Deleterious Mutations and the Evolution of Sex

See allHide authors and affiliations

Science  13 Oct 2000:
Vol. 290, Issue 5490, pp. 331-333
DOI: 10.1126/science.290.5490.331


It has been suggested that sexual reproduction is maintained because it reduces the load imposed by recurrent deleterious mutations. If rates of deleterious mutation per diploid genome per generation (U) exceed 1, and mutations interact synergistically, then sexuals can overcome their inherent twofold disadvantage. We have tested this hypothesis by estimating genomic point mutation rates for protein-coding genes in a range of animal taxa. We find a positive linear relationship between U and generation time. In species with short generation times, U is predicted to be far below 1, suggesting that sex is not maintained by its capacity to purge the genome of deleterious mutations.

Almost all species indulge in at least a little sex, the exchange of genetic material between individuals. In higher plants and animals, this generally involves anisogamy, the production of gametes of different sizes, and in its most derived form, the production of two sexes. Anisogamy incurs a cost, since resources directed towards male gametes, or male offspring, could be directed towards female gametes, or female offspring. For example, a parthenogenetic female who only produces female offspring will have a twofold advantage over her sexual conspecifics. It is this cost of sexual reproduction which has so troubled evolutionary biologists; why do so many species, what we will call “obligate sexuals,” produce two different types of gametes, or different sexes, every generation?

Sexual populations can have two principal advantages over asexuals: they can adapt more rapidly to changing environments and they are less prone to the accumulation of deleterious mutations (1). Sexuals accumulate fewer deleterious mutations because they avoid Muller's ratchet (2), but the ratchet is unlikely to maintain obligate sexuality, since it operates very slowly in large populations and can be avoided with very little sex. However, much attention has been focused on the potential benefits of sex in reducing the mutation load if mutations interact synergistically (3), for three reasons. First, the effect is independent of population size, a property which has led to the theory being called the “the mutational deterministic” (MD) hypothesis (1). Second, the benefits of sex can be large; if U exceeds 1, then sexual populations can overcome their twofold disadvantage over pure asexuals. This is a minimum condition, since U needs to be rather greater than 1 if stochastic factors are taken into account (4,5), or if obligate sexuals are to have an advantage over facultative sexuals. And third, the hypothesis is testable; if many obligate sexual populations have deleterious mutation rates below 1, then obligate sex is unlikely to be maintained by selection against deleterious mutations. Here, we estimate deleterious mutation rates for a variety of organisms in an attempt to test the MD hypothesis.

Laboratory mutation accumulation experiments provide estimates for rates of mutations with moderate phenotypic effects, but such experiments cannot be used to test the MD hypothesis, since mutations of small effect are missed (6, 7). A radically different approach is to compare the genomes of related species (8). In a neutral segment of DNA, the rate of nucleotide substitution is expected to equal the mutation rate. The substitution rate in randomly sequenced stretches of DNA is expected to be lower than the neutral rate, due to selective constraint on functional elements. The difference between the substitution rates in truly neutral and randomly sequenced segments therefore could estimate the rate of selective elimination of mutations from the randomly sequenced segments (8). Here, we apply a simplified approach that restricts the analysis to protein-coding gene sequences (9). Under the assumption that synonymous mutations are neutral, the nucleotide mutation rate can be estimated from the synonymous substitution rate (K s). The amino acid mutation rate can then be estimated from the product ofK s and the number of sites in the gene that, if changed, lead to an amino acid substitution. The number of selectively eliminated amino acid mutations is the difference between the estimated amino acid mutation rate and the observed number of amino acid substitutions. The estimate can be scaled up to the whole genome, if the number and average length of protein-coding genes are known, and expressed per generation, if estimates are available for the evolutionary divergence time and generation interval.

We obtained estimates for the genomic amino acid mutation rate (M) and genomic deleterious mutation rate (U) for six pairs of vertebrate species and three pairs of insects for which divergence dates can be inferred from the fossil record, biogeographical data, or a locally calibrated molecular clock (10). We corrected for the effect of selection on synonymous codon use in Drosophila by regressingK s against codon bias (11). ForD. melanogaster/D. simulans, the regression coefficient is nonsignificant, and this procedure has little effect on our estimate ofU; for D. picticornis/D. silvestris, there are insufficient data, but for D. melanogaster/D. pseudoobscura, the regression is significant (P < 0.05), and the corrected estimate is approximately 1.5-fold greater as a consequence. Other evidence (12) suggests that selection on synonymous codon bias is sufficiently weak in Drosophila, at least, to have little effect on the synonymous substitution rate.

Surprisingly, the estimates of mutation rates for the protein-coding genome per year are approximately constant across a broad range of taxa (Table 1). With the exception of mouse/rat, for which there is debate about the divergence time (13, 14), there is just over threefold variation in estimates of M per year (approximately sixfold inU per year). This does not just reflect the molecular clock, because the total length of the protein-coding sequence in insects is considerably smaller than that in vertebrates. Instead, it seems to reflect an inverse proportionality between the nucleotide mutation rate per cell division and the total length of protein-coding sequences in the genome (15). However, the estimates of M andU per generation vary by ∼90-fold and ∼50-fold, respectively (Table 1). Rates for M and Uwithin mammals also vary substantially by ∼11-fold and ∼6-fold, respectively. The amino acid mutation rates agree with independent estimates of M obtained from the rate of appearance of new electrophoretic mutations [2.2 and 0.09 mutations per genome per generation in humans and Drosophila, respectively (6)].

Table 1

Estimates of K s, the synonymous divergence; M′ and U′, genomic amino acid and deleterious mutation rates, respectively, per year; andM and U, genomic amino acid and deleterious mutation rates, respectively, per generation. We assume 80,000 genes, averaging 1500 bp in length, for vertebrates (34) and 13,600 genes, averaging 1770 bp in length, for insects (20).

View this table:

The constancy of the mutation rate per year leads to a highly significant positive correlation (P < 0.001) between both M and U per generation and generation time (Table 1 and Fig. 1); the relationship appears to be linear on both an untransformed and a log-log scale. This parallels a positive relationship between new mutational variance for quantitative traits per generation (expected to be linear in M) and generation time (7). The relationships between U andM and generation time (Fig. 1) are insensitive to generation time estimates, because U and M are expressed per generation. In mammals with long generation times, and particularly primates, U is sufficiently high to maintain obligate sexuality, if they were able to become asexual, but genomic imprinting probably prevents this. However, many obligate sexuals have generation times on the order of 1 month or less [e.g., many insects (16), copepods (17), and nematodes (18)], implying deleterious amino acid mutation rates of the order of 0.05, some 20 times lower than the MD hypothesis requires. The deleterious mutation rate in protein-coding sequences is therefore predicted to be so low in many obligate sexuals that the MD hypothesis can be seriously questioned.

Figure 1

Relationship between the genomic deleterious mutation rate per generation (U) and generation time with best-fitting linear regression line. The linear regression is also significant for the mammals plus bird subset (P < 0.02).

The estimates for U (Table 1 and Fig. 1) are for point mutations in protein-coding genes, but mutations in noncoding DNA and insertion/deletion mutations also contribute to the deleterious mutation rate. However, these additional sources do not appear to be sufficient to increase U to levels which would be consistent with the MD hypothesis. Currently, the only estimate for the level of genome-wide selective constraint in noncoding DNA comes from a comparison of the nematodes Caenorhabditis elegans andC. briggsae (19). Mutation events in noncoding DNA seem to contribute a similar total number of deleterious mutations as mutations that occur in coding DNA. If this finding is general, then the estimates of U in Table 1 need to be multiplied by 2 to account for mutations in noncoding DNA. U is less than 1 inDrosophila, even if we assume that all point mutations throughout the genome are deleterious {U = 0.8, obtained by multiplying the average estimate for the synonymous substitution rate per generation [corrected (11)],Ks = 2.2 × 10−9, by 3.6 × 108, the number of nucleotides in the diploidDrosophila genome (20)}.

Small insertion and deletion mutations are relatively infrequent in all taxa that have been studied and are estimated to contribute an additional ∼10% to the estimates of U (21,22). However, transposable element (TE) insertions are a major source of mutation in some organisms. For example, it has been estimated that spontaneous TE movement in Drosophilaproduces up to 0.2 new insertions per diploid per generation (23,24). Most of these appear to be deleterious, because few TE insertions are fixed within Drosophila species (24). However, it is thought that most TE insertions are deleterious because of deletions caused by ectopic recombination, and not because they disrupt gene sequences (24). If meiosis is suppressed in new asexual lineages, as often appears to be the case (25), then ectopic exchange will be eliminated, and an asexual lineage will gain a benefit relative to sexuals. Thus, the deleterious effects of TE insertions brought about by ectopic exchange may favor asexuals over sexuals. Furthermore, the rate of transposition is expected to evolve to a lower rate in asexual lineages, because of the elimination of horizontal transfer.

The variation across taxa in the amino acid mutation rate per year is a little less than the variation in the deleterious mutation rate per year, reflecting a significant (Spearman's rank correlation coefficient r s = –0.854, P< 0.02) negative correlation between the fraction of amino acid mutations eliminated by selection (U/M) and generation time (Fig. 2). The low value ofU/M for human/chimpanzee genes is partly due to unusual properties of the genes that have been sequenced (6,26); however, the rank correlation remains significant when the human/chimpanzee point is excluded (r s = –0.790, P < 0.04). The negative relationship between U/M and generation time may reflect higher rates of fixation of slightly deleterious mutations in organisms with small population sizes (27), as population size is negatively correlated to generation time (28).

Figure 2

Relationship between estimated fraction of amino acid mutations eliminated by selection (U/M) and generation time.

The MD hypothesis has stood out as one of the few hypotheses for the evolution of obligate sexuality that could be tested simply (1). However, based on our deleterious mutation rate estimates, the hypothesis is not supported as a general mechanism. While this leaves many other hypotheses to test, they all share a common feature: it is adaptive evolution that principally drives the evolution of sex, perhaps in combination with other mechanisms (29).


View Abstract

Navigate This Article