Table 1 lists all loci for which information on dominance was found. Table I also indicates whether the mutations were dominant, semidominant, or recessive. The results are unambiguous: most mutations in Chlamydomonas are recessive (Tables 1 and 2). Indeed, no fully dominant mutations were found. -Most important, the distribution of dominance effects found in Chlamydomonas does not differ from that Fisher (1) found in Drosophila (Table 2; log-likelihood ratio test G[21 = 2.25; P = 0.33). These Drosophila data, of course, provided the original motivation for Fisher's theory. If we exclude lethal mutations from the Chlamydomonas data (see below), as Fisher did for the Drosophila data, the distributions become even more similar (G121 = 0.65; P = 0.75). Thus, mutations are recessive just as often in a typically haploid eukaryote as in a typically diploid species. It is particularly interesting to note that in Chlamydomonas, just as in Drosophila (4, 8), lethal mutations are usually recessive. Among auxotrophic or severe photosyntheticdefective mutations, which would almost surely be lethal in nature (i.e., ac, arg, cr, nit, pab, and thi mutations), 13 of 16 are recessive. (Note that this does not reflect a selective bias toward recovering recessives: under permissive conditions, dominant auxotrophs can be recovered and maintained just as readily as recessives.) Fisherian selection to modify the dominance of lethals, while possible in bacteria and fingi, is, of course, impossible in a monokaryotic haploid: lethal mutations are immediately lost each time they arise regardless of their dominance in artificial diploids. It is difficult to see, therefore, how the recessivity oflethals in Chiamydomonas could result from natural selection.
One could object, however, that recessivity of Chlamydomonas mutations results from Fisherian selection in its brief diploid zygotic stage. Although unlikely, one can eliminate this possibility almost entirely. Many characters are simply not expressed in the zygote; indeed almost al of the mutations listed in Table 1 have no discernible phenotypic effect among zygotes, explaining why dominance tests must be performed in artificial, not meiotic, diploids. Perhaps the clearest and best studied of these "haploid-limited" characters is the flagellum. The mature zygote does not possess flagella; moreover, flagellar phenotype is determined solely by the haploid genotype (14). One can therefore compare the dominance of flagellar mutations in Chlamydomonas (some unknown fraction of which block some metabolic pathway) with all mutations in Drosophila. Although the sample size here is obviously smaller, the result is clear: these mutations are recessive just as frequently as those from a diploid organism despite the fact that they have never been subjected to selection in diploids (Table 2; G121 = 0.18; P = 0.91). Ibis result seems fatal to Fisher's theory of dominance. Interestingly, these data also falsify two other theories of dominance, both offered by Haldane (28, 29). In his first theory, which is closely related to Fisher's, Haldane (28) argued that natural selection provides a safety net against the heterozygous effects of mutations by replacing wild-type alleles that produce "just enough" enzyme with those producing "too much" enzyme. In short, the evolution of dominance may involve substitutions at the locus suffering the heterozygous effects of recurrent mutations, not substitutions at other modifier loci as Fisher had maintained. In his second theory, Haldane (29) suggested that, during a favorable mutation's sweep to fixation, modifiers might accumulate that render the new mutation dominant to the allele it is replacing. Both of these theories, like Fisher's, require selection on heterozygotes and so cannot explain dominance in haploid organisms.
Thus, neither Fisher's theory nor Haldane's theories of dominance can account for the recessivity of mutations in a haploid species. Indeed, the recessivity of mutations in Chlamydomonas falsifies any theory of dominance that invokes modification of heterozygotes by natural selection. In short, these data imply that most mutations are, from the beginning, recessive. Although natural selection might alter the precise activity of wild-type enzymes in particular cases, the present result lends strong support to the notion that the recessivity of mutations is a simple consequence of metabolism, as Wright suggested.
I thank B. Charlesworth, J. Coyne, J. Crow, J. Gillespie, C. Langley, T. Prout, and M. Turelli for comments and criticism. This work was supported by a Sloan Postdoctoral Fellowship in Molecular Evolution to H.A.O. and by National Institutes of Health Grant GM 38462 to J. Coyne.
Answers to all your Biology Questions
