Abstract
The traits that affect evolvability are subject to indirect selection, as these traits affect the course of evolution over many generations rather than the direct replicative fitness of an individual. However, the evolution of evolvability-determining traits is often difficult to study because putative evolvability alleles often have confounding direct fitness effects of unknown origin and size. Here, we study theoretically and experimentally the evolution of mutation rates in proofreading polymerases with orthogonal control of direct and indirect selection. Mutagenic DNA polymerases enjoy a long-time fitness advantage by enhancing the rate of acquiring beneficial mutations. However, this is offset by a short-time fitness penalty, which we trace to a counterintuitive trade-off between mutation rates and activity in proofreading polymerases. Since these fitness effects act on different timescales, no one number characterizes the fitness of a mutator allele. We find unusual dynamic features in the resulting evolutionary dynamics, such as kinetic exclusion, selection by dynamic environments, and Rock-Paper-Scissors dynamics in the absence of ecology. Our work has implications for the evolution of mutation rates and more broadly, evolution in the context of an anti-correlation between mutation rates and short term fitness.
Darwinian evolution is often seen as a tinkering algorithm that incrementally increases the fitness of an organism. As algorithms go, it is remarkably simple: relying on only a few key ingredients — heritable variation and natural selection — to adapt organisms to their environment [1].
However, in contrast to human tinkering, a unique aspect of biological evolution is that the parameters of the algorithm are themselves subject to evolution [2, 3]. The nature and amount of heritable variation available to natural selection is itself under the control of biological factors — ‘modifier’ genes that control the rate at which new variants arise. Examples include genes that affect mutation rates [4–7] and spectra [8, 9], recombination rates [10], horizontal gene transfer, and ‘buffers’ that reduce the impact of deleterious mutations [11–13]. Variation at these loci can influence the rate at which future beneficial alleles arise, thereby altering the dynamics of future evolution.
A particular simple and universally relevant source of variation is the mutation rate, which is set primarily by the biophysics of DNA replication and repair [14]. Genetic changes in the machinery of replication can lead to ‘mutator alleles’: variants with an enhanced mutation rate. Mutator alleles have been found in both laboratory and natural populations. For instance, six of the twelve long-term evolution experiment (LTEE) lineages have fixed a mutator phenotype [7, 15]. In the wild, sampling of bacterial and fungal clades has revealed a large dispersity of mutation rates [16–18]. Even within the restricted timescales of somatic evolution, mutator alleles can have outsize effects, as many cancers are thought to be driven by recurrent genetic instability [19, 20].
Mutator alleles have been the subject of recent theoretical and experimental interest, as a model system to understand when evolution tunes its own parameters. However, determining when and how mutator alleles are favoured by natural selection is challenging. For example, if mutation rates are found to increase in a stressful environment [7, 21, 22], careful work must establish whether this was an inevitable biophysical consequence of the stress, accidental fixation of the mutator alleles, or actually due to the enhanced mutation rate via the acquisition of beneficial mutations. Unfortunately, the evolution of mutation rates is thought to depend on factors that are often difficult to measure in experiments, such as the balance between available deleterious and beneficial mutations [23–32]. As a consequence, it has been difficult to design systematic, well-controlled experiments.
Here, we take advantage of a novel experimental system to study the evolution of mutator alleles. We study the selection of mutagenic alleles of a DNA polymerase in an orthogonal replication system in yeast: the killer DNA virus [33–35]. Here, a DNA polymerase copies only a short cytosolic plasmid contained within the yeast cell. By choosing genes with known fitness landscapes, we control the number and strength of beneficial mutations available under direct selection, and study their consequences on the indirect selection of mutator alleles in the polymerase.
We find that mutator evolution is determined not only by the long-term benefit of mutations that it induces, but also by a short-term cost. This short-term cost — which we trace to a counterintuitive biophysical tradeoff between DNA polymerase speed or processivity and mutation rate [36] — grows with the long-term benefit and thus imposes a kinetic barrier to mutator fixation. Even when strongly beneficial mutations are available, mutators can be driven extinct by this short-term cost.
Combining population genetics theory with an engineered, two-gene selection scheme, we map out a phase diagram that predicts when mutators fix. Our results indicate that no one number specifies the ‘fitness’ of a mutator allele when its impact on short-term and long-term fitness are in conflict since they act on different timescales. Accordingly, we find that its population genetics can show ‘rock-paper-scissor’ dynamics. Further, we identify conditions under which ramping up selection strength (e.g., drug levels) over time can select for a mutator, even if no fixed selection strength can do so. We indicate how our results, though established in a concrete experimental system, generalize to evolvability-determining modifier alleles[32, 37, 38] at large, and highlights the fundamentally dynamic nature of fitness when direct and indirect selection are in conflict.
RESULTS
A. A minimal experimental system for the evolution of mutator alleles
Prior work on the evolution of mutator alleles has used a small number of mutator alleles that act on the entire genomes of bacteria or yeast [39–41]. Consequently, the number and strength of available beneficial and deleterious mutations — and how they vary in different environments — is unknown and difficult to manipulate experimentally.
We sought a model system in which we could exert rational control over these parameters. We turned to the DNA killer virus found in the yeast species K. lactis [33, 42, 43]. This system consists of two cytosolic ‘killer’ plasmids pGKL1 (p1) and pGKL2 (p2), each copied by a dedicated, orthogonal DNA polymerase. Recently, this system was adapted into a directed evolution platform, Orthorep [34, 35], by placing a gene of interest on p1 and linking that gene’s function to host fitness.
Here, we turn this directed evolution platform on its head by using it as an indirect selection platform on the underlying DNA polymerases. We encode a gene with known fitness landscapes (e.g. well-studied antibiotic resistance genes) on p1, select directly on that gene and study the resulting indirect evolution of the DNA polymerase.
Co-opting the OrthoRep directed evolution platform to study the indirect evolution of DNA polymerases provides for several advantages. We can manipulate the strength and availability of beneficial mutations since the p1 plasmid contains a chosen set of genes. The WT DNA polymerase, tp-dnap1, achieves very low mutation rates rates[34] by exploiting kinetic proofreading[44, 45], a broad error-correcting mechanism found across the Central Dogma [46]. As a consequence, mutants of this polymerase can differ in proofreading activity or other aspects of fidelity, resulting in a wide range of mutator alleles, with mutation rates ranging from 10−7 substitutions per basepair per generation (sbp) to 10−4 sbp [35, 47]. Note that at the mutation rates and timescales studied here, deleterious load is likely to be negligible over the kilobase length-scale of the p1 plasmid.
We first studied the indirect selection of proofreading polymerases during directed evolution of the metabolic gene dihydrofolate reductase (DHFR) from P. falciparum. PfDHFR activity is sensitive to the anti-marial drug pyrimethamine (pyr), against which it can acquire resistance mutations that have been well-characterized[48, 49]. To carry out this study, we deleted the dihydrofolate reductase (DHFR) gene from the yeast genome and encoded its homolog PfDHFR on p1. The concentration of pyr in our growth media acts as a tune-able selection pressure that acts on a known landscape of beneficial mutations.
We calibrated the system by measuring the strength of a pyr-resistant mutation as a function of pyr concentration. The selection coefficient ranges from without pyrimethamine to at 200 μM pyr, Fig. 1b. Using our experimentally measured data, we simulated a competition between strains with two different mutation rates at different levels of pyrimethamine, at a total population size . Consistent with prior theory [27, 31], the simulation predicts that, at high concentrations of pyr, the mutator allele acquires the beneficial mutation and outcompetes the non-mutator, Fig. 1c.
FIG. 1. A minimal experimental system for mutator evolution reveals mutators can be driven extinct by a short-time cost that increases with the size of beneficial mutations.
a In the killer DNA virus that underlies the Orthorep system, a dedicated DNA polymerase (‘dnap’) copies a cytosolic plasmid (‘p1’) independent of the host yeast replication machinery. Indicated are chosen mutator and non-mutator dnaps used in this paper. b We encoded the metabolic enzyme dihydrofolate reductase from P. falciparum (PfDHFR) onto p1, whose function is inhibited by the drug pyrimethamine. Drug resistant mutations in PfDHFR confer a fitness benefit whose strength varies with pyrimethamine concentration (right). c (Left) A simulated competition between a mutator and a non-mutator strain, varying the strength of the beneficial mutation. (Centre + Right) Experimental competition between dnaps 318 and 611 (centre), and 633 and RR5 (right), varying pyrimethamine levels. Two replicates are shown for each pyr level. d Definition of the short-term cost (stc) and long-term benefit (ltb). e Quantification of stc and ltb as a function of pyrimethamine concentration, for each of the pairs of polymerases competed together in c.
We then performed the same competition in experiment, choosing a pair of polymerases with mutation rates of 10−7 and 10−5 sbp (termed 318 and 611, respectively; see Fig. 1a). Strains expressing each polymerase were differentially labelled with fluorescent markers mKate and mVenus, mixed 1:1 and passaged daily into fresh media at varying concentrations of pyr. The population composition was monitored by flow cytometry at each dilution.
In contrast to simulation, we observed that the mutator strain 611 was never able to fix in the population. In fact, at higher pyr concentrations, the mutator was outcompeted faster, Fig. 1c. We confirmed that beneficial mutations are available to the mutator by competing the pyr-mutant in the mutator 611 background against the non-mutator 318. In this case, the pyr-mutant rapidly fixed in the population at high concentrations of pyr, confirming that selection can act on the beneficial mutation.
We then repeated the experiment with a different pair of polymerases: RR5 and 633. The non-mutator RR5 and mutator 633 had mutation rates of approx. 10−6 sbp and 10−4 sbp, respectively, similar to the last pair. However, unlike before, we found the mutator does eventually acquire the beneficial mutation and sweep the population at high pyr, Fig. 1c. However, even here the mutator first decreases in frequency before later sweeping the population. The contrast between the two pairs of polymerases, and between the experimental and simulation results, indicated that the mutation rate and the strength of the beneficial mutation are not sufficient to determine the eventual evolutionary outcome.
B. Short-term costs impose a kinetic barrier to mutator evolution
Why does the mutator 611 lose to 318 even when beneficial mutations are available? We noted that, in all competitions, the mutator population declined at early times. We quantified this ‘short-term cost’ as the early time selective advantage of the non-mutator, Fig. 1d. In addition, we characterised separately the selective benefit of a beneficial PfDHFR mutation induced by the mutagenic polymerases. Both quantities were found to increase with pyrimethamine, though the short-term cost was markedly stronger for the pair of polymerases in which the mutator eventually lost, Fig. 1e.
We traced the molecular origin of the short-term cost to the processivity of the DNA polymerase. p1 is a multicopy plasmid, with expression of p1-encoded genes increasing with copy number. We hypothesised that our mutagenic polymerases were unable to maintain p1 at a high copy number, and consequently express p1-encoded genes at lower levels with a corresponding effect on fitness, Fig. 2a. Consistent with this, flow-cytometry measurements of a fluorescent protein encoded on p1 showed a sharp reduction in fluorescence when p1 is replicated by the mutagenic polymerases, Fig. 2b.
FIG. 2. Short-term costs, arising from dnap activity, impose a kinetic barrier to mutator fixation.
a The activity of the dnap sets the copy number of p1, which determines the expression level of p1-encoded genes. b Fluorescence of p1-encoded mKate2 for the four dnaps, measured by flow cytometry. Grey dashed line indicates background fluorescence. Data shows that 318 is a more active dnap than 611, while 633 is comparable with RR5. c We calculate the fixation probability of a mutator allele as a function of its short-term cost, , as well as the long-term benefit gained from beneficial mutations, . d A schematic phase diagram, highlighting the kinetic exclusion regime in which mutators lose even though beneficial mutations are sufficiently strong to overcome the short-term-cost. e Replicate Wright-Fisher simulations in different regimes of the phase diagram in (d); at , mutator fixation is stochastic, indicating a cross-over rather than a sharp transition. f The mutator fixation probability is a universal function of , as shown by a data collapse of many parameter values (grey points) onto a master curve (black).
Differences in expression of PfDHFR also explain why both and grow with pyrimethamine. As pyrimethamine reduces the catalytic rate of the PfDHFR, its effects can be counteracted by a higher expression of the enzyme. Poor expression leads to a higher susceptibility to the drug, and thereby a larger advantage granted by drug-resistance mutations.
Does the short term cost explain the results of our competition experiments? Naively, no. The overall selective advantage of a mutator after it has picked up a beneficial mutation is . As for both pairs of polymerases at sufficiently high pyr, we would expect the mutator to have fixed in both cases.
However, this argument neglects the kinetics of the process. As the mutator is being outcompeted by the non-mutator, it only has a finite time in which to produce a mutant that survives drift and fixes in the population. We refer to this a ‘kinetic barrier’ to mutator fixation, and find that mutators overcome the kinetic barrier when,
| (1) |
This result is plotted as a phase boundary in the plane in Fig. 2d, and rationalizes the outcomes of our PfDHFR experiments. In one pair of the polymerases (318 vs 611), the high short-term cost paid by the mutator drives the competition into the kinetically excluded regime. Consequently, the mutator is unable to produce a winning mutant before it is driven extinct. In contrast, in the other pair of polymerases (RR5 vs 633), the mutator overcomes this kinetic barrier at high levels of pyrimethamine and is thereby able to fix in the population. (For simplicity, we assumed a zero mutation rate for the non-mutator in this analysis; we relax this assumption later.)
In Fig. 2e we present replicate Wright-Fisher simulations in different regimes of the plane, calculating the probability that the mutator eventually fixes at different values of and . Replotting the simulated data as a function of we observe a good data-collapse onto a master curve, confirming the validity of Eq. 1.
C. Orthogonal control of short term costs and long-term benefits
Thus far, our theory and experimental results make two key predictions. First, that the molecular origin of the short-term cost is the activity of the DNA polymerase. Second, that varying the short-term cost can shift the outcome from the mutator being driven extinct (due to the kinetic barrier), to one in which the mutator is able to acquire beneficial mutations and fix in the population.
These predictions are hard to test in many systems[7, 40] since short and long term effects are hard to disentangle in experiments. In our system, we sought to gain orthogonal control over short-term costs and long-term benefits via a two-gene strategy: one primarily responsible for the short-term cost and the other for the long-term benefit. We integrated the metabolic genes his3 and ura3 on-to p1, with an early stop codon in the open reading frame of ura3 (amino acid 29: caa → taa). The background strain was auxotrophic for histidine and uracil due to disabled his3 and ura3 genes in the genome, and requires expression of these genes from p1 to grow in media without histidine and uracil, respectively (Fig. 3a).
FIG. 3. Orthogonal experimental control of short-term costs and long-term benefits reveals different regimes of selection for mutators.
a Schematic of the his3-ura3* construct on the p1 plasmid, which allows for orthogonal control of short-term costs and long-term benfits by varying 3-AT and uracil levels in the media, respectively. b Doubling time of DNA polymerase monocultures in varying levels of 3-AT, showing that the mutagenic, but less active polymerases are more susceptible to the drug. c Measured short-term costs and long-term benefits at varying levels of uracil and 3-AT. Solid lines show the effect of varying only uracil or 3-AT. d Results from competition between 318 and 611, in depleted uracil media (3.6 g/L) at varying levels of 3-AT.
We reasoned that, in media lacking histidine and supplemented with 3-amino-1,3,4-triazole (3-AT, an inhibitor of his3 function, [50]), the growth-rate of the strain would be a function of his3 expression level, and thereby p1 copy number and polymerase activity. Conversely, removing uracil from the media provides a selective advantage for reversion of the stop codon in ura3. These were borne out in experiment, Fig. 3b. The strain therefore provides orthogonal control over short term costs and long-term benefits by varying the concentration of 3-AT and uracil, respectively.
We confirmed this relationship by measuring the dependence of and on 3-AT and uracil, varying the former between 0 and 50 mM and the latter between 10% and 100% of its usual concentration in our synthetic complete media (i.e. 24mg/L). As expected, and in contrast to our pyrimethamine results, we could now orthogonally tune and by varying uracil or 3-AT concentrations, respectively, Fig. 3c.
With this system in hand, we returned to the polymerases 318 and 611. With PfDHFR as the primary gene, 611 was unable to obtain a beneficial mutation in competition against 318, which we attributed to kinetic exclusion arising from the short-term cost paid by 611. Varying only the concentration of 3-AT, we now observed a transition from a regime in which 611 loses to one in which it picks up a beneficial mutation and fixes in the population, Fig. 3d. We conclude that the observed short term cost is due to differences in the processivity of the polymerase, and that varying the short-term cost is sufficient to alter the eventual fate of the mutator.
Occasionally, at an intermediate value of 3-AT, the mutator’s frequency goes down, up, and finally down again. Such dynamics is likely due to the non-mutator picking up a beneficial mutation at late times and outcompeting the mutator-mutant. Consistent with this, similar dynamics could be observed in simulations with a non-zero mutation rate for the non-mutator. This highlights the complex, non-monotonic population genetics possible with second-order selection.
D. An inverted speed-accuracy trade-off constrains the evolution of proofreading polymerases
Our results so far result from a conflict between short-term and long-term fitness effects in mutator evolution. However, this conflict may or may not be relevant to a broad class of mutator alleles. In the particular case of DNA polymerases, 611 and 633 might not be representative mutagenic alleles in that other mutagenic variants might not pay an activity cost. We therefore turned to studying a larger number of DNA polymerase variants, to see if the conflict between short and long-term fitness plays a role more broadly in polymerase evolution.
In Fig. 4 we plot data on 213 polymerase variants from the largest mutational screen of a DNA polymerase accomplished to date, carried out during the development of the OrthoRep platform [35]. Shown is the mutation rate against the copy number of the p1 plasmid supported by the polymerase — a measure of polymerase activity. Over a broad range, the data shows an inverse correlation between mutation rate and activity: highly mutagenic polymerases are less active, and would be expected to incur a short-term cost against a less-mutagenic, highly-active polymerase such as the wildtype. This result contradicts many prior theoretical expectations [51–55] since proofreading activity[44, 45] is expected to reduce mutation rates at the cost of speed or activity of the polymerase; see [36] for recent work on a potential biophysical explanation for this counterintutive tradeoff.
FIG. 4. An inverted speed-accuracy trade-off governs the evolution of mutation rates in proofreading DNA polymerases.
(a) Activity and mutation rate of the killer virus DNA polymerase variants characterized in [35] (b) We competed a range of mutant polymerases against the wildtype. Shown is the final fraction of the mutant in the population for each competition, ordered by (i) the mutation rate of the mutant, (ii) the activity of the mutant, and (iii) our criterion, Eq. 1 which correctly rank-orders the competition results.
Setting aside mechanistic origin, here, we focus on the implications for evolution of mutators. The counterintuitive results in Fig. 4 suggest that the evolution of mutagenic polymerases is strongly constrained by short-term costs. If a mutagenic polymerase were to arise in a population, it may be driven extinct despite its higher rate of acquiring beneficial mutations.
To test these ideas, we performed pairwise competition of several polymerase mutants against the wildtype. Six variants were chosen with a range of mutation rates and activities, confirmed by fluctuation assay and p1-encoded fluorescence, respectively. Polymerases were introduced into a strain harboring the his3-ura3* p1 construct and competed against a strain expressing the wildtype polymerase in low-uracil media, thereby selecting for reversion of the stop codon in ura3*.
Fig. 4 shows the frequency of the polymerase mutants in the population after ~ 40 generations, at which point each mutant is either less than 1% of the population or greater than 99% of the population. We rank order the polymerases by either mutation rate or activity (as assayed by p1 copy number, measured by flow cytometry), to see if either trait can explain the competition results.
Neither trait is predictive on its own. While several of the mutagenic polymerases fix against the wildtype, the most mutagenic polymerase is consistently driven extinct. Mutation rate is therefore a poor predictor of mutator fixation. Conversely, activity is also a poor predictor: polymerases 611 and 633-p drive the expression of a p1-encoded gene to similar levels, yet have strongly divergent outcomes in competition against the wildtype polymerase.
In contrast, ordering by the expression Eq. 1 — which takes into account both the polymerase activity and the mutation rate — appropriately separates polymerases that eventually outcompete the wildtype from those that do not – see SI for details. We conclude that the evolutionary fate of mutagenic polymerases depends not only on the long-term benefit of acquiring beneficial mutations, but also on the short-term cost paid in terms of polymerase activity.
E. ‘Rock-paper-scissors’ dynamics in pairwise competition
Our results so far invoke a tension between short-term costs and long-term benefits in the evolution of mutators. The former acts against the mutator at short times, while the latter favours it at longer times. However, from the perspective of population genetics, can these two factors be combined into a single number that denotes the overall effective ‘fitness’ of a mutator allele?
If such a parameter existed, we could rank-order DNA polymerases by that number into a strict hierarchy (Fig. 5a, left) that predicts outcomes of pairwise competition. In contrast, if instead we can find three mutator alleles whose outcomes in pairwise competition follows a ‘rock-paper-scissors’ dynamic (Fig. 5a, right), no such rank-ordering can exist.
FIG. 5. Conflicting short and long term fitnesses lead to unusual evolutionary dynamics of mutator alleles.
a When genotypes have a well-defined notion of fitness, they can be strictly rank-ordered by pairwise competition outcomes. In contrast, a cyclic ‘rock-paper-scissors’ outcome is inconsistent with any one number characterizing fitness. b Simulated pairwise competitions between three strains (Left): a non-mutator (:, ), a medium-mutator (:, ), and a mutator (:, ). and can acquire a beneficial mutation of strength . (Right) Pairwise competitions results, with fraction of simulations with indicated outcome written as a percentage. Panels show four representative trajectories from each competition. c Schematic showing a permissive, low-selection environment in gray, and three different harsh, high-selection environments in orange, black, and green. All environments are in the regime where (as defined in Eq. 1), in which the mutator loses in a static environment. d Schematic of dynamics in static and dynamic environments. In a static permissive environment, the mutator loses as no beneficial mutations are available. In a static harsh environment, mutators are outcompeted due to short-term cost before they can generate beneficial mutations. In a dynamic environment, however, mutators can accrue mutations in the permissive environment that then sweep when they environment becomes harsh. e Probability that the mutator fixes in a dynamic environment, as a function of time spent in the permissive environment. Points show simulated results for three different harsh environment with increasing (as indicated by colour in c). Solid lines are Eq. 2.
In Fig. 5b we present simulations of pairwise competitions between three strains that vary in mutation rate as well as short-term cost (Fig. 5b, left) picked from the scatter plot in Fig. 4a. As we observe in real DNA polymerases, the short-term cost grows with the mutation rate. For each strain, a beneficial mutation of fixed strength is accessible. When mutator is competed against the non-mutator , is driven extinct by the short-term cost before it can acquire a beneficial mutation. Similarly, though has a higher mutation rate than , both strains are eventually able to acquire the beneficial mutation. As pays a higher short-term cost than , loses to .
As has lost to , and has lost to , we would expect to lose to . However, we observe that mutation rate is so high that it rapidly acquire a winning beneficial mutation and wins against . Thus, , , and are an example of a ‘rock-papers-scissors’ trio, and the outcome of their pairwise competition cannot be captured by a single ‘fitness’ value. We conclude that no one number summarizes the fitness of a mutator allele.
F. Mutators are favored by ramping up selection pressure over time
We find that a mutator strain (say strain A) can be selected over a non-mutator (strain B) if the selection pressure on beneficial mutations is ramped up over time, even if strain B is selected over A at any fixed value of selection pressure.
As shown in Fig. 5c, we consider an environments with either fixed low selection or fixed high selection or environments that transition between a low-selection and a high-selection environment. In Wright-Fisher simulations, we find that the mutator loses to the non-mutator if both the fixed low-selection environment and fixed high-selection environment but wins if the environment transitions from low to high selection on a specific timescale ; see Fig. 5d.
This unusual situation arises from the fact that short-term cost and long-term benefit have effects that are separated in time. The short-term cost is paid immediately by the mutator, while the long-term benefit is reaped only when (and if) the mutator acquires a beneficial mutation.
As seen in our PfDHFR experiments (Fig. 1), environments in which the long-term benefit of a mutator allele are appreciable are also those in which the short-term costs are high. Consequently, in low selection environments, short-term costs are weak in the low-selection environment, but so are the long-term benefits. In high selection environments, both costs and benefits are high but mutators are driven to extinction by the short term costs before they can accrue a winning beneficial mutation.
By switching environments at the right timescale , mutators have the opportunity to establish a standing genetic variation of mutants in the low selection environment; these mutants can then be strongly selected in the changed high selection environment. Within a simplified analytical framework, we compute the probability that a mutator fixes as a function of time spent in the low-selection environment as
| (2) |
where , is the short-term cost in the permissive, low-selection environment, and is the fitness of a mutator-mutant in the harsh, high-selection environment.
Eq. 2 is plotted on top of simulation results in Fig. 5d, showing a good agreement and demonstrating that the ‘optimal’ is . We conclude that the temporal separation between short-term costs and long-term benefits allows the population genetics of the mutator to ‘resonate’ with environmental timescales, highlighting the unusual dynamics of second-order selection.
DISCUSSION
Our results show that that the evolution of mutation rates in polymerases is determined by a conflict between two traits - mutation rate and replication activity (e.g., speed or processivity). However, unlike other common cases of pleiotropy where one allele affects multiple traits, the two traits here manifest at different timescales. One trait — the mutation rate — is subject only to second-order selection that acts on the lineage, not the individual. The other trait – replication activity — is subject to first order selection, as it determines the replication speed of the individual’s genetic material.
If the two traits had been in concert, the above distinction would not have a qualitative impact. While our case study involved proofreading polymerases, the kind of pleiotropy seen here — negative direct costs associated with positive first order costs — should be expected to be seen just as often as the converse. For example, codon usage can change the availability of beneficial mutations and mutational robustness, a second order effect; but codon usage also typically has first order direct effects in terms of protein translational efficiency. These effects could be linked in conflicting directions.
Similarly, deleterious load, while mechanistically distinct from the direct cost in replication activity studied in this work, could effectively play a similar role as seen here[27, 31, 56]. In particular, if the positive and negative tails of distribution of fitness effects tend to scale together, higher mutation rates will be linked to higher deleterious load and more strongly beneficial mutations, much like in the work here. Consequently, the counterintuitive evolutionary consequences such as rock-paper-scissors and environmental timescale-dependent selection of mutators might be seen in a range of contexts beyond proofreading polymerases.
Our work suggests the intriguing possibility that mutation rates could ‘ratchet’ or lock in place at unusually low mutation rates. The trade-off in Fig. 4a suggests drift towards higher mutation rate would be selected against due to loss of activity which appears to be significant for the WT polymerase here. As a consequence, drift dynamics would be ratchet-like, with many more non-deleterious mutations available towards lower than towards higher . In fact, this possibility could offer an alternative explanation for why the mutation rate of the killer DNA virus exploited here is so low, adding a layer to the drift-barrier hypothesis[3, 28]. The WT mutation rate is substitutions per base while the WT p1 plasmid is of length bases. The resulting deleterious load is weak; our quantitative results in this study (e.g., Fig.4) suggest that the need for sufficient replication activity exerts a substantially stronger selective pressure than deleterious load in maintaining such unusually low mutation rates.
Supplementary Material
ACKNOWLEDGMENTS
We thank the Kavli Institute for Theoretical Physics, University of California, Santa Barbara (UCSB), where some of this work was done. AM acknowledges support from the NIGMS of the NIH under award number R35GM151211 and the NSF Physics of Living Systems (PHY-2310781). BHG is a Chan Zuckerberg Biohub - San Francisco Investigator, and acknowledges support from the NIGMS of the NIH (award number R35GM146949) and the Alfred P. Sloan Foundation (grant FG-2021-15708). This work was supported by the National Science Foundation through the Center for Living Systems (grant no. 2317138).
References
- [1].Losos J. B. The Princeton guide to evolution (Princeton University Press, 2017). [Google Scholar]
- [2].Cairns J., Overbaugh J. & Miller S. The origin of mutants. Nature 335, 142–145 (1988). [DOI] [PubMed] [Google Scholar]
- [3].Lynch M. et al. Genetic drift, selection and the evolution of the mutation rate. Nat. Rev. Genet. 17, 704–714 (2016). [DOI] [PubMed] [Google Scholar]
- [4].Matic I. et al. Highly variable mutation rates in commensal and pathogenic escherichia coli. Science 277, 1833–1834 (1997). [DOI] [PubMed] [Google Scholar]
- [5].LeClerc J. E., Li B., Payne W. L. & Cebula T. A. High mutation frequencies among escherichia coli and salmonella pathogens. Science 274, 1208–1211 (1996). [DOI] [PubMed] [Google Scholar]
- [6].Loeb L. A. Human cancers express mutator phenotypes: origin, consequences and targeting. Nature Reviews Cancer 11, 450–457 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [7].Sniegowski P. D., Gerrish P. J. & Lenski R. E. Evolution of high mutation rates in experimental populations of e. coli. Nature 387, 703–705 (1997). [DOI] [PubMed] [Google Scholar]
- [8].Sane M., Diwan G. D., Bhat B. A., Wahl L. M. & Agashe D. Shifts in mutation spectra enhance access to beneficial mutations. Proceedings of the National Academy of Sciences 120, e2207355120 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [9].Couce A., Guelfo J. R. & Blázquez J. Mutational spectrum drives the rise of mutator bacteria. PLoS genetics 9, e1003167 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [10].Keightley P. D. & Otto S. P. Interference among deleterious mutations favours sex and recombination in finite populations. Nature 443, 89–92 (2006). [DOI] [PubMed] [Google Scholar]
- [11].Rutherford S. L. & Lindquist S. Hsp90 as a capacitor for morphological evolution. Nature 396 (1998). [DOI] [PubMed] [Google Scholar]
- [12].Johnson M. S. & Desai M. M. Mutational robustness changes during long-term adaptation in laboratory budding yeast populations. eLife 11, e76491 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [13].Woods R. J. et al. Second-order selection for evolvability in a large escherichia coli population. Science 331, 1433–1436 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [14].Kunkel T. A. & Bebenek K. DNA replication fidelity 1. Annu. Rev. Biochem. 69, 497–529 (2000). [DOI] [PubMed] [Google Scholar]
- [15].Barrick J. E. & Lenski R. E. Genome dynamics during experimental evolution. Nat. Rev. Genet. 14, 827–839 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [16].Drake J. W., Charlesworth B., Charlesworth D. & Crow J. F. Rates of spontaneous mutation. Genetics 148, 1667–1686 (1998). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [17].Zhu Y. O., Siegal M. L., Hall D. W. & Petrov D. A. Precise estimates of mutation rate and spectrum in yeast. Proceedings of the National Academy of Sciences 111, E2310–E2318 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [18].Jiang P. et al. A modified fluctuation assay reveals a natural mutator phenotype that drives mutation spectrum variation within saccharomyces cerevisiae. Elife 10, e68285 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [19].Turajlic S., Sottoriva A., Graham T. & Swanton C. Resolving genetic heterogeneity in cancer. Nature Reviews Genetics 20, 404–416 (2019). [DOI] [PubMed] [Google Scholar]
- [20].Coelho M. C., Pinto R. M. & Murray A. W. Heterozygous mutations cause genetic instability in a yeast model of cancer evolution. Nature 566, 275–278 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [21].Shor E., Fox C. A. & Broach J. R. The Yeast Environmental Stress Response Regulates Mutagenesis Induced by Proteotoxic Stress. PLOS Genetics 9, e1003680 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [22].Habig M., Lorrain C., Feurtey A., Komluski J. & Stukenbrock E. H. Epigenetic modifications affect the rate of spontaneous mutations in a pathogenic fungus. Nature Communications 12, 5869 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [23].Kimura M. On the evolutionary adjustment of spontaneous mutation rates. Genetics Research 9, 23–34 (1967). [Google Scholar]
- [24].Liberman U. & Feldman M. W. Modifiers of mutation rate: a general reduction principle. Theoretical population biology 30, 125–142 (1986). [DOI] [PubMed] [Google Scholar]
- [25].Dawson K. J. Evolutionarily stable mutation rates. Journal of theoretical biology 194, 143–157 (1998). [DOI] [PubMed] [Google Scholar]
- [26].Lynch M. The cellular, developmental and population-genetic determinants of mutation-rate evolution. Genetics 180, 933–943 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [27].Desai M. M. & Fisher D. S. The balance between mutators and nonmutators in asexual populations. Genetics 188, 997–1014 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [28].Lynch M. The lower bound to the evolution of mutation rates. Genome biology and evolution 3, 1107–1118 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [29].Soderberg R. J. & Berg O. G. Kick-starting the ratchet: The fate of mutators in an asexual population. Genetics 187, 1129–1137 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [30].James A. & Jain K. Fixation probability of rare non-mutator and evolution of mutation rates. Ecology and Evolution 6, 755–764 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [31].Good B. H. & Desai M. M. Evolution of mutation rates in rapidly adapting asexual populations. Genetics 204, 1249–1266 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [32].Ferrare J. T. & Good B. H. Evolution of evolvability in rapidly adapting populations. Nature Ecology & Evolution (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [33].Gunge N. & Sakaguchi K. Intergeneric transfer of deoxyribonucleic acid killer plasmids, pgkl1 and pgkl2, from kluyveromyces lactis into saccharomyces cerevisiae by cell fusion. Journal of bacteriology 147, 155–160 (1981). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [34].Ravikumar A., Arrieta A. & Liu C. C. An orthogonal DNA replication system in yeast. Nat. Chem. Biol. 10, 175–177 (2014). [DOI] [PubMed] [Google Scholar]
- [35].Ravikumar A., Arzumanyan G. A., Obadi M. K. A., Javanpour A. A. & Liu C. C. Scalable, continuous evolution of genes at mutation rates above genomic error thresholds. Cell 175, 1946–1957.e13 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [36].Ravasio R. et al. A minimal scenario for the origin of non-equilibrium order (2024). URL http://arxiv.org/abs/2405.10911.
- [37].Payne J. L. & Wagner A. The causes of evolvability and their evolution. Nature Reviews Genetics 20, 24–38 (2019). [DOI] [PubMed] [Google Scholar]
- [38].Barnett M., Zeller L. & Rainey P. B. Experimental evolution of evolvability. bioRxiv 2024–05 (2024). [DOI] [PubMed] [Google Scholar]
- [39].Desai M. M., Fisher D. S. & Murray A. W. The speed of evolution and maintenance of variation in asexual populations. Current biology 17, 385–394 (2007). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [40].McDonald M. J., Hsieh Y.-Y., Yu Y.-H., Chang S.-L. & Leu J.-Y. The evolution of low mutation rates in experimental mutator populations of saccharomyces cerevisiae. Current Biology 22, 1235–1240 (2012). [DOI] [PubMed] [Google Scholar]
- [41].Wei W. et al. Rapid evolution of mutation rate and spectrum in response to environmental and population-genetic challenges. Nature communications 13, 4752 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [42].Stark M. & Boyd A. The killer toxin of Kluyveromyces lactis: characterization of the toxin subunits and identification of the genes which encode them. The EMBO Journal 5, 1995–2002 (1986). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [43].Klassen R. & Meinhardt F. Linear protein-primed replicating plasmids in eukaryotic microbes. In Microbial Linear Plasmids, 187–226 (Springer, 2007). [Google Scholar]
- [44].Hopfield J. J. Kinetic proofreading: a new mechanism for reducing errors in biosynthetic processes requiring high specificity. Proceedings of the National Academy of Sciences 71, 4135–4139 (1974). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [45].Ninio J. Kinetic amplification of enzyme discrimination. Biochimie 57, 587–595 (1975). [DOI] [PubMed] [Google Scholar]
- [46].Phillips R. The molecular switch: Signaling and Allostery (Princeton University Press, 2020). [Google Scholar]
- [47].Rix G. et al. Continuous evolution of user-defined genes at 1-million-times the genomic mutation rate (2023). URL https://www.biorxiv.org/content/10.1101/2023.11.13.566922v1. [DOI] [PMC free article] [PubMed]
- [48].Lozovsky E. R. et al. Stepwise acquisition of pyrimethamine resistance in the malaria parasite. Proceedings of the National Academy of Sciences 106, 12025–12030 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [49].Sirawaraporn W., Sathitkul T., Sirawaraporn R., Yuthavong Y. & Santi D. V. Antifolate-resistant mutants of plasmodium falciparum dihydrofolate reductase. Proceedings of the National Academy of Sciences 94, 1124–1129 (1997). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [50].Fields S. The two-hybrid system to detect protein-protein interactions. Methods 5, 116–124 (1993). [Google Scholar]
- [51].Bennett C. H. Dissipation-error tradeoff in proofreading. BioSystems 11, 85–91 (1979). [DOI] [PubMed] [Google Scholar]
- [52].Banerjee K., Kolomeisky A. B. & Igoshin O. A. Elucidating interplay of speed and accuracy in biological error correction. Proceedings of the National Academy of Sciences 114, 5183–5188 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [53].Mallory J. D., Kolomeisky A. B. & Igoshin O. A. Tradeoffs between error, speed, noise, and energy dissipation in biological processes with proofreading. The Journal of Physical Chemistry B 123, 4718–4725 (2019). [DOI] [PubMed] [Google Scholar]
- [54].Midha T., Mallory J. D., Kolomeisky A. B. & Igoshin O. A. Synergy among pausing, intrinsic proofreading, and accessory proteins results in optimal transcription speed and tolerable accuracy. The Journal of Physical Chemistry Letters 14, 3422–3429 (2023). [DOI] [PubMed] [Google Scholar]
- [55].Murugan A., Huse D. A. & Leibler S. Speed, dissipation, and error in kinetic proofreading. Proceedings of the National Academy of Sciences 109, 12034–12039 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
- [56].Raynes Y., Wylie C. S., Sniegowski P. D. & Weinreich D. M. Sign of selection on mutation rate modifiers depends on population size. Proceedings of the National Academy of Sciences 115, 3422–3427 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.





