence interval is provided in parentheses for Rho estimates over all loci. Estimate/ Thule Air Base Thule Air Base. Thule Air Base Scoresbysund Scoresbysund S. Strømfjord Total pop. loci −Scoresbysund −S. Strømfjord −Julianehåb −S. Strømfjord −Julianehåb −Julianehåb G. diff. cph3 0.0013 S 0.1314 NS 0.0000 S 0.0020 S 0.0019 S 0.0000 S 0.0000 S cph6 0.0000 S 0.0000 S 0.0000 S 0.5661 NS 0.0000 S 0.0002 S 0.0000 S cph9 0.0000 S 0.0421 NS 0.0082 S 0.0001 S 0.0000 S 0.0256 NS 0.0000 S cph15 0.4252 NS 0.1323 NS 0.0095 S 0.4848 NS 0.2088 NS 0.5428 NS 0.1108 NS cph16 0.6412 NS 0.0040 S 0.0000 S 0.0058 S 0.0000 S 0.0099 S 0.0000 S cph18 0.0000 S 0.0000 S 0.0000 S 0.0000 S 0.0003 S 0.0054 S 0.0000 S Rho cph3 0.283 0.059 0.633 0.052 0.092 0.306 cph6 -0.019 -0.027 0.250 -0.025 0.157 0.219 cph9 0.067 0.036 0.157 0.236 0.349 0.035 cph15 -0.030 -0.010 -0.030 -0.015 -0.027 -0.010 cph16 -0.006 0.123 0.151 0.057 0.092 -0.015 cph18 0.058 0.246 0.115 0.332 0.215 -0.026 All loci 0.070** 0.065* 0.226*** 0.117*** 0.165*** 0.089** 0.128*** (0.018; 0.180) (0.028; 0.182) (0.178; 0.336) (0.057; 0.232) (0.093; 0.285) (0.053; 0.191) (0.107; 0.206) Fst 0.198*** 0.175*** 0.262*** 0.096*** 0.169*** 0.070*** 0.159*** 81Meinke et al.: Polar Research 20(1), 75–83 parison the Nm values calculated from the Rho estimates ranged from 0.86 between Thule Air Base and Julianehåb to 3.61 between Thule Air Base and Søndre Strømfjord, but, as with the Rho estimates, not one of the values was significantly different from another according to the bootstrap results. The Nm values calculated from the Fst estimates in the pairwise population compari- sons showed about the same range: from 0.70 (Thule Air Base−Julianehåb) to 3.34 (Søndre Strømfjord−Julianehåb). Discussion Four randomly mating populations The heterozygote deficiency found in the total Greenlandic sample can probably be ascribed to a Wahlund effect created when pooling the four populations. These HWE results give a prelimi- nary indication that the population is subdivided in spite of the Arctic fox’s capacity for long-dis- tance movements. Furthermore, the HWE results showed that each of the four localities can be seen as a randomly mating population (no deviation from HWE). The fact that two loci (cph15 and cph18), one each in two different populations, showed a significant deviation from HWE, look- ing across all populations, cannot be explained by null alleles: if the two loci had null alleles we would expect it to affect the other populations as well, which was not the case. It is, of course, possible that non-detectable alleles causing het- erozygote deficiency of cph15 and cph18 only were present in Thule Air Base and in Søndre Strømfjord, respectively, but since all but one private allele in the total data set had very small allele frequencies (p ≤ 0.05; Table 2), this appears unlikely. We cannot explain these signif- icant results since explanations such as assorta- tive mating or pooling of samples (resulting in the Wahlund effect) would be expected to affect the other loci of the populations as well and that is not the case. The two colour morphs There seems to be a positive correlation between the period of time an area is snow-covered and the proportion of white foxes in this area (Braestrup 1941; Vibe 1967; Hersteinsson 1989). If this is due to a sharply defined difference in habitat, inter- breeding between the morphs would only exist in a small hybrid zone, and we would expect to find a difference in allele frequencies between the morphs and certainly not random mating between them. Kapel et al. (1996) found no significant differ- ence between the prevalence of Trichinella nativa infections in the blue and the white foxes, respec- tively, which were also used for the study reported here. T. Nativa is an extraintestinal nematode which is most prevalent in foxes in areas where polar bears are hunted and sledge dogs are common (Kapel 1995). The nematode is thought to be transmitted to the Arctic fox through scav- enging on carcasses of these animals or through cannibalism. The fact that Kapel found no dif- ference in the prevalence of this parasite accord- ing to colour of the foxes could indicate that the two morphs share the same habitat and diet. The results of this study indicate random mating within the three populations of mixed colours and no allele frequency difference between the two colour morphs. It is important to note that the sta- tistical power may have been small due to small sample sizes. However, our data suggest that the two colour morphs of the Greenlandic Arctic fox share the same habitat, at least during the mating season. The investigation of differentiation within the blue and the white fox groups (Table 5) yields no indication that the blue fox is more stationary than the white one. On the contrary, Rho estimates failed to show significant difference between the Estimate/ Thule Air Base Thule Air Base Thule Air Base Scoresbysund Scoresbysund S. Strømfjord Total pop. loci −Scoresbysund −S. Strømfjord −Julianehåb −S. Strømfjord −Julianehåb −Julianehåb Nm (Rho) 3.34 3.61 0.86 1.89 1.26 2.54 1.71 (1.07; 12.0) (1.11; 8.58) (0.49; 1.15) (0.82; 3.91) (0.61; 2.42) (1.05; 4.41) (0.96; 2.07) Nm (Fst) 1.01 1.18 0.70 2.35 1.23 3.34 1.32 Table 8. Migration rate. Nm values correspond to the Rho and Fst estimates shown in Table 6. 95 % confidence interval is pro- vided in parentheses for the Nm corresponding to Rho estimates. 82 Genetic differentiation of populations of Greenlandic Arctic fox blue foxes of Søndre Strømfjord and Julianehåb while significant differentiation (p < 0.01) was found between the white foxes from the same localities. These results indicate that the differ- ence Berg (1993) found with respect to morpho- logical differentiation of Greenlandic blue and white foxes was a result of ecological rather than genetic differences, as Berg also discusses. The different sampling seasons Two of the four populations (Thule Air Base and Scoresbysund) in this study were sampled in both winter and spring/summer seasons (Fig. 1). The- ories have been proposed that the foxes migrate in November-January and return to their “home” in February−March to breed—the “homing theory” (Russian studies reviewed by Wrigley & Hatch 1976). If this is the case, sampling in winter will lead to sampling of migrants that will eventually return to their own locality and therefore have no genetic impact on the locality in which they were found. Winter populations will be a mixture of mating populations and we would expect to find a difference in allele frequencies between a winter and a summer sample from the same locality. We tried to test this theory by testing for genic differ- entiation between the sub-samples of Thule Air Base and Scoresbysund. However, in testing the homing theory this way, we have a problem in the transition between winter and mating popula- tions. Scoresbysund was artificially divided into two sub-samples: one on each side of 1 March 1994. The results of Thule Air Base neither clearly supported nor disproved the theory and we have to conclude that our data set cannot be used to test the homing theory. Samples from more popula- tions, collected in both winter and summer, are needed. Population structure Distinct geographically differences in the compo- sition of the parasitic helminth fauna of Arctic fox in Greenland have been found (Kapel et al. 1996; Kapel & Nansen 1996). The diversity of the sur- rounding fauna and thereby the food items available for the foxes seemed important influ- ences on the diversity of the helminth species (Kapel & Nansen 1996). These results indicate limited mixing of the studied populations. The present study supports this indication. All three approaches to analysing population differentia- tion (Fisher ś exact test, Fst estimates and Rho estimates) showed that the total sample of Green- landic Arctic fox was significantly differentiated into subpopulations. The Rho and the Fst esti- mates across all loci also showed significant population differentiation between all pairwise populations included in the study. Population differentiation results for individual loci are provided by Fisher ś exact test and as Rho estimates. In both tests cph15 was the only locus consistently showing no difference between any two populations, and it even showed no signifi- cant subdivision of the total population in Fish- er ś exact test (Table 7). On this basis we suggest that cph15 is not a neutral locus. Slatkin (1995a) showed that the variance of microsatellite allele size can be strongly influenced by selection at closely linked loci. Cph15 could be linked to a locus under selection or it could be involved in gene expression, resulting in a reduction in the variance of allele size of this microsatellite in a way that makes population subdivision less detectable with this locus. Migration The three different methods we have used to esti- mate average Nm between the populations of this study gave very consistent results around one to two migrating foxes between each locality every generation. The distances between the localities of this study range from 825 km between Søndre Strømfjord and Julianehåb to 1950 km between Thule Air Base and Julianehåb. From our data we cannot discern if, for instance, one fox per gener- ation travels the distance between Thule Air Base and Julianehåb or if the estimate is a result of a larger gene flow between all adjacent popula- tions along the distance between Thule Air Base and Julianehåb. In this regard it would be interest- ing to get estimates of differentiation and migra- tion between populations separated by shorter distances (including adjacent populations) and to extend the study to involve more loci and/or indi- viduals to increase statistical power. Conclusion We have found evidence of extensive genetic dif- ferentiation between four Arctic fox localities, 83Meinke et al.: Polar Research 20(1), 75–83 demonstrating that the ability of Arctic foxes to undertake long-distance movements has not resulted in a panmictic Greenlandic population. The foxes examined in this study do not show genetic differentiation between the two colour morphs within each locality. This suggests that the two colour morphs of the Greenlandic Arctic fox can be considered as members of the same populations. A test for population differentiation within each of the two colour morphs indicated that they are equally stationary. Acknowledgements.—We thank Thomas Berg for assistance during the fieldwork; Marie-Agnés Coutellec-Vreto and Sil- vester Nyakaana for statistical assistance; Michael Roy for technical advice and support; Jacob Damgaard for comments on an earlier version of the manuscript; and Lisette Rasmus- sen for linguistic assistance. We also thank Merete Fredholm and Anne Katrine Winterø for technical assistance optimizing the PCR reactions. The Commission for Scientific Research in Greenland partly supported this work. References Adalsteinsson, S., Hersteinsson, P. & Gunnarsson, E. 1987: Fox colours in relation to colours in mice and sheep. J. Hered. 78, 235–237. Anthony, R. M. 1997: Home ranges and movements of Arctic fox (Alopex lagopus) in western Alaska. Arctic 50, 147–157. Berg, T. B. 1993: Polarræven i Grønland. (The Arctic fox in Greenland.) M.Sc. thesis, Copenhagen University, Den- mark. Braestrup, F. W. 1941: A study on the Arctic fox in Greenland. Medd. Grønl. 131, 1-101. Callen, D. F., Thompson, A. D., Shen, Y., Phillips, H. A., Richards, R. I., Mulley, J. C. & Sutherland, G. R. 1993: Incidence and origin of “null” alleles in the (AC)n micros- atellite markers. Am. J. Hum. Genet. 52, 922–927. Eberhardt, L. E. & Hanson, W. C. 1978: Long-distance move- ments of Arctic foxes tagged in northern Alaska. Can. Field-Nat. 92, 386–389. Eberhardt, L. E., Garrott, R. A. & Hanson, W. C. 1983: Winter movements of Arctic foxes, Alopex lagopus, in a petroleum development area. Can. Field-Nat. 97, 66–70. Estoup, A., Garnery, L., Solignac, M. & Cornuet, J.-M. 1995: Microsatellite variation in honey bee (Apis mellifera L.) populations: hierarchical genetic structure and test of the infinite allele and stepwise mutation models. Genetics 140, 679–695. Fay, F. H. & Rausch, R. L. 1992: Dynamics of the Arctic fox population on St. Lawrence Island, Bering Sea. Arctic 45, 393–397. Frafjord, K. & Prestrud, P. 1992: Home range and movements of Arctic foxes Alopex lagopus in Svalbard. Polar Biol. 12, 519–526. Fredholm, M. & Winterø, A. K. 1995: Variation of short tandem repeats within and between species belonging to the Canidae family. Mamm. Genome 6, 11–18. Garrott, R. A. & Eberhardt, L. E. 1987: Arctic fox. In M. Novak et al. (eds.): Wild furbearer management and con- servation in North America. Pp. 395–406. Toronto: Minis- try of Natural Resources. Garza, J. C., Slatkin, M. & Freimer, N. B. 1995: Microsatellite allele frequencies in humans and chimpanzees, with impli- cations for constraints on allele size. Mol. Biol. Evol. 12, 594–603. Goodman, S. J. 1997: R ST Calc: a collection of computer pro- grams for calculating estimates of genetic differentiation from microsatellite data and determining their significance. Mol. Ecol. 6, 881–885. Hersteinsson, P. 1989: Population genetics and ecology of dif- ferent colour morphs of Arctic foxes Alopex lagopus in Ice- land. Finn. Game Res. 46, 64–78. Kapel, C. M. O. 1995: Helminths of the Arctic fox (Alopex lagopus (L.)) in Greenland. Ph.D. thesis, The Royal Veteri- nary and Agricultural University, Copenhagen. Kapel, C. M. O. 1999: Diet of Arctic foxes (Alopex lagopus) in Greenland. Arctic 52, 289–293. Kapel, C. M. O., Henriksen, S. A., Berg, T. B. & Nansen, P. 1996: Epidemiologic and zoogeographic studies on Trichinella nativa in Arctic foxes, Alopex lagopus, in Greenland. J. Helminthol. Soc. Wash. 63, 226–232. Kapel, C. M. O. & Nansen, P. 1996: Gastrointestinal helminths in Arctic foxes (Alopex lagopus) from different bioclimato- logical regions in Greenland. J. Parasitol. 82, 17–24. Kashi, Y., King, D. & Soller, M. 1997: Simple sequence repeats as a source of quantitative genetic variation. Trends Genet. 13, 74–78. Raymond, M. & Rousset, F. 1995: An exact test for popula- tion differentiation. Evolution 49, 1280–1283. Rice, W. R. 1989: Analysing tables of statistical tests. Evolu- tion 43, 223–225. Sambrook, J., Fritsch, E. F. & Maniatis, T. 1989: Molecular cloning: a laboratory manual. Second edition. New York: Cold Spring Harbor Laboratory Press. Shriver, M. D., Jin, L., Chakraborty, R. & Boerwinkle, E. 1993: VNTR allele frequency distributions under the step- wise mutation model: a computer simulation approach. Genetics 134, 983–993. Slatkin, M. 1985: Rare alleles as indicators of gene flow. Evo- lution 39, 53–65. Slatkin, M. 1995a: Hitchhiking and associative overdom- inance at a microsatellite locus. Mol. Biol. Evol. 12, 473–480. Slatkin, M. 1995b: A measure of population subdivision based on microsatellite allele freqencies. Genetics 139, 457–462. Valdes, A. M., Slatkin, M. & Freimer, N. B. 1993: Allele fre- quencies at microsatellite loci: the stepwise mutation model revisited. Genetics 133, 737–749. Vibe, C. 1967: Arctic animals in relation to climatic fluctua- tions. Medd. Grønl., Biosci., 170, 7–227. Weber, J. L. & Wong, C. 1993: Mutation of human short tandem repeats. Hum. Mol. Genet. 2, 1123–1128. Weir, B. S. & Cockerham, C. C. 1984: Estimating F-statistics for the analysis of population structure. Evolution 38, 1358–1370. Wrigley, R. E. & Hatch, D. R. M. 1976: Arctic fox migrations in Manitoba. Arctic 29, 147–158.