Mutation rates in nature have traditionally been thought of as the product of two opposing evolutionary forces: adaptability and adaptedness. As an organism becomes more adapted to its environment, the amount by which it can adapt further is diminished, a fact that is manifest genetically as a dwindling repertoire of potentially beneficial mutations. As an organism adapts to an unchanging environment, therefore, mutation in general becomes almost exclusively detrimental to the organism, and natural selection should thus be expected to favor a decrease in mutation rate. In fact, Leiberman and Feldman showed mathematically that, were it physiologically possible, the mutation rate would evolve to zero in an unchanging environment, and they called this finding the "general reduction principle". In an environment that changes over time, on the other hand, there is interminable room for improvement, manifest genetically as a continuously renewed and hence unending supply of potentially beneficial mutations. In a changing environment, therefore, it is easy to see how a certain amount of mutation might be beneficial or even vital. Under the simplest possible regime of environmental change and genetic response, Leigh calculated that the optimal mutation rate was, not surprisingly, equal to the rate of environmental change -- a result that was later shown to also be the evolutionarily stable strategy.

We tested the "general reduction principle" by measuring mutation rates of E. coli as it evolved in an unchanging laboratory environment. What we found did not corroborate the "general reduction principle"; in fact, not only was the mutation rate not reduced during evolution in this constant environment, but in three of twelve independent E. coli populations, the mutation rate spontaneously increased by roughly two orders of magnitude! So we decided that a fresh look at the underlying theory was in order. My colleague, Paul Sniegowski, first came up with the alternative theory that would explain our findings and has since been shown to be correct: variants with elevated mutation rates ("mutators") may acquire a particular beneficial mutation -- however rare it may be -- before the rest of the population, and because the beneficial mutation remains linked to the elevated mutation rate, the fixation of that beneficial mutation thus drives the elevated mutation rate to fixation. In this "mutator hitchhiking hypothesis" (P. D. Sniegowski, P. J. Gerrish, R. E. Lenski, Nature 387, (1997)), the usual causality is reversed: instead of mutation rates evolving to adjust adaptation, it is adaptation that indirectly and haphazardly adjusts mutation rate. Put differently, mutation rate evolution is a biproduct of adaptation at other loci, an evolutionary after-thought so to speak, that has little or nothing to do with any benefit or detriment that the evolved mutation rate may subsequently incur.

The mutator hitchhiking hypothesis made a lot of sense and explained our findings quite nicely. What it didn't really explain, however, was why only increases in mutation rate were observed; after all, a variant with a decreased mutation rate (an "anti-mutator") should have an evolutionary advantage because it produces fewer mutations and most mutations are detrimental. The answer to this question is two-pronged. First of all, from a purely genetical standpoint, mutator mutations are much easier to come by than anti-mutator mutations, because mutator mutations are generally loss-of-function mutations that hinder or knock out replication, proofreading or repair genes, whereas anti-mutator mutations are generally gain-of-function or compensatory mutations that improve or add functionality in these gene regions. There is thus a potentially strong mutational bias favoring increased mutation rate. Secondly, there is a population-dynamical bias favoring increased mutation rate: when a mutator acquires a new beneficial mutation, its evolutionary advantage is immediate; on the other hand, a new lineage with a decreased mutation rate has an advantage only in the long term, due to its relatively slow shedding of deleterious load. Natural selection, because it is a short-sighted process, is much more likely to respond to the immediate benefit of hitchhiking mutators than to the long-term benefit of anti-mutators, and the result is an inescapable evolutionary bias toward increasing mutation rates.

Theory (Andre & Godelle, Genetics...) and experiment (P. D. Sniegowski, P. J. Gerrish, R. E. Lenski, Nature 387, (1997)) corroborated the above logic, showing that mutation rates in asexual populations are indeed unstable and prone to ratchet-like increase. Yet a troubling question remained: when would this trend of net mutation rate increase stop? Or would it ever stop? Would it instead elevate the mutation rate to levels that are intolerable? If this were the case, it would lead to a rather absurd conclusion: it would suggest that the same adaptive process that allows a population to thrive is the process that drives the mutation rate through the roof and thereby drives the population extinct.

population extinct through mutation rate evolution. beyond those that can be tolerated to intolerable levels. In particular, would it cross the sharply defined threshold called the error threshold. If the above logic were extrapolated, however, it would lead to a rather absurd possible consequence that could not be overlooked, namely, that mutation rates will increase until they reach intolerable levels thereby driving the population extinct. This possible outcome was verbally - and I might add quite intuitively - discounted by Andre and Godelle (ref) on the basis that increased deleterious mutation and increased clonal interference would prevent it. Using analytical theory as well as individual-based simulations, however, we later found that, contrary to intuition, this absurd outcome was in fact by far the most probable outcome!

In this light, mutators

Both cancer progression and—in many cases—pathogen evolution occur via clonal
selection, and the dynamics of adaptive evolution in clonal (asexual) populations
therefore lies at an important intersection between population biology and public health.
But it has long been known that something eventually goes wrong in most clonal
populations. Truly asexual taxa are rare and short-lived. Among eukaryotes, the class
Bdelloidea is the only widely accepted example of a higher taxon that is asexual [1, 2].
Among prokaryotes and viruses, which reproduce asexually, population genetic
analyses of variation nonetheless support the inference that many recombine their
genomes through various “parasexual” processes [e.g., 3, 4]. If one could identify what
exactly goes wrong in clonal (asexual) populations, such a defect might potentially be
exploited to battle somatic or infectious diseases.

Many disadvantages of asexuality are indeed already known or have been
hypothesized in theoretical work [see reviews in 5, 6-8], but these are generally subtle
evolutionary impairments that put asexual lineages at a disadvantage relative to their
sexual counterpart. These subtle impairments do not present any real threat to an
adapting asexual population by itself.

In contrast to established theory, our preliminary work suggests that asexual systems
are not only at a disadvantage when compared to sexual systems but that, by
themselves, asexual systems are fundamentally flawed. We hypothesize that
asexuality presents a very real threat to adapting populations: as adaptive
evolution proceeds, an ever-higher mutation rate evolves, ultimately causing
extinction. This process, which we will call the “mutation rate catastrophe” or MRC,
could underlie both the comparatively rapid extinction of asexual species as well as the
progressive genomic instability of cancer lineages. Understanding the dynamics of the
process may suggest new treatment strategies for battling persistent microbial pathogen
populations within individual hosts.

Our hypothesis derives from the following observations. First, recent experimental and
theoretical work by ourselves and others has shown that genetic hitchhiking of mutators
(alleles defective in genomic repair and replication processes) occurs spontaneously in
evolving asexual populations [9-17]. A common interpretation of mutator hitchhiking has
been that it serves to confer "evolvability" on an asexual population that faces an
adaptive challenge [18-20]. In contrast, we and others have noted that, regardless of its
possible adaptive value in the short term, mutator hitchhiking poses a long term problem
for an asexual population [9, 21, 22]. Second, the process that restores low mutation
rates – the rise in frequency of antimutators (revertants of mutators or fidelity-enhancing
compensatory mutants) – rarely occurs, and is likely to be a much slower process than
the hitchhiking of mutators [9, 22]. This is because antimutators will typically arise on
backgrounds that have as many deleterious mutations as the rest of the population, and
they therefore have little or no immediate selective advantage. Third, the first two
observations, taken together, result in an intriguing phenomenon: relatively frequent
increments in mutation rate (first observation) countered only by very rare decrements
(second observation) cause a ratchet-like phenomenon that elevates mutation rate
relentlessly in an adapting population. Put differently, a finite, evolutionarily “stable”
mutation-rate for an asexual population does not exist [23], contrary to previous notions
about mutation-rate evolution [24]. Finally, in current simulation and analytical work [25]
we have shown that this upwardly biased mutation rate evolution in asexuals causes
accumulation of mutator alleles and can ultimately cause extinction when an intolerable
mutation rate evolves. The causal basis of this process is distinctly different from any
previously identified mutational basis for extinction or disadvantage of asexuality [26-
28]: paradoxically, the process is driven indirectly by positive natural selection.
There is certainly no physiological barrier to such a process: the genomic mutation rate
in organisms from viruses to eukaryotes is a quantitative trait affected by many loci
whose effects can readily cumulate to intolerable levels of mutation [29, 30]. This
proposal focuses on: i) determining the conditions under which there is also no
selective barrier to this process, and ii) understanding the absence of a selective
barrier. Preliminary studies suggest that there is no selective barrier to this process
under most reasonable conditions, and furthermore they suggest an explanation:
because the full fitness effect of increased deleterious mutation takes some time to
accumulate after a higher mutation rate has evolved, it is possible for a population to
evolve a critically high mutation rate and subsequently go extinct. Yet despite a solid
theoretical [11, 23, 25] and empirical [9, 10, 16, 22, 31-33] base, and despite
considerable advances in our preliminary studies, our understanding of the underlying
processes that allow natural selection to be so self-destructive is still quite superficial.
We propose, therefore, to continue our investigation of the MRC according to the
following three specific aims:

1) Predict under what conditions the mutation rate catastrophe will not occur,
using analytical theory and simulations. We are particularly interested in the
following questions: A) What is the critical recombination rate above which a
population is not driven extinct? B) Is there a critical beneficial mutation rate in
finite populations below which a population is not driven extinct? If so, what is it?
C) Will a large degree of fragility in replication and repair apparatuses save a
population from extinction? This last question will be especially pertinent in the
case that results from bacterial populations do not support our theoretical
findings: we suspect that substantial replication/repair fragility may offer the most
parsimonious explanation for a negative experimental outcome.

2) Understand the underlying population processes that allow natural
selection to drive a population extinct. Despite compelling theoretical and in
silico evidence that the MRC can occur, we are still not satisfied that we know
why it occurs. We already have several clues and have derived mathematical
tools to help decipher the process, but a definitive explanation remains elusive. In
our search for a definitive explanation, we will perform detailed dissections of
large-scale simulations, complemented by analytical theory and informed by
experimental data.

3) Determine the biological plausibility of such “extinction-by-naturalselection”
by testing key underlying processes with in vitro populations of
Escherichia coli. The first step of the MRC process – spontaneous hitchhiking
of mutators – has been observed in populations of E. coli [10]. Ideally, we would
like to observe the spontaneous hitchhiking of a second (and third, etc.) mutator
on top of a single mutator, until the population goes extinct. Given our limited
time-frame, however, we aim to test only the crucial building-blocks of the MRC
process through focused short-term experiments. In preliminary work, for
example, we have already observed the displacement of a single mutator by an
engineered double mutator (Preliminary Studies). In proposed work, we will
repeat this experiment in the presence of a mutagen, such that the mutation rate
of the single mutator is tolerable while that of the double mutator is intolerable. If
the double mutator still displaces the single mutator, then natural selection will
have driven the population to extinction. In addition, we will simply propagate
populations at very high mutation rate, periodically testing for positive or negative
covariance between mutation rate and fitness – a good indicator of selection on
mutation rates [34] and a robust predictor of runaway mutation rates and
extinction [25].