Temporal and Geographic Homogeneity of Gene Frequencies in the Fox Sparrow (Passerella iliaca)

Protein electrophoresis has been used frequently in the study of genetic variation within and among natural populations (Lewontin 1974, Avise and Aquadro 1982, Nevo et al. 1984). A goal of protein electrophoretic studies is to describe patterns of genetic variation and deduce historical processes (such as fragmentation of populations and gene flow) that have produced such patterns (Felsenstein 1982). A second, but relatively little-studied, aspect of the geography of genetic variation is the temporal stability of patterns of gene frequencies. There are both systematic and evolutionary implications of temporal variation in allozymic frequencies. If samples are taken from different sites in different years, temporal variation might confound systematic interpretation of geographic patterns (Zink 1983). Many investigators have implicated natural selection in the maintenance of enzyme polymorphisms (Mitton and Grant 1984, Mueller et al. 1985, Barker et al. 1986; but see Zink et al. 1985). A correlate of genetic polymorphisms influenced by selection is temporal change in gene frequencies at a site, resulting from directional natural selection. Thus, it is of interest to examine the temporal stability of gene frequencies. Previous studies of temporal variation in gene frequencies have either considered relatively few generations, limited geographic areas, or few loci (Baker and Fox 1978, Gyllensten 1985, Henderson 1977, McCauley et al. 1988, Redfield 1974). In this study, we assessed temporal and geographic variation in gene frequencies in population samples of the Fox Sparrow (Passerella iliaca). A previous allozymic survey (Zink 1986), based on samples taken in 1978-1980 in California, Oregon, and Nevada, revealed essentially no geographic differentiation in gene frequencies; the FST value (derived from 31 samples of individuals, 38 presumptive genetic loci, and a total of 619 individuals) was 0.0135 ? 0.0033 (SE). At some of these sites, Zink (1983) detected morphometric differentiation between samples taken in the mid-1920s and 19781980, a span of 50 years. We used protein electrophoresis to compare levels and patterns of genetic variation in seven samples (from California and Nevada) of Fox Sparrows taken in 1988 with those taken in 1978-1980, a span of 8-10 generations. In addition, we sampled two new sites (Table 1) geographically distant from the original sample sites to assess geographic differentiation on a larger scale. Approximately one half of our aqueous tissue extracts were obtained during preparation of tissues for mitochondrial DNA (mtDNA) analysis (see Avise and Zink 1988), and the others were prepared as outlined by Zink (1986). Methods for protein electrophoresis are given in Zink (1986). For the samples collected in 1988, we surveyed only the loci found to be polymorphic in the earlier study (Table 1). Because sample sizes for our recent samples are relatively small (1120) and because Zink (1986) showed that the same allele was the common allele at all sites for each locus, we compared only the frequencies of common alleles among sites and across generations; noncommon alleles were pooled. We performed 2 x 2 G-tests (Sokal and Rohlf 1981) for each locus and site by tallying the number of common and noncommon alleles in the "old" (1978-1980) and "new" (1988) samples. Such multiple testing requires adjustment of significance values. We followed Rice's (1989) suggestion and divided the alpha level of 0.05 by the number of G-tests (72) to get a significance level (0.0007) appropriate for our multiple G-tests. The 19 sites at which the common allele was fixed in samples taken in both years were not tested, which explains why we tested 72 values and not 91 (13 loci x 7 sites). Rice (1989) noted that, in a series of tests such as ours, if the smallest probability value is not significant, then all entries in the table are judged nonsignificant, regardless of the significance levels of individual tests. We recognize that this is a conservative test, and one that requires perhaps larger samples than we have to detect significant effects. Therefore, we computed a G-value by summing all 72 tests to test for a table-wide effect of temporal variation. We report the observed directcount heterozygosity value for each sample (see Zink 1986). We also computed separate FsT values (each corrected for sampling a finite number of individuals (Wright 1978)) for the old samples, new samples, and the new samples plus those from Wyoming and Colorado to evaluate temporal and geographic variation. Temporal variation.-G-tests revealed no significant values (at the corrected P-value) in the occurrence of common alleles between old and new samples (Table 1). We found that 7 of the 72 (10%) tests produced significant G-values at the 0.05 level when each test was evaluated independently; hence, correction for multiple testing influences interpretation (Rice 1989). We also tested for locus and site effects by summing G-values across loci and across sites, and found only one significant value, that for the locus Lgg (Table 1). We noted during our study that gel patterns at Lgg were difficult to interpret for individuals prepared for mtDNA analysis relative to standard preparations, an apparent artifact of the mtDNA buffer. It is unclear, therefore, if the marginally significant value for Lgg deserves special explanation, both because of possible