Mutation rate often increases with environmental temperature, but establishing causality is complicated. Asymmetry between physiological stress and deviation from the optimal temperature means that temperature and stress are often confounded. We allowed mutations to accumulate in two species of Caenorhabditis for approximately 100 generations at 18°C and for approximately 165 generations at 26°C; 26°C is stressful for Caenorhabditis elegans but not for Caenorhabditis briggsae. We report mutation rates at a set of microsatellite loci and estimates of the per-generation decay of fitness (ΔMw), the genomic mutation rate for fitness (U) and the average effect of a new mutation (E[a]), assayed at both temperatures. In C. elegans, the microsatellite mutation rate is significantly greater at 26°C than at 18°C whereas in C. briggsae there is only a slight, non-significant increase in mutation rate at 26°C, consistent with stress-dependent mutation in C. elegans. The fitness data from both species qualitatively reinforce the microsatellite results. The fitness results of C. elegans are potentially complicated by selection but also suggest temperature-dependent mutation; the difference between the two species suggests that physiological stress plays a significant role in the mutational process.
The relationship between metabolic rate, mutation and molecular evolution has generated much interest [1–3]. Competing hypotheses attribute the relationship to (i) generation time, or (ii) mutagenic by-products of cellular metabolism. However, other factors covary with generation time and metabolic rate, including body size, life history, population size and temperature, all of which potentially influence rate of evolution for reasons not causally related to metabolic rate or generation time. In particular, the mutagenic effects of high temperature are well-documented . However, many studies that identify a relationship between temperature and mutation may confound temperature with physiological ‘stress’. Several lines of evidence suggest that physiological stress is mutagenic [5–9] and that an upward deviation from an optimum temperature is often more stressful than an equivalent downward deviation . Thus, temperature dependence of mutation rate may be an indirect effect of stress rather than a direct effect or an effect mediated by metabolism.
To begin to disentangle the direct effects of temperature from those of its correlates, we allowed mutations to accumulate under relaxed selection (‘mutation accumulation’, MA) in two species of nematodes, Caenorhabditis briggsae and Caenorhabditis elegans, at 18°C and 26°C, for 103 and approximately 165 generations, respectively. The different temperatures are differently stressful for the two species; absolute fitness of C. briggsae at 18°C is about 60 per cent of that at 26°C, whereas absolute fitness of C. elegans at 26°C is about one-third of that at 18°C (see the electronic supplementary material, table S3). Mutation rate was assessed directly by genotyping a set of microsatellite loci chosen for their predicted high mutation rate. The cumulative effects on fitness were assessed by comparing MA lines to the cryopreserved common ancestor(s) at both temperatures. This design controls for selection mediated by population size and body size, and generation times and times of divergence are known, as is the relative degree of physiological stress. If the sole effect of temperature on the evolutionary process is via mediation of generation time, the per-generation mutation rate should not differ between the two MA temperatures. Conversely, if temperature affects the mutation process in other ways, the per-generation rate may differ between the two MA temperatures. If physiological stress is important, the relationship of mutation with temperature should differ predictably between the two species, with C. elegans having the higher mutation rate at 26°C.
2. Material and methods
(a) Mutation accumulation and fitness assay
The MA protocol follows Baer et al. ; see the electronic supplementary material, text S1. We initiated two sets of 192 replicate MA lines from the N2 strain of C. elegans and from the PB800 strain of C. briggsae; 96 lines were kept at 18°C and 96 at 26°C. Lines were maintained by transfer of a single hermaphrodite for 103 generations at 18°C for each strain and for 164 generations at 26°C in N2 and 171 generations in PB800.
Fitness assays also follow Baer et al. ; see the electronic supplementary material, text S2. Fitness was assayed in two blocks; 30 MA lines from each strain/MA temperature were randomly selected for each block, along with ancestral controls. Fifteen thawed worms were picked from each control and used to establish replicate control ‘pseudolines’. From each line, seven replicates were assayed for lifetime reproduction at 26°C and five at 18°C.
(b) Microsatellite genotyping
Sixteen AG(n) loci ≥ 9 repeats were selected from the upper 5 per cent of the length distribution in each species and matched for repeat number as closely as possible. DNA extraction, amplification and genotyping follow Phillips et al. . All surviving MA lines and their ancestral controls were genotyped at all loci. We found no cases of putative heterozygotes in either ancestor. Homozygous genotypes different from wild-type were re-amplified and re-genotyped for confirmation. Details of locus choice, primer design and genotyping are given in the electronic supplementary material, text S3 and table S1.
(c) Data analysis
Mutation rate is calculated as μ = n/lt, where n is the number of mutations, l is the number of MA lines and t is generations of MA . Because number of generations differs between the two MA temperatures, comparisons must be of mutation rates rather than of numbers of mutations. Within species, mutation rates at 18°C and 26°C were compared by paired-sample Wilcoxon signed-rank test; each locus at the two temperatures provides the paired observations. Indel spectra between temperatures and species were compared via a 2 × 2 contingency table, pooling over loci.
Relative fitness (w) is defined by the following equation: where lxmx is the product of survivorship and fecundity at day x and r0 is the mean intrinsic rate of increase of the G0 control, calculated by solving the following equation: using the average lx and mx values of all control lines in an assay block. The per-generation change in the trait mean, ΔMw = U × E[a], where U is the genome-wide mutation rate and E[a] is the average effect of a mutation on the trait . Ancestral relative fitness w0 is defined equal to 1, so
We generated 1000 bootstrap replicates (resampling lines) to estimate ΔMw for each strain/MA temperature/assay temperature group, maintaining block structure and averaging over blocks. ΔMw is considered to differ significantly between groups if the empirical 95 per cent confidence limits of the groups do not overlap. See the electronic supplementary material, text S4 for details.
The ratio of (twice) the squared change in the trait mean (ΔM) to the per-generation increase in the among-line variance (the mutational variance, VM) provides a downwardly biased estimate of the genomic mutation rate U and VM/2ΔM provides an upwardly biased estimate of the average mutational effect, E[a], the ‘Bateman–Mukai’ (B–M) estimators Umin and E[a]max . VM, Umin and E[a]max were calculated from the resampled data described earlier; details and some caveats are provided in the electronic supplementary material, text S5 and S6. The limitations of the B–M method are well-appreciated .
The complete data are in the electronic supplementary material, table S2. In C. briggsae, the mutation rate per-generation does not differ significantly between MA18 and MA26 treatments (table 1) and is very similar to a previous estimate from the same set of loci from PB800 MA lines propagated at 20°C (2.13 × 10−4 per generation; ). By contrast, in C. elegans, the per-locus mutation rate in the MA26 lines is greater than in the MA18 lines (one-tailed p < 0.006). One C. elegans in MA26 line (line 421) had an atypically high number of mutations (9/15 loci; the next highest number of mutations per line is 3/15). With line 421 omitted, the difference between the two MA treatments is smaller (1.96 : 1 versus 2.50 : 1) but remains significant (one-tailed p < 0.05). Because the sets of loci are not orthologous in the two species and were not chosen at random, formal comparison between species is inappropriate. The approximately twofold greater mutation rate in C. briggsae than C. elegans in the MA18 treatment (2.65 : 1) is consistent with a previous estimate in which mutations accumulated at 20°C (2.27 : 1; ).
In C. briggsae deletions are more common than insertions, whereas in C. elegans insertions predominate (table 1). The difference between species in the direction of indel bias is similar at both temperatures, is highly significant (likelihood-ratio χ2, p < 0.001) and is consistent with previous results .
Results are summarized in table 2; detailed results are presented in the electronic supplementary material, tables S3 and S4. There are three relevant two-way comparisons: between MA treatments within an assay temperature/species; between assay temperatures within a MA treatment/species; and between species within a MA treatment/assay temperature. The species evolve qualitatively differently: in C. briggsae, on average, MA26 lines decline in fitness (ΔMw) about twice as fast as MA18 lines, and the result is consistent across assay temperatures. By contrast, in C. elegans, ΔMw is about two times larger when fitness is assayed at 26°C than at 18°C. However, there is a substantial variation between blocks, and the differences between groups approach significance (p ∼ 0.05) only between the C. briggsae MA18 and MA26 lines when assayed at 26°C.
Two broad patterns emerge from the B–M estimates. First, in both species at both assay temperatures, Umin is greater in MA26 than MA18, and second, in every case E[a]max is smaller when assayed at 26°C than at 18°C, although the differences between assay temperatures are trivial in the C. briggsae MA26 lines.
Three features of the microsatellite results are consistent with previous findings: (i) at cool, non-stressful temperatures (20°C in Phillips et al.  and 18°C here), the per-locus mutation rate of C. briggsae is about twice that of C. elegans; (ii) those rates are similar in the two studies; and (iii) the two species differ consistently in the direction of indel bias. The significantly higher per-generation mutation rate in the MA26 lines compared to the MA18 lines in C. elegans, coupled with the lack of a similar difference between the two MA temperatures in C. briggsae, has two implications. First, the difference between species in the relationship between MA temperature and mutation rate strongly suggests that there is not a universal relationship between metabolic rate and mutation rate as implied by proponents of a ‘global molecular clock’ . Second, physiological stress is implicated as a cause of elevated mutation rate in C. elegans. Agrawal and his co-workers [6,9] have convincingly demonstrated that Drosophila melanogaster in poor condition accumulate mutations more rapidly than flies in good condition, and that a likely cause is the preferential use of an error-prone DNA repair mechanism by individuals in poor condition. Moreover, Muller  observed an almost identical, twofold difference in lethal mutation rate between D. melanogaster maintained at 26.5°C—near the upper limit of D. melanogaster's thermal tolerance—and flies maintained at 19°C.
Taken as a whole, the fitness data suggest that (i) mutations do accumulate at least a little faster at high temperature, and (ii) more mutations with smaller effects contribute to the mutational decay of fitness at 26°C than at 18°C. Thus, there is evidence that high temperature per se is mutagenic to some extent. The pattern is much more pronounced in C. elegans than in C. briggsae, which further suggests that physiological stress either exacerbates the mutagenic effects of temperature or is itself mutagenic. These results are quite consistent with the microsatellite data.
Two additional observations suggest that some aspect of the mutational process in C. elegans is temperature-dependent. First, many more C. elegans MA26 lines were lost during the MA phase than in the other groups (19/96 versus ≤3/96). Second, Ne in the C. elegans MA26 lines was ≈ 2, whereas in the other three groups Ne ≈ 1, the result of having to use backups more frequently owing to much higher mortality (see the electronic supplementary material, text S1). These results suggest that the larger Ne in the MA26 lines leads to a class of mutations that are effectively neutral (4Nes < 1) in the MA18 lines that are purged by selection in the MA26 lines. Mutations in this window (12.5% < s < 25%) contribute significantly to the mutational decay of fitness in C. elegans ; note that in both species, E[a]max in MA18 assayed at 18°C is approximately 13 per cent (table 2). Taken together, the evidence suggests that the mutation rate is much more strongly temperature-dependent in C. elegans than in C. briggsae, which in turn suggests a predominant role for physiological stress in the mutational process.
We thank the Baer Laboratory Worm Crew for picking and counting, especially F. Cadavid, S. Lewis, L. Sylvestre, B. Tabman and A. Upadhyay; we thank J. Gillooly for spirited discussion. Support was provided by NIH (grant no. R01GM072639) to C.F.B. and D. Denver and NSF (grant no. DEB-0717167) to C.F.B.
One contribution to a special feature on ‘Experimental evolution’ organized by Paul Sniegowski, Thomas Bataillon and Paul Joyce.
- Received April 10, 2012.
- Accepted July 10, 2012.
- This journal is © 2012 The Royal Society