Uploaded by Nelson Sarmiento

The Evolution of Mutation Rates from Elsevier

advertisement
Review articles
The evolution of mutation rates:
separating causes
from consequences
Paul D. Sniegowski,1* Philip J. Gerrish,2 Toby Johnson,3 and Aaron Shaver1
Summary
Natural selection can adjust the rate of mutation in a
population by acting on allelic variation affecting processes of DNA replication and repair. Because mutation
is the ultimate source of the genetic variation required for
adaptation, it can be appealing to suppose that the
genomic mutation rate is adjusted to a level that best
promotes adaptation. Most mutations with phenotypic
effects are harmful, however, and thus there is relentless
selection within populations for lower genomic mutation
rates. Selection on beneficial mutations can counter this
effect by favoring alleles that raise the mutation rate, but
the effect of beneficial mutations on the genomic mutation rate is extremely sensitive to recombination and is
unlikely to be important in sexual populations. In
contrast, high genomic mutation rates can evolve in
asexual populations under the influence of beneficial
mutations, but this phenomenon is probably of limited
adaptive significance and represents, at best, a temporary reprieve from the continual selection pressure to
reduce mutation. The physiological cost of reducing
mutation below the low level observed in most populations may be the most important factor in setting the
genomic mutation rate in sexual and asexual systems,
regardless of the benefits of mutation in producing new
adaptive variation. Maintenance of mutation rates higher
than the minimum set by this ``cost of fidelity'' is likely
only under special circumstances. BioEssays 22:1057±
1066, 2000. ß 2000 John Wiley & Sons, Inc.
Introduction
The genetic variation that is acted on by natural selection is a
product of two processes: mutation and recombination.
1
Department of Biology, University of Pennsylvania, Philadelphia, PA
19104.
2
Theoretical Biology and Biophysics, Los Alamos National Laboratory.
3
Institute of Cell, Animal, and Population Biology, University of
Edinburgh.
Funding agencies: P.D.S. is supported by award 98-4-3-ME from the
Alfred P. Sloan Foundation and by NSF DEB 9981518; P.J.G. is
supported by a Director's Fellowship Award at Los Alamos National
Laboratory; T.J. was supported by BBSRC Studentship number
97/B1/G/03163; A.C.S. is supported by a Howard Hughes Predoctoral
Fellowship.
*Correspondence to: Paul Sniegowski, Department of Biology,
University of Pennsylvania, Philadelphia, PA 19104. Email:
paulsnie@sas.upenn.edu
BioEssays 22:1057±1066, ß 2000 John Wiley & Sons, Inc.
Analysing how natural selection affects rates of mutation and
recombination is a complex task, because these processes
affect both the fitnesses of individuals and the evolutionary
potential of populations.(1) A large body of literature has been
devoted to the evolution of sexual recombination, and debate
persists as to the major factors explaining its evolutionary
maintenance(2,3) (see also the commentaries accompanying
Ref. 3). The evolution of mutation rates has received less
attention, but significant progress has been made and an
outline of the major evolutionary factors affecting the mutation
rate has been presented.(4) Here we review these factors and
attempt to evaluate their relative importance. We emphasize
the importance of logically separating the causes of mutation
rate evolution from the subsequent adaptive consequences of
altered mutation rates.
There is ample evidence for genetic variation in the general DNA repair and replication processes that affect mutation,(5 ±10) and thus there can be little doubt that the genomic
rate of mutation in a population can be influenced by natural
selection. Because mutation is required for long-term adaptive
evolution, it might seem obvious that the need for new
beneficial mutations is of primary importance in setting the
genomic mutation rate. The problem with this viewpoint is that
the vast majority of mutations that have any phenotypic effect
are likely to be deleterious to individual fitness.(11) The influx of
deleterious mutations maintains continual selection pressure
within populations in favor of lower genomic mutation rates.
Indeed, it is likely that this selection to decrease mutation rates
was the cause of some major evolutionary transitions such as
the use of DNA instead of RNA as the hereditary molecule and
the evolution of complex enzymatic proofreading and repair
systems.(12) As we discuss in more detail below, selection to
decrease the deleterious mutation rate is likely to be much
stronger than selection to increase the beneficial mutation rate
under most circumstances.
Nonetheless, mutation persists in populations. A.H. Sturtevant thus highlighted the essential problem of mutation rate
evolution in 1937 when he asked ``Why does the mutation rate
not evolve to zero?''(13) In the simplest possible analysis, a
modern answer to Sturtevant's question relies on either (1)
physicochemical or physiological constraints on evolutionary
reductions in mutation rate to below the observed levels, or (2)
selection for an increased rate of production of beneficial
BioEssays 22.12
1057
Review articles
Figure 1. Evolutionary forces affecting the genomic mutation rate. Deleterious mutations (red arrow) decrease fitness, resulting in
continual selection for a lower mutation rate. Two possible selective forces constrain the mutation rate from evolving to zero: (1) the
increased probability of acquiring beneficial mutations under a higher mutation rate (green arrow), and (2) constraints on the fidelity of
replication (blue arrow). A: Intuition might suggest that the mutation rate is set by a tradeoff between the effects of deleterious and
beneficial mutations (green line), since mutation is required for long-term adaptation. Recombination, however, weakens the effect of
beneficial mutations on mutation rate modifiers (see Fig. 2 and Box 1) and thus the mutation rate in many populations may instead be set
by a tradeoff between deleterious mutations and constraints on the fidelity of replication (blue line). B: Little is known about the cost of
fidelity. In both sexual and asexual populations, this cost may be sufficient to set the prevailing mutation rate higher than the value that
would be set by beneficial mutations, since decreases in mutation would be costly to individual fitness.
mutations. The challenge in understanding mutation rate
evolution lies in evaluating the relative importance of these
factors, as illustrated in Fig. 1.
Theoretical progress in understanding the genetical evolution of mutation rates has been achieved by explicitly considering the effect of natural selection on the frequencies of
alleles that modify the mutation rate (mutation rate modifiers)
in populations. Selection on a mutation rate modifier can be
classified as either direct or indirect. Direct selection is
theoretically straightforward and depends on the effect (if
any) of the modifier allele on fitness through factors other than
its effect on mutation. Indirect selection, in contrast, depends
on nonrandom association (termed ``linkage disequilibrium'')
between the modifier allele and alleles at other loci affecting
fitness. Because linkage disequilibrium in a population is
rapidly eroded by recombination, the efficacy of such indirect
selection on the mutation rate is highly dependent on the
recombination rate. In particular, because beneficial mutations are expected to be rare compared with deleterious
mutations, indirect selection to increase the mutation rate is
greatly weakened by recombination (see Fig. 2), whereas indirect selection to decrease the mutation rate is less affected.
The mutation rate that evolves in a population thus depends
on direct and indirect selection on modifier alleles, and the
strength of indirect selection depends on the rate of recombination in the population. In the remainder of this review, we
consider in more detail how these factors affect the mutation
rate. We begin with a discussion of the evolution of equilibrium
1058
BioEssays 22.12
Figure 2. Indirect selection on the mutation rate. A modifier
that increases the mutation rate (red circle) tends to be
preferentially associated with (in positive linkage disequilibrium with) a beneficial allele arising by mutation (green
square). A: With complete linkage between the two loci, the
modifier can hitchhike(75) along with the beneficial allele as it
sweeps to fixation in the population. B: Recombination
disrupts the association between the modifier and the
beneficial allele and decreases the probability of hitchhiking.
Note that deleterious mutations are not shown; their
prevalence creates a continual indirect selection in favor of
reduced mutation rates, as described in the text and in Box 1.
Review articles
genomic mutation rates in both sexual and asexual populations; this is the area in which most theoretical work has been
concentrated. Next we consider the sporadic evolution of high
mutation rates in asexual populations; most empirical studies
have been concentrated in this area. Then we briefly discuss
some proposed ways in which the adaptive advantage of a
high beneficial mutation rate may be achieved without a
substantial increase in the rate of deleterious mutation. Finally,
we suggest avenues for further research.
Equilibrium genomic mutation rates
As mentioned above, most theoretical work has concentrated
on calculating the genomic mutation rate (hereafter mg)
expected at equilibrium in an evolving population. This focus
Box 1: Indirect Selection on a Modifier Allele
Deleterious alleles in a two-locus model
The simplest model is a two-locus, two-allele model. The first locus is the modifier locus with alleles m and M, which have no
direct effects on fitness. Indirect selection at the modifier locus is caused by linkage disequilibrium between it and a second
locus, the selected locus with alleles a and A. Let the fitness of an Aa diploid (or a haploid) individual, relative to an AA (or A)
individual, be (1 ÿ sd). We assume that the a allele is rare so that aa diploids can be ignored and that mutation from a to A is
negligible. If the M allele modifies the rate of mutation from A to a by a small amount Dm then the indirect selection coefficient
(c) for the M allele is approximately ÿ Dm sd /(sd ‡ r) where r is the recombination rate between the two loci.(14,15) A modifier
allele that increases the mutation rate is disfavoured (c < 0), and a modifier allele that decreases the mutation rate is
favoured (c > 0). This illustrates the more general ``reduction principle'', which states that when there is an unchanging
viability selection regime in an infinite random-mating population, a rare modifier can only invade if it causes a net reduction
in the mutation rate.(71±73)
When r ˆ 0 the two-locus model can be interpreted as a model of a single locus (with 4 alleles) which can evolve a higher
or lower mutability. The strength of indirect selection on mutability is given by c ÿ Dm and is independent of the strength of
direct selection. Since per-locus mutation rates are small (10 ÿ 5 or less), this force is weak and is likely to be overwhelmed by
drift in small populations or by selection at linked loci in asexual populations.
Deleterious alleles in a multilocus model
For a modifier of mg with small effect, approximate results can be obtained by multiplying the indirect selection over loci. Since
only loci under selection are relevant, consider a modifier that increases the genomic deleterious mutation rate from U to
U ‡ DU. In asexual populations the modifier has marginal fitness 1 ‡ c exp( ÿ DU) relative to the wild type (but see Ref.
38). This is independent of the selection coefficients at the selected loci, because in a sufficiently large population, each
deleterious allele is ultimately removed by selection, and a modifier unable to escape by recombination will suffer the same
fate.
For a sexual population, consider deleterious mutations occuring at random positions over n chromosomes each 100M
centimorgans long (with nM > 1). In this case, the modifier has marginal fitness approximately
ln ‰1=2 sd Š
1 ‡ c exp ÿU sd 1 ‡
nM
relative to the wild type.(28)
Comparison with indirect selection due to novel beneficial mutations
Let us now consider occasional beneficial mutations with selection coefficient sb in a population of effective size Ne. When
the indirect selection on a weak modifier is c ˆ (B DU ‡ D DU ) due to beneficial and deleterious mutations respectively, then
B ˆ D(U*/U ) where U* is the evolutionary equilibrium mutation rate in the absence of any cost of fidelity.(28) For his model,
Johnson(28) estimated that (for nM > 1)
U K
sb
1
sd ln ‰4Ne sb Š nM ‡ ln‰1=2 sd Š†
which suggests that for most populations U* K where K is the rate of selective sweeps per population per generation. From
nucleotide substitution data,(74) we can be reasonably confident that K U and therefore (U*/U ) 1 and B D. Therefore,
the cost of fidelity must maintain U at the observed levels.
BioEssays 22.12
1059
Review articles
Figure 3. Estimates of mutation rates per-genome (mg, upper points) and per-base-pair (mbp, lower points), plotted against genome
size on a log-log scale. The data are taken from Drake and Holland(76) and Drake et al.(4) Multiple symbols are drawn where independent
estimates from different loci in the same organism were available. No error bars are shown but the errors are probably large. RNA
viruses are shown in pink as follows: rv, rhinovirus; pv, poliovirus; vsv, vesicular stomatitis virus; mv, measles virus. Bacteriophages
(DNA-based) are shown in red according to their usual epithets M13, T2, T4, and l. E. coli (Ec) is shown in orange; Saccharomyces
cerevisiae (Sc) and Neurospora crassa (Nc) are shown in green. Higher eukaryotes are shown in blue: Ce, C. elegans; Dm, D.
melanogaster; Mm, Mus musculus; Hs, Homo sapiens. Outliers thought to reflect estimates from non-representative loci (as classified by
Drake 1991, Ref. 77) are not shown. Drake's observation of a conserved mg amongst DNA-based microbes is highlighted with dashed
lines.
on genomic mutation rates was motivated by early analyses(14,15) that suggested that selection would be too weak for
most individual loci to evolve specific mutation rates (see Box
1). Empirical motivation for the theoretical interest in genomic
mutation rates has come from the remarkable observation of
Drake(4,16) that in a range of DNA-based microbes with
genome sizes spanning almost four orders of magnitude,
estimated mg varied by considerably less than one order of
magnitude (Fig. 3). This pattern suggests an evolutionary
equilibrium value of mg that is independent of genome size in
these taxa. In eukaryotes, there is no evidence for a constant
value of mg (see Fig. 3), but it seems reasonable to suppose
that mg could evolve to a characteristic equilibrium value in
each eukaryotic species under one of the general scenarios
illustrated in Fig. 1.
In a seminal early paper, Kimura(14) proposed two
hypotheses for how, in the absence of beneficial mutations,
the ultimate evolution of mg to zero under indirect selection due
to deleterious mutations could be prevented. These hypotheses are: (1) that further reductions of mg (below prevailing
values) are physicochemically impossible, and (2) that further
reductions are physiologically costly and hence would impose
on individuals a prohibitively high direct selective cost. Neither
1060
BioEssays 22.12
hypothesis can be ruled out for all taxa at present. While the
relatively narrow range of per-base-pair mutation rates
observed in eukaryotes across a large range of genome sizes
(Fig. 3) might indicate that mg has evolved to some universal
physicochemical minimum in these taxa,(4) there is no clear
theoretical prediction as to what that minimum should be, and
some empirical evidence (see below) suggests that further
reductions in per-base-pair mutation rate remain possible. The
observation that per-base-pair mutation rates vary widely
among microbial species (Fig. 3) clearly suggests that lower
values of mg are physically possible in at least some prokaryotic
taxa. The isolation of antimutator mutants (bearing alleles that
lower mutation rates) in bacteriophage T4(7) and in Escherichia coli (9,10) also hints that further reductions in mutation rate
may be possible. The T4 antimutators have been shown to
decrease mutation rates along certain mutational pathways
while actually increasing rates along other pathways, yielding
no net reduction in mg.(17) At least one E. coli antimutator,
however, appears to decrease mg approximately twofold below
the prevailing wild-type level.(10)
Limited available theory suggests that the physiological
cost of fidelity in DNA replication (in terms of time and energy)
should increase monotonically as the mutation rate ap-
Review articles
proaches zero,(18,19) thus potentially imposing a direct fitness
cost on individuals bearing lower mutation rates. In vitro
studies of polymerase enzyme purified from antimutator and
wild-type strains of phage T4(20) have shown that increased
fidelity incurs both time and energy costs, but there has been
no measurement or estimation of the effects of such costs on
fitness in any system. There is only indirect evidence suggesting in general that the fitness cost of increasing fidelity
could be substantial. In E. coli, there is a weak negative
relationship between growth rate on minimal medium(21) and
genome size.(22) Indeed, the rarity of noncoding DNA in the
genomes of many microbes may reflect pervasive selection to
increase the rate of genome replication. These observations,
however, do not demonstrate that the further increased cost of
an antimutator phenotype would have a substantial effect on
fitness. In Drosophila, an experiment in which populations
were exposed to different levels of X-irradiation for up to 600
generations documented stable evolutionary decreases in the
rate of X-ray-induced mutation, followed by an evolutionary
return to wild-type mutation rates in some populations after
irradiation was stopped.(23) This result suggests that the level
of investment in repair of radiation damage is set by a tradeoff
between deleterious mutational effects and the cost of repair.
In mice, the X chromosome appears to have a lower mutation
rate than the autosomes as estimated from data on neutral
substitution rates.(24) This is consistent with a tradeoff between the expected greater deleterious effect of recessive
mutations expressed on the (hemizygous) X chromosome in
males and the cost of fidelity on the X chromosome.
The cost of fidelity is thus a credible factor for preventing the
evolution of mg to zero, although it remains possible that some
taxa are already at or near the physicochemical limit to fidelity
of replication. What about the need for new beneficial mutations to facilitate evolutionary adaptation: can it too prevent the
evolution of mg to zero? A fundamental point to bear in mind
when considering this question is that mutation cannot be
maintained in a population for the sake of its future adaptive
utility; natural selection can only maintain mutation as a
consequence of its past adaptive utility.
The fundamental population genetic process underlying
evolutionary adaptation is the rise in frequency of a beneficial
allele. As selection increases the frequency of a beneficial
allele, it indirectly increases the frequencies of linked alleles;
this process is referred to as ``genetic hitchhiking''. A modifier
that increases the mutation rate has an enhanced probability
of association with a beneficial allele (that is, it is more likely to
exist in linkage disequilibrium with a beneficial allele), and
hence it has an enhanced probability of hitchhiking to fixation.
However, this hitchhiking process is extremely sensitive to
recombination, which acts to randomize the associations
between alleles in a population (see Fig. 2; Box 1). A major
conclusion that has emerged from studies of modifier models
is that hitchhiking is unlikely to be an important factor in
adjusting mg in sexual populations because of the strong effect
of recombination in eroding linkage disequilibrium between
mutation rate modifiers and fitness alleles.(14,25) Although
fluctuating selection on fitness loci can generate indirect
selection for increased mg even with recombination,(15,26,27)
Leigh(15) showed that this effect is weak in comparison with
indirect selection due to deleterious mutations. Leigh's infinite
population model ignored the strong linkage disequilibrium
generated by the occurrence of unique new beneficial
mutations;(1) a more recent analysis of a finite population
model,(28) however, supports Leigh's conclusion (see Box 1).
In sexual populations, therefore, it is likely that mg is not
affected by the occurrence of beneficial mutations, but is
instead determined by a tradeoff between indirect selection
due to deleterious mutations and the direct selective cost of
increasing the fidelity of replication. Equilibrium mutation rates
can be calculated under such a tradeoff,(29,30) but no
quantitative predictions are possible because there are no
data relating increased fidelity directly to fitness.
In asexual populations, beneficial mutations can have a
strong effect on the fate of mutation rate modifiers because of
linkage (Fig. 2); indeed, the evolution of mg in asexual
populations is a matter of competition among clonal lineages
with different mutation rates. R.A. Fisher appears to have been
the first to articulate the possibility that an equilibrium mutation
rate can evolve in an asexual population under the influence of
deleterious and beneficial mutations, although his published
treatment of this subject was purely verbal.(11) In essence,
Fisher's approach considers an ensemble of asexual clones
distinguished by their respective values of mg. It is assumed
that each clone attains a distribution of fitness under the
influence of deleterious and beneficial mutations as if it were
an independent population at equilibrium and the clone with
the mutation rate conferring highest mean fitness prevails.
Fisher's conjecture was that this process would yield an
optimal compromise between deleterious and beneficial mutation in an asexual system.
Numerous mathematical models have subsequently been
developed for the evolution of an optimal equilibrium mg in
asexual populations.(14,25,31±34) Some of these models have
the potential to explain Drake's observation of conserved mg in
microbes. For example, Kimura(14) considered a model in
which the rate of substitution of beneficial mutations (hereafter, K) was determined by the rate of environmental change
and showed that the optimal mutation rate is equal to K; he
tentatively suggested that competition between evolving
species might result in a roughly constant value of K across
species. Recently Orr(34) analysed a model in which K was
determined by the rate at which beneficial mutations arise and
by their fixation probabilities, which are strongly influenced by
the frequencies of segregating deleterious mutations.(11,35,36)
Orr found that the optimal mutation rate is proportional to the
harmonic mean of the selection coefficients of deleterious
BioEssays 22.12
1061
Review articles
mutations. It is conceivable that the distribution of deleterious
mutational effects is roughly constant across microbes, and
thus Orr's theory provides another potential explanation for
Drake's observation. Finally, Dawson(29,37) has modeled the
situation in which only deleterious mutations and the cost of
fidelity influence mg in an asexual population. His model assumes that the cost is a per-nucleotide cost and that fitness is
determined by the genomic replication rate. Dawson's analysis
identifies a plausible cost function under which mg does not depend on genome size or on the number of origins of replication
and thus may evolve to a constant observed value in microbes.
Several models thus provide insight into the evolution of mg
in asexual populations, and some of these models are candidates for explaining Drake's observation. Their success in
accounting for observed mutation rates in asexual populations, however, is likely to be limited for several reasons. (1) It
is clear that genetic exchange is not completely absent in many
asexually reproducing populations.(38,39) Low levels of genetic
exchange may be sufficient to disrupt interclone selection and
prevent the evolution of an optimal mutation rate predicted in
Kimura's and Orr's models. (2) Johnson(40) has argued that,
even for strictly asexual populations, there is a problem with all
of these analyses. He has shown that, in a simple model of
competition between a mutator clone (one with a high mg) and a
wild-type clone, mutation from wild type to mutator disrupts
interclone selection and may weaken the indirect selection due
to deleterious mutations by up to an order of magnitude. (3) As
noted by Orr,(34) if the temporal intervals between beneficial
mutations are sufficiently long, then no components of the
models of Kimura and Orr would stop mg from evolving to a
suboptimal level in the interim as a consequence of indirect
selection due to deleterious mutations. (4) There is evidence
that hitchhiking of modifiers in asexual populations may be
more likely to raise mg far above its optimum value than to finetune mg by making small adjustments (see below).
The complete linkage between mutation rate modifiers and
fitness loci in asexual populations thus may produce a situation
in which mg is unlikely to settle into a long-term equilibrium
value. When beneficial mutations are not occurring, selection
drives mg toward a minimal value constrained by the cost of
fidelity, regardless of what the long-term optimal value might
be; when beneficial mutations are occurring, however, hitchhiking of mutation rate modifiers sporadically elevates the
mutation rate above the optimal level. The intuitive picture that
emerges is one in which the mutation rate is continually
buffeted about by the effects of indirect selection on modifiers.
In the next section, we consider the evidence supporting this
nonequilibrial view of mutation rates in asexuals.
Evolution of high mutation rates in
asexual populations
Most experimental work on the evolution of mutation rates has
been conducted in laboratory populations of bacteria, whose
1062
BioEssays 22.12
large population sizes and short generation times facilitate
direct observation of long-term evolutionary phenomena.
Numerous studies have documented the evolution of high
mutation rates in such experimental populations.(41 ±48)
Mutator phenotypes have also been observed at substantial
frequencies among natural bacterial isolates(49 ±53) and in
some cancerous somatic cell lineages,(54) indicating that the
evolution of high mutation rates in asexual populations is not
merely an artifact of laboratory conditions. None of the
experimental studies has uncovered evidence that alleles
responsible for high mutation rates (mutator alleles) confer
direct fitness benefits, and hence their increase in populations
is most likely a result of hitchhiking. Results of an early
experiment(44) suggested that most mutator alleles would be
unlikely to hitchhike unless initially present at improbably high
frequencies in a population (Box 2). Computer simulation
studies, however, have shown that with sufficient time even
very rare mutators can hitchhike to fixation in finite asexual
populations,(55,56) and similar results have been obtained in an
analytical study (P.J.G., R.E. Lenski and P.D.S., unpublished
data). Both of these modeling approaches indicate that
mutators of strong effect (those that elevate mg substantially)
are most likely to hitchhike, although deleterious mutations
prevent fixation of the strongest mutators. Thus it appears that
adapting asexual populations have a propensity to evolve
sharply elevated mutation rates under hitchhiking rather than
to evolve optimal mutation rates.
Because asexual populations cannot generate variation by
recombination, it might be hypothesized that high mutation
rates in asexual populations are an adaptation for generating
variation. A mutator phenotype that has hitchhiked to fixation is
properly regarded as a consequence of adaptation, however,
not an adaptation itself. The adaptive value of a mutator
phenotype depends on whether it increases the rate of
subsequent adaptation in an asexual population, and the
circumstances under which this is likely are limited. On one
hand, computer simulation studies(56) have shown that, if a
small number of beneficial mutations is to be substituted in a
finite asexual population during adaptation, then under some
circumstances a population that fortuitously substitutes a
mutator allele by hitchhiking early in the bout of adaptation will
reach this goal sooner. On the other hand, an analytical model
of adaptive evolution in asexual populations(57) predicts a
diminishing increase in rate of adaptation with increased
mutation rates. At higher rates of beneficial mutation, the
speed of adaptation becomes increasingly limited by the rate
of selective sorting among clones bearing different beneficial
mutations rather than by the rate at which new beneficial
mutations arise in the population. Experimental work has
shown that this ``clonal interference'' effect constrains the
adaptive usefulness of a high mutation rate to situations in
which beneficial mutations are extremely infrequent, such as
when population sizes are very small (perhaps due to
Review articles
Box 2. Frequency of Mutator Alleles in Asexual Populations
Mutator alleles arise in asexual populations as a product of recurrent mutation and are removed by indirect selection.
Assuming that the deleterious mutation rate is proportional to the total genomic mutation rate, the selection coefficient
against a mutator that raises the genomic mutation rate by a factor m 1 relative to the wild-type rate mg is approximately
mU, where U is the genomic deleterious mutation rate in the wild type. On the simplest possible model, the equilibrium
frequency of such a mutator under mutation and indirect selection is approximately mmut/mU, where mmut is the rate of
mutation from wild type to mutator alleles. Substitutions occurring on the wild type background in the population will tend
to keep the mutator frequency lower than this equilibrium mutation-selection balance (P.J.G. et al., unpublished results);
in contrast, the influx of unloaded new mutator genomes arriving by mutation from the wild type subpopulation will reduce
the strength of indirect selection and tend to elevate mutator frequencies relative to the simple mutation-selection
balance.(40)
If the rate of beneficial mutation is proportional to the genomic mutation rate, then a subpopulation of mutator individuals
in a finite asexual population is expected to increase in frequency by hitchhiking whenever N 0 /N > 1/m, where N 0 and N are
the mutator and wild-type population numbers. This dependence of mutator hitchhiking on mutator number relative to wildtype number was first illustrated in experimental E. coli populations by Chao and Cox in 1983.(44) Given reasonable
estimates of mutation rates, the critical threshhold frequency 1/m for mutator hitchhiking is unlikely to be exceeded under the
dynamics of mutation and selection described above, and thus mutators are not expected to be in the process of hitchhiking
in most asexual populations. With sufficient time, however, a rare mutator subpopulation can by chance acquire a beneficial
mutation that is destined for fixation in a finite population. This stochastic turn of events has been modeled extensively in
computer simulation,(55,56) and also confirmed analytically (P.J.G. et al., unpublished results) by extension of a model
for adaptation in asexuals.(57) Both modeling approaches predict that mutator alleles of substantial effect will have the
highest probability of hitchhiking and thus suggest that mutator hitchhiking need not fine tune mutation rates to equilibrium
levels.
bottlenecks) or when a population is initially well-adapted.(58)
High mutation rates can substantially accelerate adaptation in
asexual populations, but do not necessarily do so.
The available empirical evidence is consistent with this
cautious view of the adaptive significance of mutators. In a
study of twelve replicate experimental populations of E. coli
propagated in the laboratory for 10,000 generations,(47) three
populations evolved mutator phenotypes but there was no
evidence that these mutator populations adapted faster than
the remaining nine populations. Some, but not all, cancer cell
lineages exhibit a mutator phenotype,(54) and simulation
studies show that clonal selection without elevated mutation
rates can be sufficient to promote carcinogenesis.(59) A survey
of Salmonella and E. coli isolates classified as pathogenic or
nonpathogenic found a suggestive (but nonsignificant) association between mutator phenotype and pathogenicity,(51)
hinting at a selective advantage to mutator phenotypes under
the rigors of host immune surveillance. A later survey,
however, failed to document such an association.(52) Finally,
a recent study documented a high frequency (20%) of
mutators among Pseudomonas aeruginosa strains infecting
the lungs of cystic fibrosis (CF) patients.(53) This finding
provides circumstantial evidence that mutator phenotypes
have a selective advantage in colonizing the CF lung.(60) The
alternative view that selection for colonization ability merely
increases the probability of a mutator hitchhiking event,
however, remains a possibility; the study does not compare
the rates of adaptation of mutator and wild-type strains after
colonization.
Even if a high mutation rate increases the rate of adaptation
in an asexual population in the short term, over the
evolutionary long term it is clear that indirect selection due to
deleterious mutations eventually favors a decrease in mutation rates. Otherwise, mutator phenotypes would be the rule
rather than the exception in asexual populations, which is
clearly not the case. Given the propensity for asexual
populations to substitute mutator alleles by hitchhiking, there
is a need for more theory and experimentation on how low
mutation rates are restored. Three processes could potentially
reduce mutation rates within an asexual population that is fixed
for a mutator allele. (1) The most obvious is outright reversion
of the mutator allele and substitution of the new wild-type allele
back into the population. (2) Compensatory evolution at
additional modifier loci could also reduce the mutation rate,
as observed in experimental populations of E. coli that were
fixed for a very strong mutator allele.(61) (3) Rare horizontal
genetic exchange events could replace the mutator allele with
its wild-type counterpart.
BioEssays 22.12
1063
Review articles
A major alternative to the reduction of mutation rates by
selection acting within clones is the possibility that mutator
clones are evolutionary dead ends which, despite a possible
short-term adaptive advantage, are destined to be outcompeted by their wild-type counterparts. Theoretical studies
have explored the evolutionary ``meltdown'' of finite asexual
populations under deleterious mutation pressure,(62±64) and
there is direct genetic evidence for accelerated accumulation
of deleterious mutations within mutator clones of E. coli.(65)
The interaction of mutator hitchhiking and deleterious mutation
accumulation in asexual populations is an interesting area for
further research.
Evolving high mutation rates without
high mutational load
Throughout this review, we have taken the position that high
genomic mutation rates are generally disfavored in both
sexual and asexual populationsÐdespite their potential
usefulness for adaptionÐbecause they increase the load of
deleterious mutations. Nonetheless, there are two biologically
plausible situations in which the adaptive advantage of a high
mutation rate may be acquired without large increases in
mutational load: (1) by restricting elevation of the mutation rate
to certain loci, and (2) by restricting elevation of the genomic
mutation rate to times when environmental stress prevents
further growth and reproduction unless mutation occurs.
Increasing the mutation rate only in a single locus or genetic
region reduces the force of indirect selection against a high
mutation rate considerably by confining deleterious mutations
to a small fraction of the genome.(25) The best evidence for
such locus-specific evolution of high mutation rates is the
existence of hypermutable ``contingency loci'' in pathogenic
bacteria.(66,67) Such loci interact with the environment in highly
unpredictable ways (for example, via surveillance by the host
immune system) that place a premium on variability rather
than on conservation of specific function. Most examples
invoke the replicative instability of tandem DNA repeats as the
basis for locus-specific high mutation rates.(66) It is interesting
to note that the evolution of locus-specific high mutation rates
appears to require the presence of at least some recombination in a bacterial population, since without recombination
genetic drift or selective events at other loci are likely to swamp
the forces adjusting mutation rates on a per-locus basis (Box
1). Hypermutable contingency loci currently provide the best
examples of adaptively high mutation rates.(68)
Episodes of environmental stress that prevent growth and
reproduction may put a high premium on the production of new
variation. A mechanism that allowed genomic mutation rates
to be elevated during such times of stress might be favored by
selection, particularly in asexual populations in which a
modifier allele that increased mutation only under stress could
remain linked to beneficial mutations. Recent evidence
(reviewed by Foster in this volume) suggests that at least
1064
BioEssays 22.12
some cells in bacterial cultures under starvation stress have
elevated genomic mutation rates. This phenomenon has
acquired the name ``adaptive mutation'' and has been linked to
various proposed mechanisms of mutagenesis unique to
nonreplicating cells (Foster, this volume). The adaptive
mutation hypothesis is intuitively appealing because, by
restricting high mutation rates to stress episodes, the general
problem of increased mutational load that would arise if
mutation were continually elevated is avoided. To date,
however, there has been little rigorous modeling of the
evolutionary processes by which stress-induced mutation
might arise and be maintained in populations, and alternative
explanations such as unavoidable increases in error rates or
decreases in repair rates due to the direct effects of
physiological stress have not been ruled out.(68,69)
Progress and future directions
In answer to his question ``why does the mutation rate not
evolve to zero?'' Sturtevant could only offer the following in
1937:(13) ``No answer seems possible at present, other than
the surmise that the nature of genes does not permit such a
reduction. In short, mutations are accidents, and accidents will
happen.'' Indeed, mutations are accidents,(69) but since 1937
our understanding of the mechanistic and evolutionary causes
of the mutational accident rate has improved in many ways.
Fundamental studies on the mechanisms of replication and
repair have amply validated the view that mutation rates are
genetically variable and thus potentially subject to adjustment
by natural selection, and evolutionary increases and decreases in mutation rate have been studied directly in laboratory populations. Modifier models have shown that beneficial
mutations are unlikely to affect the mutation rate in sexual
populations, while experiments and related theoretical work
have shown that high mutation rates can evolve sporadically
under the influence of beneficial mutations in asexual
populations.
The available evidence suggests that it is the cost of
increasing fidelity and not absolute physical constraints that
prevents the evolution of lower mutation rates in most species.
Little is known about the magnitude of this cost, however, and
this is probably the most important area for further empirical
research on the evolution of genomic mutation rates. The
discovery of general antimutator alleles in E. coli (10) may
provide an opportunity to study the fitness effects of reducing
the mutation rate in this organism in an experimental setting.
Results of such experiments may allow discrimination between the theories(29,34,37) proposed to explain Drake's
observation of conserved mg in microbes.
Indeed, if the genomic mutation rate in asexuals is set by
the cost of fidelity and not by the effect of beneficial mutations,
then it is possible to make a prediction concerning equilibrium
values of mg in sexual and asexual populations. All else being
equal, the greater strength of indirect selection due to
Review articles
deleterious mutations should cause an asexual population to
evolve a lower mg than a sexual population.(30) In this regard, it
is interesting to note that a recent study(70) comparing ancient
asexual bdelloid rotifers with facultatively sexual monogonont
rotifers found no significant difference in mutation rates per
year between the two taxa as estimated by synonymous
nucleotide substitution rates. This result appears not to uphold
the theoretical prediction of lower mutation rates in asexuals,
but its interpretation is complicated by possible differences in
generation time between the taxa and by the potentially
confounding effects of selection on the somatic mutation rate.
Further data of this kind comparing unicellular sexual and
asexual taxa would be helpful.
The ultimate fate of asexual mutator phenotypes is an
important area for further research. Circumstances in which
mutators can have a short-term adaptive advantage have
been recognized, but mutators must be selected against over
the long term; otherwise, mutators would be the rule rather
than the exception in asexual populations. Locus-specific
hypermutability and stress-induced hypermutability in bacteria
are also emerging as important areas for further evolutionary
modeling and critical testing. Knowledge of the factors
affecting the evolution of genomic mutation rates provides a
solid foundation on which to build in all of these areas.
Conclusion
At the beginning of this review, we drew a comparison between
mutation and recombination: both can affect the fitnesses of
individuals and the evolutionary potential of populations. As
we have discussed above, recombination also profoundly
influences the rate of mutation that can evolve and be
maintained in a population. Where recombination is substantial, indirect selection to raise the mutation rate is likely to be
ineffective. Where recombination is minimal or absent, indirect
selection can raise the mutation rate, but this does not
necessarily result in a mutation rate that maximizes or even
increases the long-term rate of adaptation. Mutation itself,
however, is likely to be a key factor in the maintenance of
sexual recombination; major alternative theories for the
evolution of sex invoke the clearance of deleterious mutations
and the combining of beneficial mutations as ultimate causal
factors.(2) Seen in this light, the adjustment of mutation rates
by natural selection is but one in a number of intricately related
evolutionary processes.
Note added in proof
A paper published while this review was in press (O. Tenaillon,
H. Le Nagard, B. Godelle, and F. Taddei, Proc Natl Acad Sci
USA 2000;97:10465±10470) uses computer simulation to
analyse the fate of mutator alleles in bacterial populations with
rare genetic exchanges. The authors conclude that rare
genetic exchange can inhibit both the hitchhiking and long
term persistence of mutators in bacterial populations.
Acknowledgments
We thank N. Barton, B. Charlesworth, J.A. Coyne, J.W. Drake,
A. Dunham, W.J. Ewens, and G. Kienitz for comments on all or
parts of the manuscript, and we thank L. Chao, R.E. Lenski,
H.A. Orr, S. Otto, and R. Redfield for helpful discussions.
References
1. Maynard Smith J. The Evolution of Sex. Cambridge, UK: Cambridge
University Press; 1978.
2. Barton NH, Charlesworth B. Why sex and recombination? Science 1998;
281:1987±1990.
3. West SA, Lively CM, Read AF. A pluralist approach to sex and recombination. J Evol Biol 1999;12:1003±1012.
4. Drake JW, Charlesworth B, Charlesworth D, Crow JF. Rates of spontaneous mutation. Genetics 1998;148:1667±1686.
5. Miller JH. Spontaneous mutators in bacteria: Insights into pathways of
mutagenesis and repair. Ann Rev Microbiol 1996;50:625±643.
6. Friedberg EC, Walker GC, Siede W. DNA Repair and Mutagenesis.
Washington, DC: ASM Press; 1995.
7. Drake JW, Allen EF, Forsberg SA, Preparata R-M, Greening EO. Genetic
control of mutation rates in bacteriophage T4. Nature 1969;221:1128±
1132.
8. Horst JP, Wu T, Marinus MG. Escherichia coli mutator genes. Trends
Microbiol 1999;7:29±36.
9. QuinÄones A, Piechocki R. Isolation and characterization of Escherichia
coli antimutators. Mol Gen Genetics 1985;201:315±322.
10. Schaaper RM. Antimutator mutants in bacteriophage T4 and Escherichia
coli. Genetics 1998;148:1579±1585.
11. Fisher RA. The Genetical Theory of Natural Selection. Oxford: Clarendon
Press; 1930.
12. Maynard Smith J, SzathmaÂry E. The Major Transitions in Evolution.
Oxford: Freeman/Spektrum; 1995.
13. Sturtevant AH. Essays on evolution. I. On the effects of selection on
mutation rate. Q Rev Biol 1937;12:467±477.
14. Kimura M. On the evolutionary adjustment of spontaneous mutation
rates. Genet Res 1967;9:23±34.
15. Leigh EG. The evolution of mutation rates. Genetics (Suppl) 1973;73:
1±18.
16. Drake JW. A constant rate of spontaneous mutation in DNA-based
microbes. Proc Natl Acad Sci USA 1991;88:7160±7164.
17. Drake JW. General antimutators are improbable. J Mol Biol 1993;229:
8±13.
18. Fersht A. Enzyme Structure and Mechanism. New York: W.H. Freeman;
1985.
19. Kirkwood TBL, Rosenberger RF, Galas DJ. Accuracy in molecular
processes: Its control and relevance to living systems. London: Chapman and Hall; 1986.
20. Bessman M, Muzyczka N, Goodman M, Schnaar R. Studies on the
biochemical basis of spontaneous mutation. II. The incorporation of a
base and its analogue into DNA by wild-type, mutator, and anti-mutator
DNA polymerases. J Molec Biol 1974;88:409±421.
21. Mikkola R, Kurland CG. Is there a unique ribosome phenotype for
naturally occurring E. coli? Biochimie 1991;73:1061±1066.
22. Bergthorsson U, Ochman H. Distribution of chromosome length variation
in natural isolates of Escherichia coli. Molec Biol Evol 1998;15:6±16.
23. NoÈthel H. Adaptation of Drosophila melanogaster populations to high
mutation pressure: Evolutionary adjustment of mutation rates. Proc Natl
Acad Sci USA 1987;84:1045±1049.
24. McVean GT, Hurst LD. Evidence for a selectively favourable reduction in
the mutation rate of the X chromosome. Nature 1997;386:388±392.
25. Leigh EG. Natural selection and mutability. Am Nat 1970;104:301±305.
26. Gillespie JH. Mutation modification in a random environment. Evolution
1981;35:468±476.
27. Ishii K, Matsuda H, Iwasa Y, Sasaki A. Evolutionarily stable mutation
rate in a periodically changing environment. Genetics 1989;121:163±
174.
28. Johnson T. Beneficial mutations, hitchhiking and the evolution of
mutation rates in sexual populations. Genetics 1999;151:1621±1631.
BioEssays 22.12
1065
Review articles
29. Dawson KJ. The dynamics of infinitesimally rare alleles, applied to the
evolution of mutation rates and the expression of deleterious mutations.
Theor Pop Biol 1999;55:1±22.
30. Kondrashov AS. Modifiers of mutation-selection balance: General
approach and the evolution of mutation rates. Genet Res 1995;66:53±70.
31. Eshel I. Clone selection and optiimal rates of mutation. J Appl Prob
1973;10:728±738.
32. Painter PR. Mutator genes and selection for the mutation rate in bacteria.
Genetics 1975;79:649±660.
33. Woodcock G, Higgs P. Population evolution on a multiplicative singlepeak fitness landscape. J Theor Biol 1996;179:61±73.
34. Orr HA. The rate of adaptation in asexuals. Genetics 2000;155:961±968.
35. Barton NH. Linkage and the limits to natural selection. Genetics 1995;
140:821±841.
36. Peck JR. A ruby in the rubbish: beneficial mutations, deleterious mutations and the evolution of sex. Genetics 1994;137:597±606.
37. Dawson KJ. Evolutionarily stable mutation rates. J Theor Biol 1998;194:
143±157.
38. Milkman R. Recombination and population structure in Escherichia coli.
Genetics 1995;146:745±750.
39. Maynard Smith J, Smith NH, O'Rourke M, Spratt BG. How clonal are
bacteria? Proc Natl Acad Sci USA 1993;90:4384±4388.
40. Johnson T. The approach to mutation-selection balance in an infinite
asexual population, and the evolution of mutation rates. Proc Roy Soc
Lond B 1999;266:2389±2397.
41. Gibson TC, Scheppe ML, Cox EC. Fitness of an Escherichia coli mutator
gene. Science 1970;169:686±688.
42. Nestmann ER, Hill RF. Population changes in continuously growing
mutator cultures of Escherichia coli. Genetics (Suppl) 1973;73:41±44.
43. Cox EC, Gibson TC. Selection for high mutation rates in chemostats.
Genetics 1974;77:169±184.
44. Chao L, Cox EC. Competition between high and low mutating strains of
Escherichia coli. Evolution 1983;37:125±134.
45. TroÈbner W, Piechocki R. Competition between isogenic mutS and mut ‡
populations of Escherichia K12 in continuously growing cultures. Mol
Gen Genet 1984;198:175±176.
46. Mao EF, Lane J, Lee J, Miller JH. Proliferation of mutators in a cell
population. J Bacteriol 1997;179:417±422.
47. Sniegowski PD, Gerrish PJ, Lenski RE. Evolution of high mutation rates in
experimental populations of E. coli. Nature 1997;387:703±705.
48. Boe L, Danielson M, Knudsen S, Petersen JB, Maymann J, Jensen PR.
The frequency of mutators in populations of Escherichia coli. Mutation
Res 2000;448:47±55.
49. Jyssum K. Observations on two types of genetic instability in Escherichia
coli. Acta Pathol Microbiol Scand 1960;48:113±120.
50. Gross MD, Siegel EC. Incidence of mutator strains in Escherichia coli
and coliforms in nature. Mutation Res 1981;91:107±110.
51. LeClerc JE, Li B, Payne WL, Cebula T. High mutation frequencies among
Escherichia coli and Salmonella pathogens. Science 1996;274:1208±
1211.
52. Matic I, Radman M, Taddei F, Picard B, Doit C, Bingen E, Denamur E,
Elion J. Highly variable mutation rates in commensal and pathogenic E.
coli. Science 1997;277:1833±1834.
53. Oliver A, CantoÂn R, Campo P, Baquero F, BlaÂzquez J. High frequency of
hypermutable Pseudomonas aeruginosa isolates in cystic fibrosis lung
infection. Science 2000;288:1251±1253.
1066
BioEssays 22.12
54. Loeb KR, Loeb LA. Significance of multiple mutations in cancer.
Carcinogenesis 2000;21:379±385.
55. Taddei F, Radman M, Maynard Smith J, Toupance B, Gouyon P-H,
Godelle B. Role of mutator alleles in adaptive evolution. Nature 1997;387:
700±703.
56. Tenaillon O, Toupance B, Le Nagard H, Taddei F, Godelle B. Mutators,
population size, adaptive landscape and the adaptation of asexual
populations. Genetics 1999;152:485±493.
57. Gerrish PJ, Lenski RE. The fate of competing beneficial mutations in an
asexual population. Genetica 1998;102/103:127±144.
58. de Visser JAGM, Zeyl CW, Gerrish PJ, Blanchard JL, Lenski RE.
Diminishing returns from mutation supply rate in asexual populations.
Science 1999;283:404±406.
59. Tomlinson IPM, Novelli MR, Bodmer WF. The mutation rate and cancer.
Proc Natl Acad Sci USA 1996;93:14800±14803.
60. Rainey PB, Moxon ER. When being hyper keeps you fit. Science 2000;
288:1186±1187.
61. TroÈbner W, Piechocki R. Selection against hypermutability in Escherichia
coli during long term evolution. Mol Gen Genet 1984;198:177±178.
62. Lynch M, Gabriel W. Mutation load and the survival of small populations.
Evolution 1990;44:1725±1737.
63. Lynch M, BuÈrger R, Butcher D, Gabriel W. The mutational meltdown in
asexual populations. J Hered 1993;84:339±344.
64. Gabriel W, Lynch M, BuÈrger R. Muller's ratchet and mutational
meltdowns. Evolution 1993;47:1744±1757.
65. Funchain P, Yeung A, Stewart JL, Lin R, Slupska MM, Miller JH. The
consequences of growth of a mutator strain of Escherichia coli as
measured by loss of function among multiple gene targets and loss of
fitness. Genetics 2000;154:959±970.
66. Moxon ER, Rainey PB, Nowak MA, Lenski RE. Adaptive evolution of
highly mutable loci in pathogenic bacteria. Curr Biol 1994;4:24±33.
67. Field D, Magnasco MO, Moxon ER, Metzgar D, Tanaka MM, Wills C,
Thaler DS. Contingency loci, mutator alleles, and their interactions. Ann
NY Acad Sci 1999;870:378±382.
68. Metzgar D, Wills C. Evidence for the adaptive evolution of mutation rates.
Cell 2000;101:581±584.
69. Sniegowski PD, Lenski RE. Mutation and adaptation: the directed
mutation controversy in evolutionary perspective. Ann Rev Ecol Syst
1995;26:553±578.
70. Welch DM, Meselson M. Evidence for the evolution of bdelloid rotifers
without sexual reproduction or genetic exchange. Science 2000;288:
1211±1214.
71. Holsinger K, Feldman MW. Modifiers of mutation rate: evolutionary
optimum with complete selfing. Proc Natl Acad Sci USA 1983;80:
6772±6784.
72. Karlin S, McGregor J. Towards a theory of the evolution of modifier
genes. Theor Pop Biol 1974;5:59±103.
73. Liberman U, Feldman M. Modifiers of mutation rate: a general reduction
principle. Theor Pop Biol 1986;30:125±142.
74. Eyre-Walker A, Keightley PD. High genomic deleterious mutation rates in
hominids. Nature 1999;397:344±347.
75. Maynard Smith J, Haigh J. The hitch-hiking effect of a favourable gene.
Genet Res 1974;23:23±35.
76. Drake JW, Holland JJ. Mutation rates among RNA viruses. Proc Natl
Acad Sci USA 1999;96:13910±13913.
77. Drake JW. Spontaneous mutation. Ann Rev Genet 1991;25:125±146.
Download