Separate DNA Pol II- and Pol IV-Dependent Pathways of Stress-Induced Mutation during Double-Strand-Break Repair in Escherichia coli Are Controlled by RpoS

Abstract

Previous work showed that about 85% of stress-induced mutations associated with DNA double-strand break repair in carbon-starved Escherichia coli result from error-prone DNA polymerase IV (Pol IV) (DinB) and that the mutagenesis is controlled by the RpoS stress response, which upregulates dinB. We report that the remaining mutagenesis requires high-fidelity Pol II, and that this component also requires RpoS. The results identify a second DNA polymerase contributing to stress-induced mutagenesis and show that RpoS promotes mutagenesis by more than the simple upregulation of dinB.

Stress-induced mutagenesis is a collection of mechanisms observed in bacterial, yeast, and human cells in which mutation pathways are activated in response to adverse conditions, such as starvation or antibiotic stresses, under the control of stress responses (18). The coupling of mutagenesis to stress responses generates genetic diversity upon which natural selection can act specifically when cells are maladapted to their environment, i.e., are stressed. These mechanisms are potentially important models for mutagenesis that drives pathogen-host adaptation (5, 44), antibiotic resistance (8, 9, 18, 31, 35), cancer progression, and resistance (3).

Perhaps the most well-characterized mechanism of stress-induced mutagenesis is mutagenesis associated with DNA double-strand break (DSB) repair in carbon-starved Escherichia coli cells, as characterized in the E. coli Lac assay (7) and related assays. In the Lac assay, cells with an F′-borne lac +1-bp frameshift allele are starved on lactose medium on which they accumulate Lac⁺ reversion mutations (7), chromosomal tet frameshift (6), or ampD base substitution and frameshift mutations (41). Starvation in stationary phase without lactose produces a similar effect (42). The Lac reversions occur either by compensatory frameshift (point) mutations (15, 48) or by the amplification of the weakly functional lac gene to 20 to 50 copies (24, 43). Stress-induced point mutagenesis in lac (33, 34), ampD (41), and tet (42), as well as stress-induced lac amplification (34), require the RpoS-controlled general stress response. The point mutagenesis mechanism is addressed here.

The point mutagenesis mechanism appears to be a switch from high-fidelity to error-prone DSB repair under stress, and it is controlled by the SOS DNA damage and RpoS stress responses. Point mutagenesis requires the induction of the SOS (36), RpoS (33, 34), and σ^E-controlled extracytoplasmic unfolded protein (20) stress responses, the proteins of DSB repair by homologous recombination (16, 22, 23, 28), either the F-encoded TraI single-strand endonuclease (42) or a DSB delivered near lac by I-SceI endonuclease expressed in vivo (42), and the DinB error-prone DNA polymerase IV (Pol IV) (13, 37), which SOS (10, 30) and RpoS (33) upregulate transcriptionally. The sole role of the SOS response in point mutagenesis is the upregulation of Pol IV (17). The mutations occur in acts of DSB repair that become mutagenic under the influence of the SOS and RpoS responses as follows. First, the delivery of an I-SceI-made DSB near lac increases mutagenesis ∼6,000-fold (above TraI⁻ levels), completely bypassing the requirement for TraI single-strand endonuclease and stimulating mutation an additional 70-fold (42). TraI-generated single-strand nicks are thought to become DSBs upon replication (32, 46, 47). Second, the I-SceI-promoted mutations have the same sequences as those formed normally (42) and also require DSB repair proteins RecA, RecB, and RuvABC, an inducible SOS and RpoS response, and DinB, indicating that they form by the same mutation pathway as that used when I-SceI cuts are not given. Third, I-SceI-generated DSBs provoked mutation 6,000-fold when made near lac but only 3-fold when made in another DNA molecule, supporting the idea that the mutations form in acts of DSB repair. Finally, I-SceI-generated DSBs and their repair are not always mutagenic and do not always use Pol IV, but they become mutagenic via Pol IV either in stationary phase (when RpoS is expressed normally) or in unstressed log-phase cells if RpoS is expressed artificially (42). These data showed that RpoS somehow licenses the use of Pol IV in acts of DSB repair either during stress or if RpoS is expressed artificially in unstressed cells. Precisely how RpoS allows Pol IV into acts of DSB repair is unknown, as is whether that is its only role in stress-induced point mutagenesis. Here, we show that RpoS plays at least one other role.

Although Pol IV is required for ∼85% of stress-induced point mutagenesis, 15% remained that was Pol IV independent (37). These remaining point mutations also were −1-bp deletions, many of them in simple repeat sequences, and so they appeared likely to be DNA polymerase errors made by a different DNA polymerase (37).

Here, we show that the Pol IV-independent stress-induced mutagenesis requires the polB gene, encoding DNA Pol II, a relatively high-fidelity DNA polymerase that plays roles in lagging-strand replication (2, 19), some DNA repair processes (reviewed in reference 19), and replication restart after DNA damage (45). Additionally, we show that RpoS is required for both the Pol IV- and Pol II-dependent mutagenesis. The data imply that either Pol IV or II can generate the errors during DSB repair that become point mutations and indicate that the role of RpoS in mutagenesis is more than the simple upregulation of DinB.

MATERIALS AND METHODS

Bacterial strains and media.

E. coli K-12 strains used in this study are listed in Table 1. New genotypes were constructed using standard phage P1-mediated transduction (38) or phage λ Red-mediated recombineering methods (11). Transductants and integrants were selected on Luria-Bertani-Herskowtiz (LBH) plates (51) containing antibiotics at the following concentrations (in μg/ml): kanamycin, 30; chloramphenicol, 25; tetracycline, 10; rifampin, 100; and ampicillin; 100. Sodium citrate (20 mM) was added for transductions. LBH plates used for the construction of strains containing the P_BADI-SceI transcriptional fusion contained 0.1% (wt/vol) glucose to repress I-SceI gene transcription (42). M9 minimal medium (38) contained 10 μg/ml thiamine and 0.1% (wt/vol) glucose, glycerol, or lactose. Relevant genotypes were confirmed via antibiotic resistance, PCR followed by gel electrophoresis, catalase activity, and/or UV light sensitivity.

TABLE 1.

E. coli strains used in this study

Strain	Relevant genotype	Reference or source
FC29	Δ(lac-proAB)XIIIara thi [F′ ΔlacIZ proAB+]	7
FC40	Δ(lac-proAB)XIIIara thi Rifr [F′ lacIΩlacZ33 proAB+]	7
SMR3661	FC40 polBΔ1::ΩSm-Sp	25
SMR4562	Independent construction of FC40	36
SMR5830	FC40 dinB10 [F′ dinB10]	37
SMR5833	SMR4562 [pKD46]	Laboratory stock
SMR6276	SMR4562 ΔaraBAD567 Δattλ::PBADI-SceI	42
SMR6308	SMR6276 [F′ codA21::mini-Tn7Kan(I-SceI site) ΔtraI::dhfr]	42
SMR6509	SMR6308 Δ(srlR-recA)306::Tn10	42
SMR6541	SMR4562 rpoS::Tn10	34
SMR6755	SMR6308 dinB10 [F′ dinB10]	42
SMR8949	SMR5830 polBΔ1::ΩSm-Sp	25
SMR11437	SMR6308 rpoS::Tn10	SMR6308 × P1(SMR6541)
SMR11439	SMR6755 rpoS::Tn10	SMR6755 × P1(SMR6541)
SMR12282	SMR5833 ΔpolB802::FRTcatFRT	SMR5833 × FRTcatFRT from pKD3 (11)
SMR12367a	SMR6308 ΔpolB802::FRTcatFRT	SMR6308 × P1(SMR12282)
SMR12369a	SMR6755 ΔpolB802::FRTcatFRT	SMR6755 × P1(SMR12282)
SMR12371a	SMR11437 ΔpolB802::FRTcatFRT	SMR11437 × P1(SMR12282)
SMR12374a	SMR11439 ΔpolB802::FRTcatFRT	SMR11439 × P1(SMR12282)

^aThese strains do not carry the ΔaraBAD567 allele, because the ara⁺ genes were introduced linked with ΔpolB.

Stress-induced mutagenesis assays.

The Lac assay was performed as described for both I-SceI-free (23) and I-SceI-endonuclease-producing (42) strains with the following modifications. Cells were cultured in liquid medium at 32°C for 3 days instead of 37°C for 2 days prior to plating on M9 lactose. As observed previously, overall stress-induced mutation rates are higher with the lower-temperature preincubation, but mutations that decrease mutagenesis do so similarly at both temperatures (42). The relative viability was monitored as described previously (23). Lac⁺ colonies originate either by a −1-bp frameshift point mutation (15, 48) or by the amplification of the leaky lac allele to multiple copies (24, 43). Because amplification accounts for a small portion of the Lac⁺ colonies arising on day 5 and earlier (24), we have not corrected for levels of amplification. Data shown represent the means ± standard errors of the means (SEM) of at least three and usually four independent cultures, each plated on at least two independent M9 lactose plates.

Statistical methods.

Statistical analyses were performed using the Statplus software package.

RESULTS

DNA polymerase II is required for Pol IV-independent stress-induced point mutagenesis.

McKenzie et al. (37) showed that the loss of dinB caused the loss of ∼85% of stress-induced Lac⁺ point mutagenesis. We find that DNA Pol II, encoded by polB, is required for the residual stress-induced mutagenesis remaining in dinB cells (Fig. 1). As reported previously (14, 21), the deletion of polB increased reversion (Fig. 1A and B). However, the deletion of polB in dinB cells strongly reduced the remaining Lac reversion (Fig. 1A and B). polB dinB cells produced fewer mutants than dinB cells. We conclude that DNA Pol II can account for much of the Pol IV-independent mutation. Because the effect of polB mutation adds to that of dinB mutation (it is not epistatic), we conclude that Pol II and Pol IV promote mutations independently of each other, not by concerted action. polB has no effect on gene amplification (50), supporting the conclusion that it is the remaining point mutagenesis that requires Pol II.

FIG. 1.

Pol II is required for Pol IV-independent stress-induced mutagenesis. (A) Representative experiment. The dinB10 mutation (□; strain SMR5830) removes about 85% of point mutagenesis (37) compared to pol⁺ SMR4562 cells (○). Whereas the deletion of polB alone (▵; SMR3661) increases mutagenesis (12, 14), apparently due to the competition of the higher-fidelity Pol II with Pol IV (25), polB dinB SMR8949 cells (⋄) show decreased stress-induced mutagenesis compared with that of dinB cells, indicating that Pol II is required for the DinB-independent mutagenesis. In this and all figures, error bars represent 1 standard error of the means (SEM) and, where not visible, are smaller than the symbol. (B) Quantification of fold differences between isogenic strains from four experiments. Here, and in Fig. 2D and Table 2, mutation rates are mutants per 10⁸ cells per day, determined as the change between days 3 and 5. Values listed are fold differences between the two strains indicated and were derived from each independent experiment done in parallel and averaged from four experiments.

RpoS is required for DinB-independent stress-induced mutagenesis stimulated by I-SceI endonuclease cuts made in vivo.

RpoS transcriptionally upregulates dinB about 2-fold (33), which might have been its sole contribution to DSB-associated stress-induced point mutagenesis. We wished to determine whether RpoS controls all or only the DinB-dependent fraction of stress-induced point mutagenesis. One way to address this is to determine whether the loss of RpoS knocks down mutagenesis to the same extent as dinB mutation does or to the greater extent observed in dinB polB double mutants. However, because the mutagenesis defects caused by dinB, rpoS, or polB dinB double mutations are so large (34, 37) (Fig. 1), greater reductions might be difficult to quantify accurately, so we performed these experiments in strains with the mutation rate elevated by the provision of a DSB near lac using I-SceI endonuclease expressed in vivo (per the method of reference 42). This sensitizes the assay for the detection of mutagenesis defects. The presence of an I-SceI-mediated DSB increased the Lac reversion rate ∼30-fold over the level for the enzyme-only (no-cut site) control (Fig. 2 D). In these strains with I-SceI-generated DSBs, the loss of dinB decreased the stress-induced mutagenesis rate 10-fold compared with that of the pol⁺ control (Fig. 2A and D), whereas the inactivation of rpoS depressed mutagenesis 25-fold compared to the wild-type level, to an additional statistically significant 2.5-fold decrease of the rate in the dinB strain (Fig. 2B and D). Also, the dinB rpoS double mutant shows a significantly lower mutation rate, 17-fold lower than the level seen in dinB single-mutant cells (Fig. 2D). These data show that RpoS is required not only for the DinB-dependent component of stress-induced mutagenesis but also for the mutagenesis remaining in dinB cells. The results imply that the mechanism of the RpoS upregulation of mutagenesis is more than the simple transcriptional upregulation of dinB.

FIG. 2.

RpoS is required for Pol IV-dependent and Pol IV-independent I-SceI-enhanced double-strand-break repair-associated stress-induced mutagenesis. (A) I-SceI-promoted stress-induced Lac reversion requires DinB and RpoS (as previously described [42]); shown is a representative example. Strains were the following: DSB (carrying both the chromosomal inducible I-SceI gene and an I-SceI cut site near lac; ○; SMR6308), DSB dinB (♦; SMR6755), DSB rpoS (▵; SMR11437), and DSB dinB rpoS (⋄; SMR11439). (B) Data from panel A shown with an expanded y axis. (C) Relative viable cell counts of the Lac⁻ cells on the M9 lactose medium normalized to day 1 counts. (D) Quantification of fold differences between isogenic strains from six experiments. Mutation rates and fold differences are the same as those described for Fig. 1B. Mutagenesis is significantly lower in the rpoS than the dinB strain (P = 0.0004 by two-tailed Student's t test) and significantly lower in the dinB rpoS than the dinB strain (P = 0.0003 by two-tailed Student's t test).

The data shown in Fig. 2D show that the vast majority, 96%, of the I-SceI-provoked mutagenesis is RpoS dependent. There is a small (4%) fraction that is RpoS independent, and the comparison of the rpoS mutant to the dinB rpoS double mutant shows that DinB contributes to 4/5 of that small fraction of RpoS-independent mutations. Thus, although the vast majority of DinB-dependent mutagenesis also requires RpoS, a very small component of DinB-dependent mutagenesis does not.

Independence of the Pol II-dependent and Pol IV-dependent pathways during I-SceI-stimulated mutagenesis.

We tested further the apparent independence of the Pol IV- and Pol II-dependent DSB repair-associated stress-induced mutagenesis (Fig. 1) in assays with I-SceI cleavage near lac, a more sensitive assay than that using spontaneous events (Fig. 3 and Table 2). As seen in non-I-SceI-inducing cells (14, 21) (Fig. 1), the loss of polB increased I-SceI-promoted reversion above pol⁺ levels (Fig. 3A) by 6- ± 2-fold (Table 2). The polB dinB double mutant showed a significant decrease to 0.15-fold of the mutagenesis seen in dinB single-mutant cells (Fig. 3B and Table 2), indicating that, in this more-sensitive genetic background, Pol II-dependent mutagenesis is an independent pathway that does not require Pol IV.

FIG. 3.

RpoS promotes both the Pol IV- and Pol II-dependent pathways of I-SceI-promoted DSB repair-associated stress-induced mutagenesis. (A and B) In I-SceI-stimulated stress-induced mutagenesis, as in spontaneous events (14, 21), the loss of polB increases dinB-dependent mutagenesis. I-SceI-inducing DSB strains were the following: pol⁺ (○; SMR6308), polB (▿; SMR12367), dinB (♦; SMR6755), and dinB ΔpolB (⋄; SMR12369). (C) Stress-induced I-SceI-promoted Lac reversion rates are similar between dinB polB rpoS (+; SMR12374) and dinB polB (⋄; SMR12369) cells. (D) Reversion rates of dinB polB rpoS (+; SMR12374) and dinB rpoS (•; SMR11439) cells are similar, and recA (□; SMR6509) cells had a stronger mutation defect than either of those strains, indicating that the assay is capable of detecting greater defects in mutagenesis than occur in the dinB polB rpoS triple mutant. Representative examples are shown in each panel. The quantification of multiple experiments is shown in Table 2.

TABLE 2.

Both Pol IV- and Pol II-dependent I-SceI DSB-provoked mutagenesis pathways require RpoS^a

Relevant genotype of DSB-producing strains		Fold difference from control in mutation rateb (mutants per 108 cells per day; means ± SEM)	Pc
Mutant	Isogenic control
polB	pol+	6.0 ± 1.6	0.002
dinB	pol+	0.14 ± 0.01	0.05
dinB polB	dinB	0.15 ± 0.04	0.05
polB rpoS	rpoS	1.2 ± 0.4	0.7
dinB polB rpoS	dinB polB	0.76 ± 0.06	0.4
dinB polB rpoS	dinB rpoS	1.6 ± 0.4	0.3
dinB polB rpoS	polB rpoS	0.22 ± 0.06	0.06
dinB	recA	68 ± 15	0.04
rpoS	recA	36 ± 13	0.04
dinB polB	recA	9.4 ± 1.8	0.007
dinB rpoS	recA	4.9 ± 1.3	0.02
polB rpoS	recA	39.6 ± 16.7	0.05
dinB polB rpoS	recA	7.0 ± 1.1	0.03

^aStrains used were the following: polB, SMR12367; pol⁺, SMR6308; dinB, SMR6755; polB dinB, SMR12369; rpoS, SMR11437; polB dinB rpoS, SMR12374; recA, SMR6509; dinB10 rpoS, SMR11439; and polB rpoS, SMR12371.

^bMutation rates and fold differences per Fig. 1B from three independent experiments.

^cTwo-tailed Student's t test for strains compared in parallel in the same experiments.

RpoS is required for both Pol IV- and Pol II-dependent I-SceI-promoted stress-induced mutagenesis.

To examine whether the RpoS requirement for mutagenesis is via both the Pol IV- and Pol II-dependent pathways, we compared mutagenesis in dinB polB rpoS triple mutants with that of each of the isogenic double mutant strains to test for additivity (separate pathways) or epistatic (same pathway) effects (1). First, the dinB polB rpoS triple mutant shows a mutation rate that is not significantly different from that of the highly mutagenesis-defective dinB polB double mutant (Fig. 3C and Table 2). This indicates that all or nearly all of the effect of RpoS in mutagenesis is via the DinB- and Pol II-dependent pathways, with little effect in any other pathway that might be operative. Second, it is not simply that the mutation assay is not sensitive enough to register any greater loss of mutagenesis, because the dinB polB rpoS triple and dinB polB double mutants showed significantly higher mutation rates than that seen with severely defective recA cells (Fig. 3D and Table 2), showing that these experiments were performed within the detection limits of this assay; lower-level mutagenesis phenotypes could have been detected had they occurred. Interestingly, this implies that some DSB repair-associated mutagenesis is Pol II and Pol IV independent. The nature of these mutations has not yet been explored. Third, the dinB polB rpoS triple mutant also is indistinguishable in mutation rate from dinB rpoS cells (Fig. 3D, Table 2), indicating that essentially all polB-dependent mutagenesis already was eliminated from the dinB rpoS mutant; that is, there is no RpoS-independent component to Pol II-dependent mutagenesis; both Pol IV and Pol II pathways already were inactivated in the dinB rpoS strain. These data show that RpoS promotes mutagenesis by both the DinB- and Pol II-dependent pathways. How RpoS might promote Pol II-dependent mutagenesis is not known, but a model is presented below. Fourth, whereas in RpoS⁺ cells the loss of Pol II greatly increases mutagenesis (Fig. 3A, Table 2), there is no increase in mutagenesis caused by polB mutation in rpoS cells (Table 2), supporting previous conclusions that the DinB-dependent mutagenesis pathway that increases in ΔpolB RpoS⁺ cells requires RpoS and thus is mostly unavailable in rpoS mutants. Although there appears to be a slightly greater mutagenesis defect in dinB polB rpoS cells than in polB rpoS double mutants (Table 2), on these data this effect is not quite significant (P = 0.06). If real, this would imply that some small component of the DinB-dependent mutagenesis is independent of (and is additive with) the RpoS-dependent mutagenesis. Overall, these data show that both the Pol IV- and Pol II-dependent stress-induced mutagenesis pathways require RpoS, and that RpoS promotes little or no point mutagenesis outside the Pol IV- and Pol II-dependent pathways.

DISCUSSION

Roles of DNA polymerases in point mutagenesis and gene amplification.

DSB repair-associated mutagenesis includes stress-induced point mutagenesis (15, 48) and gene amplification (42, 50), a process of recombination between microhomologies that produces genome rearrangements (26). Whereas point mutagenesis was shown to be 85% dependent on Pol IV (37), here we show that the remainder requires Pol II whether the mutagenesis is promoted by nicks made by TraI in F′, which are thought to become DSBs upon replication (32, 42, 46, 47) (Fig. 1), or by I-SceI-induced DSBs (Fig. 3, Table 2). A requirement for Pol II but not Pol IV was reported previously for RpoS-dependent stress-induced mutagenesis in an E. coli natural isolate (4). Amplification, by contrast, requires DNA Pol I (27, 50), which is not required for point mutagenesis (27, 50), showing that these two truly are independent pathways (i.e., amplification is not the precursor of point mutagenesis, as had been suggested previously [49]). Amplification also is independent of Pol IV (37) and Pol II (50).

Although Pol II produces a fraction of the point mutants, the loss of Pol II causes a Pol IV-dependent increase in mutagenesis (Fig. 1 and 3A) (14, 21). This could be explained by models in which both Pol II and IV compete for a spot at the DSB repair replisome under stress (14, 21, 25), and that when higher-fidelity Pol II wins, fewer mutations result because of the exclusion of lower-fidelity Pol IV. This interpretation has been supported by recent results showing that the loss of Pol II does not indirectly increase Pol IV-dependent mutagenesis via the upregulation of the SOS response, which upregulates Pol IV (25). Moreover, in that study we provided evidence that high-fidelity DNA Pol I, II, and III all compete with Pol IV for synthesis during stress-induced mutagenesis (25), with mutation rates ultimately hinging on whether a high-fidelity enzyme or the error-prone Pol IV wins. Altering the outcome of this competition might be an important regulatory step in mutagenesis.

Could RpoS promote genome instability by directly or indirectly inhibiting high-fidelity Pol III?

Whereas previously it was possible to imagine that RpoS promoted point mutagenesis solely by its 2-fold transcriptional upregulation of dinB (33), our data showing a Pol II-dependent and RpoS-dependent component of point mutagenesis (Fig. 1 to 3, Table 2) imply a larger or different role for RpoS. We suggest here that RpoS licenses the use of all DNA polymerases, except Pol III, in DSB repair replication events by inhibiting the ability of Pol III to compete at the replisome. It could do this by the transcriptional downregulation of the dnaE gene encoding the Pol III catalytic subunit (reported previously [29]) or it might, for example, upregulate a factor that inhibits Pol III in DSB repair-associated replication directly or indirectly (both ideas are illustrated in Fig. 4). Supporting the basic model that RpoS promotes stress-induced mutagenesis by tilting a DNA polymerase competition away from Pol III (Fig. 4), we note that stress-induced mutagenesis under the carbon starvation conditions used here encompasses various genome instability outcomes dependent on all of the five DNA polymerases except Pol III. First, RpoS-dependent stress-induced frameshift mutagenesis in lac (37) and the F′-borne (42) and chromosomal tet (6) or ampD (41) gene requires Pol IV, and at least at lac, Pol II also can contribute to a minority of the frameshift mutations (Fig. 1 and 3, Table 2). Second, when RpoS-dependent loss-of-function mutations in the chromosomal ampD gene are selected in these carbon-starved cells, both base substitution and frameshift mutations abound, and the mutagenesis required Pol IV and partially required Pol V (41). The data implied that Pol IV was required for substitution and frameshift, whereas Pol V promoted only the substitution mutations (41). Finally, RpoS-dependent stress-induced gene amplifications (34) and other genome rearrangements caused by replication and recombination between DNA microhomologies (26) require Pol I (27, 50). An economical hypothesis is that all of these genome instability events become possible and dependent upon RpoS, because RpoS somehow decreases the likelihood of Pol III winning the competition at the DSB-repair replisome (Fig. 4). The many possible mechanisms by which RpoS might achieve this will be interesting to explore in the future.

FIG. 4.

Model for RpoS promotion of various mutagenic events that require DNA Pol IV, I, II, and V by inhibiting the ability of Pol III, the major replicative DNA polymerase, to compete at a DSB repair replisome. Parallel lines, strands of DNA; dashed lines, newly synthesized DNA; DSE, double-strand end; Xs, DNA polymerase errors that become mutations. In this model, the repair of DSEs by homologous recombination is high fidelity and nonmutagenic in unstressed log-phase cells (left half of the diagram), because it is proposed to rely mainly on DNA Pol III, which we propose outcompetes other more-mutagenic DNA polymerases for a spot at the DSE repair replisome (diagram on the left; Pol III is large and in boldface because it dominates). DSE repair was shown to be of high fidelity and nonmutagenic in unstressed cells (42) and was shown to use Pol III previously (39). Under stress (the right half of the diagram) and dependently on the RpoS and SOS responses, repair was shown to switch to a mutagenic mode using Pol IV (42). In this model, we propose that RpoS causes this switch by directly or indirectly inhibiting Pol III, so that other more-mutagenic DNA polymerases can then outcompete Pol III at the DSE repair replisome. SOS is induced in most or all cells experiencing a DSB (40) and transcriptionally upregulates Pol IV, II, and V (arrow; the larger typeface indicates Pol IV, II, and V upregulation), but this is insufficient for stress-induced mutation unless RpoS also is expressed (42). Pol IV also is transcriptionally upregulated ∼2-fold by RpoS (33) (arrow from RpoS to Pol IV). Pol IV is in boldface on the right because it is the largest contributor to DSB repair-associated stress-induced mutagenesis. We suggest that RpoS inhibits the ability of DNA Pol III to compete with the remaining DNA polymerases (bar from RpoS to Pol III) and thereby causes Pol IV- and Pol II-dependent frameshift reversions associated with DSB repair (Fig. 1 to 3 and Table 2); Pol IV- and Pol V-dependent stress-induced base substitution mutations (41); Pol I-dependent deletions, duplications, and inversions (26); and gene amplifications (27, 50) under stress.