Isozyme Polymorphisms in Daphnia

Temporal sequences of allele frequencies in natural populations of Daphnia are analyzed to obtain the mean and variance of the selection coefficient for both asexual and sexual phases. In general, the alleles at enzyme loci appear to be quasi-neutral. Although significant variation exists for the estimated selection coefficients, the means are in all cases close to zero. Estimates of the variance of selection intensity are applied to existing models to demonstrate the implications of fluctuating selection for the spatial and temporal distribution of gene frequencies in Daphnia. The empirical and analytical results are shown to provide a possible solution to some previously puzzling aspects of Daphnia population genetic surveys. Neither genetic drift nor diversifying selection are necessary conditions for the local diversification of gene frequencies. OLLOWING the early work of WRIGHT (1948) F and KIMURA (1 954), great advances have been made in the theory of fluctuating selection intensity (GILLESPIE and LANGLEY 1974; KARLIN and LIEBERMAN 1974; COOK and HARTL 1975; FELSENSTEIN 1976; HARTL and COOK 1976; HEDRICK, GINEVAN and EWINC 1976; GILLESPIE 1976, 1978; MATSUDA and GOJOBORI 1979; TAKAHATA and KIMURA 1979; TAKAHATA 1981; TIER 1981). Models now exist for both discrete and overlapping generations, for asexuality and sexuality, and for finite and infinite effective population sizes. Mutation has been incorporated into some models, as has serial autocorrelation in selection intensity. One of the primary reasons that the problem of temporally variable selection has received so much attention is that it appears to help maintain genetic polymorphisms in some cases, thereby providing a potential explanation for levels of genetic variance observed in natural populations. Another reason is that the incorporation of variable selection into the theory of rates of molecular evolution is expected in some cases to provide a more satisfactory fit to the data than a purely neutral model (MATSUDA and GOJOBORI 1979; NEI 1980; KIMURA 1982; NEI and GRAUR 1984). WRIGHT (1948) and KIMURA (1954) first showed that, under appropriate conditions, random variation in selection intensity could cause a drift-like phenomenon. Although an analytical error in this work was pointed out by GILLESPIE (1973a) and JENSEN (1973), the qualitative conclusions have been upheld by many subsequent studies. Random variation in selection can result in substantial divergence in gene frequencies among isolated populations. Indeed, TIER (1 98 1) has Genetics 115: 657-669 (April, 1987) demonstrated that, for certain parameter values, the expected gene frequency distribution for a random environment model will be virtually indistinguishable from that for a random drift-mutation balance model. Although it is reasonable to expect selection coefficients on individual loci in natural populations to be temporally variable, the quantitative significance of variable selection remains an enigma. Genetic studies of natural populations exceeding a few generations are very rare, and for those that do exist, it is generally difficult to determine the relative contributions of drift, sampling error, and selection to observed gene frequency changes. The most familiar empirical study bearing on the problem of temporally variable selection, FISHER and FORD’S (1947) eight-generation census of the medionigra gene in the moth Panaxia dominula, drew immediate criticism from WRIGHT ( 1 948). A subsequent analysis of the same data (TUCKWELL 1976) has also been questioned (TURELLI 1977). Recently, MUELLER et al. (1985) and MUELLER, BARR and AYALA (1985) have developed statistical tests for detecting natural selection from observed temporal changes in gene frequencies. Their application of the tests to date on natural populations of Drosophila pseudoobscura and the butterfly Euphydryas editha led to the conclusion that significant selection was operating on several enzyme loci. Unfortunately, while the analytical procedures engaged by MUELLER, BARR and AYALA (1985) and MUELLER et al. (1985) assume a closed population, the possibility that their local study populations were subject to significant immigration from pools of individuals with different gene frequencies cannot be ruled out (M. TURELLI, personal communication). Consequently, the shifts in