A study of roundworms shows that genomic mutations occur surprisingly frequently, and that the kinds of changes involved differ from those predicted. Are genomes inherently less stable than previously suspected?
DNA carries the coded information that specifies the size, shape, body plan and many other basic characteristics of most organisms. To transmit these characteristics faithfully, DNA must pass from generation to generation with relatively few mutations. But mutations do happen, and can have profound consequences. These include inherited diseases, cancer and drug-resistant infections, but also the genetic differences among individuals that, through natural selection, drive evolution.
Until now, mutations seemed to be relatively rare and to occur in a characteristic spectrum. But such observations are challenged in the paper by Denver and colleagues on page 679 of this issue1. These authors used a particularly powerful way to hunt for mutations in the roundworm Caenorhabditis elegans (Fig. 1) — and found at least ten times more mutations, and a different assortment, than anticipated.
Traditional ways of estimating mutations are indirect2, involving either phylogenetic studies of wild organisms or phenotypic methods in the lab. Phylogenetic studies involve comparing DNA sequences between species and estimating the number and kinds of changes that have occurred since the species diverged. Phenotypic methods rely on the ability of some mutations to change a trait (phenotype) of an organism. After a defined number of generations, rare mutants carrying the new trait are quantified, and mutation rates are calculated and then extrapolated to predict rates for the whole genome. This extrapolation takes into account the genome's size and the fraction of mutations that has been estimated to produce phenotypic change (about one-third)2.
However, both approaches probably underestimate the inherent mutation rate and skew the variety of mutations found. For instance, some mutations are harmful, and so the organisms that carry them are less likely to contribute to the next generation (they are ‘selected against’), both in the wild and in large cultures. And the fraction of mutations that produces no phenotypic change might be larger than imagined.
Denver et al.1 bypassed the pheno-type-bias problem by directly sequencing randomly chosen stretches of DNA in laboratory-grown worms. They also minimized selection against harmful mutations by maintaining many lines of worms, separating a single worm from each progeny and allowing it to produce the next generation by self-fertilization, without competing with other worms. Rapid and severe loss of fitness occurs in these worms because, when their numbers are reduced to one repeatedly, random mutations become fixed — a phenomenon known as Muller's ratchet3.
From these pampered worms, Denver et al. sequenced four million base pairs of DNA, and found 30 new mutations compared with the original animals. This equates to a rate of 2.1 mutations per genome per generation. This rate is at least ten times higher than those reported previously in worms and other DNA-based organisms, which curiously maintain constant predicted mutation rates per genome, irrespective of genome size2. Moreover, the kinds of mutations differ from those previously seen in many organisms, even worms. Why so many and such different mutations?
Denver et al. suggest that the crucial difference lies in reducing the genetic selection that biases the accumulation of mutations. Specifically, they find that insertions of one to a few bases of DNA are more common than previously reported, and argue that these findings, when compared with the results of phylogenetic studies, suggest that larger genomes are selected against in the wild. This might be true. But Denver et al. also found more insertions than are seen in lab-based studies2, in which selection for small genome size should be minimal.
We suggest two other explanations. First, previous estimates of how many mutations give rise to phenotypic changes might be too low. Phenotypes might be masked by variables not factored in, such that mutation rates are higher than predicted from lab-based data — as Denver et al. observe. For example, phenotypes of altered proteins are known to be masked, or ‘buffered’, by molecular chaperones, molecules that spruce up distorted proteins into better approximations of their functional shapes. Loss of available chaperones — in response to ‘heat shock’, for instance — reveals many new phenotypes, produced by previous mutations, that are invisible with chaperones present4. Also, the numbers and kinds of mutations to which specific DNA sequences are prone might be an evolved trait5; perhaps most genes are actually less prone to mutations that cause phenotypic changes. So, correction factors for the number of genomic mutations per phenotypically detectable mutation may need to be revised upward. Denver and colleagues' data could help to provide an empirical correction factor.
A second, not mutually exclusive idea is that a state of increased mutability (a ‘mutator’ state) is induced in the worms specifically because they have accumulated harmful mutations through Muller's ratchet. We suggest that these mutations provoke cellular stress responses that, in turn, cause further mutations. The idea that Muller's ratchet can provoke general stress responses is supported by studies of the bacterium Buchnera, which endures severe population ‘bottlenecks’ and induces the production of a protein involved in the heat-shock/protein-stress response6. And in yeast, most gene-inactivating single mutations cause decreased fitness and altered patterns of expression of other genes7, perhaps reflecting the activation of stress responses.
Furthermore, at least two stress responses have been documented to cause mutations: the so-called SOS response to DNA damage in bacteria8; and the bacterial general-stress response, controlled by the RpoS protein9,10, which increases the activity of some 50 genes in response to starvation and to temperature, pH, osmotic and oxidative stresses. One of these genes encodes a DNA-synthesizing enzyme11 whose error-prone nature might underlie some of the mutation-inducing effects of this stress response9,10. Moreover, as mentioned above, the heat-shock stress response exposes phenotypic variation4 — but might also cause genetic variation by increasing the mutation rate. We predict that the heat-shock response will indeed be found to cause mutations, and that many other general-stress responses will do so too.
It makes sense for stress responses to cause mutations; it may be a ‘selected’ feature that increases genetic variation, thus increasing ‘evolvability’ under stress when organisms are suboptimally adapted to their environments. Most of the mutations would be harmful or neutral, but rare adaptive mutations would also occur. Rare individuals in large stressed populations could thereby flourish, then, later, turn off their stress responses and readjust their mutation rates downwards to achieve greater genetic stability. Mutation-inducing stress responses might also underlie most cancers, which do not acquire mutator mutations, but still accumulate surprisingly high numbers of tumour-promoting alterations. Curiously, non-mutational gene-silencing events are common early in tumour progression, whereas mutations are more common later12. Perhaps this is because early gene-inactivating events cause mutation-inducing stress responses. Denver and colleagues' method and findings may provide windows on both of these important processes.
The nature of the alterations found by Denver et al.1 supports the idea of a connection between mutations and stress responses. They strikingly resemble the mutations seen in cells lacking DNA ‘mismatch repair’13, which corrects replication errors and is debilitated during a mutational stress response. Both the worms and mismatch-repair-defective cells show 10–100 times more mutations than normal worms or cells; more small sequence insertions and deletions than base substitutions; more insertions and deletions in single-base repeats; and more ‘transitions’ than ‘transversions’ among base substitutions. We suspect that loss of mismatch-repair capacity is a general feature of mutation-promoting stress responses, as observed in starved bacteria14, in which a special error-prone DNA polymerase enzyme, induced by the SOS response8 and starvation11, causes increased mutation15. In general, whenever excessive replication errors occur, mismatch repair should become exhausted14,15,16. So Denver and colleagues' creatures may have wormed their way into genetic instability. Whatever the mechanism that accounts for their altered DNA, they are telling us something important about mutation.
About this article
Nature Reviews Genetics (2004)