Processing of reversed replication forks is required for the resolution of replication-transcription conflicts

Abstract

DNA replication and transcription occur simultaneously on the same template, leading to conflicts between the two machineries. Conflicts stall replication forks, lead to genome instability, mutagenesis, and the formation of deleterious R-loop structures. Despite many studies across various organisms, how the replication fork structure is impacted by these encounters remains unclear. Here, we performed an unbiased genetic screen using a transposon library to identify factors essential for surviving deadly conflicts in Bacillus subtilis. Our screen identified three hits: RNase HIII, AddA, and AddB. Our prior work had shown that RNase HIII, which processes R-loops, is essential for surviving conflicts. However, AddA and AddB, which function together as a complex to process blunt DNA ends, have not been previously identified as essential conflict resolution factors. Through follow-up genetic and biochemical analyses, we found that the helicase activity but not the nuclease activity of this complex is required for conflict resolution. Based on the fundamental properties of DNA, our work collectively indicates that upon conflicts, the nascent strands form a reversed fork structure, which is then unwound by AddAB, leading to re-annealing to the parental strands. This process re-establishes an intact replication fork that can be used to restart replication after conflicts.

Graphical Abstract

Graphical Abstract.

Introduction

DNA replication and transcription share a common template and can occur simultaneously, leading to conflicts between the replication and transcription machineries. These conflicts occur either co-directionally when a gene is encoded on the leading strand, or head-on when a gene is encoded on the lagging strand. While both types of conflicts can be detrimental for cells and lead to genome instability [1], head-on conflicts have been shown to have more adverse effects, which include replication stalling [2, 3, 4], replication restart [3], chromosome breaks [5, 6], and higher mutation rates [7, 8]. Accordingly, across highly divergent bacteria, most genes are encoded on the leading strand, and this is especially true for those that are highly expressed [9]; meanwhile, in human cells, origins of replication tend to coincide with transcription start sites, which has been proposed as a way to ensure co-directionality of replication and transcription, at least for protein-coding genes [10].

However, both bacteria and humans have plenty of genes in the head-on orientation, which require specialized cellular pathways that help resolve head-on replication-transcription conflicts. For instance, head-on conflicts lead to the preferential formation of R-loops [5, 11], three-stranded nucleic acid structures that form when the nascent RNA reanneals to its template DNA and displaces the non-template DNA strand. R-loops can lead to genome instability, mutagenesis, replication stress, and chromosomal breaks, and need to be resolved for proper cellular function [5, 11, 12]. Other mechanisms that promote genome stability at head-on conflicts are recombinational repair [4, 13], resolution of topological stress [14], and coupling between transcription and mRNA processing [15]. Alternatively, transcription-coupled repair promotes mutagenesis at head-on conflict regions in both prokaryotes and eukaryotes [16, 17].

Using the Gram-positive model bacterium Bacillus subtilis, we set out to determine what other factors prevent cell death due to head-on conflicts. To address this question, we used a transposon mutagenesis screen. Our screen revealed three genes that promote cell survival specifically in cells containing an engineered head-on conflict: rnhC, addA, and addB. While the role of rnhC, which codes for RNase HIII, in head-on conflict resolution is well understood [11], an essential role of addA and addB, which generate the AddAB heterodimer, in conflict resolution has not been previously described.

AddAB is a helicase-nuclease that is involved in the first step of recombinational repair in Gram-positive bacteria and is a functional analog of the Escherichia coli RecBCD helicase-nuclease complex [18]. AddAB recognizes a blunt or nearly blunt double-stranded DNA end [19]. Our data show that these DNA ends are generated by replication fork reversal in the context of head-on replication-transcription conflicts.

Reversed replication forks are four-way junctions that can form when replication is perturbed, such as by uncoupling of the leading and the lagging strands [20]. The formation of reversed replication forks was first proposed in E. coli, as a mechanistic explanation for the observed role of RuvAB in creating double-strand breaks in RecBCD-deficient cells [21]. They can be formed spontaneously (non-enzymatically), likely due to accumulation of positive supercoiling in front of the fork [22], or enzymatically, by recombinases such as RecA or Rad51 [23, 24]. Because positive supercoiling accumulates at head-on conflict regions and can lead to genome instability [14], we suggest that these reversed forks are formed spontaneously. These reversed forks need to be restored, either by unwinding or by degradation. Here, we determined that the helicase but not the nuclease activity of the AddAB complex is required for processing of stalled replication forks post-conflicts. Taken together, our results show that fork remodeling, specifically re-annealing of the nascent strands, is a key process for continuing DNA replication when it stalls due to conflicts with transcription.

Materials and methods

Bacterial culture

Bacillus subtilis was cultured in lysogeny broth (LB). Bacterial plates were grown overnight at 37°C with the following antibiotics when appropriate: 100 µg/ml spectinomycin, 500 µg/ml erythromycin, and 12.5 µg/ml lincomycin (MLS), 5 µg/ml (B. subtilis) or 50 µg/ml (E. coli) kanamycin, 10 µg/ml chloramphenicol, and 100 µg/ml carbenicillin. Liquid cultures were started from single colonies and grown with aeration (260 rpm). A list of all strains used in this study can be found in Supplementary Table S1, and a list of plasmids can be found in Supplementary Table S2.

Strain and plasmid construction

The parental strain for all B. subtilis strains used in this study is HM1 (same as AG174, originally named JH642) [25, 26]. Gene deletions that are marked with MLS or kanamycin resistance were obtained from [27], and recombined into the HM1 chromosome as in [28] except for ΔaddAB, which was made in the HM1 background [29]. When indicated, antibiotic-resistant cassettes were excised by transforming the strains with a plasmid expressing the Cre recombinase (pDR244, BGSCID: ECE274) purified from Rec + E. coli cells (AG1111) [30] with the QIAprep Spin Miniprep Kit (QIAGEN), generating markerless strains.

For the transposon screen, the spectinomycin resistance gene in pWX642 (a gift from Dr. David Rudner) [31] was switched with the kanamycin resistance gene from mfd::kan B. subtilis cells (BGSCID: BKK00550) to generate pHM785. To do this, the kanamycin resistance gene, including the promoter and ribosome-binding site (RBS), and pWX642 without the spectinomycin resistance gene were PCR amplified (Q5 High-Fidelity DNA Polymerase, NEB) with primers containing 20 nucleotides of overlap. The PCR products were cloned using NEBuilder® HiFi DNA Assembly Master Mix (NEB) for 1 h at 50°C. All primers used in this study can be found in Supplementary Table S3.

For complementation experiments, the coding region of addB or addA was amplified from HM1 gDNA (Supplementary Table S3) and cloned using NEBuilder® HiFi DNA Assembly Master Mix into pCAL838 [32] (Supplementary Table S2). For addB, the PCR product included the native RBS, while the following transcription terminator was included in the reverse primer GTCCCATTCATAAGGA; for addA, the native transcription terminator was included in the PCR product, and the following RBS and linker were included in the forward primer: AAGGAGGTATACAT. To make the catalytic mutants, these plasmids were amplified with primers containing the appropriate mutation and 20 nt of overlap (Supplementary Table S3), and circularized using NEBuilder® HiFi DNA Assembly Master Mix (Supplementary Table S2). Plasmids were linearized using KpnI and transformed into competent cells containing a deletion of the appropriate gene.

For tagging rnhC with the SsrA degron, approximately 500 bp at the 3′ end of the gene were amplified (Supplementary Table S3) and cloned EcoRI-XbaI into pGCS [33] (Supplementary Table S2). This plasmid, purified from AG1111 cells, was transformed into competent B. subtilis cells.

For AddAB purification, the coding sequence of addB was amplified (Supplementary Table S3) and cloned into pET28a (Thermo) (Supplementary Table S3) by HiFi (NEB) to generate a 6XHis N-terminal tagged protein. The coding sequence of addA was amplified (Supplementary Table S3) and cloned into pET21a (Thermo) (Supplementary Table S3) by HiFi (NEB) to generate a 6XHis C-terminal tagged protein. To make the catalytic mutants, these plasmids were amplified with primers containing the appropriate mutation and 20 nt of overlap (Supplementary Table S3), and circularized using NEBuilder® HiFi DNA Assembly Master Mix (Supplementary Table S2).

Transposon screen

HM2195 [11] cells were transformed with pHM785 purified from AG1111 cells, plated on MLS, and incubated at 30°C for 40 h. For each biological replicate (three biological replicates were done in parallel), four colonies were inoculated into four tubes with 2 ml of LB supplemented with 5 µg/ml of kanamycin, and were grown for 22 h at 25°C with aeration. The four cultures were mixed thoroughly, and 100 µl of cells were mixed with 100 µl of 50% glycerol and frozen at −80°C. The next day, the frozen cells were thawed in a 37°C water bath, added to 50 ml of LB supplemented with 5 µg/ml of kanamycin, and grown for 5 h at 37°C with aeration. Around 500 µl of these cells were added to two 50 ml flasks with LB supplemented with 5 µg/ml of kanamycin, one of which had 1 mM IPTG (conflict on condition) while the other did not (conflict off condition). These cultures were grown to an OD600 of approximately 0.5, after which they were spun down, and the pellets were frozen at −20°C. The pellets were then resuspended in 1 ml of solution A (10 mM Tris–HCl, pH 8.0, 20% w/v sucrose, 50 mM NaCl, 10 mM EDTA, 1 mg/ml lysozyme) and incubated at 37°C for 30 min. Then, cells were lysed with 1 ml lysis buffer (100 mM Tris pH 7.0, 10 mM EDTA, 2% triton X-100, 300 mM NaCl) supplemented with 40 µg/ml of RNase A, DNase and protease-free (Thermo) and incubated for 30 min at room temperature, followed by addition of SDS to a concentration of 0.25% and 0.4 mg/ml of proteinase K, recombinant, PCR grade (Thermo) and incubation at 37°C for 30 min. The genomic DNA was then purified by standard phenol:chloroform extraction and ethanol precipitation, and resuspended overnight with 300 µl of TE buffer pH = 8. 10 ug of gDNA were digested with 6 units of MmeI (NEB) for 3 h at 37°C, after which 10 units of Quick CIP (NEB) were added, and the DNA was incubated for another hour at 37°C. The digested DNA was purified using the GeneJET PCR Purification Kit (Thermo) and ligated to a DNA duplex made of HM7071 and HM7039 (Supplementary Table S3) for 16 h at 16°C using 400 units of T4 DNA ligase (NEB). The ligation products were cleaned with AMPure XP Beads (Beckman Coulter), and 900 ng of DNA were PCR amplified using HM7115 and HM7079 as primers for 15 cycles (Supplementary Table S3). The PCR products were cleaned with AMPure XP Beads, and one fourth of the eluted product was PCR amplified using an i7 and an i5 primer from the NEBNext Multiplex Oligos for Illumina (Dual Index Primers Set 2) (NEB) for six cycles. The PCR products were pooled, purified using the GeneJET PCR Purification Kit (Thermo), and submitted for sequencing in a NovaSeq XP Plus (2 × 150 flow cell) at the Vanderbilt Technologies for Advanced Genomics (VANTAGE), obtaining an average of 9.7 million reads per library.

For sequencing analysis, low-quality treads and bases were trimmed with fastp [34], and the resulting reads were aligned to the HM1 genome using bowtie2 [35]. Using a custom R script, all unique alignments that either start or end with TA (the target sequence of the Himar1 transposase) were obtained, and the coordinate of the start of the read was assigned to the corresponding gene or non-coding region. During analysis, an outlier for the rnhC gene was removed from the minus IPTG condition.

Survival assays

A single colony of the indicated strain was inoculated into 2 ml of LB and grown at 37°C with aeration to an OD600 of > 0.5. This culture was then diluted to an OD600 of 0.3, and then 10-fold serially diluted in 1x Spizizen’s Salts (15 mM ammonium sulfate, 80 mM dibasic potassium phosphate, 44 mM monobasic potassium phosphate, 3.4 mM trisodium citrate, and 0.8 mM magnesium sulfate). Around 5 µl of each dilution was plated onto LB plates with the indicated concentration of IPTG and incubated at 30°C overnight. Plates were imaged with a BioRad Gel Doc XR + Molecular Imager, and colonies were counted after either 24 or 48 h of incubation.

Genome copy number analysis

A single colony of the indicated strain was inoculated into 5 ml of LB and grown to an OD600 of > 0.5. This culture was then diluted to an OD600 of 0.05 in LB supplemented with the indicated amount of IPTG and grown to an OD of approximately 0.5. The cultures were then centrifuged, washed with 1X PBS, and frozen at −20°C. The pellets were then resuspended in 1 ml of solution A at 37°C for 30 min. Then, cells were lysed with 1 ml lysis buffer, incubated on ice for 30 min, and sonicated four times at 30% amplitude alternating 10 s of sonication and 10 s of rest and centrifuged at 7000 G and 4°C for 15 min. 0.4 mg/ml of proteinase K was then added, and the samples were incubated at 37 C for 30 min. The genomic DNA was then purified by standard phenol:chloroform extraction and ethanol precipitation, and resuspended overnight with 100 µl of TE buffer pH = 8. 1 µg of DNA was used to make sequencing libraries using the NEBNext Ultra II DNA Library Prep Kit for Illumina (NEB). The libraries were submitted for sequencing in a NovaSeq XP Plus (2 × 150 flow cell) at the Vanderbilt Technologies for Advanced Genomics (VANTAGE), obtaining an average of 4.5 million reads per library. Low-quality treads and bases were trimmed with fastp [34], and the resulting reads were aligned to the appropriate strain using bowtie2 [35].

AddAB purification

Both pHM803 and pHM804 were transformed together into BL21(DE3)pLysS Competent Cells (Agilent), and a single colony was inoculated into 5 ml of LB and grown overnight in LB containing kanamycin and carbenicillin. 2 ml of culture were then inoculated in 1 L of LB containing kanamycin and carbenicillin and grown until an OD600 of 0.3, when 1 mM IPTG was added to the media and the cells were moved to 25°C, were they were grown for an additional 3 h, centrifuged for 15 min at 6000G, resuspended in 10 ml of 50 mM Tris (pH 7.5), 10% sucrose and stored at −80°C. Around 10 ml of CelLytic B cell lysis reagent (Sigma) with 2 µl of Benzonase (Sigma), 10 mM imidazole, and one cOmplete™, EDTA-free Protease Inhibitor Cocktail tablet (Millipore) were added, and the mixture was shaken gently at RT for 10 min. The lysate was centrifuged at 20000G at 4°C for 30 min, the supernatant was mixed with an equal volume of equilibration buffer (20 mM sodium phosphate, pH 7.4, 300 mM sodium chloride, 10 mM imidazole), and run twice through 10 ml of equilibrated HisPur™ Ni-NTA Resin (Thermo) at 4°C. The resin was washed with 50 ml of wash buffer (20 mM sodium phosphate, pH 7.4, 300 mM sodium chloride, 40 mM imidazole) and eluted with 10 ml of elution buffer (20 mM sodium phosphate, pH 7.4, 300 mM sodium chloride, 250 mM imidazole). The protein was dialyzed with a 15 ml Slide-A-Lyzer G3 Dialysis Cassette G2 20 000 MWCO (Thermo) against 10 mM Tris pH 8, 50 mM NaCl, 5% glycerol, 0.1 mM DTT, 0.1 mM EDTA overnight at 4°C and concentrated with two Amicon Ultra-4 Centrifugal Filter Units 100 000 NMWL (Millipore) to a final concentration of 0.9 mg/ml measured by Qubit (Thermo). The catalytic dead mutants were purified in the same way as the WT, to a concentration of 0.7 mg/ml (helicase dead) and 1.4 mg/ml (nuclease dead).

AddAB digestion assay

AddAB digestion was tested in 10 µl reactions containing 1X rCutSmart buffer (NEB), 1 mM ATP (when indicated), 1.5 nM DNA substrate, and 1.5 µM ET SSB (NEB). The reactions were started by adding the indicated concentration of AddAB, and they were incubated at 37°C for 10 min, after which they were stopped with 10 μL of 98% formamide and 10 mM EDTA. DNA was denatured at 95°C for 5 min and run in a 10% Mini-PROTEAN® TBE-Urea Gel, 10 wells, 30 µl (Biorad). Gels were scanned in a ChemiDoc imaging system (BioRad). A ladder with three Cy5-labeled 46, 29, and 20 nt oligos was included in all gels.

AddAB unwinding assay

AddAB unwinding was tested in 10 µl reactions containing 1X rCutSmart buffer (NEB), 1 mM ATP (when indicated), and 1.5 nM DNA substrate. The reactions were started by adding the indicated concentration of AddAB, and they were incubated at 37°C for 10 min, after which glycerol was added to a final concentration of 6.6%. The reactions were immediately run in a native 7.5% polyacrylamide gel.

Results

A screen to find genes required for replication-transcription conflict resolution

Using a transposon mutagenesis screen, we set out to find genes that are involved in the resolution of replication-transcription conflicts in B. subtilis. We generated transposon libraries in cells where we inserted the luxABCDE operon in the head-on orientation to replication, into the amyE locus, under the control of the inducible promoter Pspank(hy). This promoter is turned on by the addition of isopropyl β-D-1-thiogalactopyranoside (IPTG). Conflicts between transcription of the luxABCDE operon and replication lead to cell death in the absence of conflict resolution factors [11]. This method cannot identify essential genes, such as the helicase PcrA, which mitigates replication-transcription conflicts [36].

To generate the transposon library, we used a plasmid harboring a hyperactive mariner-Himar1 transposase and a kanamycin resistance gene flanked by the invert repeats that this transposase recognizes. Because this plasmid has a temperature-sensitive origin of replication in B. subtilis, once the transposon has been mobilized to the chromosome in the permissive temperature, the rest of the plasmids can be eliminated by growing the cells in the restrictive temperature (see Methods for more details) [31, 37]. The transposon library was then grown separately in cultures either without IPTG (conflict off) or with IPTG (conflict on). The cells were then harvested, and the chromosomal regions adjacent to the transposon insertion were sequenced (Fig 1A).

Figure 1.

A transposon mutagenesis screen identifies AddAB as a head-on replication-transcription conflict resolution factor. (A) Schematic of the transposon screen. A whole-genome transposon library was made in cells expressing an inducible gene in the head-on orientation, and this library was grown in parallel in media without (Conflict OFF) or with 1 mM IPTG (Conflict ON), followed by high-throughput sequencing of the transposons. (B) Volcano plot showing the log₂ of the ratio of transposon insertions per gene (Conflict OFF/ON) on the x-axis and the -log₁₀ of the P-value (two-tailed t-test) on the y-axis. The three main hits (addA, addB, and rnhC) are highlighted. (C) Number of transposon insertions per kilobase of DNA found in addA and addB in the conflict OFF and ON conditions. Statistical significance was assessed by two-way ANOVA with multiple comparisons. *P < 0.05. Each replicate represents an independent Tn-seq experiment. (D and E) Location of the transposon insertions found in addA and addB in the conflict OFF and ON conditions.

We determined the number of transposon insertions per coding sequence by aligning the sequence reads to the B. subtilis genome. We obtained an average of approximately 11 transposon insertions per transcription unit. We plotted the resulting data in a volcano plot, with the log₂ of the ratio of transposon insertions per gene without IPTG over the transposon insertions per gene with IPTG on the x-axis, and the -log₁₀ of the P-value (two-tailed, unpaired t-test) on the y-axis (Fig. 1B). Our data revealed three genes that were strongly depleted specifically when the conflict is on: rnhC, addA, and addB (Fig. 1B–D). We previously described that the role of rnhC, which encodes for the RNase HIII enzyme, is essential in head-on conflict resolution [11], confirming that our screen was highly efficient and can identify critical conflict resolution factors efficiently. However, since addA and addB have not been described as crucial conflict resolution factors, we decided to follow up on these hits. These two genes encode the proteins AddA and AddB, which form a heterodimer and can only function when they are in a complex together [38].

The AddAB complex is required for survival during head-on conflicts

To validate the results of our screen, we plated serial dilutions of WT cells and cells with either the addA or addB gene deleted, on plates that either had the conflict off (without IPTG) or on (with 1 mM IPTG). While without induction we observed a small decrease in the viability of the mutant cells (Fig. 2A), inducing the conflict led to a large decrease in survival (>90%), specifically in ΔaddA and ΔaddB cells (Fig. 2A-C). The surviving colonies in the mutants are very small (micro-colonies) and only visible after 48 h of growth, highlighting the importance of these genes for surviving replication-transcription conflicts. Importantly, expression of the WT version of each gene, respectively, from a different chromosomal location (thrC locus) fully rescued the viability of the cells when the conflict was on (Fig. 2A–C). Moreover, cells lacking both addA and addB (ΔaddAB) had a similar loss in viability as the single mutants (Fig. 2A–C). This indicates that, as expected, AddA and AddB work together to promote cell viability in the presence of severe head-on replication-transcription conflicts.

Figure 2.

AddAB is required for cell viability in cells experiencing a severe head-on conflict. (A and B) Serial dilutions of cells of the indicated genotype with the luxABCDE conflict integrated into amyE in the head-on orientation were plated in a plate without IPTG (A) or with 1 mM IPTG (B). (C). Fraction of surviving cells of the indicated genotype with the luxABCDE conflict integrated into amyE in the head-on orientation, calculated as the ratio of the number of colonies when the conflict is ON over the number of colonies when the conflict is OFF. (D and E) Survival assay of cells of the indicated genotype with the lacZ conflict integrated into thrC in the head-on orientation. (F and G) Survival assay of cells of the indicated genotype with the luxABCDE conflict integrated into amyE in the co-directional orientation. Fraction survival was calculated as in Fig. 2C. (H and I) Fraction of reads that map to the indicated coordinates (5 kb window) of the B. subtilis genome, smoothed using a second-order polynomial using 10 neighbors on each side in WT (H) or ΔaddAB (I) cells growing exponentially when the luxABCDE conflict is OFF (no IPTG) or ON (1 mM IPTG).

To control for sequence bias and genomic location, we repeated these experiments with a different conflict gene, lacZ. We integrated the lacZ gene in the head-on orientation and under the control of the same promoter as the luxABCDE conflict, into the thrC locus. We observed a similar phenotype, albeit milder (Fig. 2D-E). This is consistent with previous data indicating that conflicts within the luxABCDE operon are more severe than the lacZ gene [11]. This difference in severity is likely due to the difference in gene length, as the luxABCDE operon is roughly 5.5 kb long, whereas lacZ is 3 kb long, and the longer the gene, the more conflicts can occur. As an additional control, we tested the role of AddAB in the resolution of co-directional conflicts using luxABCDE and observed no difference in viability between WT and ΔaddAB (Fig. 2F-G). This result is expected given that co-directional conflicts are readily resolved and are generally not nearly as severe as head-on conflicts [3]. The absence of a phenotype different from WT with the co-directional conflict is either because we do not have the resolution to detect a role for AddAB at these regions, or that AddAB specifically resolves head-on conflicts.

Head-on conflicts have been shown to completely stall replication in cells lacking conflict resolution factors [11]. Therefore, we used genome copy number analysis to test whether this is also the case for ΔaddAB mutans. To do this, we high-throughput sequenced the genomes from either exponentially growing WT or ΔaddAB cells, containing the engineered luxABCDE operon, under conditions where the gene is either not induced (conflict off) or is induced (conflict on), after which we aligned them to the B. subtilis genome. Because replication initiates at the origin prior to cell division, when cells are in the exponential phase, the sequences surrounding the origin have more reads that align to them, with the number of reads progressively decreasing along each chromosome arm to a minimum that overlaps with the terminus. Moreover, due to multi-fork replication, a phenomenon observed when bacteria grow in rich media, the ratio of reads that align to the origin and reads that align to the terminus is higher than 2 (approximately 4.5 in our case) (Fig 2H-I). In WT cells, when the conflict was on, we observed a small decrease in sequencing reads right after the conflict region, indicating a mild slowdown of the replication fork (Fig. 2H). However, in ΔaddAB cells, upon induction of the conflict, we observed a sharp decrease in the number of reads aligning to the conflict region, which continues to the terminus (Fig. 2I). This indicates that in the absence of AddAB, replication is unable to continue past the conflict region, which by definition will lead to cell death as described above.

The role of AddAB in conflict resolution is independent of R-loop formation

In addition to addA and addB, we had another hit in the screen, rnhC, which codes for RNase HIII. We have established in the past that R-loop processing by RNase HIII underlies its requirement for survival when cells contain a severe head-on replication-transcription conflict [11]. We set out to determine if AddAB and RNase HIII function in the same genetic pathway during conflict resolution. If so, this would suggest that AddAB might be involved in R-loop formation or processing. This experiment required the generation of an addA addB rnhC triple mutant. However, we were unable to make this strain, suggesting that deletion of addAB might be synthetic lethal with rnhC, as is the case with other genes involved in recombinational repair, such as recA, in B. subtilis [39]. Moreover, the functional analog of AddAB in E. coli, RecBCD, is synthetic lethal with the functional analog of RNase HIII, RNase HI [40, 41, 42]. We confirmed this synthetic lethal interaction by using a conditional degron system in which expression of the SspB protein targets any protein with a C-terminus SsrA tag for ClpXP-dependent degradation [33]. We added the SsrA tag to the C-terminus of RNase HIII in a strain that expresses SspB under the control of the IPTG-inducible (Pspank) promoter, which leads to RNase HIII degradation upon expression of sspB (Supplementary Fig. S1A). Consistent with rnhC and addAB being synthetic lethal, we observed no surviving colonies when we turned on the expression of sspB with IPTG, degrading RNase HIII (Supplementary Fig. S1B and C).

To determine whether AddAB and RNase HIII are in the same genetic pathway during conflict resolution, we incorporated the head-on luxABCDE conflict into this system and induced the expression of both the luxABCDE operon and the sspB gene with 100 µM IPTG. Even with reduced luxABCDE expression, we still observed a strong decrease in survival of the ΔaddAB cells that have WT levels of RNase HIII (Supplementary Fig. S1D and E; second column), yet smaller than when using 1 mM IPTG (70% lethality versus >90% in Fig. 2C). When RNase HIII levels are decreased by induction of sspB in otherwise WT cells, we also observed a large yet not complete decrease in viability (Supplementary Fig. S1D-E; third column). These intermediate phenotypes allowed us to observe an additive effect of decreasing RNase HIII levels in ΔaddAB cells, as we did not detect any survival in ΔaddAB cells with decreased levels of RNase HIII and the head-on conflict (Supplementary Fig. S1D and E; fourth column). To confirm that these cells are dying due to the head-on conflict and not due to the synthetic lethality we just described, we used the luxABCDE operon in the co-directional orientation. We observed that ΔaddAB cells with decreased levels of RNase HIII and a co-directional conflict are able to survive in 100 µM IPTG (Supplementary Fig. S1D and E; fifth column), indicating that the reduced levels of RNase HIII protein are enough to sustain viability in ΔaddAB cells in the absence of a head-on conflict. This indicates that conflict resolution by AddAB and RNase HIII occurs in different genetic pathways. Additionally, overexpression of RNase HIII, which removes R-loops at head-on conflict regions [11], does not rescue the viability of ΔaddAB cells upon induction of the conflict (Supplementary Fig. S1F and G). This result also supports that AddAB does not have a role in R-loop formation or processing.

AddAB’s role in conflict resolution is independent of double-strand break repair

AddAB’s best understood function is in recombinational repair, the main double-strand break repair pathway in most bacteria. At the beginning of this pathway, AddAB generates the 3′ ssDNA tails that allow RecA, the main protein that initiates recombinational repair in bacteria, to load onto the DNA and perform its strand exchange function [18]. Because of this, and even though RecA did not come out as a significant hit in the screen (Fig. 3A), we tested the role of RecA in conflict resolution. In agreement with our screen, we observed no difference in survival between WT and ΔrecA cells when the luxABCDE conflict was on (Fig. 3B and C). Moreover, the replication profile of ΔrecA cells with the conflict showed a similar mild replication slowdown as in WT cells, not the complete replication stalling that we observed in ΔaddAB cells (Fig. 3D). This suggests that the role of AddAB in conflict resolution is not related to recombinational repair.

Figure 3.

RecA is not involved in the resolution of head-on replication-transcription conflicts. (A) Number of transposon insertions per kilobase of DNA found in recA in the conflict OFF and ON conditions. Statistical significance was assessed by an unpaired t-test. NS: P > 0.05. Each replicate represents an independent Tn-seq experiment. (B and C) Survival assay of cells of the indicated genotype with the luxABCDE conflict integrated into amyE in the head-on orientation. Fraction survival was calculated as in Fig. 2C. (D) Fraction of reads that map to the indicated coordinates of the B. subtilis genome in ΔrecA cells growing exponentially when the luxABCDE conflict is OFF (no IPTG) or ON (1 mM IPTG).

Because B. subtilis can repair double-strand breaks by non-homologous end joining, we reasoned that perhaps this pathway was masking RecA’s role in conflict resolution. To test this, we introduced the luxABCDE conflict into cells lacking the main proteins known to be involved in this pathway in B. subtilis: Ku, which is involved in DNA-end-binding, and LigD, the double-strand DNA ligase [43], and made double mutants with RecA. We did not observe a decrease in cell viability when the conflict was on in either single or double mutants (Supplementary Fig. S2A-B). This is consistent with previous findings showing that, in B. subtilis, the non-homologous end joining pathway is restricted to germinating spores [43]. These data indicate that these pathways are not required for conflict resolution.

Additionally, RecBCD has been shown to be involved in replication termination, through a mechanism that is independent of RecA and dependent on the nuclease complex SbcCD [44–46, 47]. Moreover, in E. coli, replication-transcription conflicts trigger SbcCD-dependent degradation in RecBCD-deficient cells [48]. To rule out that an analogous mechanism is responsible for AddAB-dependent conflict resolution, we tested the role of SbcC and SbcD in conflict resolution and did not observe any decrease in viability (Supplementary Fig. S2C-D), consistent with the fact that these genes also did not come up in the screen.

The AddAB helicase, but not the nuclease activity, is essential for conflict resolution

To determine the precise mechanism behind AddAB conflict resolution in vivo, we set out to test which AddAB biochemical activities are involved in this process. In vitro, AddAB has three activities: AddA has an ATP-dependent 3′ to 5′ helicase activity and a 3′ exonuclease activity, whereas AddB only harbors a 5′ exonuclease activity [49]. Combined, these activities lead to double-stranded DNA degradation that is initiated by DNA unwinding, as a helicase-deficient AddA is unable to degrade DNA [49]. Upon encountering the B. subtilis Chi site (5′-AGCGG-3′) [50], the 3′ exonuclease activity of AddA is suppressed, rendering a 3′ ssDNA tail as the product of the reaction [49].

We complemented both the ΔaddA and ΔaddB strains with addA and addB separation of function mutants that specifically lack one biochemical activity: AddB^N− (D961A), which lacks 5′ exonuclease activity, AddA^N− (D1172A), which lacks 3′ exonuclease activity, and AddA^H− (K36A), which lacks helicase activity (Fig. 4A) [49]. We then introduced the mutants into cells carrying the luxABCDE head-on conflict. We observed that the 5′ exonuclease activity of AddA is dispensable for conflict resolution, that the 3′ exonuclease activity of AddB is partially required, and that the helicase activity of AddA is absolutely required (Fig. 4B).

Figure 4.

The AddAB helicase activity is essential for head-on conflict resolution. (A) Schematic of AddB and AddA with the nucleases and Walker A (AddA) motifs highlighted. The mutated amino acids to create the separation of function mutants are indicated. (B and C) Survival assay of cells of the indicated genotype with the luxABCDE conflict integrated into amyE in the head-on orientation. Fraction survival was calculated as in Fig. 2C.

This is in contrast to E. coli, as the role of RecBCD in the resolution of replication-transcription conflicts depends on its nuclease activity [6, 51].

Moreover, when we overexpressed the helicase dead mutant in a WT background, we observed a partial decrease in cell viability (Fig. 4C), indicating that this mutant has a dominant negative effect. This is likely because this mutant is able to bind the substrate that AddAB is acting on, blocking the WT enzyme, although the exact mechanism requires further investigation.

AddAB can unwind reversed replication forks

To initiate DNA degradation, AddAB requires a blunt or almost-blunt DNA end, as it is unable to degrade ssDNA that is bound by single-stranded binding protein (SSB) in vitro [19]. The lack of a role for RecA in conflict resolution led us to hypothesize that AddAB might be acting at reversed replication forks, four-way junctions that are formed by annealing of the nascent strands and re-annealing to the parental strands. These structures have been observed in many organisms under conditions that lead to replication stress and provide a double stranded DNA end that AddAB can act on [20]. Moreover, reversed forks have been observed by electron microscopy in B. subtilis cells in which the inducible lacZ conflict had been integrated into the amyE locus in the head-on orientation [52]. Additionally, the genetic requirement for the resection machinery, but not RecA, in head-on conflict resolution has been observed in E. coli, and the formation of reversed replication forks has been proposed to explain this [6]. AddAB could be unwinding the annealed nascent strands, which also allows for the fork to be restored. However, it is unknown if AddAB can perform this reaction.

To test this, we purified B. subtilis WT AddAB as well as helicase dead (AddA^H−B) and nuclease dead (AddA^N−B^N−) (Supplementary Fig. S3A). First, by running the products of the reactions in a denaturing polyacrylamide gel, we confirmed that, as has been previously shown, purified AddAB can digest dsDNA but not ssDNA when SSB is included in the reaction (Supplementary Fig. S3B and C) [19, 49], and that it requires ATP for digestion (Supplementary Fig. S3D) [49]. We also confirmed that no degradation is observed in either the helicase dead or the nuclease dead AddAB (Supplementary Fig. S3E). Running the products on a native gel allowed us to distinguish DNA unwinding by the nuclease dead enzyme from no changes to the substrate in the helicase dead enzyme (S3F). We also confirmed that the unwinding reaction is dependent on ATP (Supplementary Fig. S3G).

We then tested whether AddA^N−B^N− can unwind a reversed fork (Fig 5A). We used this mutant rather than the WT enzyme, so we only detected the helicase activity, and not degradation of the substrate. We observed robust, concentration-dependent unwinding (Fig. 5B and C). When we tested the effect of all three versions of AddAB on the reversed fork substrate, we observed, as expected, degradation by the WT protein, no effect of the helicase dead, and unwinding by the nuclease dead (Supplementary Fig. S3H).

Figure 5.

AddAB can unwind a reversed fork. (A) Schematic of the reaction of reversed fork unwinding by AddA^N−B^N−. (B) Representative 7.5% native polyacrylamide gel with the indicated substrates after incubation at 37°C with AddA^N−B^N− in the presence of ATP for 10 min. The highest concentration of AddAB is 150 nM, and the lanes to the right are at twofold dilutions. (C) Fraction of ssDNA and reversed fork in reactions performed as in B. n = 3 biological replicates, except for the lowest concentration of protein (n = 2).

The combination of our genetic (Fig. 4) and biochemical (Fig. 5) data suggests that AddAB’s role in the restoration of reversed replication forks is to unwind and re-anneal the nascent strands (Fig. 6, left). Such a mechanism would be analogous to Human RECQ1-dependent fork restoration [53], and is different from fork reset [54] by nascent strand degradation observed in E. coli (Fig. 6, right). However, we cannot rule out that the helicase activity of AddAB is required for unwinding the annealed nascent strands, which would allow other ssDNA exonucleases in the cell to digest them.

Figure 6.

Schematic of the proposed model for replication-transcription conflict resolution by replication fork reversal. Upon encountering a highly expressed gene in the head-on orientation, the replication fork reverses, and AddAB/RecBCD is restored by different mechanisms. On the left, the helicase activity of B. subtilis AddAB re-anneals the nascent strands. On the right, the helicase and nuclease activities of E. coli RecBCD cooperate to degrade the nascent strands. Both mechanisms restore replication forks and allow replication to continue.

Discussion

Transcription is a major threat to genome stability [55], and one of the main reasons for this is the fact that replication and transcription, two fundamental cellular processes, co-occur in time and space, leading to conflicts between the two machineries. These conflicts are detrimental to cells in both prokaryotes and eukaryotes. Therefore, cells have pathways that resolve these conflicts to promote genome stability and survival [1].

In this study, we set out to identify genes that are essential for the resolution of replication-transcription conflicts in B. subtilis using an unbiased approach, a whole-genome transposon mutagenesis screen. The results of this screen, which is limited to non-essential genes, identified rnhC, addA, and addB as genes that encode for proteins necessary for conflict resolution (Fig. 1). We previously described the role of rnhC, which encodes for RNase HIII, in promoting survival of cells experiencing a severe head-on conflict [11]. This finding, in essence, served as a positive control for the screening system we constructed. AddAB, on the other hand, had not been previously shown to be essential for conflict resolution. Our follow-up studies confirmed that AddAB is indeed required for survival in cells expressing the luxABCDE operon in the head-on orientation, which creates severe conflicts (Fig. 2A–C). Our genome copy number experiments showed that this is due to replication not being able to progress past the conflict region (Fig. 2H and I).

AddAB is a helicase-nuclease complex best known for carrying out the first step of recombinational repair [18]. Surprisingly, recA was not identified in our screen as an essential conflict resolution factor (Fig. 3A), and we confirmed this in follow-up experiments (Fig. 3B–D). Therefore, the role of AddAB in conflict resolution is unlikely to depend on its function in recombinational repair. Similarly, the other double-strand break repair pathway known in B. subtilis, non-homologous end joining, was also not required for conflict resolution.

When we tested the role of the different biochemical activities of AddAB in promoting cell survival upon induction of the conflict, we observed that the helicase activity of AddAB is absolutely required for survival, yet the nuclease activities are not. The 5′ exonuclease has a partial role, although this activity is dispensable.

In vitro, AddAB requires a blunt or almost blunt DNA end to function, and we reasoned that the only process that can generate this type of DNA end and be repaired independently of recombinational repair or non-homologous end joining is replication fork reversal [6]. Importantly, reversed replication forks have been observed following replication stalling by transcription in B. subtilis cells [52]. Approximately 10% of replication forks are reversed in WT B. subtilis cells experiencing a head-on conflict. This number increases to over 20% in ΔrnhC cells, supporting our conclusion that AddAB and RNase HIII are in different genetic pathways (Supplementary Fig. S1). It is important to point out that the cells used for the experiments, in which reversed forks were observed, contained AddAB, the low number of cells with these structures is expected.

The data in Fig. 4 led us to propose a model in which AddAB’s role in reversed fork restoration is through re-annealing of the nascent strands (Fig. 6, left pathway) and that degradation of the nascent strands might be less favored (Fig. 6, right pathway). Importantly, this reannealing needs to be coupled with elimination of the positive supercoiling that accumulates at the conflict, likely by type II topoisomerases [14]. Alternatively, the helicase activity might be promoting the unwinding of the nascent strands that can then be degraded by other exonucleases in the cell. Interestingly, a recent study observed a similar requirement for only the helicase activity of AddAB and not the exonucleases in the repair of replication forks stalled by a nickase [56], suggesting that reversed fork restoration by AddAB might not be restricted to head-on conflicts.

While the activities of AddAB are well established in vitro [19, 49], it remained unknown if this complex could unwind a reversed fork, leading to restoration of replication forks so they can continue replication. We determined that AddAB can indeed unwind this substrate in vitro, using purified nuclease-dead AddAB, which allowed us to investigate its helicase activity in depth.

Although we have not directly demonstrated that AddAB acts on reversed forks, its functional analog in E. coli, RecBCD, has been shown to act on these structures when they form at replication-transcription conflict regions [6, 21]. This, combined with our data showing that AddAB’s role in conflict resolution is independent of recombinational repair (Fig. 3) and non-homologous end joining (Supplementary Fig. S2), strongly suggests this is the case.

Replication fork reversal is a highly conserved phenomenon. In human cells, reversed forks can be restored by both re-annealing and degradation of the nascent strands [20]. The RECQ1 helicase has been shown to perform the re-annealing while the DNA2 and WRN nucleases degrade the extruded nascent strand duplex and promote replication restart [57]. This indicates that, mechanistically, reversed fork restoration is conserved across the tree of life.

It is interesting to note that, after the reversed fork is reset by AddAB, it must become “immune” to the head-on transcription barrier, so replication is able to proceed. This is likely because another factor removes the RNA polymerases, allowing replication to continue. In E. coli, this is done by the accessory replicative helicases Rep and UvrD [58, 59]. In B. subtilis, the homolog of these helicases, PcrA, facilitates replication through highly transcribed genes [36]. Therefore, we hypothesize that PcrA provides this “immunity.” However, this possibility remains to be formally tested.

In the past, we observed slightly different genetic requirements for the repair of head-on conflicts when using an engineered conflict with the lacZ gene that is constitutively expressed at high levels in the head-on orientation [13]. This is in contrast to our system, in which we can turn the conflict on by inducing transcription in an acute manner. In the constitutively expressed conflict, we only observed a role for AddAB in the absence of RecO. Because this result is in direct conflict with the results in this manuscript, we went back and sequenced the strains from our previous manuscript. We found that only the ΔaddAB strain with the head-on replication-transcription conflict used in that study had a unique 192 base pair insertion 87 nucleotides into the coding sequence of the lacZ gene. This insertion leads to a 96 bp inverted repeat, which likely decreases lacZ transcription levels, and therefore the severity of the conflict. We think this is why we did not see an effect of AddAB by itself in that work.

Interestingly, in the constitutively expressed conflict system, survival is dependent on RecA, and therefore recombinational repair, which is an interesting difference with the system used here, and that we cannot explain without further investigation.

There is functional enrichment for stress response genes in the head-on orientation [60, 61]. In contrast to our previous conflict system mentioned above (constitutive expression), the system used in this study, where we suddenly induce the conflict, is more similar to a cellular response to environmental stress. Many head-on genes are commonly not expressed or expressed at low levels until they are induced in response to changes in the environment [11].

The requirement of the resection machinery for replication-transcription conflict resolution has been observed in E. coli [6, 51, 62, 63]. One of these studies proposed that head-on conflicts lead to replication fork reversal [6]. Another study proposed whole replicon degradation and fork reset [51]. Our data shows that, in B. subtilis, DNA degradation by AddAB is not required for replication fork processing after replication-transcription conflicts, highlighting potentially a key difference in how these two species resolve conflicts or differences between the interpretation of experimental results.

Notably, we have shown in the past that AddAB does not promote mutagenesis due to head-on replication-transcription conflicts [29], suggesting that replication fork reversal and subsequent re-annealing is an error-free mechanism that suppresses DNA repair pathways that promote transcription-dependent mutagenesis, specifically, transcription-coupled repair [64, 65].

Overall, we provide evidence that, upon encountering a highly transcribed gene in the head-on orientation, two nucleic acid structures are formed. On the transcription side, the annealing of the mRNA to its template DNA forms an R-loop [11]. Meanwhile, on the replication side, forks can reverse, creating four-way junctions. It is the resolution of both structures, the R-loop by RNase HIII and the reversed fork by AddAB, that allows replication to continue and cells to overcome head-on replication-transcription conflicts.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

None declared.

Funding

This work was supported by the National Institute of Environmental Health Sciences [grant number: K99ES037493 to JC-G] and Vanderbilt University School of Medicine Department of Biochemistry startup funds to HM.

This research was also supported by the Vanderbilt School of Medicine Basic Sciences, Department of Biochemistry, Destination Biochemistry Advanced Postdoctoral Scholar Award funded by the Armstrong Family to J.C-G. Funding to pay the Open Access publication charges for this article was provided by: Destination Biochemistry Advanced Postdoctoral Scholar Award to J.C-G.

Data availability

All high-throughput sequencing data have been deposited in GEO and made public at the time of publication (Tn-seq data: GSE327290, Genome copy number analysis GSE32729).