Report

High Frequency of Cryptic Deleterious Mutations in Caenorhabditis elegans

See allHide authors and affiliations

Science  10 Sep 1999:
Vol. 285, Issue 5434, pp. 1748-1751
DOI: 10.1126/science.285.5434.1748

Abstract

Deleterious mutations with very small phenotypic effects could be important for several evolutionary phenomena, but the extent of their contribution has been unknown. Fitness effects of induced mutations in lines of Caenorhabditis elegans were measured using a system for which the number of deleterious point mutations in the DNA can be estimated. In fitness assays, only about 4 percent of the deleterious mutations fixed in each line were detectable. The remaining 96 percent, though cryptic, are significant for mutation load and, potentially, for the evolution of sex.

The prevalence of sexual reproduction may be explicable by the ability of sexual populations to sustain higher rates of deleterious mutation than asexuals (1). However, it has been argued that the accumulation of deleterious mutations could lead to fitness loss and ultimately to population extinction, even in sexual species, if effective population size is small (2). Mutation accumulation experiments inDrosophila suggest that mutations with harmful effects of the magnitude required for this fitness loss (∼1%) occur frequently (3), but joint estimates of the spontaneous genomic deleterious mutation rate, U, and the mean deleterious mutation effect, s, from these experiments have been questioned (4, 5). An alternative molecular constraint approach to estimate U compares rates of nucleotide substitution in functional and neutral regions of the genome (6), but does not give direct information on the magnitude of mutation effects. Here, we combine the molecular constraint and mutation accumulation approaches to infer the distribution of fitness effects of point mutations in the nematodeC. elegans.

We exposed the N2 (wild type) strain of C. elegans to 50 mM ethylmethane sulfonate (EMS) for 4 hours, a dosage for which the number of mutations induced at the DNA level has been calibrated (7). About 92% of the mutations produced by this treatment are G/C → A/T transitions; the remainder are other kinds of point mutations, plus a low frequency of small deletions (7). Data from experiments to measure forward mutation rates (8) and suppressor-induced reversion mutation rates (9–11) under EMS mutagenesis provide a mean estimate for the rate of transitions of 6.2 × 10−6 [95% confidence interval (CI) of (Table 1) 4.0 × 10−6 to 8.4 × 10−6] per G/C base pair. The haploid genome of C. elegans is 9.7 × 107 base pairs, comprising 36% G/C (12), so the mutagenesis generated an expected number of 220 transition mutations per haploid. The fraction of the genome in exons is 27% (12), and we infer from tables of codon usage in >15,000 C. elegans genes (13) that 30.6% of nucleotides in exons are G/C and generate an amino acid change if mutated to A/T. Therefore, we estimate that the EMS treatment generated an average of (6.2 × 10−6) × (9.7 × 107) × 0.27 × 0.306 = 50 (95% CI of 32 to 68) transition mutations per haploid that changed an amino acid in a protein-coding gene. Protein-coding sequences are under strong selective constraint in C. elegans: Data on synonymous and nonsynonymous substitution rates in C. elegans and its relative C. briggsae (14) suggest that ∼90% of amino acid mutations are removed by natural selection. Thus, we estimate that the minimum number of mutations induced per haploid genome that are deleterious in natural conditions is ∼45, but the number could be substantially higher because some noncoding DNA is selectively constrained (15). A comparison of the relative levels of selection against amino acid changes [using a log-odds (Dayhoff) matrix (16)] arising from G/C → A/T transitions to other types of point mutations (A/T → G/C transitions and transversions) suggests that the severity of individual amino acid changes generated by EMS is similar to that of spontaneous mutations (17).

Table 1

Rates of G/C → A/T transition mutations induced by EMS. The frequency of mutations is adjusted to a dosage of 50 mM EMS for 4 hours [as used here and in (9)] by assuming the EMS dose-response curve of (21) [(10, 11); 25 mM EMS, 4 hours], or assuming mutation rate is linear with time [(8); 50 mM EMS, 3.5 hours].

View this table:

After the mutagenesis, we bred 60 independent EMS lines toward homozygosity by selfing. In the absence of selection, the number of mutations fixed is expected to be the number of mutations induced per haploid genome. During the inbreeding, conditions were made as favorable as possible to avoid selection. However, some backup cultures were used, and four lines were lost, implying the action of natural selection (Fig. 1) and the selective loss of some mutations (such as recessive lethals). To estimate the magnitude of this loss, we performed computer simulations of lines undergoing multiple generations of selfing, analogous to our experimental design (18). The mean proportion of plates producing at least one progeny (surviving cultures) among the EMS-treated lines in our experiment was 0.74, with a 95% CI of 0.68 to 0.80 (Fig. 1). Under the parameter values simulated, only mutation effects of less than 10−3 are predicted to lead to viabilities within this range (Fig. 1). In our simulations, if lines had 45 heterozygous mutations at the start of inbreeding, such mutation effects led to the loss of 6% of mutations or fewer, on average, to selection (Fig. 1). This suggests that at most three mutations, two of which were minor-effect mutations, were selectively lost (on average) per line. Simulations assuming that fitness declines faster than exponentially with increasing numbers of mutations (synergistic epistasis) led to similar or lower estimates of the number of mutations selectively lost (17).

Figure 1

The N2 strain of C. elegans was obtained from the Caenorhabditis Genetics Center. Cultures were maintained using standard techniques (31) at 20°C on 3.5-cm agar plates, seeded with a suspension of E. colistrain OP50, and allowed to grow overnight. A large, synchronous population of young adults was divided into two subpopulations. One population was mutagenized with 50 mM EMS in M9 buffer for 4 hours at 20°C according to the protocol of (7). The control population was treated in an identical manner, but worms were instead maintained in M9 buffer. Worms were allowed to recover on plates for 24 hours; any eggs present during the mutagenesis would therefore have been expelled and hatched. The worms were then removed by washing with M9 buffer and treated with alkaline hypochlorite to harvest eggs, which were then transferred to fresh plates. Two days later, small numbers of worms were distributed onto individual plates to start 60 mutant and 40 control lines. Mutant and control lines were then selfed for 10 generations to fix mutations. Lines were propagated in parallel by the transfer of one hermaphrodite, chosen from a random position on each plate. Backup plates (two per EMS line, one per control) were used in cases where the parent failed to produce progeny. If no replicate produced offspring, a worm was substituted from the plate of the previous generation (kept at 14°C to avoid starvation), or up to a maximum of three generations prior. After 10 generations, lines were cryopreserved (31) at–80°C until the reproductive output assays. (A) Mean proportion of plates surviving per generation of selfing in control (solid line) and EMS (dashed line) lines. (B) simulation results for constant-effects model,h = 0; dashed line is the mean proportion of plates surviving per line (v); solid line is the proportion of mutations selectively lost (m l), as functions of mutation effect s m. Gray bar represents the 95% CI on v from the data, so values ofs m for which simulated v falls within it are compatible with the data; the range of proportion of mutations that may have been lost to selection can be inferred from them l values associated with theses m values. Dotted gray line corresponds to the maximum parameter values consistent with data: s < 10−3, m l < 0.06. Simulations run under a variety of dominance coefficients (0 ≤h ≤ 0.5) or assuming γ-distributeds m yield almost identical limits on sand m l.

We measured lifetime reproductive output of individual worms from the 56 surviving EMS lines and 40 control lines. This fitness measure includes the viability of the parents, their fertility, and the viability of offspring until the L3 to young adult stage. The EMS treatment reduced mean productivity by 34% (±4%), and it produced a large increase in the between-line variance (V G) and obvious changes in the distribution of line means that are consistent across the three replicates carried out (Fig. 2). Qualitatively similar results are obtained if age-specific reproductive output is converted to an intrinsic growth rate fitness measure (17). As a positive control for mutagenesis, we also measured the mutation rate at the unc-22 locus (19). Our estimated rate is 7.6 (±1.0) × 10−4 per haploid, a ∼1250-fold increase over the spontaneous background rate of 6 × 10−7(20).

Figure 2

(A) Mean productivity for the three assays. (B) Distribution of line means, averaged over replicates, for productivity. After mutagenesis, the cryopreserved lines were thawed, then individual replicates within each line were maintained for three generations. Productivity was measured contemporaneously in three replicates from each mutant and control line. In each replicate, four worms were allowed to lay eggs on a plate for ∼3 hours, then removed. After 48 hours, a single randomly picked worm was transferred onto a new plate, then transferred 48 hours later and at 24-hour intervals for the entire reproductive period. The offspring were counted manually 48 to 72 hours after the parental transfer. Each counter assayed one replicate per line, and the entire assay was performed three times. Lines were randomized and counters were unaware of line identity. Analysis of variance was used to estimate the between-line variance for productivity; additional effects fitted were measurer, assay number, and their interaction, which was nonsignificant. The between-line variance component was highly significant for the EMS lines (P < 0.001) but not the controls (P > 0.6). The EMS-induced mutational heritability wasV M/V E = 1.0 (±0.22), where V M =V G/2 and V E is the environmental variance (2426 worms2). The control and EMS mean productivities were 248.5 and 163.2 worms, respectively.

Estimates of U I, the EMS-induced mutation rate per haploid genome, and s, the homozygous effect on productivity, can be obtained from phenotypic data under a model of equal mutation effects, for which the overall change in mean fitness is ΔM = U I s and the increase in between-line variance from mutation isV G =U I s 2 (3). From our productivity data, these estimates areÛ I = ΔM 2/V G = 1.4 (±0.36) and ŝ =V GM = 24% (±3.9%). The number of mutations detected is therefore only slightly higher than the number of recessive lethal mutations induced by 50 mM EMS, estimated at ∼1 per haploid genome (21, 22). This contrasts sharply with our inference that each line contains an average of >45 deleterious point mutations. Clearly, the assumption made in these calculations that mutations have equal effects is violated.

To find distributions of mutation effects giving an improved fit to the data, we used maximum likelihood (ML) (23). Surprisingly, the best-fitting γ distribution is the equal-effects model (β → ∞); a γ distribution in which there are many mutations with small effects and a diminishing tail of larger effects gives a poorer fit than equal effects if the mutation rate is a variable in the model (Table 2). This result implies that the underlying distribution of mutation effects may be multimodal, because the γ distribution does not simultaneously allow a high frequency of slightly deleterious mutations and a “spike” of strongly deleterious mutations. The best-fitting model we obtained under ML, assuming our a priori estimate of U I = 45, is two classes of deleterious mutation effects, the first accounting for an average of 1.6 mutations per line (these had effects of 22%, as for the single-class equal-effects model under ML), whereas the remaining 43.4 mutations have virtually zero effect (Table 3). The maximum fitness effect that this class of weakly deleterious mutations can take is only 0.07% (Table 3). The distribution of mutation effects is therefore multimodal. If a number of deleterious mutations other than 45 is assumed, the estimated number and mean effect of strongly deleterious mutations remains the same (Û I1 = 1.6,ŝ 1 = 22%), whereas the maximum effect of the weakly deleterious mutations changes proportionately (for example, U = 20, ŝ 2< 0.17%; U = 80, ŝ 2< 0.04%).

Table 2

Estimates of mutation rates and effects withU I variable.

View this table:
Table 3

Estimates of mutation rates and effects, and fit of models to the data, with U I fixed at 45 and ML analysis (23).

View this table:

Distributions of EMS-induced and spontaneous (24) mutation effects on productivity therefore appear to be remarkably similar: The model of equal effects fits the data better than a γ distribution model in both cases, and ŝ was 21% (±4%) for spontaneous mutations (24) compared to 22% (±2%) here. Our results imply that the spontaneous genomic deleterious mutation rate U, 0.0026 (±0.0012) per haploid (24), could have been underestimated by a factor of at least 28 (that is, 45/1.6), and that U is therefore closer to ∼0.07. Our estimate for the mean mutation effect under an equal effects model is also nonsignificantly different from the positive mean mutational effect on productivity measured in a more recent spontaneous mutation accumulation experiment in C. elegans (25), that is, s = −24% ± 23%.

Diploid U must be greater than 1 for sexual reproduction to be maintained by deleterious mutations (1, 26). Because C. elegans reproduces primarily as a self-fertilizing hermaphrodite, it does not pay the full twofold cost of sex. However, our corrected estimate for the spontaneous deleterious mutation rate specific to protein-coding genes (∼0.14 mutations per diploid per generation) leads to the prediction that related sexual species would have higher mutation rates if deleterious mutations explain sex (although a maximum corrected U, assuming all 220 EMS-induced nucleotide changes in the genome are deleterious, is 0.72 per diploid). The high frequency of mutations of very small effect also has implications for the rate of fitness loss due to mutation accumulation. It has been assumed that mutations with fitness effects on the order of 1% are common (2, 26, 27), but our finding that >96% of mutations are undetectable in the laboratory, and have fitness effects of less than 0.07%, brings the validity of this assumption into question. We have inferred that there is a large class of deleterious mutations with tiny effects in the laboratory. Whether their effects are magnified in harsher natural environments remains to be determined; evidence for strong interactions between deleterious mutation effects and environmental conditions is equivocal (5, 28). It is clear, however, that mutation accumulation experiments may substantially underestimate mutation rates, and this will be undetected unless the number of events at the DNA level can be estimated.

  • * To whom correspondence should be addressed. E-mail: p.keightley{at}ed.ac.uk

REFERENCES AND NOTES

View Abstract

Navigate This Article