Identification of recG genetic interactions in Escherichia coli by transposon sequencing

ABSTRACT Maintaining the integrity of the genome is of utmost importance for cell division and propagation. In Escherichia coli, the RecG protein has been implicated in processing branched recombination intermediates during DNA repair processes, but the primary cellular role(s) of RecG and the repair pathways in which it acts have been difficult to define. To gain additional insight into RecG function, we employed transposon sequencing to identify recG genetic interactions and reveal complementary or redundant functions. The strongest hits from the screen were the dam, uvrD, rnhA, radA, and rep genes. The conditional importance of these five genes in cells lacking recG was confirmed using a plasmid-based assay, revealing synthetic lethal interactions for most double deletion strains. Several of the synthetic lethal gene combinations (with uvrD, rep, and radA, but not rnhA or dam) were suppressed by deletion of recF or recO, indicating that their genetic relationships involved roles in post-replication gap repair. Additionally, loss of the RecG/SSB interaction phenocopied a recG deletion when combined with dam, uvrD, radA, or rnhA deletions but not with rep. The results reinforce the idea of RecG as a general genome guardian. RecG has at least two functions. It plays an important role in the resolution of joint molecules behind the fork, formed during post-replication gap repair. RecG is also required to suppress genome over-replication caused by unscheduled replication initiation at R-loops, at double-strand breaks caused by dam inactivation, and during replication termination. IMPORTANCE DNA damage and subsequent DNA repair processes are mutagenic in nature and an important driver of evolution in prokaryotes, including antibiotic resistance development. Genetic screening approaches, such as transposon sequencing (Tn-seq), have provided important new insights into gene function and genetic relationships. Here, we employed Tn-seq to gain insight into the function of the recG gene, which renders Escherichia coli cells moderately sensitive to a variety of DNA-damaging agents when they are absent. The reported recG genetic interactions can be used in combination with future screens to aid in a more complete reconstruction of DNA repair pathways in bacteria.

One member of these repair pathways, recG, was first identified in a screen for mutations that render E. coli recombination deficient (4).The RecG protein is a 76-kDa superfamily (SF) 2 helicase found in almost all bacterial species (5) and has been implicated in several DNA repair roles and pathways.Loss of RecG in E. coli results in modest sensitivity to DNA-damaging agents such as mitomycin C, ultraviolet (UV) light, and ionizing radiation (6,7).Cells lacking recG struggle to segregate chromosomes after DNA replication and frequently form filaments (8).
RecG binds and remodels several types of DNA substrates in vitro, but its preferred substrates are branched DNAs, such as three-strand structures and Holliday junctions (9,10).A structure of RecG from Thermotoga maritima has been solved bound to a synthetic three-way junction, revealing three domains: a "wedge" domain that is involved in DNA junction binding and unwinding and two RecA-like helicase domains responsible for ATP hydrolysis activity and motor function (11).RecG can also unwind D-loops and R-loops, sites where constitutive stable DNA replication (cSDR) can be initiated (12,13).E. coli strains lacking recG display cSDR activity, and loss of recG results in overamplification of DNA in the terminus region of the chromosome and around double-strand breaks (DSBs) (14)(15)(16)(17).These findings suggest an important role for RecG in preventing a lethal DNA replication cascade resulting from continuous reloading of the DnaB helicase by the replication restart protein PriA (15,18).This cascade is negated when PriA helicase activity is eliminated (18), and suppressors of the ΔrecG phenotype often manifest as missense mutations within PriA helicase motifs (19)(20)(21).In agreement with these observations, the ΔrecG phenotype is also suppressed by deletion of the replication restart gene priB in strains containing additional mutations in genes encoding major subunits of RNA polymerase (rpoA, rpoB, or rpoC) (22).
RecG interacts with the single-stranded DNA (ssDNA) binding protein (SSB) (23,24).A recent investigation of the interaction with purified proteins revealed that the interaction is mediated by the highly conserved SSB-Ct interaction motif, which comprises the eight C-terminal residues of SSB (25).The SSB-Ct docks onto a highly conserved surface on the RecG helicase domain 2, framed by residues Arg474, Arg614, and Arg467.Perturbations of any of the three Arg residues result in loss of interaction with SSB and partial loss of RecG function in vivo, as demonstrated by cell UV sensitivity, SOS induction, and filamentation (25).
Genetic interaction between recG and the radD (formerly yejH) gene has been explored (19,20).RadD is a putative SF2 helicase and is important for E. coli survival after ionizing radiation treatment (7) and for accelerating the rate of RecA-mediated strand exchange in vitro (26).Deletion of both recG and radD results in a severe growth defect that leads to the formation of spontaneous suppressor mutations in priA helicase motifs and the recA promoter (19).The ΔrecGΔradD phenotype is also suppressed by deletion of either the recombination mediator recF or recO, both of which encode proteins that help load RecA onto SSB-coated ssDNA gaps to promote recombination.Characterization of the recG-radD interaction provides further evidence for a role for both proteins in alleviating toxic DNA situations, likely resulting from an accumulation of recombination intermediates.
Several additional genetic interactions between recG and DNA metabolism genes have been identified.Synthetic lethality has been reported for strains lacking recG and uvrD (27), rnhA (14,18), polA (14,28), or priA (29).Similar to a ΔrecG ΔradD strain, ΔrecG ΔuvrD cells experience a "death by recombination" phenotype that can be suppressed by the deletion of recF or recO, as well as other genes known to be involved in promoting recombination or an accumulation of recombination intermediates, including recQ (27).The synthetic lethality of cells lacking recG and rnhA (which encode RNase HI) likely stems from the toxicity associated with constitutive stable DNA replication, or cSDR (14).RecG is also required for cell viability when 3′ ssDNA exonucleases are absent, and this lethality is a consequence of PriA activity (18).Although PriA helicase activity can be detrimental when recG is absent, deletion of recG is synthetically lethal with priA-null mutations, as PriA is necessary for DnaB loading in the absence of RecG (29).
The RecF, RecO, and RecR proteins serve to load the RecA protein into post-replication gaps created by replisome lesion-skipping (36).The RecA protein thus loaded generates recombination intermediates, or joint molecules, that link together the two recently replicated chromosome copies.Failure to resolve the joint molecule can result in cell death since the linkage prevents chromosome segregation at cell division (36).If one or more of the RecFOR proteins are missing, the joint molecules are not generated.Thus, deleterious effects that are suppressed by inactivating RecF, RecO, or RecR can be traced to a role in the resolution of joint molecules behind the replication fork, a role that is no longer needed if the joint molecules are not generated.
The previous identification of recG genetic interactions summarized above has been driven by targeted approaches involving deletion of recG in combination with a second gene of interest to study the resulting phenotype(s).While informative, such approaches are not systematic and have provided a limited scope of the entirety of genetic interactions in recG.Previously, we employed transposon sequencing (Tn-seq) (37,38), an untargeted genetic screening approach, to screen for genes involved in the survival of ionizing radiation in E. coli and to link a DNA replication restart pathway to DSB repair (7,39).Here, we used Tn-seq to systematically evaluate the presence and relative strength of interactions between recG and every known nonessential gene in E. coli.We identify genes that become conditionally essential or conditionally important in the absence of the recG gene, confirm several of these genetic interactions with additional methods, evaluate suppression of synthetic lethality with recF or recO deletions, and test whether loss of the interaction between RecG and SSB contributes to lethality.

Bacterial strains and primers
Strains used in this study are E. coli K12 MG1655 and its derivatives (Table 1).Chromoso mal gene knockouts were generated by λ-Red recombination and/or added to strains by P1 transduction, as previously described (19,40).Deletion of recG was added to strains after transformation of each strain with pJJ100, which harbors a wild-type copy of recG.Deletions of recO or recF were added last to strains due to the impaired recombination ability of the resulting strains.PAGE-purified primers were purchased from Integrated DNA Technologies (Table 2), and terminal phosphorothioate bonds are denoted with an asterisk.
The preparation of electrocompetent cells for transposition was performed as previously described (7).Briefly, cells were cultured in 1 L of Super Optimal broth with Catabolite repression (SOC) medium (2% tryptone, 0.5% yeast extract, 10 mM NaCl, 2.5 mM KCl, 10 mM MgCl 2 , 20 mM glucose) at 37°C with shaking to an OD 600 nm of 0.4 to 0.6, chilled at 4°C, pelleted, and washed three times with cold 10% glycerol.Cells were resuspended in 3 mL of cold glycerol-yeast extract medium [10% glycerol (vol/vol), 0.125% yeast extract, 0.25% tryptone] and stored at −80°C.One hundred microliter aliquots of cell suspension were mixed with 5 µL of transposome and electroporated at 2.5 kV.Cells were recovered in 1 mL of SOC medium at 37°C for 1 h, spread on multiple 15 mm by 150 mm Super Optimal Broth (2% tryptone, 0.5% yeast extract, 10 mM NaCl, 2.5 mM KCl, 10 mM MgCl 2 , 10 mM MgSO 4 ) round plates supplemented with 40 µg/mL kanamycin, and incubated overnight.Total colony counts were estimated by counting the number of colonies on a one-third section of each plate.Colonies were pooled in Luria Broth (LB) and stored with 10% dimethyl sulphoxide at −80°C.Transpositions were repeated to create triplicate pools of transposon insertion mutants in an MG1655 (wild-type) strain and an MG1655 ΔrecG (EAW505) strain, with completed libraries totaling an estimated of >450,000 mutants per strain.

Library sample preparation, sequencing, and data analysis
Each Tn-seq library replicate was used to inoculate 100 mL of LB to an OD 600 nm of 0.02 and grow overnight at 37°C with shaking.Genomic DNA was isolated from each culture (Promega Wizard gDNA extraction kit) and mechanically sheared to an average size between 400 and 500 bp.The preparation of the fragment library for sequencing was performed as described using the NEBNext Ultra II DNA Library Prep Kit for Illumina (#E7645), USER Enzyme (New England BioLabs), and AxyPrep Mag PCR Clean Up Kit.Custom primers oAM103, oAM104, oAM105, oAM106, oAM068, or oAM069 and Tn-enrich were used for the PCR amplification step so that only sequences flanking transposons were amplified (Table 2).Tn-enrich was the forward primer and is complementary to the adaptor.Custom reverse primers oAM068, -069, and -103-106 contained different indexes (bolded in Table 2) for multiplexing.The amplified DNA fragment libraries were sequenced on a single-end Illumina NextSeq full flowcell for 75 cycles with 25% PhiX DNA and primers Read 1 and Index 1. Data were analyzed using a previously described method (7,39,42) with slight variations.Sequencing data were checked for transposon sequence enrichment in the first 10 bases using the sequence TAAGAGACAG.The number of reads containing the transposon sequence was divided by the entire read count to ensure data quality.The data were analyzed using a two-sample analysis.Five percent of the start and end of each gene was discarded in the analysis, and only insertions with greater than five sequencing reads were counted.Potential screen hits were chosen based on a greater than 4 log 2 -fold decrease in insertion sequencing reads in a particular gene in the ΔrecG library as compared to the wild-type library.Genes with significant P-values (<0.05) for log-fold decreases in read count or number of insertions were sorted and chosen for further validation.

Mini-F synthetic lethality assay
The pJJ100 plasmid is a recG + derivative of the mini-F low-copy lac + pRC7 plasmid and was a generous gift from Christian Rudolph, constructed as previously described (43,44).The synthetic lethality assay was performed as previously described (19).Briefly, strains were transformed with pJJ100 before the chromosomal recG deletion was added.Cultures of each strain with the pJJ100 plasmid and 50 µg/mL ampicillin were grown overnight.The following day, 5 mL of LB without antibiotic was inoculated with a 50-µL overnight culture.Cultures were grown to an OD 600 nm of 0.2, immediately serially diluted in 1× phosphate buffered saline (PBS) buffer (137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , 1 mM CaCl 2 , and 0.5 mM MgCl 2 ), and spread on LB plates supplemented with 500 µM isopropyl β-D-1-thiogalactopyranoside (IPTG) and

DNA damage sensitivity assays
Strains MG1655, ΔrecG, recG-R474E, recG-R614E, and recG-R484E were grown overnight.Saturated overnight cultures were diluted 1:100 in fresh LB, grown to an OD 600 nm of 0.2, and immediately serially diluted in 1× PBS.Ten microliters of each dilution was spotted onto square LB plates supplemented with ciprofloxacin, mitomycin C, hydrox yurea, trimethoprim, or nitrofurazone.Once the spots had adsorbed into the plates, the plates were incubated at 37°C overnight under foil and photographed on an iBright CL1000 (Thermo Fisher Scientific) imaging system the following morning.The plates were poured fresh the day of the experiment and kept under foil to avoid degradation of the DNA-damaging agents.

Growth competition assay
Strain fitness was assessed using a modified growth competition assay (45,46) descri bed previously (25).Briefly, saturated cultures of ΔaraBAD and araBAD + strains were mixed at equal volumes and used as inoculum (1:100) into fresh LB.Mixed cultures were grown at 37°C with shaking and reinoculated into fresh medium every 24 h.At 0, 24, 48, and 72 h of incubation, cultures were serially diluted in 1 × PBS and spread on tetrazolium arabinose plates (1% tryptone, 0.1% yeast extract, 0.5% NaCl, 1% arabinose, 0.005% 2,3,5-triphenyltetrazolium chloride, 1.6% agar).Plates were incubated at 37°C overnight before counting white (araBAD + ) and red (ΔaraBAD) colonies.Growth competition graphs represent data from biological triplicate experiments, with the average percentage of red and white colonies reported.One standard deviation from the mean is represented by error bars.

Tn-seq is performed to identify recG genetic interactions
Tn-seq libraries consisting of more than 450,000 insertion mutants were generated in E. coli K12 MG1655 (wild-type) and an MG1655 derivative lacking the recG gene (EAW505) (Fig. 1).Transformants were pooled together in groups of ~150,000 to form three library replicates per strain.Each pooled library replicate was subjected to a competitive outgrowth phase.Genomic DNA was extracted and prepared for sequencing using commercially available kits, and the prepared libraries were sequenced using Illumina Next Generation Sequencing technology.
Each library replicate generated 20 to 30 million sequencing reads, with the total reads per library amounting to over 75 million reads (Table 3).Over 97% of sequencing reads contained perfect barcode matches, and over 85% of sequencing reads contained the transposon sequence (Table 3).An average of one insertion every 24-25 bases per replicate was calculated for each sample, suggesting a high-density Tn-seq library.
Transposon insertion sites were evaluated in parallel between the wild-type and ΔrecG libraries.Parallel sequencing results revealed genes that, when disrupted by transposon insertions in the absence of recG, caused cells to become inviable or less viable than when disrupted in the wild-type strain.The number of transposon insertion sequencing reads per gene in each library was evaluated and is represented in Fig. 1B across the E. coli chromosome as a log 2 -fold change.Negative log-fold change values represent genes in which transposon insertion reads mapped less frequently in the ΔrecG library than in the wild-type library, and positive log-fold change values represent the opposite effect.
Genes with the most negative log-fold changes in insertion sequencing reads are considered the strongest screen "hits." The gene with the most negative log-fold change in sequencing reads (−11.2) was recG, which is due to its presence and tolerance to disruption in the wild-type strain and its removal in the ΔrecG strain.To evaluate genetic interactions with nonessential genes only, genes with less than 12 average unique insertions across replicates were deemed essential and excluded from further evaluation.This threshold identified 549 genes out of 4,314 genes (12.7%) as essential (Table S1).Several methods have been employed to evaluate gene essentiality in E. coli with varying results (47,48).Factors such as medium and growth conditions, the presence and duration of competitive outgrowth, and an experimental approach impact essentiality.Transposon insertion studies typically overestimate the number of  essential genes because library outgrowth can eliminate slow-growing mutants from the population (7,47,49).However, transposon insertions can also be detected in essential genes, especially in regions such as protein termini or flexible interdomain regions that allow a partially functional protein to be produced (50).Of the 3,765 genes deemed nonessential, the largest log-fold changes in transposon insertion sequencing reads were observed in dam (−7.6,DNA adenine methyltransfer ase), uvrD (−6.7,DNA helicase), rnhA (−5.5, ribonuclease HI), radA (−4.7,DNA recombina tion protein), and rep (−4.1,DNA helicase).The presence of insertions in these five genes in the wild-type library population indicates that the genes are nonessential under these conditions and in this strain (Fig. 2A).However, the substantial decrease in both insertion sites and insertion sequencing reads across these five genes in the ΔrecG background indicates that disruptions to any of these genes result in cells with greatly reduced viability.Such strong effects suggest potential synthetic lethal interactions between the identified genes and recG.Synthetic lethality has been previously reported for recG dam, recG uvrD, and recG rnhA double deletion mutants (18,27,28), and a strong synergistic effect has been observed for cells lacking both recG and radA (31).Finally, rep represents a new recG genetic interaction.

Full-Length Text
Other genes have been implicated in performing distinct but partially overlapping functions as recG, including the radD (19), rarA (33), and ruvABC (33) genes.The log-fold changes for radD and rarA in the recG-deletion strain were −2.2 and −1.4,while the log-fold changes for ruvABC were −0.5, −1.4, and −0.3, respectively.Of these genetic interactions, the interaction between radD and recG is the strongest; viability in double deletion mutants is greatly reduced and only sustained for multiple generations with the accumulation of suppressor mutations (19).Cells lacking recG and rarA or recG and ruvB have synergistic sensitivity to DNA-damaging agents (33), but these strains are viable.However, deletion of recG, rarA, and ruvB together is not possible (33).Transposon insertion profiles for radD, rarA, and ruvABC support these previously reported observa tions (Fig. S1), where insertions are greatly reduced in radD when recG is absent and more modestly reduced in the rarA and ruvABC genes.
Additional modest recG interactions have been reported, including with the uvrA and recD genes (6), where double deletion mutants have reduced viability or increased sensitivity to DNA damage.Transposon insertion profiles show that transposon insertions in uvrA and recD are still well tolerated in cells lacking recG, but strong reductions in the number of sequencing reads are observed.The log-fold changes for these genes in the recG-deletion strain were −1.6 and −0.8, respectively.This is consistent with double deletion mutants being viable but displaying reduced fitness.Indeed, ΔrecG ΔuvrA and ΔrecG ΔrecD strains are outcompeted by a ΔrecG strain (Fig. S2), confirming fitness defects for the double mutants.The genes with log-fold decreases in sequencing reads between dam and uvrA are listed in Table S2, and going forward, we will focus on the five strongest screen hits (dam, uvrD, rnhA, radD, and rep).

Validation of recG synthetic lethal interactions
To independently validate and examine the strongest Tn-seq screen hits, a plasmidbased lethality assay was used.recG under the control of its native promoter cloned into the unstable mini-F lacZ + pRC7 plasmid (plasmid pJJ100, a kind gift from Christian Rudolph) (19,44) was transformed into ΔlacIZYA strains also lacking dam, uvrD, rnhA, radD, or rep.Retention of the recG + plasmid was evaluated in cultures grown with antibiotic selection pressure withheld.Cultures were plated on LB plates supplemented with IPTG and X-gal, which allowed colonies that retained the mini-F lacZ + plasmid (blue) to be distinguished from colonies that lost the plasmid (white).Each of these strains (which had the wild-type chromosomal recG gene) readily lost the plasmid-encoded copy of recG (Fig. 2B and C).When the plasmid was present in a ΔrecG background (EAW1102), the plasmid was lost at a similarly high rate (37%; Fig. 2C).This is identical to a result seen previously for this strain (19).The loss of the recG-expressing plasmid was somewhat higher (49.8%) when it was introduced into a ΔlacIZYA strain with an otherwise wild-type background (EAW408; recG + ), using this same protocol (data not shown).
To evaluate synthetic lethality of double deletion strains in the ΔlacIZYA background, recG was subsequently deleted from the transformed strains lacking dam, uvrD, rnhA, radA, or rep.The resulting new set of strains was tested in the plasmid retention assay.Unlike the initial set of strains, which readily lost the recG + plasmid, deletion of recG in strains already lacking dam, uvrD, rnhA, radA, or rep resulted in significantly higher rates of plasmid retention (Fig. 2B and C).For the ΔrecG Δrep and ΔrecG ΔradA strains, a small percentage of white colonies were observed.These white colonies appeared much smaller in size than the blue colonies.Attempts to re-streak these small white colonies resulted in streak plates with no growth or the growth of more small colonies, ruling out the possibility that these colonies are a result of suppressor mutations.Altogether, these results indicate that recG is synthetically lethal with dam, uvrD, and rnhA.Furthermore, recG is conditionally important but not quite synthetically lethal in cells lacking radA and rep.

Suppression by deletions of recF or recO
Suppression of the synthetic lethality or synergistic sensitivity phenotypes of recG and other genes (radD, ruvB, and rarA) by loss of recF or recO has been reported (19,33).This is thought to be the case because the RecFOR pathway results in branched DNA intermediate formation in post-replication gaps that utilizes RecG (and other proteins) for processing.To evaluate whether deletion of recF or recO restores the viability of ΔrecG Δdam, ΔrecG ΔuvrD, ΔrecG ΔrnhA, ΔrecG ΔradA, and ΔrecG Δrep strains, the plasmid retention assay was repeated with sets of strains also lacking recF or recO.We note that, despite several attempts, we were not able to construct ΔrecG ΔrnhA ΔrecF or ΔrecG ΔrnhA ΔrecO strains.
In agreement with previous observations, loss of recF or recO restores the viability of ΔrecG ΔuvrD cells (Fig. 3), which have been reported to experience a death by recombina tion (27).However, there do appear to be slight differences in the extent of suppression between a recF or recO deletion in ΔrecG ΔuvrD cells.Loss of recO results in a consistently greater number of white colonies (loss of plasmid) in the plasmid retention assay, and the white colonies appear slightly larger in size than the white colonies that appear in strains lacking recF.The opposite effect appears in ΔrecG ΔradA cells.While the white colonies in ΔrecG ΔradA ΔrecF and ΔrecG ΔradA ΔrecO experiments appear similar in size, there were consistently more white colonies appearing in the plasmid retention experiments with the ΔrecG ΔradA ΔrecF strain (Fig. 3).It has been previously reported that loss of recF improves the viability of a ΔrecG ΔradA strain (31), but the effect of loss of recO has not been reported.These results indicate that loss of recF or recO does not always produce identical outcomes in people with the same genetic backgrounds.In the case of ΔrecG Δrep cells, loss of recF or recO improved the viability of the strain to similar extents.
Interestingly, the synthetic lethality of ΔrecG Δdam is not suppressed by deletion of recF or recO, supported by the lack of white colonies present in plasmid retention assays (Fig. 3).This indicates that the mechanism of lethality in ΔrecG Δdam cells is not RecFORdependent.

Effects of a mutation that abates RecG interaction with SSB
It was recently shown that an interaction with SSB is important for RecG function in vivo, with RecG residues R474 and R614 playing essential roles in the interaction between RecG and the SSB C-terminus (21).A recG-R474E mutation abates the interaction between RecG and SSB, and a recG-R474E strain displays UV sensitivity, increased basal SOS induction, and cell filamentation, albeit not to the same extent as a complete loss of recG (25).To test whether the recG-R474E mutation produces similar effects to a complete loss of recG in the plasmid retention assay, the mutation was added to strains lacking dam, uvrD, rnhA, radA, or rep in the ΔlacIZYA background.As a negative control, we also tested the effects of a recG-R484E mutation, which mutates a residue that is near but not in the SSB binding pocket on RecG, in the plasmid retention assay as a negative control.Neither the recG-R474E nor the recG-R484E mutation alone led to higher levels of recG + plasmid retention in recG + , as expected (Fig. 4).
When combined with deletions in dam, uvrD, radA, or rnhA, the recG-R474E mutation was synthetically lethal, mirroring the effects of a complete recG deletion (Fig. 4).In contrast, loss of the RecG/SSB interaction did not affect Δrep cells, which differed from the synthetic lethality observed with deletion of recG in Δrep cells.This difference could help to explain why the recG-R474E does not phenocopy ΔrecG mutation in E. coli (21).The results were different for recG-R484E.In all cases except for one (rnhA), the recG-R484E mutation did not impact plasmid retention or strain viability, highlighting the dependence of the genetic interactions on the ability of RecG to interact with SSB.
To more broadly examine whether loss of the RecG/SSB interaction produced general or specific DNA damage sensitivity phenotypes, the sensitivity of the recG-R474E and recG-R614E binding pocket mutant strains to DNA-damaging agents was examined alongside the wild-type (MG1655), ΔrecG, and recG-R484E control strains.recG-R614E encodes a RecG variant that is also incapable of SSB binding, and both recG-R474E and recG-R614E mutant strains are sensitive to UV exposure (25).We found that the recG-R474E and recG-R614E mutations render cells sensitive to a variety of DNA-damaging agents, including ciprofloxacin, hydroxyurea, trimethoprim, mitomycin C, and nitrofura zone, but the sensitivity was not to the same extent as complete loss of recG (Fig. 5).
Altogether, combining the recG-R474E mutation with deletions in dam, uvrD, rnhA, radD, or rep and surveying the sensitivity of the mutant to a panel of DNA-damaging agents reveal a more general role in repair and recombination for the RecG/SSB interac tion.It does not appear that loss of the interaction affects a specific RecG repair pathway or function.Instead, loss of interaction generally diminishes but does not entirely eliminate the capacity of RecG to access the sites where it is to function.

DISCUSSION
This study used Tn-seq to identify several recG genetic interactions and gain insight into recG function in E. coli.The strongest interactions identified in the screen were the dam, uvrD, rnhA, radA, and rep genes.These genes were either synthetically lethal with ΔrecG or conditionally important in cells lacking recG.The screen results were confirmed with a plasmid-based lethality assay.Evidence for strong interactions between recG and dam (28,34), uvrD (27,28), rnhA (14), and radA (31) has appeared previously, while the interaction we identified with rep has not been previously reported.The set of genes listed in Table S2 represents a complete accounting of the strongest recG interactions.
We also found differing patterns in suppression of synthetic lethality/conditional importance with deletions of recF or recO.Deletions in recF or recO improved the viability of recG uvrD, recG radA, and recG rep mutant strains, but no suppression of synthetic lethality was observed for recG dam mutants.We were unable to construct a strain lacking the recG, rnhA, and recF or recO genes simultaneously, so we were unable to test for recFO suppression directly.Loss of recO appeared to restore the health of recG uvrD to a greater extent than loss of recF, based on qualitative colony size observations.In contrast, recG radA viability appeared to be more consistently improved by the loss of recF over recO.Additionally, we evaluated the effect of the recG-R474E mutation on gene interactions and found that loss of the RecG/SSB interaction produced similar effects as a complete loss of recG when paired with several other gene deletions.Loss of the RecG/SSB interaction appears to produce general effects that affect RecG activity in some but not all repair pathways it is involved with, along with the response to different types of DNA damage.Altogether, the work is consistent with prior studies indicating an important role for RecG in multiple DNA repair pathways.Here, we develop RecG roles in the repair of post-replication gaps and in the suppression of over-replication by aberrant replication initiation.
The absence of dam, uvrD, rnhA, radA, or rep is associated with an increase in double-stranded breaks, stalled replication forks, and/or an increase in recombination or accumulation of recombination intermediates (31,39,(51)(52)(53)(54)(55)(56)(57).The functions of these genes appear to become especially important, if not essential, when RecG is absent.Interestingly, differences in the mechanisms leading to the conditional importance of these genes in the absence of recG are starting to emerge.We have identified that some (recG rep, recG uvrD, recG radA) genetic interactions are suppressed by recF or recO deletions, while others are not or probably not (recG dam, recG rnhA), and that recF or recO deletion can produce minor differences in suppression (in recG radA and recG uvrD mutants).The suppression by recF and recO reflects an important role for RecG in the recombinational repair of post-replication gaps (19,33).
UvrD modulates recombination by dismantling RecA nucleoprotein filaments (54,(58)(59)(60).Loss of UvrD results in an accumulation of recombination intermediates that require RecG Holliday junction resolution activity.The toxicity caused by this accumulation of recombination intermediates is negated when recombination is suppressed by the loss of recF or recO (27).Recombination intermediate accumulation is also likely the cause of ΔrecG ΔradA cell lethality, as RadA has been implicated in performing similar activities to RecG in processing branched DNA intermediates (32).The cause of lethality with rep, which encodes a DNA helicase responsible for clearing replication roadblocks, is not immediately clear and warrants further investigation.The suppression of the synthetic lethality of Δrep ΔrecG strains by deletion of the genes encoding RecF or RecO suggests that Rep may be yet another helicase (along with RadA, RadD, UvrD, RecQ, RuvB) with a role in preventing the formation of or resolving joint molecules in post-replication gaps in a manner that complements the role of RecG.
Unlike uvrD, rep, and radA, loss of recF or recO did not suppress the lethality of ΔrecG Δdam, and a recF or recO deletion could not be introduced to ΔrecG ΔrnhA cells.Dam is responsible for DNA methylation, which helps direct mismatch repair.Loss of dam is associated with double-strand breaks, the repair of which could trigger unscheduled DNA replication behind the fork (51,61).The dam recG results suggest that RecG plays an equally essential role in repairing the DSBs associated with dam removal.Dam is also implicated in suppressing cSDR originating from R-loops, as are the RecG and RNase HI proteins (62).The overall theme is that RecG appears to be needed to prevent genome over-replication.When Dam is absent, RecG may be required to act at D-loop intermediates generated by double-strand break repair so as to prevent an unscheduled replication restart that would lead to over-replication of segments of the genome.Similarly, when RNase HI, which degrades RNA:DNA hybrids, is absent, RecG R-loop unwinding may be required to prevent inappropriate replication initiation at these sites.Collectively, the results are also consistent with in vitro studies on RecG remodeling of branched structures (10,63) and in vivo work showing that DNA amplification or over-replication is a detrimental problem in cells lacking recG, especially at sites of DSBs or replication termination (15,17).
The in vivo function of RecG is dependent to a significant degree on its interaction with SSB (25).However, mutations in recG that affect the SSB binding pocket produce phenotypes that are not as severe as a complete loss of the gene.This has led us to question whether the interaction with SSB is important only for some, but not all, RecG functions.To answer this question, we combined the SSB-binding-deficient recG-R474E mutation with deletions in dam, uvrD, rnhA, radD, or rep and surveyed the sensitivity of the mutant to a panel of DNA-damaging agents.The recG-R474E mutation led to similar synthetic lethality patterns and DNA damage sensitivity as ΔrecG for dam, uvrD, rnhA, and radA, suggesting that RecG/SSB interaction is important for coordinated RecG function with the proteins encoded by these genes.In contrast, the recG-R474E mutation did not affect viability in Δrep cells, whereas deleting recG from Δrep cells was lethal.This indicates that RecG interaction with SSB is not required for its coordinated activities with Rep.
The results are summarized in Fig. 6.In general, RecG plays important roles in at least two kinds of DNA transactions.Post-replication gap repair, where it plays a role in resolving some, if not many, joint molecules constructed behind the replication fork by RecA, is one.A second involves a role in the suppression of inappropriate replication initiation events at R-loops, double-strand breaks arising from dam inactiva tion, or during replication termination.The function of RecG complements different additional enzymes in these two capacities.However, each role of RecG represents a critical contribution to cell survival.Each separate set of double mutants can result in cell death.Many, but not all, of these transactions require the interaction of RecG with SSB.
We note that while the top five screen hits were the subject of this investigation, Tnseq allowed us to evaluate the comparative importance of every nonessential gene in wild-type and ΔrecG E. coli.Genes with known (but weaker) interactions with recG (such as radD, ruvABC, rarA, recD, uvrA) were also observed to be of greater importance in cells lacking recG in the Tn-seq data.It is likely that there are additional genes with which recG may have a functional relationship identified in the Tn-seq screen (listed in Table S2), and this presents a significant opportunity for future investigation.
Ιn addition to providing insight on recG function by studying recG genetic interactions and recG-R474E phenotypes, we hope that this data set will be useful to the science community for a variety of applications.These applications include evaluating gene essentiality, identifying nonessential domains within essential genes, and identifying regions of genes of interest that may tolerate manipulation, such as the addition of tags or fusion to fluorescent proteins for imaging.The insertion data can also be viewed in programs such as MochiView (64) or Integrative Genomics Viewer (65) to evaluate tolerance of transposon insertions in noncoding or intergenic regions of the genome.

FIG 1
FIG 1 Tn-seq schematic and overview of insertion changes across the E. coli genome.(A) Schematic showing the general process used to carry out the Tn-seq screen, which involved generating single gene knockout libraries in wild-type and ΔrecG cells, growing libraries, gDNA isolation of the cultures, and Illumina sequencing.An example of the insertion profiles expected for nonessential, essential, and conditionally essential genes is drawn.(B) Circos plot showing the log 2 -fold change in sequencing reads between wild-type and ΔrecG libraries for all genes in E. coli (light gray) and top screen hits (red).The positions of oriC and terC are marked in black to orient them to the chromosome.

FIG 3
FIG 3 Suppression of synthetic lethality by deletion of recF or recO.Representative photos of plasmid retention assay blue and white colonies with strains lacking recF (A) or recO (B).Small regions of each plate are shown at 2.5× zoom to highlight the presence of small, white colonies.(C) Quantification of plasmid retention in strains from assays shown in panels A and B, with bar graphs displaying the mean and error bars representing one SD away from the mean.All experiments were performed in triplicate.P-value designations (*, **, and ***) are the same as listed in Fig. 2 legend.

FIG 4
FIG 4 The effects of recG-R474E mirror those of ΔrecG in strains lacking radA, rnhA, uvrD, and dam, but not rep.(Α) Photos depicting blue and white colonies from plasmid retention assays where deletions in rep, radA, rnhA, uvrD, or dam are combined with the recG-R474E or recG-R484E mutation (control).(B) Quantification of plasmid retention from experiments depicted in panel A. All experiments were performed in triplicate.Bar graphs plot the mean value, with error bars representing one SD from the mean.P-value designations (*, **, and ***) are the same as listed in Fig. 2 legend.

TABLE 1
Strains used in this study

TABLE 2
Oligonucleotides used in this study 60 µg/mL X-gal (stored in the dark).Plates were incubated at 37°C for 16-18 h before blue and white colonies were counted, and plates were photographed.All experiments were repeated in biological triplicate with similar results.Bar graphs were generated in GraphPad Prism, where statistical tests (Brown-Forsythe and Welch analysis of variance) were conducted to determine significance.