Essential Role for an Isoform of Escherichia coli Translation Initiation Factor IF2 in Repair of Two-Ended DNA Double-Strand Breaks

ABSTRACT

In Escherichia coli, three isoforms of the essential translation initiation factor IF2 (IF2-1, IF2-2, and IF2-3) are generated from separate in-frame initiation codons in infB. The isoforms have earlier been suggested to additionally participate in DNA damage repair and replication restart. It is also known that the proteins RecA and RecBCD are needed for repair of DNA double-strand breaks (DSBs) in E. coli. Here, we show that strains lacking IF2-1 are profoundly sensitive to two-ended DSBs in DNA generated by radiomimetic agents phleomycin or bleomycin, or by endonuclease I-SceI. However, these strains remained tolerant to other DSB-generating genotoxic agents or perturbations to which recA and recBC mutants remained sensitive, such as to mitomycin C, type-2 DNA topoisomerase inhibitors, or DSB caused by palindrome cleavage behind a replication fork. Data from genome-wide copy number analyses following I-SceI cleavage at a single chromosomal locus suggested that, in a strain lacking IF2-1, the magnitude of recombination-dependent replication through replication restart mechanisms is largely preserved but the extent of DNA resection around the DSB site is reduced. We propose that in the absence of IF2-1 it is the synapsis of a RecA nucleoprotein filament to its homologous target that is weakened, which in turn leads to a specific failure in assembly of Ter-to-oriC directed replisomes needed for consummation of two-ended DSB repair.

IMPORTANCE Double-strand breaks (DSBs) in DNA are major threats to genome integrity. In Escherichia coli, DSBs are repaired by RecA- and RecBCD-mediated homologous recombination (HR). This study demonstrates a critical role for an isoform (IF2-1) of the translation initiation factor IF2 in the repair of two-ended DSBs in E. coli (that can be generated by ionizing radiation, certain DNA-damaging chemicals, or endonuclease action). It is proposed that IF2-1 acts to facilitate the function of RecA in the synapsis between a pair of DNA molecules during HR.

INTRODUCTION

Among the most severe forms of damage to the genetic material in all life forms is the double-strand break (DSB) in DNA (1–11). In bacteria such as Escherichia coli, repair of a DSB is achieved by homologous recombination (HR), which entails a step of synapsis between the broken DNA ends and an intact DNA duplex (1–3, 8, 11, 12). RecA is the central protein of HR that mediates synapsis, and its action in DSB repair is preceded by a set of (presynaptic) reactions performed by the RecBCD complex.

RecBCD is a dual function exonuclease-helicase that acts on broken DNA ends to generate 3′-overhangs of single-stranded (ss) DNA, upon which the RecA nucleoprotein filaments are then assembled. Formation of these 3′-overhangs is facilitated by the presence of cis sites in DNA designated Chi (in E. coli, the Chi-site is an asymmetric 8-bp sequence), since the exonuclease (but not helicase) activity of RecBCD is attenuated upon its encounter with a Chi site (1, 2, 4, 11, 13).

The RecA-mediated synaptic step in DSB repair results in formation of D-loops and Holliday junctions that serve as substrates for action by the RuvABC proteins (1, 2, 14). This is followed by a process of replication restart, by which a replisome is assembled on the restored replication fork structure. The proteins PriABC, DnaT, and Rep are proposed to act through multiple alternative and redundant pathways to achieve replication restart (14–19).

As may be expected from the description above, deletion mutations in genes recA, recB, recC, ruv or priA confer marked sensitivity to DSBs in DNA (1, 2, 4). On account of redundancy in the replication restart pathways, mutants individually deficient for PriB, PriC, or the helicase activity of PriA (through a priA300 mutation) are tolerant to DSBs, but the double mutants priB-priC or priB-priA300 are sensitive (14–19).

DSBs in E. coli may be either one-ended or two-ended (2, 14). The former occur at sites of replication fork collapse or disintegration, in which one sister chromatid arm is dissociated from the remaining circular chromosome (1, 2, 14, 20). RecA- and RecBCD-mediated repair of a one-ended DSB results in a restart replisome that is directed to progress toward the chromosomal terminus region (oriC-to-Ter direction).

Two-ended DSBs occur when there is scission of both strands of a DNA molecule which is unrelated to replication, as occurs with ionizing radiation, radiomimetic agents (21, 22) such as phleomycin (Phleo) or bleomycin (Bleo), or cleavage by endonucleases such as restriction enzymes or I-SceI. The general concept has been that repair of two-ended DSBs in E. coli “can [simply] be viewed as a combination of two initially independent” events of one-ended DSB repair (1), by which a pair of restart replisomes are established that converge toward one another in oriC-to-Ter and Ter-to-oriC directions, respectively (“ends-in” replication) (2, 8, 14). In eukaryotes, distinctive mechanisms apparently exist for two-ended DSB repair (23–27).

We now report that isoforms of the translation initiation factor IF2 can specifically modulate two-ended DSB repair in E. coli. These findings are complementary to those reported by us in the accompanying paper (28), that the IF2 isoforms can influence HR through RecA and the RecFORQ-mediated presynaptic pathway. IF2 is an essential protein for translation initiation (29), which exists in three isoforms (IF2-1, IF2-2, and IF2-3) that are generated from three in-frame initiation codons 1, 158, and 165 of the infB ORF (30–33). The isoforms IF2-1 and IF2-2,3 (the latter term refers to a mixture of IF2-2 and IF2-3, which are only marginally different from one another) exist in approximately equimolar amounts within the cells (33, 34); and strains expressing just one or the other of them (at their respective physiological levels) are equally proficient for growth (33).

Nakai and coworkers had previously reported that DNA transactions, including damage repair processes are impacted upon differential expression of IF2 isoforms in the cells, and they had suggested that this is a consequence of the isoforms’ effects on different pathways of replication restart (35–37). Based on our present studies involving genome-wide copy number analyses following a two-ended DSB at a single chromosomal locus, we propose an alternative model wherein absence of isoform IF2-1 abrogates the Ter-to-oriC replication component that is needed for two-ended DSB repair.

RESULTS

Loss of IF2-1 confers profound sensitivity to two-ended DSBs in DNA.

Given our findings in the accompanying paper (28) that IF2 isoforms can modulate HR functions, as well as on earlier reports from the Nakai group (36, 37) on differential sensitivity to DNA damage of strains deficient for particular IF2 isoforms, we investigated in more detail the relationship between expression of the IF2 isoforms and DNA damage tolerance.

The following nomenclature is employed below for different alleles and constructs related to IF2 isoforms, as was also adopted in the accompanying paper (28). infB⁺ and ΔinfB refer to the wild-type and deletion alleles, respectively, at the native chromosomal location. For the set of three ectopic infB chromosomal constructs described by Nakai and coworkers (36, 37) (in each of which IF2 expression remains under the control of the natural cis regulatory elements for infB), ΔIF2-2,3 refers to that encoding only IF2-1, but not IF2-2 or IF2-3; ΔIF2-1 to that encoding both IF2-2 and IF2-3, but not IF2-1; and IF2(wt) to that encoding all three isoforms. For plasmid-encoded infB constructs under the control of the isopropyl-β-d-thiogalactoside (IPTG)-induced P_trc promoter, P_trc::IF2-2,3 refers to that expressing both IF2-2 and IF2-3, but not IF2-1; and P_trc::IF2-3 to that expressing IF2-3 alone.

Our results indicate that the ΔIF2-1 strain (deficient for IF2-1, and expressing just the IF2-2,3 isoforms) is profoundly sensitive to perturbations that unambiguously generate two-ended DSBs on the chromosome (Fig. 1). These perturbations include exposure to radiomimetic agents Phleo or Bleo, and cleavage by endonuclease I-SceI. Thus, at concentrations of Phleo or Bleo that provoke chromosomal two-ended DSBs (and so render a recA mutant inviable), the strain lacking IF2-1 was killed to at least the same extent as recA itself; on the other hand, the ΔIF2-2,3, priA300 or ΔpriB strains were as tolerant to these agents as was the IF2(wt) strain (Fig. 1A). Phleo sensitivity of the ΔIF2-1 strain could be complemented by plasmid-borne infB⁺ (Fig. 1B). It is the RecBCD presynaptic pathway that is involved in RecA-mediated recombinational repair of two-ended DSBs in DNA (1, 2, 4–6), which was confirmed also in our experiments by demonstrating that a recB but not recO mutant is sensitive to Phleo and Bleo (Fig. 1C).

FIG 1.

Demonstration by dilution-spotting experiments of sensitivity to two-ended DSBs in DNA in the absence of IF2-1. In the different panels, strains designated IF2(wt), ΔIF2-1, or ΔIF2-2,3 were also ΔinfB at the native chromosomal locus. Growth medium was LB with supplements as indicated on top; for the genotoxic agents, numbers in parentheses refer to concentrations in μg/mL, while Glu and Ara were each at (0.2%). Relevant strain genotypes/features are shown at left (strains of panel D carried the construct for Ara-inducible expression of I-SceI and the cognate cut site). Strains employed for different rows were (from top, all strain numbers are prefixed with GJ): panel A – 19844, 19193, 19194, 15494, 15495, and 19812; panel B - 15494 and 15494/pHYD5212; panel C - 15410, 19816, and 19817; and panel D – 15837, 19804, 19805, 19806, and 19818.

Likewise, the ΔIF2-1 strain (but not IF2[wt] or ΔIF2-2,3), as well as the recA and recB derivatives of IF2(wt), were markedly sensitive to two-ended DSB at an I-SceI site in the lacZ locus, that was generated by L-arabinose (Ara)-induced expression of endonuclease I-SceI (Fig. 1D; and Fig. S1 ii). These results therefore establish that loss of IF2-1 is associated with compromise of two-ended DSB repair mediated by RecA and the RecBCD presynaptic pathway. On the other hand, none of the pri mutations tested (priA300, ΔpriB or ΔpriC) had any effect on sensitivity or tolerance to I-SceI cleavage of the IF2(wt), ΔIF2-1, or ΔIF2-2,3 strains (Fig. S1, compare i with iii-v).

Interestingly, with lower concentrations of Phleo or Bleo wherein recA viability was only marginally affected, the strain without IF2-1 continued to exhibit marked sensitivity (Fig. S2, compare rows 4 and 7 of three panels at left; supported also by data in Bleo subpanel of Fig. 1A). This would suggest that at the low doses, DNA damage other than DSBs continues to occur (perhaps ss-DNA gaps), which can be repaired by RecA-independent mechanism(s) (such as by DNA polymerase I followed by DNA ligase [1]) but only so if IF2-1 is present.

Confirmation by flow cytometry of ΔIF2-1's sensitivity to two-ended DSBs.

Sensitivity of the ΔIF2-1 derivatives to Phleo and to I-SceI cleavage was demonstrated also by flow cytometry following propidium iodide staining for dead cells in cultures (Fig. 2, middle and bottom rows), wherein these strains exhibited much greater cell death (37% and 63%, respectively, for the two treatments) than did the isogenic IF2(wt) or ΔIF2-2,3 strains (each showing cell death of 5% and 3%, respectively, with the two treatments). Even in ordinarily grown cultures without any DNA damaging agent or treatment, around 15% of cells of the ΔIF2-1 strain were scored as dead, which is much higher than those for IF2(wt) or ΔIF2-2,3 derivatives (1% each) (Fig. 2, top row).

FIG 2.

Demonstration by flow cytometry of sensitivity to two-ended DSBs in DNA in the absence of IF2-1. Flow cytometry was performed following propidium iodide staining of cells in LB-grown cultures of strains whose relevant genotypes/features are indicated on top and perturbations, if any, at left; strains designated IF2(wt), ΔIF2-2,3, or ΔIF2-1 were also ΔinfB. Phleo supplementation was at 2 μg/mL, and I-SceI refers to Ara-supplementation in cultures of derivatives carrying P_ara::I-SceI and the cognate cut site in lacZ. In each subpanel, the percentage of total cells whose intensity of propidium iodide staining exceeded the threshold that was taken to demarcate dead cells (10³ arbitrary units, marked by vertical line), is indicated at top right. Strains employed were (all strain numbers are prefixed with GJ): I. derivatives without P_ara::I-SceI – IF2(wt), 19193; ΔIF2-2,3, 19194; ΔIF2-1, 15494; recA, 19844; and II. derivatives with P_ara::I-SceI – IF2(wt), 19804; ΔIF2-2,3, 19805; ΔIF2-1, 19806; and recA, 19807.

By comparison, in a recA mutant, the ratio of dead cells in cultures without any overt DNA damaging treatment was 11% (Fig. 2, top row); the proportions after Phleo exposure or I-SceI cleavage (11% and 25%, respectively) were greater than those in the rec⁺ IF2(wt) strain, and yet lower than may have been expected. Perhaps this could be on account of occurrence of anucleate cells or of cell lysis in these cultures under our experimental conditions (38, 39).

Loss of IF2-1 does not confer sensitivity to several other DSB-generating perturbations.

Notwithstanding the sensitivity of the ΔIF2-1 mutant to two-ended DSB damage, the strain was tolerant to other DSB-generating perturbations to which recA and recBC mutants are known to be sensitive (Fig. 3). Thus, type-2 DNA topoisomerase inhibitors nalidixic acid and ciprofloxacin were tolerated to equivalent extents by the strains with differential expression of the IF2 isoforms but conferred 10³-fold greater lethality upon loss of RecA or of PriB (Fig. 3A; see also Fig. 1C for sensitivity to ciprofloxacin of a recB but not recO mutant). The reversal in rank order of sensitivity between ΔpriB and ΔIF2-1 strains, to type-2 DNA topoisomerase inhibitors on the one hand and (high-dose) radiomimetic agents on the other, suggests that repair mechanisms following exposure to these two agent categories are distinct and different, although both are RecA- and RecBCD-mediated (1, 40).

FIG 3.

Tolerance of strains expressing different IF2 isoforms to other genotoxic agents and perturbations. Dilution-spotting assays were performed on LB with supplements as indicated on top; numbers in parentheses refer to concentrations in μg/mL. Cipro, ciprofloxacin; Nal, nalidixic acid; and MMC, mitomycin C. Relevant strain genotypes are shown at left. Strains whose designations include IF2(wt), ΔIF2-1, or ΔIF2-2,3 were also ΔinfB. In the strains of panel B, SbcCD expression that is induced in cells on Ara-supplemented medium leads to cleavage of the sister chromatid associated with the lagging-strand template at an engineered palindrome sequence in the lacZ gene (41). Strains used were (from top, unless otherwise indicated all strain numbers mentioned are prefixed with GJ): in panel A – 19844, 19193, 19194, 15494, 15495, and 19812; and in panel B – DL2006, 19811, 19808, 19809, and 19810.

Again, unlike a recA derivative, the ΔIF2-1 strain was not sensitive to mitomycin C (Fig. 3A), which cross-links DNA strands and generates DSBs following DNA replication. Leach and coworkers have described a model of site-specific DSB generation on one of the pair of sister chromatids immediately behind a replication fork, that occurs by SbcCD-mediated cleavage of a palindromic sequence (41, 42); to this perturbation as well, a strain deficient for IF2-1 remained tolerant whereas the recA mutant was sensitive (Fig. 3B).

Overexpression of isoforms IF2-2,3 is also correlated with sensitivity to two-ended DSBs.

In the accompanying paper (28), we have provided evidence that HR functions in E. coli are affected by an imbalance between the different IF2 isoforms, with the phenotypes of IF2-1 deficiency being recapitulated by overexpression in an infB⁺ strain of IF2-2,3 or IF2-3. With respect to DSB damage as well, our results indicate that, just as with loss of IF2-1, IPTG-induced overexpression of IF2-2,3 or of IF2-3 in the IF2(wt) strain confers sensitivity to Phleo but not to nalidixic acid (Fig. 4).

FIG 4.

Overexpression of isoforms IF2-2,3 or IF2-3 confers sensitivity to two-ended DSBs in DNA. Dilution-spotting assays were performed on LB with Amp and IPTG, along with the additional supplements as indicated on top; numbers in parentheses refer to concentrations in μg/mL. Nal, nalidixic acid. Relevant strain genotypes are shown at left (all were ΔinfB). Strains used for different rows were derivatives of GJ19193 with following plasmids (from top): vector pTrc99A, pHYD5207, and pHYD5208.

Role of IF2 isoforms in two-ended DSB repair is independent of GreA/DksA.

GreA and DksA are factors modulating transcription elongation in E. coli (43), but they also participate in DNA repair. Loss of GreA is associated with increased tolerance to DSBs in DNA (44–46), which was confirmed for Phleo in this study (Fig. S2 for 3 μg/mL). DksA is needed for DSB repair (47), and it is an apparent antagonist of GreA with respect both to DSB repair (44–46) and to other phenomena (48, 49); in our study, its loss conferred a modest sensitivity to Phleo (Fig. S2 for 0.25 and 1 μg/mL).

We examined the relationship, if any, between the roles in two-ended DSB repair of GreA and DksA on the one hand, and that of IF2-1 on the other. Our results show that sensitivity to Phleo of a strain lacking IF2-1 is reversed, but only partially so, upon loss of GreA and that it is somewhat exacerbated upon loss of DksA (Fig. S2 for 0.25 μg/mL); thus, the opposing effects of the two losses (IF2-1 and GreA) appear to be algebraically additive. These results suggest that the mechanism by which IF2 isoforms modulate DSB repair is different from that by GreA/DksA.

Comparison of DNA copy number changes following site-specific two-ended DSB in IF2(wt) and ΔIF2-1 strains.

The Herman lab (46) has previously shown by whole-genome sequencing (WGS) that following induction of synthesis of I-SceI to generate a two-ended DSB at a single genomic location in the lac operon, an equilibrium between DNA resection on the one hand, and resynthesis by recombination-dependent replication (RDR) through replication restart mechanisms on the other, is reached by 30 min. The result is a Chi-site modulated, asymmetric V-shaped dip in DNA copy number extending from ∼100 kb ori-proximal to ∼200 kb ori-distal of the DSB site. The recA mutant, on the other hand, exhibits extensive DNA copy number reduction (consequent to “reckless” RecBCD-mediated DNA degradation without any RDR), without an equilibrium being attained.

We performed similar WGS experiments to determine genome-wide DNA copy numbers in LB-grown cultures for strains carrying an I-SceI site at the lacZ locus, and in which the cognate enzyme (under P_ara control) was expressed in early exponential phase for one hr by addition of Ara as inducer; D-glucose (Glu) was used instead of Ara in the control uninduced cultures. The strains were ΔinfB at the native locus and carried the ectopic Nakai infB constructs (IF2[wt], ΔIF2-1, or ΔIF2-2,3); a recA derivative of the IF2(wt) strain was also used.

Normalized copy number distributions from the cultures were determined as described in “Materials and Methods,” and all of them exhibited a bidirectional oriC-to-Ter gradient that is expected for cells in asynchronous exponential growth in rich medium (Fig. 5). Superimposed upon this gradient distribution were several distinct features of interest that are discussed below.

FIG 5.

Chromosomal DNA copy number analysis by WGS in strains expressing different IF2 isoforms, following two-ended DSB generation at lacZ. DNA copy numbers (after normalization) are plotted as semilog graphs for overlapping 10-kb intervals across the genome for derivatives each carrying P_ara::I-SceI and the cognate cut site in lacZ, after supplementation of cultures grown in LB with 0.2% Glu (control) or Ara for 1 h (top and bottom rows, respectively). Relevant genotypes are indicated at top of each pair of panels; all strains were also ΔinfB. In these Cartesian graphical representations, the circular 4642-kb long chromosome is shown linearized at oriC, with genome coordinates on the abscissa corresponding to the MG1655 reference sequence (wherein oriC is at 3926 kb). Ordinate scales (log₂) shown at left on top and bottom rows are common for, respectively, panels i-iv and v-vii. The positions of lacZ, TerA and TerC/B are marked. Strains used were (all strain numbers are prefixed with GJ): IF2(wt), 19804; ΔIF2-2,3, 19805; ΔIF2-1, 19806; and IF2(wt) recA, 19818.

The maximum extent of copy number reduction around lacZ in each of the Ara-grown cultures is presented as a log₂ dip value, in Table 1 (for derivatives of IF2[wt] and ΔIF2-1). Following I-SceI induction with Ara (1-h exposure), the IF2(wt) strain exhibited the asymmetric V-shaped Ter-biased reduction around lacZ (Fig. 5 v, and Fig. S3 i; log₂ dip = 0.8), as had previously been reported by the Herman group (46). The ΔIF2-2,3 strain behaved similarly to IF2(wt) for copy number changes around lacZ (Fig. 5 vi and Fig. S3 ii; log₂ dip = 0.6). The recA derivative of IF2(wt) exhibited a pattern very similar to that described by the Herman group (46), that is, a very extensive degradation on either side of lacZ (Fig. 5 viii; log₂ dip = 3.2). The recA mutant also showed a small dip in lacZ region read counts in the Glu-grown culture (Fig. 5 iv and Fig. S3, compare iv and v), suggestive of I-SceI cleavage in a minor proportion of cells even under uninduced conditions which is likely efficiently repaired in the IF2(wt) strain but is lethal in recA.

TABLE 1.

Copy number reduction (log₂) at lacZ following I-SceI cleavage^a

Mutation	Strain background		Difference (IF2[wt]) − (ΔIF2-1)	Difference between mutation and no mutation in strain background
	IF2(wt)	ΔIF2-1		IF2(wt)	ΔIF2-1
Nil	0.8	0.1	0.7	NAb	NA
recA	3.2	2.6	0.6	2.4	2.5
recB	2.0	1.3	0.7	1.2	1.2
priA300	1.5	0.7	0.8	0.7	0.6
priB	2.0	0.9	1.1	1.2	0.8
priC	0.8	0.1	0.7	0	0

^aCopy number reductions at lacZ in Ara-grown cultures have been determined (as log₂ values) from the plots depicted in Fig. 5 and 6, with the average copy number value in the absence of I-SceI cleavage estimated (from plots for Glu-grown cultures in Fig. 5) as 0.1.

^bNA, not applicable.

On the other hand, the ΔIF2-1 strain exhibited only a very minimal reduction in read counts at the lacZ region (Fig. 5 vii and Fig. S3 iii; log₂ dip = 0.1). The latter finding was quite unexpected, since it was opposite to that in recA with which ΔIF2-1 shares the phenotype of pronounced sensitivity to two-ended DSBs.

The observations on copy number changes upon 1-h Ara exposure of the IF2(wt), ΔIF2-1, or ΔIF2-2,3 strains were reproducible, in that they were replicated in other independently grown cultures of these strains (Fig. S4A i-iii, respectively); the ΔIF2-1 strain cultured continuously with Ara also showed less or no dip in read counts around lacZ compared to that in the IF2(wt) strain similarly cultured (Fig. S3 and S4B, compare, respectively, subpanels vi with vii and i with ii). DNA sequence data for each of these cultures revealed no mutation in any of the candidate genes related to DNA recombination and repair.

Efficiency of cleavage at I-SceI site is uniformly high and is unaffected by the IF2 isoforms.

In theory, the absence of reduction in read counts near the I-SceI site of the ΔIF2-1 strain could be because of reduced efficiency of DSB generation in these cells following 1-h exposure to Ara. To test for this possibility, we measured viability in cultures of different I-SceI derivatives at 0 and 1 h after Ara addition; the strains used were IF2(wt), ΔIF2-1, and ΔIF2-2,3 as well as each of their recA and recB derivatives (Fig. S5).

Since DSB repair is completely blocked in the absence of RecA or RecB (1, 12), the fraction of survivors in recA and recB derivatives after 1-h Ara exposure (as measured on LB medium without Ara) can be taken as an inverse measure of efficiency of I-SceI cleavage in the cells. By this criterion, the cleavage efficiency was 0.99 or greater in both IF2(wt) and ΔIF2-2,3 strains. For the ΔIF2-1 strain, even the rec⁺ derivative exhibited >99% decrease in viability upon exposure for 1 h to Ara and this was maintained in the recA and recB mutants, once again attesting to a very high efficiency of DSB generation as well as to absence of suppressor accumulation in the cultures. We conclude that the efficiency of I-SceI cleavage in all strains is uniformly high under the conditions used for DNA copy number determinations.

Contributions of DNA resection and of RDR to the copy number changes following a site-specific two-ended DSB.

To dissect the relative contributions of DNA resection and of RDR to the copy number changes described above, we performed additional WGS experiments (following Ara-induced cleavage at the I-SceI site in lacZ) of the following single mutant derivatives of both IF2(wt) and ΔIF2-1 strains: ΔrecB, priA300, ΔpriB, or ΔpriC. A recA derivative of ΔIF2-1 was also tested, which displayed a log₂ dip in copy number of 2.6 in the vicinity of lacZ (Fig. 6A i), implying DNA degradation and thus also confirming that I-SceI cleavage does occur in derivatives lacking IF2-1.

FIG 6.

Effects of rec (A) or pri (B) mutations on DNA copy numbers in IF2(wt) or ΔIF2-1 strains following two-ended DSB generation at lacZ. Each derivative carried P_ara::I-SceI and the cognate cut site in lacZ, and cultures grown in LB were supplemented with Ara for 1 h. Representations of WGS analysis and notations used are as described in legend to Fig. 5. Ordinate scales (log₂) shown at left are common for all subpanels in that row. Strains used for the different subpanels were (all strain numbers are prefixed with GJ): (A) i, 19857; ii, 19861; and iii, 19863; and (B) i, 19864; ii, 19858; iii, 19867; iv, 19866; v, 19860; and vi, 19869.

In earlier studies from the Herman lab (46), the recB mutant had exhibited a substantial dip in copy number around lacZ; this was an unexpected finding and was attributed to the actions (in the absence of RecBCD) of RecJ and other as yet unidentified exonucleases without any RDR. In our study, the recB derivative of IF2(wt) also behaved similarly, with a log₂ dip around lacZ of 2.0 (Fig. 6A ii). The log₂ dip in the recB derivative of ΔIF2-1 strain was lower, at 1.3 (Fig. 6A iii). Thus, it is noteworthy that the values in both recA and recB derivatives of ΔIF2-1 were negatively offset from those in the corresponding IF2(wt) derivatives by about the same log₂ magnitude (0.7) as that between ΔIF2-1 and IF2(wt) themselves (Table 1).

With respect to each of the pri mutants, it was expected that the extent of additional copy number reduction if any (compared to that in the isogenic pri⁺ strain) would reflect the magnitude of RDR deficit conferred by the cognate pri mutation. The log₂ dips in copy number in the priA300, priB and priC derivatives of the IF2(wt) strain were 1.5, 2.0, and 0.8, respectively (Fig. 6B i-iii), implying a log₂ RDR deficit (obtained by subtracting the log₂ dip of 0.8 in pri⁺ IF2[wt]) of 0.7 imposed by priA300, 1.2 by priB, and nil by priC.

Just as had been noted above for the rec⁺ pri⁺ as well as the recA and recB derivatives, the log₂ dips in each of the pri mutant derivatives of the ΔIF2-1 strain were shallower than those in the isogenic IF2(wt) counterparts (Fig. 6B iv-vi, and Table 1). The values were 0.7 (priA300), 0.9 (priB), and 0.1 (priC), and hence the corresponding RDR deficits (after accounting for the log₂ dip in ΔIF2-1 of 0.1) were 0.6, 0.8, and nil, respectively.

The detailed interpretations from the WGS data near lacZ, with respect to a possible mechanism for mediation by IF2 isoforms of two-ended DSB repair, are provided in the “Discussion.”

Other features of interest from the WGS data.

In the chromosomal terminus region, there was a peak of read counts between the TerA and TerC/B boundaries that was extremely prominent for the Ara-exposed cultures of IF2(wt) and moderately so for ΔIF2-2,3 and ΔIF2-1 (Fig. 5 v-vii). As proposed earlier (50), we believe that this midterminus peak represents the algebraic sum of read counts of two major subpopulations, in which, respectively, clockwise and counterclockwise moving forks have traversed the terminus and are paused at Tus-bound TerC/B and TerA.

Another feature was the presence of sharp deep dips at several genomic positions (resembling stalactite images) even in Glu-grown cultures of several different strains. The dips were especially pronounced in the ΔIF2-1 strain, representing log₂ drops in normalized read counts of around 3 or more (Fig. 5 iii; and see maroon lines in Fig. S6A). Similar dips were observed in copy number curves for Glu-grown cultures of (in order of their prominence) recA and IF2(wt) (Fig. 5 iv and i, respectively). The positions of these dips were identical in all three cultures, to a resolution of < 2 kb; five such representative genomic locations are depicted in Fig. S6A (green and violet lines for recA and IF2[wt], respectively).

We suggest that this feature is correlated with presence in the strains of the gene encoding I-SceI, whose basal expression is perhaps associated with nickase activity (51) at specific sequences which then leads to ss-DNA gaps at these sites (since such ss-regions are not expected to be captured in Illumina WGS protocols [46]). Indeed, dips at several of the identical locations were observed upon re-analysis of the Herman lab data (46) for uninduced cultures of wild-type and recA strains carrying the I-SceI gene, with those of recA being the more prominent (Fig. S6B i-ii; and yellow and dark blue lines, respectively, in Fig. S6A). Interestingly, the dips were least distinct for a Glu-grown culture of the ΔIF2-2,3 strain (Fig. 5 ii, and light blue line in Fig. S6A); this last observation serves to exclude, as a possible explanation for these dips, sequence-specific bias in generation of read numbers during WGS.

I-SceI cleavage lethality conferred by loss of IF2-1 is suppressed by an RNA polymerase mutation.

To account for the distinctive pattern of copy number changes following two-ended DSB in the ΔIF2-1 mutant, as well as for the observation that the mutant is killed by two-ended but not one-ended DSBs, we have proposed below that D-loop assembly at the Ter-proximal end of a two-ended DSB is specifically compromised in the absence of IF2-1 (see “Discussion”). Furthermore, we have suggested that this defect may be correlated with the fact that invasion of a RecA nucleoprotein filament at the Ter-proximal DSB end is more likely to face head-on conflicts with transcription than that at the ori-proximal DSB end.

The RpoB*35 variant of RNA polymerase has been shown in several studies to destabilize the transcription elongation complex and to ameliorate the deleterious effects of transcription-replication conflicts (44, 45, 52–56). Therefore, in a genetic approach, we employed the rpoB*35 mutation to examine whether transcription might modulate two-ended DSB tolerance in the ΔIF2-1 mutant.

Our results indicate that growth sensitivity of a ΔIF2-1 strain to Ara-induced cleavage at an I-SceI site is suppressed by rpoB*35 (Fig. 7 right panel, compare rows 4 and 5). On the other hand, both rpoB⁺ and rpoB*35 derivatives of a recA mutant were equally sensitive to such cleavage (Fig. 7 right panel, compare rows 2 and 3). Furthermore, the triple mutant ΔIF2-1 recA rpoB*35 was also sensitive (Fig. 7 right panel, row 6), indicating that the efficiency of I- SceI cleavage in the ΔIF2-1 mutant is not impaired by rpoB*35.

FIG 7.

Suppression by rpoB*35 mutation of I-SceI cleavage lethality in ΔIF2-1 strain. Dilution-spotting assays, of strains carrying the construct for Ara-inducible expression of I-SceI and the cognate cut site, were performed on LB supplemented with Glu or Ara (each at 0.2%), as indicated. Relevant strain genotypes/features are shown at left. Note that the growth observed for the strains recA and ΔIF2-1 on Ara-supplemented medium is that of suppressor mutants. Strains employed for different rows were (from top, all strain numbers are prefixed with GJ): 19891, 19818, 19896, 19852, 19892, and 19897.

These data are consistent with the notion advanced in our model below that head-on transcription might impede repair at the Ter-proximal end of a two-ended DSB in strains lacking IF2-1.

DISCUSSION

In this study, we report that the ΔIF2-1 strain deficient for isoform IF2-1 is markedly sensitive to perturbations that generate two-ended DSBs in DNA (which are dependent on RecA and RecBCD for repair); a similar phenotype is elicited also upon overexpression of isoforms IF2-2,3. At the same time, the ΔIF2-1 mutant retains tolerance to other perturbations that damage DNA and whose repair is also RecA- and RecBCD-dependent (such as exposure to type-2 DNA topoisomerase inhibitors, or DSB generation on one sister chromatid behind a replication fork). The ΔIF2-1 mutant displays a distinctive pattern of DNA copy number changes following a site-specific two-ended DSB on the chromosome. These features are discussed below to formulate a new model to account for the action of IF2 isoforms in HR and DNA damage repair.

Can the findings be explained by an effect of IF2 isoforms on replication restart?

Nakai and coworkers (36, 37) have previously suggested that the IF2 isoforms differentially affect the replication restart pathways, and accordingly one possible explanation for our results from WGS experiments is that in the ΔIF2-1 mutant there is inappropriate restart and (futile) templated DNA repair synthesis at a two-ended DSB site on the chromosome. However, we believe this to be an unlikely explanation.

Thus, from among the replication restart mutants, the profound sensitivity to two-ended DSBs observed for a ΔIF2-1 mutant would be mimicked only by ΔpriA or combinations such as priB-priC or priA300-priB (14, 16). Yet, several other phenotypes of these restart-deficient mutants, including sensitivity to type-2 DNA topoisomerase inhibitors or to mitomycin C, are not observed in the IF2-1 deficient strain.

Furthermore, the Nakai group (36) has shown that, at an UV-irradiation dose which led to 80% more killing of a priB or priA300 single mutant relative to pri⁺ (thereby establishing the need for replication restart processes for damage repair under these conditions), the ΔIF2-1 strain was not UV^S. Therefore, not all categories of DNA repair requiring replication restart are compromised in the absence of IF2-1.

Second, we have reported in the accompanying paper (28) that the ΔIF2-1 constructs suppress rho-ruv and uvrD-ruv lethalities, with the epistasis results indicating that IF2-1 acts downstream of the RecFORQ presynaptic pathway and upstream of formation of D-loops or Holliday junctions. Since replication restart occurs downstream of D-loop formation, these findings also argue against a role for IF2-1 in replication restart.

Finally, as explained below, the interpretations from WGS data of different strains following I-SceI cleavage do not also support the suggestion that loss of IF2-1 interferes with replication restart pathways.

Interpretations from DNA copy number changes around two-ended DSB site in different strains.

The IF2(wt) (parent) strain suffered a log₂ dip of 0.8 following 1-h exposure to site-specific endonuclease action, which represents the equilibrium between DNA cleavage and degradation on the one hand and RDR on the other (46). The ΔIF2-1 mutant exhibited a log₂ dip of just 0.1, and the possibility was excluded that this could be because of decreased DSB generation at the I-SceI site in the mutant. Therefore, this difference from the parent must be a consequence of decreased degradation, increased RDR, or both.

We begin with a reasonable assumption that in the recA and recB derivatives, DNA synthesis makes no contribution to copy number changes following I-SceI cleavage (since D-loops cannot occur to initiate RDR). The log₂ dips in the recA or recB derivatives of the parent (IF2[wt]) strain are deeper than those in the cognate derivatives of ΔIF2-1 by the same value of around 0.7 as that seen between the rec⁺ pair itself. These results therefore suggest that following a two-ended DSB, the ΔIF2-1 strain is partially protected from DNA degradation to a log₂ extent of 0.7-fold, and furthermore, that this degree of postulated protection is itself a sufficient explanation for the observed differences between the IF2(wt) and ΔIF2-1 strains.

Next, copy number changes around the DSB site in the different pri mutants can be interpreted on the assumptions that (i) the mutants are unaffected for DNA resection by RecBCD and synapsis by RecA; (ii) the pathways for RDR are redundant; and (iii) not all pathways are abolished in any of the single mutants studied. The last of these assumptions is supported by the fact that I-SceI cleavage was not lethal in single pri mutants of the IF2(wt) strain. It may therefore be expected that at equilibrium, there would merely be a more pronounced dip in copy number than that in the parent, whose magnitude reflects the contribution to RDR of the cognate Pri function.

Our data suggest that RDR following a two-ended DSB in the wild-type strain is performed redundantly and additively by two pathways which are abolished, respectively, in the absence of PriB and of PriA helicase. Their log₂ contributions to copy number restoration are 1.2 and 0.7, respectively, whereas PriC makes no contribution to RDR in this situation.

Given that the net log₂ dip in copy number in IF2(wt) was 0.8, the gross log₂ reduction caused by DNA resection in the strain is computed (to a first approximation) to be 2.7 (since aggregate log₂ RDR from the two restart pathways is calculated as 1.9). This calculated value for DNA degradation is consistent with the observed log₂ dip in recA of 3.2 (where there is no RDR).

Finally, all of the data on copy number changes around the DSB site in the ΔIF2-1 derivatives may be explained on the postulates that, in the absence of IF2-1, DNA degradation is decreased by a log₂ factor of around 0.7 and that RDR capacity is more or less preserved. However, this RDR does not lead to repair of the DSB. Thus, in the ΔIF2-1 background, the log₂ net contributions of the PriB and PriA helicase pathways for RDR are 0.8 and 0.6, respectively, compared to the corresponding values of 1.2 and 0.7, respectively, in the IF2(wt) parent.

A model for the role of IF2 isoforms in two-ended DSB repair.

Before formulating a model, we highlight the following observations concerning the ΔIF2-1 strain: (i) it is as sensitive as recA to two-ended DSBs on the chromosome; (ii) at the same time, it is tolerant to DSBs generated by type-2 DNA topoisomerase inhibitors, mitomycin C, or palindrome cleavage on a sister chromatid behind a replication fork; and (iii) it is reasonably proficient for RDR at a two-ended DSB site, unlike the recA mutant.

We postulate that the fundamental molecular defect in the absence of IF2-1 is reduced synapsis of a RecA-bound nucleoprotein filament to a target homologous DNA duplex. One could speculate, for example, that under these conditions the reverse reaction is rendered more favorable, of dislodgement of the invading RecA-bound strand from the heteroduplex through competition with the displaced ss-DNA of the D-loop (7); if so, it remains to be determined whether this may represent a property of IF2-1 to enhance D-loop stability or of IF2-2,3 to decrease it. In this context, an earlier study (57) has demonstrated that IF2 can bind DNA, through its C-terminal domain.

We further propose that even as this handicap associated with lack of IF2-1 confers a modest deficiency in all HR reactions (both RecFOR- and RecBCD-mediated) as reported in the accompanying paper (28), it has a particularly severe effect on Ter-to-oriC directed replisome assembly and progression (which is required only during two-ended DSB repair) (Fig. 8). The assumption is that a RecA nucleoprotein filament assembled on a Ter-proximal DSB end is more likely (in comparison with its counterpart on the ori-proximal DSB end) to encounter oncoming RNA polymerase molecules when it attempts to invade into a target sister chromosome; this premise is based on the fact that a majority of heavily transcribed genes are codirectionally oriented with replication (50). In the ΔIF2-1 mutant, such encounters would serve to further majorly reduce D-loop formation. Hence, the efficiency of replisome assembly at the Ter-proximal DSB end will be greatly diminished (Fig. 8B 2); that at the ori-proximal DSB end would also be decreased, but to much smaller extent (note that the combined log₂ RDR contribution in IF2[wt] was 1.9, whereas that in ΔIF2-1 was 1.4). Thus, although RDR across the two-ended DSB site will still be accomplished in the ΔIF2-1 mutant by the oriC-to-Ter directed replisome, two-ended DSB repair remains unconsummated (Fig. 8B, 3).

FIG 8.

Model to explain failure of two-ended DSB repair in the absence of IF2-1. (A) In IF2(wt) strain, cleavage at I-SceI site in lacZ (red arrow) leads to DNA resection by RecBCD of the two ends toward oriC and Ter, respectively (interrupted lines), followed by two events of RecA-mediated strand invasion into a sister DNA duplex (subpanel 1). “Ends-in” replication is initiated by redundant restart pathways (violet arrows) that are, respectively, PriA helicase- and PriB-dependent (subpanel 2), finally resulting in successful reconstitution of two intact chromosomes (subpanel 3). (B) In ΔIF2-1 mutant, steps until initiation of RecA-mediated strand invasion are similar to those above (subpanel 1). The oriC-to-Ter directed replisome is established through the redundant restart pathways, but that from Ter-to-oriC is not assembled (subpanel 2). Copy number at lacZ is restored by the former, but two-ended DSB repair fails to be consummated (subpanel 3).

Accordingly, this model would be able to account for the observations that a ΔIF2-1 mutant is as deficient as recA for two-ended DSB repair and yet is unaffected for RDR at the DSB site. Also consistent with this proposal is our finding that lethality associated with I-SceI cleavage in the ΔIF2-1 mutant is rescued by an RNA polymerase mutation (rpoB*35) that is otherwise known to ameliorate transcription-replication conflicts (44, 45, 52–56).

Additional features in the WGS data.

To explain reduced DNA degradation at a two-ended DSB site in the ΔIF2-1 mutant, we suggest that loss of IF2-1 leads to reduced exonuclease V action (which is one component of RecBCD function [1, 6, 13]) on DSB ends. Should this postulated second effect of IF2-1 deficiency be in some way a consequence of the first (decreased strand exchange by RecA-bound DNA), it may point to existence of an interesting phenomenon of retrograde control of RecBCD nuclease function by the RecA nucleoprotein filament. A similar concept has previously been suggested for Caulobacter crescentus (whose RecBCD equivalent is AddAB) (58).

The other feature that distinguishes between IF2 isoforms is the set of sharp dips in copy number read counts that occur at specific genomic locations in strains with the I-SceI gene, even in uninduced cultures. On the assumption that the dips represent the net prevalence of ss-DNA gaps at these sites in cells of the population (that is, the balance between their generation and repair), it would appear that the magnitude of reduction in read counts (which is in the order ΔIF2-1 >  recA > IF2(wt) > ΔIF2-2,3; Fig. S5A) inversely reflects the efficiency of their repair in the different strains. Consistent with this interpretation is our finding that loss of IF2-1 confers greater sensitivity than that of RecA to low concentrations of Phleo or Bleo (Fig. S2), which may further suggest that in IF2-1's absence, RecA's nonproductive binding to ss-DNA itself interferes with successful operation of the RecA-independent repair mechanisms.

Concluding remarks.

IF2 would thus represent another example, apart from NusA (59–61), DksA (45, 47), and GreA (46), of proteins earlier characterized for other critical functions also participating in DNA repair. IF2 isoforms exist in other bacteria such as, for example, in species of the Gram-negative Salmonella, Serratia and Proteus (34) as well as in Gram-positive Bacillus subtilis (62), and their role if any in modulating HR and two-ended DSB repair may be examined in future studies.

MATERIALS AND METHODS

Growth media, bacterial strains, and plasmids.

Rich and defined growth media were, respectively, LB and 0.2% Glu-minimal A (63), and growth temperature was 37°. Concentrations used of antibiotics or Xgal were as described (28). Inducers Ara and IPTG were added at 0.2% and 0.05 mM, respectively. Genotoxic agents were added at concentrations as indicated. E. coli strains used are listed in Table S1, with the following knockout alleles sourced from the collection of Baba et al. (64): dksA greA, recA, recB, and priC; the ΔinfB knockout mutation has been described earlier (65). Alleles priA300 and ΔpriB302 were sourced from strains provided by Steve Sandler (66).

Plasmids pKD13, pKD46, and pCP20, for use in recombineering experiments and for Flp-mediated site-specific excision of FRT-flanked DNA segments, have been described by Datsenko and Wanner (67). Plasmids pHYD5207 and pHYD5208, whose construction is described in the accompanying paper (28), were used to achieve IPTG-induced expression of different IF2 isoforms from the P_trc promoter, and the parental plasmid vector pTrc99A (68) was used as control in these experiments. Plasmid pHYD5212 bearing the infB⁺ gene has also been described in the accompanying paper (28).

Copy number analysis by deep sequencing after I-SceI cleavage.

Strains employed each carried an I-SceI site in lacZ and an Ara-inducible gene construct for I-SceI enzyme (69). Following Ara-induced I-SceI cleavage, copy number determinations of the various chromosomal regions were performed by a WGS approach, essentially as described (56). After alignment of reads to the MG1655 reference sequence, the read count for each genomic region was normalized to read counts for a 600-kb region between genome coordinates 2501 and 3100 kb (which, relative to oriC, is “antipodal” to the region around lacZ on the opposite replichore and is therefore expected to be the least affected following cleavage by I-SceI at lacZ). Additional details are given in the Supplementary Text.

Other methods.

Procedures were as described for P1 transduction (70) and recombineering (67). Protocols of Sambrook and Russell (71) were followed for recombinant DNA manipulations, PCR, and transformation. Procedures for flow cytometric quantitation of dead cells by propidium iodide staining (72) are described in the Supplementary Text.

Data availability.

The data described in this work are available for full public access from the following repositories: WGS, http://www.ncbi.nlm.nih.gov/bioproject/734449 (Accession no. PRJNA734449); and flow cytometry, https://flowrepository.org/id/FR-FCM-Z442 and https://flowrepository.org/id/FR-FCM-Z5Z2.