JOURNAL OF BACTERIOLOGY, Aug. 2004, p. 4846–4852 0021-9193/04/$08.00⫹0 DOI: 10.1128/JB.186.15.4846–4852.2004 Copyright © 2004, American Society for Microbiology. All Rights Reserved.
Vol. 186, No. 15
Adaptive Mutation in Escherichia coli Patricia L. Foster* Department of Biology, Indiana University, Bloomington, Indiana 47405 figured out if this similarity is informative or merely a coincidence. The high postplating reversion rate of FC40 allowed us to eliminate a number of artifactual explanations for the phenomenon. In our first paper (10) we ruled out the possibility that the appearance of Lac⫹ mutants was due simply to proliferation of the Lac⫺ population. We demonstrated that the mutations only appeared when lactose was present, not if the cells were merely starving, and thus were adaptive. Importantly, we showed that reversion to Lac⫹ during lactose selection, but not during nonselective growth, was dependent on some function or functions of RecA. Thus, our results demonstrated that the mechanism by which adaptive mutations occur is different from the mechanism by which growth-dependent mutations occur. It was this feature of FC40—a high rate of adaptive mutation that occurs by a distinct mechanism—that made it attractive for further study.
In 1988 John Cairns, Julie Overbaugh, and Stephan Miller published a paper entitled “The origin of mutants,” in which they suggested that bacteria could choose which mutations to make (11). Shortly thereafter, Cairns and I began collaborating on a National Science Foundation-funded project to investigate the genetic basis of what was popularly called “directed mutation” (although at the time we were calling it “selectiondependent mutation,” which now seems as good a name as any). Our plan was to mutagenize an appropriate strain of Escherichia coli and identify and characterize variants defective in selection-dependent mutation. Cairns and coworkers had been investigating a strain called SM195, described in their original paper (11). However, the mechanisms of mutation in this strain had proved to be complicated, and we wanted a new experimental subject. We were setting up to investigate selection-induced activation of the cryptic bgl operon (36), when Jeffrey Miller gave us a Lac⫺ strain, called ␣45, that he said had a high rate of reversion to Lac⫹ after plating on minimum lactose medium. The lac allele in this strain is a fusion of lacI to lacZ that eliminates the lac regulatory region as well as several residues of lacI and lacZ; transcription of the fusion is constitutive, initiating at the lacI promoter (7). ␣45 has a ⫹1 frameshift mutation, lacI33, in the lacI coding region (12). Miller and coworkers have used several similar strains to study various mutational mechanisms (52, 67). As in many of the strains that originated with Jacob and Monod, proAB and lac are deleted from the chromosome and carried on an episome, F⬘128. This arrangement greatly facilitates genetic manipulations and turned out to be important to adaptive mutation. We mated the episome from ␣45 into a ⌬(lac-pro) recipient that we had made rifampin resistant and named the new strain FC40 (Foster and Cairns #40). In our first paper (10), we showed that Lac⫹ revertants of FC40 accumulate at a constant rate for about a week after the cells are plated on minimal lactose plates. Two days after plating (the first day that Lac⫹ colonies appear), the numbers of mutants among cultures has a Luria-Delbru ¨ck distribution (meaning that the mutations occurred prior to plating), whereas on subsequent days the distribution becomes Poisson (meaning that the mutations occurred after plating). After 5 days on lactose plates, there are about 100 Lac⫹ colonies per 108 cells plated. We calculated that the normal preplating mutation rate was about 10⫺9 Lac⫹ revertants per cell per generation, whereas the postplating mutation rate was about 10⫺9 per cell per h; considering that the doubling time in minimal medium is about an hour, this means that per unit time, the two mutation rates are the same. We have never
MUTATIONS ARE NOT DIRECTED Fairly early on in our studies, Cairns and I eliminated the hypothesis that mutations were “directed” toward a useful goal. The first negative evidence was obtained not with FC40, but with SM195. SM195 has an amber mutation in lacZ and so reverts both by intragenic mutations and by the creation of tRNA suppressors (11). The continued appearance of extragenic suppressors during lactose selection allowed us to dismiss the hypothesis that the selective conditions “instructed” the cell to make appropriate mutations—in the case of extragenic suppressors, there is no direct path from the phenotype (Lac⫹) to the mutated gene (encoding a tRNA) (23). Later it was shown that about two-thirds of the late-appearing Lac⫹ revertants of SM195 were due to slow-growing ochre suppressors that probably arose during growth prior to lactose selection (57). Nonetheless, the continued appearance of fast-growing amber suppressors in addition to the true revertants demonstrated that mutations appear elsewhere than in the gene directly under selection (24). The second piece of evidence against directed mutation was obtained by putting a second revertible allele, a ⫹1 frameshift in the tetA gene, close to the Lac⫺ allele in FC40. During lactose selection, Tetr revertants appeared at about the same rate as did Lac⫹ mutations and had the same genetic requirements (21). The frequency of double Lac⫹ Tetr mutants in these experiments indicated that the two events were not independent (21). Nonetheless, the occurrence of nonselected mutations during lactose selection demonstrated that the mutational mechanism was not directed at a specific gene.
* Mailing address: Department of Biology, Indiana University, Jordan Hall, 1001 East Third St., Bloomington, IN 47405. Phone: (812) 855-4084. Fax: (812) 855-6705. E-mail:
[email protected]. 4846
VOL. 186, 2004
GENETICS OF ADAPTIVE MUTATION The following summarizes what has been learned in the last decade about the mechanism of adaptive Lac⫹ mutation in FC40. (i) The Lacⴙ mutations that arise during lactose selection are different from the mutations that arise during nonselective growth. While a variety of deletions, duplications, and frameshifts revert the Lac⫺ allele during growth, adaptive Lac⫹ mutations consist almost exclusively of ⫺1-bp frameshifts in runs of iterated bases (27, 61). (ii) Adaptive but not growth-dependent reversion to Lacⴙ requires the recombination functions for double-strand break repair encoded by recA and recBCD and the DNA branch migration and resolution functions encoded by ruvAB and ruvC (10, 19, 30, 39, 40). In contrast, mutations in recG, which encodes an enzyme also involved in DNA branch migration, increase the rate of adaptive mutation up to 100-fold (30, 40). (See below for a new explanation of this phenotype.) (iii) The high level of adaptive reversion to Lacⴙ in FC40 requires that the Lac allele be on the episome. When the same allele is at its normal position on the chromosome, the rate of adaptive reversion to Lac⫹ falls about 100-fold and the mutations are no longer recA dependent (28, 58). (iv) The high level of adaptive Lacⴙ mutation also requires that one or more conjugal functions be expressed (28, 32). When conjugal functions are not expressed, the rate of adaptive mutation falls 10-fold but, unlike Lac⫹ mutations on the chromosome, 90% of the remaining Lac⫹ mutations are still dependent on RecA (28). We find that actual conjugation is not required (28, 29), although others disagree (33, 34). The production of DNA nicks at the conjugal origin is most likely the conjugal function that promotes adaptive mutation (59). (v) At least 50% of the adaptive Lacⴙ mutations are eliminated if DNA polymerase IV (Pol IV) is defective (22, 51). DNA polymerases IV and V were discovered only recently. Both are induced as part of E. coli’s SOS response to DNA damage, and both are highly error prone (35). However, Pol V is not involved in adaptive mutation (10, 50). (The roles in adaptive mutation of all of E. coli’s polymerases are discussed below.) (vi) Adaptive Lacⴙ mutation is reduced about 90% if the general stress sigma factor, RpoS, is eliminated (44, 48). Part, but not all, of this effect is due to the fact that Pol IV is positively regulated by RpoS (44). (vii) About 1% of the Lacⴙ mutants that appear after plating on lactose medium have other mutations somewhere in their genome (60, 69). These extra mutations do not appear if Pol IV is defective (68) or if the SOS response is repressed (63). MODEL FOR ADAPTIVE Lacⴙ MUTATION Our current model for adaptive mutation to Lac⫹ is as follows. When FC40 is incubating on lactose, the cells are not proliferating but occasionally replication is initiated at one of the episome’s vegetative origins, oriS or oriV. Because of the persistent nicking at the conjugal origin, oriT (31), the replication fork will have a high probability of encountering a nick and collapsing, thus creating a double-strand end. The double-
DIALOG
4847
strand end is recognized by the RecBCD enzyme, which initiates double-strand-break repair by degrading the 5⬘ strand, producing an invasive 3⬘ single strand. After RecA-catalyzed invasion of the single strand into the homologous duplex of the sister chromosome or another episome, new DNA synthesis is primed, initially requiring the actions of DNA Pol IV or II and the PriA primosome. As the polymerase (mainly Pol IV) copies the lac region, errors are made that give rise to the Lac⫹ mutants. Eventually a new replication fork incorporating the normal replicase, DNA Pol III, is established, but the fourstranded recombination intermediate (a Holliday junction) remains and must be resolved by RuvAB-catalyzed branch migration and RuvC-catalyzed strand cleavage. This model accounts for all the features and genetic requirements of adaptive mutation in FC40. The Lac⫺ allele is sufficiently “leaky” so that enough lactose is metabolized to provide the energy for episome replication and recombination. We assume that during lactose selection SOS genes are derepressed to some extent, allowing for expression of the LexArepressed genes recA, ruvAB, polB (which encodes DNA Pol II), and dinB (which encodes DNA Pol IV). In addition, Pol IV is further induced in late stationary phase under control of RpoS (44). The resulting high levels of Pol IV allow it to outcompete other more accurate polymerases, such as Pol II, for access to the DNA termini provided by recombination. While the population is incubating on lactose, a subpopulation of cells enters into a transient state of increased mutation. But only 10% of the Lac⫹ mutations arise in these hypermutators, whereas 90% of the Lac⫹ mutations arise in “normal” cells by the pathway outlined above (60) (others disagree with this conclusion; see reference 8). ROLE OF THE ERROR-PRONE POLYMERASE, Pol IV Recently we discovered that Pol IV is induced late in stationary phase under positive control of RpoS, the stationaryphase sigma factor. After induction, high levels of the protein are maintained for at least 3 days after the cells have reached stationary phase (44). Other researchers have shown that the dinB gene is transcribed in a 5-day-old culture (71). The induction of Pol IV under RpoS control helps to explain several puzzling features of adaptive mutation in FC40. SOS derepression. In a LexA-defective (Def) mutant strain the SOS response is fully derepressed, and one would expect that the levels of adaptive mutation would be high. However, in FC40 adaptive mutation is inhibited in a LexA(Def) mutant strain because psiB, an episomal gene encoding an inhibitor of SOS (2), is also induced (50). Therefore, during lactose selection there must be enough active LexA repressor to keep psiB and other LexA-repressed genes from being fully induced. However, Pol IV and other functions required for adaptive mutation that are also RpoS regulated can be induced independently of LexA. Variability of results. The level of mutation obtained in adaptive mutation experiments can vary from two- to fivefold. Since the level of Pol IV depends on RpoS, this variation may reflect the degree to which RpoS is activated. This, in turn, may depend on uncontrolled variables such as the age of the culture, the moisture content of the plates, etc.
4848
DIALOG
Role of RecG. The recG gene encodes an enzyme that can resolve Holliday junctions and rescue stalled replication forks (49). Adaptive mutation is eliminated in ruvA, -B or -C mutant strains but is increased 100-fold in a recG mutant strain (30, 40). The solution to this puzzle is that Pol IV levels are elevated two to fivefold in recG mutant strains, and all of the extra adaptive mutations in the recG mutant strain are due to Pol IV (44). Presumably Pol IV levels are elevated because the SOS response is partially derepressed in recG mutant strains (47). Role of Pol II. Pol II, encoded by the polB gene, is an accurate DNA polymerase that is also induced as part of the SOS response (4). In a polB mutant strain adaptive mutation is increased 10-fold (15). As in a recG mutant strain, levels of Pol IV are elevated in polB mutant cells (44) and all of the extra mutations are due to Pol IV (22). However, Pol II has an additional, possibly direct, role in adaptive mutation. A mutation in Pol II’s proofreader domain that leaves its polymerase domain intact increases adaptive mutations; in a polB mutant strain that has a high adaptive mutation rate due to Pol IV, that rate is further elevated if the proofreader-mutant Pol II is supplied on a plasmid (25). These results suggest that Pol II is active during lactose selection, that it competes with Pol IV and limits Pol IV’s mutagenic activity, and that when overexpressed Pol IV outcompetes Pol II. Role of Pol III. Pol IV, like other error-prone polymerases, is poorly processive (35). After a mutation is made, Pol III, the replicase, presumably is required to complete the synthesis of the error-containing molecule so that the mutant can be recovered. But Pol III may also have a direct role in adaptive mutation since some other polymerase must make the 20 to 50% of the mutations that are not due to Pol IV. The polymerase subunit of Pol III is encoded by the dnaE gene; a mutation that improves polymerase fidelity, dnaE915 (17), reduces adaptive mutation threefold (25, 38). Before the discovery of Pol IV, we assumed that this result meant that Pol III was responsible for most Lac⫹ adaptive mutations. Now we believe that Pol IV makes most of the mutations; but why then does dnaE915 result in such a strong phenotype? The dnaE915 allele also reduces adaptive mutation in recG and polB mutant strains, whose high mutation rates are due to overexpression of Pol IV. Were this because the dnaE915-encoded polymerase cannot complete replication, one would expect that dnaE915 mutant strains would hardly be viable, which is not the case. I hypothesize that the dnaE915 allele is an antimutator because it does not allow Pol IV (and possibly Pol V) access to DNA termini (22). GROWTH OF A SUBPOPULATION ⫹
The Lac mutants that appear after plating FC40 on minimal lactose medium cannot be due simply to mutations occurring as a result of proliferation of the Lac⫺ population. If Lac⫹ mutations occurred at the normal growth-dependent rate, the population would have to increase 50- to 100-fold during the course of an experiment, which it clearly does not (10, 20). First Lenski et al. (46), and then Andersson et al. (1), proposed that Lac⫹ adaptive mutations arise not in the general population, but in a subpopulation that is increasing in number. But under carefully controlled conditions, no growth or death of Lac⫺ FC40 cells can be detected (10). The amount of growth
J. BACTERIOL.
that could occur and still remain undetected would account for, at most, one mutation in 5 days at the normal growthdependent mutation rate (20). Another way of looking at these results is that, in a typical experiment when 3 ⫻ 108 FC40 cells are plated on a lactose plate and incubated for 5 days, no more than 3 ⫻ 108 cell divisions could take place and be undetected. To give rise to 100 mutants, the mutation rate would have to be at least 3 mutations per 107 cell divisions, which is 100-fold higher than the normal mutation rate. This is true whether the mutations are occurring in all the cells or in only a subpopulation of them (20). This calculation has recently been incorporated into the models presented by Roth’s group, who now propose that the mutation rate to Lac⫹ has to be at least 3.5 ⫻ 10⫺7 mutations per replication of the lac allele (42). Another argument against models that postulate that Lac⫹ mutations are produced by a growing subpopulation is that the rate at which Lac⫹ mutations appear in FC40 is constant for at least 5 days after the cells are put under lactose selection (10). After day 5 the rate at which Lac⫹ colonies appear often accelerates (see below). Confining ourselves to the first 5 days, the mutations that appear cannot be due to an increasing population of cells or an increasing number of copies of the lac allele because then the rate at which Lac⫹ mutants appear would increase (unless, bizarrely, the mutation rate per lac allele decreases to compensate) (20). Finally, a growing subpopulation presumably would give rise to microcolonies out of which true Lac⫹ revertants would arise. I looked for these microcolonies in two ways. First, I examined plates each day and marked the position of tiny colonies (which, at a magnification of ⫻30, become visible when they consist of 104 cells). If these microcolonies were giving rise to Lac⫹ mutants, then some would remain microcolonies and others would produce large colonies over the next few days. All microcolonies observed produced large colonies by the next day, suggesting that they were all composed of Lac⫹ cells (20). The second experiment was modeled on the original “Newcombe respreading experiment” (53). Each day for 3 days after plating FC40 on lactose plates, I chose a few plates without visible Lac⫹ colonies and respread them. This should have redistributed the cells in each microcolony with two possible results, depending on the mutation rate to Lac⫹. A few days after respreading, the respread plates should either have more Lac⫹ colonies than nonrespread plates or they should produce another crop of microcolonies. I observed neither (20). I conclude that either microcolonies do not exist or they have the unusual property of always producing exactly one Lac⫹ mutant. AMPLIFICATION In 1992 Cairns and I suggested that adaptive reversion of FC40 could be explained by amplification of the Lac⫺ allele (23). We were led to this hypothesis by a paper published in 1984 by Jeffrey Miller’s group (67). In this study they used a strain carrying the same lacI-lacZ fusion as that carried by FC40, but with a different mutation, X13, in the lacI gene. Several days after the X13 strain was plated on lactose plates, Lac⫹ colonies appeared that had an unstable Lac⫹ phenotype (i.e., Lac⫹ clones segregated Lac⫺ cells when picked and
VOL. 186, 2004
grown under nonselective conditions). The appearance and instability of these phenotypically Lac⫹ clones were dependent on recA⫹. The parallels to FC40 are obvious, but there is a very important difference. X13, an amber mutation, is much more “leaky” than the I33 allele, producing about eight times more -galactosidase. (FC40 produces 2 U of -galactosidase when Lac⫺ and 200 when Lac⫹.) This difference in leakiness is probably because the X13 mutation creates a translation reinitiation site (67), whereas the leakiness of the I33 mutation may depend on ribosomal frameshifting. The greater amount of -galactosidase produced by the X13 strain means that it slowly proliferates on minimal lactose plates and that the early steps of the amplification process confer large growth advantages. In this respect the Salmonella enterica serovar Typhimurium strains used by Roth’s group, which also proliferate on lactose medium (32), resemble X13 more closely than FC40. Cairns and I hypothesized that when Lac⫺ FC40 cells are plated on lactose plates, a small proportion of them amplify the lac region by a RecA-dependent process. However, the amplified copies are unstable. Among the copies, true Lac⫹ mutations occur, and as the Lac⫹ mutant cell starts to proliferate, the extra Lac⫺ copies are resolved by a RecA-dependent process. During deamplification the Lac⫹ copy that provides the selective advantage on lactose is retained and a pure Lac⫹ colony results. We further speculated that the extra DNA copies might have an increased rate of mutation (18, 23). A similar model was proposed by John Roth and Franklin Stahl in 1993 in an unpublished manuscript, later published in 1998 (1). This model differs only in proposing that cells with an amplified Lac⫺ allele grow on lactose, producing microcolonies out of which true Lac⫹ revertants arise. This hypothesis makes the same predictions as the Lenski hypothesis, discussed above, and is subject to the same criticisms. If true, the rate at which Lac⫹ revertants appear should increase with time, but that rate is constant for 5 days. Thus, if the amplification hypothesis is true, the number of cells that have amplifications, or the extent of their amplification, must be a constant unless the reversion rate per gene copy declines with time (20). As mentioned above, the rate at which Lac⫹ colonies of FC40 appear typically starts to accelerate after 5 days of incubation on lactose. About 2% of the FC40 Lac⫹ clones isolated on days 5 to 7 are composed mostly of cells that have amplified the Lac⫺ allele (20, 25). If incubation is continued for 10 days, up to 60% of the newly arising Lac⫹ colonies consist of amplifiers (41, 56). The late appearance of these colonies reflects the fact that they are slower to develop than colonies composed of true Lac⫹ revertants (41). But these results do not refute or confirm the hypothesis that true Lac⫹ revertants arise in microcolonies that initially are made up of cells that are amplifying the lac allele. For this, one needs to isolate newly arisen microcolonies and see if these contain cells that are unstably Lac⫹. When I did this experiment I found that none of 82 newly arisen microcolonies contained cells that produced sectored colonies when streaked on nonselective Lac indicator medium (20). In contrast, Hendrickson et al. (42) found that 98% of newly arisen colonies contained unstable Lac⫹ cells, although in nearly half of the colonies these were only 1% or less of the cells. One difference between our experiments is that I looked for sectoring on minimal medium plus X-Gal (5-bromo-4-chloro-3-indolyl--D-galactopyranoside) whereas
DIALOG
4849
Hendrickson et al. used rich medium plus X-Gal, suggesting that sectoring is either suppressed by minimal medium or stimulated by rich medium. Another test of the amplification hypothesis is to measure the amount of lac DNA in Lac⫺ cells incubating in lactose. The degree of amplification and the number of amplifying cells suggested by Andersson et al. (1) predicts that the amount of lac DNA among the Lac⫺ population has to increase at least 30-fold to account for the mutations that arise. We measured the amount of lac DNA in Lac⫺ cells of FC40 incubating in lactose medium over 3 days and found that it did not increase relative to that of a control gene (26). We later confirmed this result and found that the amount of lac DNA doubles between days 3 and 4 (W. A. Rosche and P. L. Foster, unpublished results). Since it takes 2 days for a Lac⫹ colony to become visible on lactose plates, these results mean there is no detectable increase in lac DNA during the time in a normal experiment when about 100 true Lac⫹ mutants arise at a constant rate. But, there is an increase in lac DNA corresponding to the point late in an experiment when colonies appear that are composed of cells that have amplified the unreverted Lac⫺ allele. Since amplification is dependent on recombination, genetic conditions that increase recombination and also increase Lac⫹ adaptive mutation should increase the frequency of Lac⫹ colonies containing unstable Lac⫹ cells. We have tested this prediction in three cases: in a recD mutant strain (26), when RecBCD function is replaced by bacteriophage ’s Red recombination function (55), and in a strain with an enzyme that produces DNA nicks targeted to the lac region (59). Although adaptive mutation was increased 10- to 100-fold in each of these backgrounds, the frequency of Lac⫹ colonies composed of unstable Lac⫹ cells was the same as in FC40. Surprisingly, the Red functions, which convey a hyperrecombinogenic phenotype, increased adaptive reversion of the lac allele when it was on the chromosome. Adaptive reversion of the chromosomal allele is not normally dependent on recombination (28, 58). About half of the Red-dependent Lac⫹ colonies contained unstable Lac⫹ cells, indicating that the Red functions can induce or stabilize amplification on the chromosome but not on the episome (55). So where are we now? A recent variant of the amplification model has been proposed by Roth’s group (42). This variant postulates that fewer cells start to amplify and the extent of their amplification is less; to compensate, the mutation rate increases ⬎35-fold due to induction and amplification of Pol IV (42, 62). How can this be reconciled with the fact that the accumulation of Lac⫹ mutants is linear with time? Roth’s group hypothesizes that a given amplified array is constantly being destroyed and reformed (63). Although not stated explicitly, they apparently believe that this process would allow for a constant rate of mutant production. But how can such instability be consistent with the fact that clones can be isolated with up to 50 copies of the lac region (1) and that some cells build up enough Lac⫺ copies to become phenotypically Lac⫹ (20)? I find it easier to believe that there are two processes occurring: a mutational process that produces true revertants and an amplification process that produces slowly growing colonies that start appearing after about 5 days of incubation. But it is interesting that the original Roth-Stahl hypothesis has
4850
DIALOG
J. BACTERIOL.
evolved so that it is now almost exactly what Cairns and I proposed years ago (18, 20, 23). HYPERMUTATION It has been well documented that a subpopulation of FC40 cells under lactose selection experience a transient state of increased mutation (34, 42, 60, 69). The same phenomenon has been observed in other strains of E. coli (3, 37) and in Salmonella (63). Transient hypermutation was predicted by Hall (37) and modeled by Ninio (54) and Cairns (9, 60). Hall hypothesized “the hypermutable state” to explain how mutations could appear to be “directed” to useful genes; the hypermutators would die unless a good mutation occurred, and thus nonselected mutations would be lost from the population. Of course, as discussed above, we now know that mutations are not directed (21). Nevertheless, Hall’s hypothesis appears to be essentially true, although the hypermutating cells do not die during the course of our experiments (60). Note that we and Rosenberg and coworkers call the state hypermutation, or transient mutation, whereas Roth and coworkers call it general mutagenesis or general hypermutability. In FC40 and in Roth’s Salmonella strains, hypermutation requires Pol IV (62, 68). In addition, mismatch repair is not active among hypermutators (60). Thus, we hypothesize that the hypermutator state is due to the combination of induced expression of Pol IV plus decreased activity of mismatch repair. We further hypothesize that the mutations in the hypermutators arise by the same recombination-dependent mechanism that produces mutations in normal cells, but the mutagenicity of the process is enhanced in the hypermutators (68). There is disagreement about what proportion of the Lac⫹ adaptive mutations arise in hypermutators. To answer this question, we need to know the size of the population of hypermutating cells and how elevated are their mutation rates. To estimate these parameters, a set of simple algebraic equations was derived using the proportions of Lac⫺ and Lac⫹ clones that have unselected mutations. When these equations were solved using data from FC40, they showed that the hypermutators are about 0.1% of the total population under lactose selection and that their mutation rate is elevated about 200-fold. From these results it is easy to calculate that only 10% of the Lac⫹ adaptive mutations arise in hypermutators and 90% arise in normal cells (60). Even without algebra it is easy to demonstrate that only a minority of Lac⫹ revertants come from the hypermutators. We assume that all cells bearing two or more mutations come from the hypermutating population (because the chance of obtaining a double mutant if the normal mutation rates apply is extremely small). For our study we used defects in motility as the major unselected phenotype. Among 3,168 Lac⫹ isolates, 210 had motility defects, but among 13 Lac⫹ isolates that had a mutation giving one of five other phenotypes, 8 had motility defects. Thus, the frequency of motility defects among Lac⫹ mutants (0.07) was nearly 10-fold lower than among Lac⫹ mutants with another mutation (0.6). Since the double mutants are all from the hypermutators, the single mutants cannot be because the single mutants have a lower mutation rate. The data generated by Torkelson et al. (69) supports the same
conclusion. They found that the frequency of Mal⫺ mutants among Lac⫹ mutants was 36-fold lower than the frequency among Lac⫹ mutants with another mutation. Slechta et al. (62) have recently questioned whether the hypermutator state is induced in response to environmental stress or is simply an accident. They hypothesize that hypermutation occurs only when the dinB gene is amplified with the lac allele, which happens in about one-fifth of the amplifying cells. This hypothesis is consistent with our calculations. However, it is inconsistent with the fact that hypermutators also appear when a strain with the lac allele on the chromosome is under lactose selection (60); Lac⫹ adaptive mutations in this strain normally are not RecA dependent and thus are not due to amplification (28). Recent evidence strongly suggests that both adaptive mutation and hypermutation are induced responses to stress. As discussed above, several components of adaptive mutation in FC40 are part of the SOS response (16), which is at least partially induced during lactose selection. Some components of adaptive mutation, in particular DNA Pol IV, are also positively regulated by RpoS (44), which is the regulator of the general stress response (43). In addition, key components of mismatch repair are down-regulated under the control of RpoS (70). Although mismatch repair is active during lactose selection (23), the fact that mismatch repair proteins are in low supply may mean that in some cells the pathway is saturated or components are not present (as suggested by Ninio [54]). Thus, we hypothesize that in most cells during lactose selection, Pol IV is induced but in the hypermutator population, Pol IV is further induced and mismatch repair is also down-regulated, so the mutation rate is elevated about 200-fold (44). However, dividing the population into high and low mutators is a simplification; in fact, there may be many populations of cells with many different mutation rates. EVOLUTIONARY SIGNIFICANCE There are several aspects of adaptive mutation in FC40 that may be important in evolution. The first is recombinationdependent mutation. In our strains this mechanism is particularly active on the F⬘ element that carries the lac allele, probably because of the persistent nick induced at oriT. But the same mechanism can be expected to occur at some frequency whenever a nick is encountered during DNA replication. This may not be a major source of variation in growing organisms when other mutational mechanisms are active, but it might become significant in static populations. Many of the components of this system—RecA, Pol IV, and RuvAB—are induced as part of the SOS response to DNA damage. The SOS response is also induced in aging colonies (65) and at the end of growth in rich medium (14). In addition, Pol IV is positively regulated by the general stress sigma factor RpoS (44) and is expressed in starving cells (44, 71). Similar inducible systems may exist in other organisms, allowing mutations to occur when genetic variability may be advantageous. We have argued that the mutation rate of genes on the F episome is unusually high. Part of the evidence for this has been dismissed because one target, a Tets element, is close to the lac allele and so its reversion to Tetr may not be independent of reversion to Lac⫹. However, under nonselective con-
VOL. 186, 2004
DIALOG
ditions reversion of this Tets allele on the episome is higher than when it is on the chromosome (our group’s unpublished results). In addition, another episomal site, the codAB operon, has a high mutation rate in our experiments (60). The F plasmid is unusual in that it can recombine with the chromosome and pick up chromosomal genes. Thus, it could be an agent of horizontal gene transfer. About 20% of natural isolates of E. coli carry F or related conjugal plasmids (6). These plasmids show evidence of extensive recombination and horizontal transfer among diverse E. coli strains and between E. coli and Salmonella (5, 6). If chromosomal genes are picked up by F, exposed to a high mutation rate, and then recombined back onto the chromosome, F would indeed be a potent evolution machine. A number of experimental and theoretical studies have shown that individual organisms with high mutation rates can have a selective advantage in a changing environment (13, 45, 64). Indeed, models predict that the random appearance of a mutator allele can accelerate the adaptive evolution of an entire population (66). The transient mutator state described above would be even more advantageous. When confronted with an adverse situation, only a very small proportion of the population become hypermutators and most cells have a normal mutation rate. If the current problem can be solved with a single advantageous mutation, it is likely to appear in the cells with normal mutation rates. They will proliferate and carry no extra mutational burden. However, if advantageous mutations are rare, or if more than one mutation is needed, the hypermutating cells will succeed and proliferate. They will carry extra mutations, but because the hypermutable state is transient, their mutation rates return to normal, minimizing the genetic burden carried by their progeny. ACKNOWLEDGMENTS I thank John Cairns for his continuing collaboration and Jeffrey H. Miller for strains and advice. Work in my laboratory is supported by grant MDB-9996308 from the National Science Foundation and Public Health Service grant GM065175 from the National Institutes of Health. REFERENCES 1. Andersson, D. I., E. S. Slechta, and J. R. Roth. 1998. Evidence that gene amplification underlies adaptive mutability of the bacterial lac operon. Science 282:1133–1135. 2. Bagdasarian, M., A. Bailone, M. M. Bagdasarian, P. A. Manning, R. Lurz, K. N. Timmis, and R. Devoret. 1986. An inhibitor of SOS induction, specified by a plasmid locus in Escherichia coli. Proc. Natl. Acad. Sci. USA 83:5723– 5726. 3. Boe, L. 1990. Mechanism for induction of adaptive mutations in Escherichia coli. Mol. Microbiol. 4:597–601. 4. Bonner, C. A., S. Hays, K. McEntee, and M. F. Goodman. 1990. DNA polymerase II is encoded by the DNA damage-inducible dinA gene of Escherichia coli. Proc. Natl. Acad. Sci. USA 87:7663–7667. 5. Boyd, E. F., and D. L. Hartl. 1997. Recent horizontal transmission of plasmids between natural populations of Escherichia coli and Salmonella enterica. J. Bacteriol. 179:1622–1627. 6. Boyd, E. F., C. W. Hill, S. M. Rich, and D. L. Hartl. 1996. Mosaic structure of plasmids from natural populations of Escherichia coli. Genetics 143:1091– 1100. 7. Brake, A. J., A. V. Fowler, I. Zabin, J. Kania, and B. Mu ¨ller-Hill. 1978. -Galactosidase chimeras: primary structure of a lac repressor--galactosidase protein. Proc. Natl. Acad. Sci. USA 75:4824–4827. 8. Bull, H. J., G. J. McKenzie, P. J. Hastings, and S. M. Rosenberg. 2000. Response to John Cairns: the contribution of transiently hypermutable cells to mutation in stationary phase. Genetics 156:925–926. 9. Cairns, J. 1998. Mutation and cancer: the antecedents to our studies of adaptive mutation. Genetics 148:1433–1440. 10. Cairns, J., and P. L. Foster. 1991. Adaptive reversion of a frameshift mutation in Escherichia coli. Genetics 128:695–701.
4851
11. Cairns, J., J. Overbaugh, and S. Miller. 1988. The origin of mutants. Nature 335:142–145. 12. Calos, M. P., and J. H. Miller. 1981. Genetic and sequence analysis of frameshift mutations induced by ICR-191. J. Mol. Biol. 153:39–66. 13. Chao, L., and E. C. Cox. 1983. Competition between high and low mutating strains of Escherichia coli. Evolution 37:125–134. 14. Dri, A.-M., and P. L. Moreau. 1994. Control of the LexA regulon by pH: evidence for a reversible inactivation of the LexA repressor during the growth cycle of Escherichia coli. Mol. Microbiol. 12:621–629. 15. Escarceller, M., J. Hicks, G. Gudmundsson, G. Trump, D. Touati, S. T. Lovett, P. L. Foster, K. McEntee, and M. F. Goodman. 1994. Involvement of Escherichia coli DNA polymerase II in response to oxidative damage and adaptive mutation. J. Bacteriol. 176:6221–6228. 16. Fernandez de Henestrosa, A. R., T. Ogi, S. Aoyagi, D. Chafin, J. J. Hayes, H. Ohmori, and R. Woodgate. 2000. Identification of additional genes belonging to the LexA regulon in Escherichia coli. Mol. Microbiol. 35:1560–1572. 17. Fijalkowska, I. J., R. L. Dunn, and R. M. Schaaper. 1993. Mutants of Escherichia coli with increased fidelity of DNA replication. Genetics 134: 1023–1030. 18. Foster, P. L. 1992. Directed mutation: between unicorns and goats. J. Bacteriol. 174:1711–1716. 19. Foster, P. L. 1993. Adaptive mutation: the uses of adversity. Annu. Rev. Microbiol. 47:467–504. 20. Foster, P. L. 1994. Population dynamics of a Lac⫺ strain of Escherichia coli during selection for lactose utilization. Genetics 138:253–261. 21. Foster, P. L. 1997. Nonadaptive mutations occur on the F⬘ episome during adaptive mutation conditions in Escherichia coli. J. Bacteriol. 179:1550–1554. 22. Foster, P. L. 2000. Adaptive mutation in Escherichia coli. Cold Spring Harbor Symp. Quant. Biol. 65:21–29. 23. Foster, P. L., and J. Cairns. 1992. Mechanisms of directed mutation. Genetics 131:783–789. 24. Foster, P. L., and J. Cairns. 1998. Adaptive mutation of a lacZ amber allele. Genetics 150:1329–1330. 25. Foster, P. L., G. Gudmundsson, J. M. Trimarchi, H. Cai, and M. F. Goodman. 1995. Proofreading-defective DNA polymerase II increases adaptive mutation in Escherichia coli. Proc. Natl. Acad. Sci. USA 92:7951–7955. 26. Foster, P. L., and W. A. Rosche. 1999. Increased episomal replication accounts for the high rate of adaptive mutation in recD mutants of Escherichia coli. Genetics 152:15–30. 27. Foster, P. L., and J. M. Trimarchi. 1994. Adaptive reversion of a frameshift mutation in Escherichia coli by simple base deletions in homopolymeric runs. Science 265:407–409. 28. Foster, P. L., and J. M. Trimarchi. 1995. Adaptive reversion of an episomal frameshift mutation in Escherichia coli requires conjugal functions but not actual conjugation. Proc. Natl. Acad. Sci. USA 92:5487–5490. 29. Foster, P. L., and J. M. Trimarchi. 1995. Conjugation is not required for adaptive reversion of an episomal frameshift mutation in Escherichia coli. J. Bacteriol. 177:6670–6671. 30. Foster, P. L., J. M. Trimarchi, and R. A. Maurer. 1996. Two enzymes, both of which process recombination intermediates, have opposite effects on adaptive mutation in Escherichia coli. Genetics 142:25–37. 31. Frost, L. S., and J. Manchak. 1998. F⫺ phenocopies: characterization of expression of the F transfer region in stationary phase. Microbiology 144: 2579–2587. 32. Galitski, T., and J. R. Roth. 1995. Evidence that F plasmid transfer replication underlies apparent adaptive mutation. Science 268:421–423. 33. Godoy, V. G., and M. S. Fox. 2000. Transposon stability and a role for conjugational transfer in adaptive mutablility. Proc. Natl. Acad. Sci. USA 97:7393–7398. 34. Godoy, V. G., F. S. Gizatullin, and M. S. Fox. 2000. Some features of the mutability of bacteria during nonlethal selection. Genetics 154:49–59. 35. Goodman, M. F. 2002. Error-prone repair DNA polymerases in prokaryotes and eukaryotes. Annu. Rev. Biochem. 71:17–50. 36. Hall, B. G. 1988. Adaptive evolution that requires multiple spontaneous mutations. I. Mutations involving an insertion sequence. Genetics 120:887– 897. 37. Hall, B. G. 1990. Spontaneous point mutations that occur more often when they are advantageous than when they are neutral. Genetics 126:5–16. 38. Harris, R. S., H. J. Bull, and S. M. Rosenberg. 1997. A direct role for DNA polymerase III in adaptive reversion of a frameshift mutation in Escherichia coli. Mutat. Res. 375:19–25. 39. Harris, R. S., S. Longerich, and S. M. Rosenberg. 1994. Recombination in adaptive mutation. Science 264:258–260. 40. Harris, R. S., K. J. Ross, and S. M. Rosenberg. 1996. Opposing roles of the Holliday junction processing systems of Escherichia coli in recombinationdependent adaptive mutation. Genetics 142:681–691. 41. Hastings, P. J., H. J. Bull, and S. M. Rosenberg. 2000. Adaptive amplification: an inducible chromosomal instability mechanism. Cell 103:723–731. 42. Hendrickson, H., E. S. Slechta, U. Bergthorsson, D. I. Andersson, and J. R. Roth. 2002. Amplification-mutagenesis: evidence that “directed” adaptive mutation and general hypermutability result from growth with a selected gene amplification. Proc. Natl. Acad. Sci. USA 99:2164–2169.
4852
DIALOG
43. Hengge-Aronis, R. 2002. Signal transduction and regulatory mechanisms involved in control of the s (RpoS) subunit of RNA polymerase. Microbiol. Mol. Biol. Rev. 66:373–395. 44. Layton, J. C., and P. L. Foster. 2003. Error-prone DNA polymerase IV is controlled by the stress-response sigma factor, RpoS, in Escherichia coli. Mol. Microbiol. 50:549–561. 45. Leigh, E. G. 1970. Natural selection and mutability. Am. Natural. 104:301– 305. 46. Lenski, R. E., M. Slatkin, and F. J. Ayala. 1989. Mutation and selection in bacterial populations: alternatives to the hypothesis of directed mutation. Proc. Natl. Acad. Sci. USA 86:2775–2778. 47. Lloyd, R. G., and C. Buckman. 1991. Genetic analysis of the recG locus of Escherichia coli K-12 and of its role in recombination and DNA repair. J. Bacteriol. 173:1004–1011. 48. Lombardo, M.-J., I. Aponyi, and S. M. Rosenberg. 2004. General stress response regulator RpoS in adaptive mutation and amplification in Escherichia coli. Genetics, 166:669–680. 49. McGlynn, P., and R. G. Lloyd. 2002. Genome stability and the processing of damaged replication forks by RecG. Trends Genet. 18:413–419. 50. McKenzie, G. J., R. S. Harris, P. L. Lee, and S. M. Rosenberg. 2000. The SOS response regulates adaptive mutation. Proc. Natl. Acad. Sci. USA 97:6646–6651. 51. McKenzie, G. J., P. L. Lee, M.-J. Lombardo, P. J. Hastings, and S. M. Rosenberg. 2001. SOS Mutator DNA polymerase IV functions in adaptive mutation and not adaptive amplification. Mol. Cell 7:571–579. 52. Miller, J. H. 1985. Mutagenic specificity of ultraviolet light. J. Mol. Biol. 182:45–65. 53. Newcombe, H. B. 1949. Origin of bacterial variants. Nature 164:150–151. 54. Ninio, J. 1991. Transient mutators: a semiquantitative analysis of the influence of translation and transcription errors on mutation rates. Genetics 129:957–962. 55. Poteete, A. R., H. R. Wang, and P. L. Foster. 2002. Phage Red-mediated adaptive mutation. J. Bacteriol. 184:3753–3755. 56. Powell, S. C., and R. M. Wartell. 2001. Different characteristics distinguish early versus late arising adaptive mutations in Escherichia coli FC40. Mutat. Res. 473:219–228. 57. Prival, M. J., and T. Cebula. 1996. Adaptive mutation and slow-growing revertants of an Escherichia coli lacZ amber mutant. Genetics 144:1337– 1341. 58. Radicella, J. P., P. U. Park, and M. S. Fox. 1995. Adaptive mutation in Escherichia coli: a role for conjugation. Science 268:418–420.
J. BACTERIOL. 59. Rodriguez, C., J. Tompkin, J. Hazel, and P. L. Foster. 2002. Induction of a DNA nickase in the presence of its target site stimulates adaptive mutation in Escherichia coli. J. Bacteriol. 184:5599–5608. 60. Rosche, W. A., and P. L. Foster. 1999. The role of transient hypermutators in adaptive mutation in Escherichia coli. Proc. Natl. Acad. Sci. USA 96:6862– 6867. 61. Rosenberg, S. M., S. Longerich, P. Gee, and R. S. Harris. 1994. Adaptive mutation by deletions in small mononucleotide repeats. Science 265:405– 407. 62. Slechta, E. S., K. L. Bunny, E. Kugelberg, E. Kofoid, D. I. Andersson, and J. R. Roth. 2003. Adaptive mutation: general mutagenesis is not a programmed response to stress but results from rare coamplification of dinB with lac. Proc. Natl. Acad. Sci. USA 100:12847–12852. 63. Slechta, E. S., J. Liu, D. I. Andersson, and J. R. Roth. 2002. Evidence that selected amplification of a bacterial lac frameshift allele stimulates Lac(⫹) reversion (adaptive mutation) with or without general hypermutability. Genetics 161:945–956. 64. Sniegowski, P. D., P. J. Gerrish, and R. E. Lenski. 1997. Evolution of high mutation rates in experimental populations of E. coli. Nature 387:703–705. 65. Taddei, F., I. Matic, and M. Radman. 1995. cAMP-dependent SOS induction and mutagenesis in resting bacterial populations. Proc. Natl. Acad. Sci. USA 92:11736–11740. 66. Taddei, F., M. Radman, J. Maynard-Smith, B. Toupance, P. H. Gouyon, and B. Godelle. 1997. Role of mutator alleles in adaptive evolution. Nature 387:700–702. 67. Tlsty, D. T., A. M. Albertini, and J. H. Miller. 1984. Gene amplification in the lac region of E. coli. Cell 37:217–224. 68. Tompkins, J. D., J. E. Nelson, J. C. Hazel, S. L. Leugers, J. D. Stumpf, and P. L. Foster. 2003. Error-prone polymerase, DNA polymerase IV, is responsible for transient hypermutation during adaptive mutation in Escherichia coli. J. Bacteriol. 185:3469–3472. 69. Torkelson, J., R. S. Harris, M.-J. Lombardo, J. Nagendran, C. Thulin, and S. M. Rosenberg. 1997. Genome-wide hypermutation in a subpopulation of stationary-phase cells underlies recombination-dependent adaptive mutation. EMBO J. 16:3303–3311. 70. Tsui, H.-C. T., G. Feng, and M. E. Winkler. 1997. Negative regulation of mutS and mutH repair gene expression by the Hfq and RpoS global regulators of Escherichia coli K-12. J. Bacteriol. 179:7476–7487. 71. Yeiser, B., E. D. Pepper, M. F. Goodman, and S. E. Finkel. 2002. SOSinduced DNA polymerases enhance long-term survival and evolutionary fitness. Proc. Natl. Acad. Sci. USA 99:8737–8741.
