Replication fork collapse at a protein-DNA roadblock leads to fork reversal, promoted by the RecQ helicase

There are numerous impediments that DNA replication can encounter while copying a genome, including the many proteins that bind DNA. Collapse of the replication fork at a protein roadblock must be dealt with to enable replication to eventually restart; failure to do so efficiently leads to mutation or cell death. Several prospective models have been proposed that process a stalled or collapsed replication fork. This study shows that replication fork reversal (RFR) is the preferred pathway for dealing with a collapsed fork in Escherichia coli, along with exonuclease activity that digests the two nascent DNA strands. RFR moves the Y-shaped replication fork DNA away from the site of the blockage and generates a four-way DNA structure, the Holliday junction (HJ). Direct endo-nuclease activity at the replication fork is either slow or does not occur. The protein that had the greatest effect on HJ processing/RFR was found to be the RecQ helicase. RecG and RuvABC both played a lesser role, but did affect the HJ produced: mutations in these known HJ processing enzymes produced longer-lasting HJ intermediates, and delayed replication restart. The SOS response is not induced by the protein-DNA roadblock under these conditions and so does not affect fork processing. Author Summary To transfer genetic material to progeny, a cell must replicate its DNA accurately and completely. If a cell does not respond appropriately to inhibitors of the DNA replication process, genetic mutation and cell death will occur. Previous works have shown that protein-DNA complexes are the greatest source of replication fork stalling and collapse in bacteria. This work examines how the cell deals with replication fork collapse at a persistent protein blockage, at a specific locus on the chromosome of Escherichia coli. Cells were found to process the DNA at the replication fork, moving the branch point away from the site of blockage by replication fork reversal and exonuclease activity. Our data indicate that it is the RecQ helicase that has the main controlling role in this process, and not the proteins RecG and RuvABC, as currently understood. RecQ homologs have been shown to be involved in replication fork processing in eukaryotes and their mutation predisposes humans to genome instability and cancer. Our findings suggest that RecQ proteins could play more important role in replication fork reversal than previously understood, and that this role could be conserved across domains.


22
Abstract 23 There are numerous impediments that DNA replication can encounter while copying a genome, including the 24 many proteins that bind DNA. Collapse of the replication fork at a protein roadblock must be dealt with to enable 25 replication to eventually restart; failure to do so efficiently leads to mutation or cell death. Several prospective 26 models have been proposed that process a stalled or collapsed replication fork. This study shows that replication 27 fork reversal (RFR) is the preferred pathway for dealing with a collapsed fork in Escherichia coli, along with 28 exonuclease activity that digests the two nascent DNA strands. RFR moves the Y-shaped replication fork DNA 29 away from the site of the blockage and generates a four-way DNA structure, the Holliday junction (HJ). Direct 30 endo-nuclease activity at the replication fork is either slow or does not occur. The protein that had the greatest 31 effect on HJ processing/RFR was found to be the RecQ helicase. RecG and RuvABC both played a lesser role, but 32 did affect the HJ produced: mutations in these known HJ processing enzymes produced longer-lasting HJ 33 intermediates, and delayed replication restart. The SOS response is not induced by the protein-DNA roadblock 34 under these conditions and so does not affect fork processing. During DNA replication, the replication machinery (replisome) can arrest due to impediments on the DNA such 51 as lesions or nucleoprotein blockages. Removal of bound proteins that the replisome itself fails to displace can 52 be carried out by accessory helicases: in E. coli these are Rep, UvrD and/or DinG (1, 2). However, even with the 53 full complement of these helicases, protein roadblocks are still found to be the most common obstacle to 54 replisome progression, especially RNA polymerase (3). Encounter with a protein roadblock can lead to the 55 dissociation of the replisome, and the frequency with which this happens is indicated by the central role of 56 PriA/PriC restart pathways in bacterial cell viability (4). If the blockage is not removed, DNA replication is not 57 able to continue to completion and the cell will not survive. The processing of these stalled forks is likely to be 58 relatively frequent with most or all replisomes predicted to stall during the cell cycle (5, 6). Bacterial replisome 59 dissociation has recently been reported to be occur at a frequency of about 5 events per replisome, per cell cycle 60 (7).

61
The DNA at a replication fork that has stalled upon encounter with a protein block can be processed by a number 62 of possible pathways. Endonucleases can cut the forked DNA, producing a double strand break, followed by 63 homologous recombination that restores DNA integrity (8). Alternatively, exonucleases can act to degrade the 64 nascent leading and lagging strands, moving the Y-shaped branch point away from the site of blockage (9).

65
Finally, replication fork reversal (RFR) can occur, whereby the leading and lagging nascent strands separate from 66 their respective template strands and anneal to each other, concurrent with the two template strands also re-67 annealing (Reviewed in (10)). This leads to the formation of a four-way DNA structure called a Holliday junction 68 (HJ) that is the substrate for proteins in homologous recombination pathways. This HJ has one arm that has free 69 DNA ends that itself can be acted upon by exonucleases whilst the other 3 arms are continuous DNA.

70
DNA breakage or the recruitment of RecA to ssDNA at the stalled fork may lead to SOS induction (11). However, 71 the SOS response is only expected to be a major influence in incidences of prolonged DNA damage, high levels 72 of unresolved ssDNA and when the recombination pathways are ineffective at processing blocked forks 73 (reviewed in (12)). RecA acts in the SOS response as a co-protease to cleave the LexA repressor, inducing SOS 74 (reviewed in (13)). The RecA protein has been proposed to play a role in RFR, employing its strand exchange 75 capacity; this is most readily explained as a RecA filament, bound to ssDNA on the lagging strand template of a 4 76 replication fork, which catalyses invasion into the leading strand duplex, displacing the nascent strand and 77 forming a reversed fork (14). RecA

109
One of the major impediments to studying RFR in vivo has been the combination of a large chromosome and a 110 rapid replisome: E. coli contains a genome of ~4.6 Mbp and the replisome moves at a rate close to 1kb/s (4). A 111 system has been developed whereby a site-specific road block to DNA replication can be induced at a known 112 position in the chromosome using a transcriptional repressor (TetR) bound to an array of operator sites within 113 the E. coli chromosome (44, 45). Addition of a temperature sensitive allele of the replicative helicase (DnaBts) 114 allows the rapid inactivation of the replisome across a population of cells, and the replication forks are processed 115 by the cell within 5 minutes (44). Furthermore, the disappearance of the forked DNA structure from the array 116 region coincided with the visualisation of HJs upstream of the array suggesting RFR had taken place in a sizeable 117 proportion of cells (44). The disappearance of Y-shaped DNA will be used here as a definition of replication fork 118 collapse; whether or not this is accompanied by partial or complete dissociation of the replisome is unknown.

119
The regression of the replication fork away from the site of DNA damage or protein block allows repair proteins 120 or accessory helicases to access the DNA and resolve the problem. The upstream HJ that is formed may be 121 processed in a recombination-dependent or independent manner to restore a replication fork (reviewed by 122 (10)). If the blockage is removed, replication is able to restart using the reformed fork structure onto which 123 replisome reloading takes place (45), most likely in a PriA-dependent manner. Indeed, replication was observed 124 to restart and proceed through the array in the vast majority of cells within 5 minutes of the addition of release 125 of the block by addition of anhydrotetracycline.

126
In this report a site-specific replication block was established and characterised at the lac locus, approximately 127 midway round the right replichore on the E. coli chromosome. A dnaBts allele was introduced to be able to 128 trigger replication fork collapse by shifting to a non-permissive temperature. This site-specific replication 129 blockage system was used to examine the process of RFR, and the relative contributions of candidate proteins.

130
RFR is seen to occur as a major pathway used in cells to deal with a persistent roadblock to replication, with little 131 or no evidence for endonuclease action at the fork. Some exonuclease activity is seen as well, that could either 6 132 be from action directly at the Y-shaped fork, or at the HJ produced by RFR. Deletion of recQ is seen to have the 133 greatest effect on RFR with the process being highly inefficient in its absence, while both ruvABC and recG play 134 a more minor role in RFR. However, the data suggest that RecG is involved in migrating the HJ formed by RFR 135 back into a forked DNA structure that can be used for replication restart; in the absence of RecG HJs are much 136 more prominent and are seen to persist. The induction of the SOS response was only observed to begin after 4 137 hours of replication fork blockage, implying that it does not affect fork processing during the standard 138 experimental timeframe used here.

156
To assess whether the established replication block could be removed to allow the resumption of replication, 7 159 the two copies had segregated from each other (Fig. 1A). Given the position of the array, the 10 minute AT 160 exposure does not allow time for a newly formed fork to start from oriC and progress to the array position. Thus, 161 the multiple foci are most likely produced exclusively from restarted forks. The viability of these cells was 162 restored to initial untreated levels, confirming that replication was able to restart throughout the population 163 (Fig. 1B). These results confirm that the tetO array, incorporated at lac, is an effective blockage to replication 164 when the repressor is overproduced, and that the roadblock is reversible with addition of AT. 165 dnaBts was incorporated into the tetO array-carrying strain and assessed in comparison to a WT strain for its 166 ability to block and restart replication (Fig. 1C, 1D). It has recently been proposed that the replication fork 167 present at the TetR-YFP roadblock is not stable but has a half-life of 3-5 minutes (44). An equilibrium exists 168 between forks that are stalled at the block and ones that have collapsed and are being processed and will 169 subsequently restart leading to their collision with the replication roadblock once again. DnaBts is utilised here

177
Following validation of the replication block in both strains, the effect of a shift to 42°C, a non-permissive 178 temperature for DnaBts, was investigated. Cells were grown at 30°C and replication was blocked by 179 overproduction of TetR-YFP as before. Each strain was then transferred to 42°C for one hour, to inactivate DnaB 180 in the temperature sensitive strain. The ability of DNA replication to restart in each strain was then compared 181 by transferring cells back to the permissive temperature of 30°C for 10 minutes, with or without the addition of 182 AT. In the WT strain a similar pattern was seen to cells grown only at 30°C; the majority of the population had a 183 single fluorescent focus following the temperature shifts indicating replication was still being efficiently blocked, 184 and addition of AT led to ~70% of cells showing two or more foci within 10 minutes of addition of AT 185 demonstrating effective replication restart (Fig. 1A). Similarly, viability was almost fully recovered by addition of 186 AT. Therefore, the temperature shift to 42°C had little noticeable effect upon the replication block or replication 8 187 restart in the WT strain. In the dnaBts strain the 1 hour at 42°C did not affect the integrity of the replication 188 blockage as judged by the high proportion of cells with a single focus and the drop in viability when arabinose is 189 present (Fig. 1C, 1D). Upon addition of AT when the cells were returned to 30°C viability was also seen to recover 190 to WT levels. This demonstrates that transient inactivation of DnaBts is well tolerated and cells recover fully 191 when returned to a permissive temperature and the replication blockage is removed. However, the proportion 192 of cells with 2 or more foci after 10 minutes of AT treatment is much lower than seen with WT. This likely reflects 193 the extra time this strain requires to restart replication that may involve the refolding of inactive DnaBts protein 194 or the novel synthesis of the protein. It has been shown that the majority of locations on the E. coli chromosome, 195 including lac, require 7-10 minutes to visibly segregate from the sister DNA following replication (46, 47). Thus, 196 even a short delay in restarting replication in the dnaBts strain may be sufficient to prevent the replicated sister 197 duplexes from separating from each other, within the resolution limit of the microscope.

198
The DNA structures present within each strain during the replication block experiment were visualised using 2-199 D neutral-neutral agarose gels (Fig. 2). DNA from each condition was extracted, digested and electrophoresed.

200
A radiolabelled probe was used to detect either a 5.8 kb array region including 0.6 kb upstream from the 201 beginning of the array, or a 3.3 kb region from 0.5 kb to 3.8 kb upstream of the array ( determined that ~60% of DNA was seen to be Y-shaped at 30°C; ~70% was as a Y-structure after the hour at 42°C 208 ( Fig. 2B). The region upstream of the array showed that upon replication blockage a clear signal for HJ and Ys 209 could be seen (Fig. 2C), and the levels of these signals did not vary with changes in temperature in the WT strain, 210 suggesting that the replication forks maintain a relatively constant fork turnover rate.

211
In contrast, the dnaBts strain displayed a vast decrease from ~50% to 12% in forked DNA signal at the array 212 following the shift to 42°C ( Fig. 2A and B).

219
When the dnaBts cells were returned to the permissive temperature of 30°C for 10 minutes the forked DNA 220 signal at the array increased above the level seen at 42°C, consistent with restoration of DnaB function, restart 221 of replication and subsequent collision with the block (Fig. 2A). However, the amount of Y-shaped DNA was 222 lower than the level seen prior to the shift to 42°C, possibly due to the time taken to reactivate/re-synthesise 223 DnaB and then to restart the replication fork. This is in agreement with the microscopy data shown in Fig. 1.

225
It was possible that the SOS response may have been activated in these cells after prolonged replication fork 226 stalling events (~2 hours), as used in this study. In order to address this concern, the timing and extent of SOS 10 242 S2). 4 hours after induction of the replication fork block, each strain still had a single focus across the population 243 confirming replication could proceed through the array, and cells had begun to elongate ( Fig.3 and Fig. S2). At 244 this time point, some cells showed delocalised mCherry fluorescence by microscopy, and the population showed 245 a slight increase in red fluorescence detected by the plate-reader (2250 RFU in WT cells, slightly lower than uvrD 246 by 250 RFU, Fig. 3 and Fig. S2). This indicated that the SOS response was active in some cells. By 6 hours, the 247 majority of the cell population were seen to display a red fluorescence signal, and correspondingly an RFU of 248 7300 was detected using the plate reader (similar to uvrD). The fluorescence detected at 24 hours post fork-249 blockage had increased dramatically (22,000 RFU) though plasmid loss and cell death were observed alongside 250 cells with increased mCherry intensity (Fig. 3). This result was similar to the uvrD 24-hour sample (Fig. S2). As 277 Y-shaped DNA remained from the 50% seen initially (Fig. 4B). The DNA upstream was mostly linear with a low 278 level of Y and HJ DNA, and remained so over the hour (Fig. S1, Fig. 2C).

279
Following the same procedure to block DNA replication in dnaBts recG cells, a strong Y-shaped forked DNA signal 280 was observed indicative of the replication roadblock, but at a significantly lower amount than dnaBts (Fig. 4A   281 and B). Upstream of the array a strong HJ signal is visible, and is reproducibly stronger than in other genetic 282 backgrounds (Fig. 4C). At the permissive temperature for DnaBts, the steady turnover of forks and their 283 processing and subsequent restart should be occurring, yet without RecG there is a marked accumulation of 284 unresolved HJ signal, indicating a role for RecG in the processing of these HJs. Shifting the dnaBts recG cells to 285 42°C resulted in a reduction in Y-DNA signal within the array region as fork-collapse occurs to an extent similar 286 to that seen in WT (Fig. 4A); after 15 minutes, 20% of the signal was present as forked DNA in the array, which 287 further reduced to 13% by an hour (Fig. 4B), levels that are slightly higher than seen for dnaBts. This was 288 accompanied by the loss of HJ and Y-shaped signals upstream over the 42°C time-course (Fig. 4C), but this 289 occurred more slowly than in dnaBts where by 60 minutes almost all the signal had disappeared (Fig. S1). It is 290 plausible that the slightly reduced level of Y-shaped DNA seen at the array upon replication blockage is a result 291 of a subtle alteration in the equilibrium between blocked replication forks at the array, and the collapsed forks 292 undergoing processing. The higher level of HJ seen may reflect that RFR takes longer to be resolved in a recG 293 mutant with DNA existing as a HJ for a longer time, even in the presence of RuvABC. It is noteworthy that the 294 position of the HJ signal in the 2-D gels is that of a so-called "X spike", which is distinctive of a HJ made from two 295 dsDNAs of full length. If degradation on one arm had occurred, or if the HJ was formed by homologous 12 296 recombination using a broken strand within this region, then it would migrate in the "cone signal" area and not 297 as seen. The HJ is, therefore, distinctive of a RFR event.

298
The dnaBts ruvABC mutant produced a replication block at a similar level to dnaBts at 30°C (Fig. 4A). Inducing 299 fork collapse at 42°C led to a rapid drop in Y-DNA signal as seen in dnaBts (Fig. 4B). However, after an hour at 300 42°C a slightly higher percentage of DNA remained as a fork structure, 20% compared to 12% for dnaBts. The 301 prominent HJ seen upstream in the recG mutant was not present in dnaBts ruvABC and the non-linear DNA 302 structures that were present were largely processed in the course of the hour at 42°C, similar to dnaBts (Fig. 4C; 303 S1). It is noteworthy that this processing of the upstream DNA appeared efficient even in the absence of RuvABC.

304
The greatest persistence of Y-shaped DNA following a shift to 42°C was seen in the dnaBts recQ mutant. Upon 305 establishment of a replication block at 30°C, 65% of DNA was Y-shaped, a significantly larger proportion than in 306 dnaBts ( Fig. 4A and B). After 15 minutes at 42°C the DNA remained largely in the forked-DNA structure (Fig. 4A).

307
Quantitatively, forked DNA now made up 44% of the total DNA, signifying that only a third of the initial Y-shaped 308 DNA had been processed. The fork signal persisted even after an hour at 42°C, with 33% of the total DNA still 309 being in a Y-structure ( Fig. 4A and B) i.e. roughly half the initial Y-shaped DNA remained after 1 hour, compared 310 to 23% of initial forks in dnaBts (Fig. 2B). Upstream of the replication block processing intermediates were 311 present at all times; this could reflect that processing of the upstream DNA was also slower in recQ mutants, or 312 that there was a constant low level of processing of the Y-shaped structures to generate more upstream signals 313 that continued over the full hour (Fig. 4C). The slowed processing of the stalled fork could explain why a higher 314 proportion of Y-structured DNA was initially seen at the array: it is the result of an alteration of the equilibrium 315 between fork collapse/processing and restart similar to what was proposed for the RecG results above (Fig. 2B).

316
The majority of processing of the Y-shaped DNA in these experiments occurs by RFR rather than direct nuclease 317 cleavage of the Y-shaped DNA, as RecQ has no nuclease function. Following inactivation of DnaBts it has been 318 seen that both ExoI and RecJ can contribute to degradation of the nascent DNA strands (50), and RecJ activity 319 has been shown to be stimulated by RecQ activity that provides a suitable substrate (51). Therefore, inactivation 320 of RecQ may inhibit processing that leads both to RFR and exonuclease digestion to yield Y-shaped structures 321 upstream.

322
These findings demonstrated that recG, ruvABC and recQ mutants exhibit a diminished ability to reverse 323 replication forks, implying that all three proteins are involved in RFR. It also implies that very little direct 324 endonuclease digestion of the Y-shaped DNA occurs, as it is difficult to envisage this being restored to a stable 325 Y-structure at the non-permissive temperature for DnaBts. However, the most extensive deficiency to RFR was 326 seen in dnaBts recQ indicating that RecQ is key to the majority of RFR under these conditions.

327
RecG and RuvABC act in a distinct pathway from RecQ to process stalled forks

328
To determine whether RecG, RuvABC and RecQ act synergistically to perform RFR, we investigated strains with 329 combinations of these gene knockouts. If these proteins act independently, the absence of multiple proteins 330 should be an additive effect, and the signal indicative of forked DNA that accumulated in the array region during 331 induction of the replication blockage will persist at a higher proportion than in any of the single mutants alone 332 upon DnaBts inactivation.

333
In the dnaBts recGrecQ mutant, the proportion of forked-DNA at the array was significantly higher than in a 334 dnaBts mutant; it remained at a high intensity in the array region 15 minutes after the shift to 42°C when it 335 reduced to 37% from 64% ( Fig. 5A and B). The intensity of this Y-DNA signal was unchanged at the 60-minute 336 time point (37%, Fig. 5A and B). Therefore, 57% of the initial forks within the array region failed to be processed 337 in the dnaBts recGrecQ mutant even after one hour, marginally more impaired than a recQ mutant. The HJs and

342
Analysis of the dnaBts recQruvABC strain bore a close resemblance to that of dnaBts recGrecQ and dnaBts recQ.

343
Once shifted to 42°C the intensity of the forked DNA signal remained relatively high at both the 15-and 60-344 minute marks (Fig. 5A). 43% of the total DNA remained forked at the block site after 15 minutes with a small 345 decrease to 35% by 60 minutes. Therefore, of the original Y-shaped signal 54% remained at 60 minutes (Fig. 5B).

346
Upstream DNA signals were also akin to those seen in the dnaBts recQrecG mutant, as they were maintained at 347 a high intensity upstream of the block once fork collapse was initiated and observed at 15 minutes, but became 348 fainter after an hour (Fig. 5C). The similarity in the signals seen in both recGrecQ and recQruvABC could mean 14 349 that that RecG and RuvABC contribute equally to RFR, or could act in the same pathway together. However, 350 RecQ's contribution to RFR appears to far outweigh either RecG or RuvABC.

351
When dnaBts recGruvABC cells were grown at 30°C and the replication roadblock was induced, ~45% of the total 352 DNA was seen to be Y-shaped. This percentage is similar to that seen with dnaBts recG (46%), and is lower than  (Fig. 6). Additionally, the signal patterns detected upstream over the hour of fork collapse in the triple 366 mutant were consistent with the single and double mutants, where the non-linear DNA signals were strong at 0 367 and 15 minutes but reduced by an hour (Fig. 6). This again verified that RecQ contributes greatly to RFR.

369
It has been previously established that the TetR-YFP replication roadblock can be relieved with the addition of 370 AT, resulting in the restart of replication (Fig. 1). The addition of AT for 10 minutes following 2 hours of replication 371 roadblock restored viability to levels seen in isogenic cells that never had arabinose to induce replication stalling. 372 2-D gel analysis of this replication resumption in dnaBts saw the disappearance of forked DNA (Fig. 7), leaving 373 solely linear DNA, in both the upstream and array regions after 10 minutes AT exposure, which correlates with 374 the restoration of viability (Fig. 1). The absence of forked DNA meant that all the intermediates previously seen 375 (Fig. 2) had been processed within this 10 minute period. The most likely explanation is that replication restart 15 376 had occurred across the population of cells. Importantly, once the cells had undergone the temperature shift to 377 42°C for one hour and then back to 30°C to allow re-establishment of the replisome, again only linear DNA 378 resided in the 2-D gels after 10 minutes AT exposure (Fig. 7A) (compare to signals in Fig. 2 where AT was not 379 added). As previously noted (Fig. 1) the counting of cell numbers with multiple foci for dnaBts revealed that 380 there was a slower release of the block under these conditions, however, this delay wasn't seen in the 2-D gels.

381
This suggests that the replication forks have restarted within 10 minutes and the array has been copied, but 382 there has not been sufficient time for the two daughter copies to segregate from each other.

383
The dnaBts recG strain produced extremely prominent HJ and Y-shaped signals upstream of the replication 384 roadblock (Fig. 4C). The ability of these cells to restart replication was examined by removal of the TetR block by 385 addition of AT for 10 minutes. It was seen that substantial levels of Y-shaped DNA and HJ were present both at 386 30°C and after temperature shift to 42°C (1 hour) and back to 30°C, in both the array and upstream regions (Fig.   387 7A). However, the former spot on the Y-arc, indicative of replication blockage, is now absent showing successful 388 release of the protein roadblock. The absence of RecG left a proportion of HJs and forks that were yet to be 389 processed and/or migrated in order to restart replication in a timely manner. Despite the inability of the cells to 390 resolve these intermediates within the 10 minute window, they were able to eventually recover their viability 391 upon release of the block (Fig. 7B), indicating that this was a delay rather than a failure to restart DNA replication.

392
To ensure the DNA structures visible in the 2-D gels were not an artefact of the presence of DnaBts, the same 393 assay was performed on a recG strain with WT DnaB (Fig. S3). The resulting 2-D gel showed similar signals as for 394 the dnaBts equivalent.

395
The addition of AT to the dnaBts ruvABC mutant resulted in barely visible HJ and Y-arc signals upstream of the 396 array, both at 30°C and after temperature shifting. In the array region, the HJ and Y-arc signal were faint, but 397 clearly present, though at a lower level than in the equivalent recG mutant (Fig. 7A). This may also suggest a 398 slight delay in replication restart, and longer persistence of HJ intermediates in the absence of RuvABC. Overall, 399 the majority of HJs were able to be processed or resolved in the absence of RuvABC (compare Fig. 4 to Fig. 7).

400
Following addition of AT to allow replication to proceed through the array, the number of foci within each cell 401 was similar to that seen in dnaBts; 72% of cells had two or more foci at 30°C but following the temperature shift 402 to 42°C only 21% showed multiple foci in the presence of AT (Fig. 7C). Cell viability was also completely restored 403 by 10 minutes treatment with AT (Fig. 7B).

404
Linear DNA was the only DNA structure visualised by 2-D gels in both the array and upstream regions of a dnaBts 405 recQ mutant following the release of the roadblock via AT addition at 30°C (Fig. 7A). Replication restart was also 406 seen to be similar to dnaBts in terms of cell viability and the number of foci observed per cell as replication 407 restarted ( Fig. 7B and C). However, following the temperature shift, lingering intermediate signals remained in 408 both regions (Fig. 7A). Cell viability was seen to be fully restored in the dnaBts recQ following addition of AT (Fig.   409 7B) suggesting replication restart does eventually occur across the population. Furthermore, the double and 410 triple mutants did not affect replication restart, as their viabilities were shown to be restored to levels not 411 significantly different from untreated (Fig. S4).

466
RuvAB has been shown to be able to regress a fork in vitro to form a HJ, but it does so with low efficiency, 467 preferring to unwind DNA in the opposite direction from that required to form a HJ (60

488
RFR is only one possible mechanism to deal with a collapsed replication fork; another proposal is that direct 489 endonuclease action on an arrested fork can lead to breakage of one arm at the fork, which can then be repaired

507
The other possible processing event at the collapsed replication fork is the action of exonucleases that digest 508 the two nascent DNA strands, an activity that has been previously reported when DnaBts is inactivated and is 509 attributed to RecJ and ExoI (50

537
Although the effect of deletion of recG upon RFR was modest, it was clear that in the absence of RecG a much 538 stronger HJ signal was seen in the region upstream of the replication roadblock. This implicates RecG in the 539 timely processing of HJs. The simplest and least recombinogenic mechanism to deal with a HJ produced by RFR 540 is to branch migrate the junction back into a Y-shaped fork, upon which the replisome can be reloaded, and this 541 may be a role played by RecG. recG mutants also showed an accumulation of HJ structures during replication 21 542 restart ( Fig. 7 and Fig. S3) suggesting that HJs persist even after the replication roadblock has been removed in 543 these cells, although the viability data suggests that the delay to recovery is not fatal and is eventually overcome.

544
The deletion of ruvABC was also seen to increase the level and persistence of HJs in the upstream DNA region, 545 consistent with the known role of the complex in branch migration and resolution of HJs.

546
It is noteworthy that HJs are seen to be present upstream of the replication roadblock in both recGruvABC and 547 recQrecGruvABC backgrounds, and that the levels of these intermediates decline over time, along with a 548 reduction in Y-shaped DNA at the block. There is clearly another process able to either migrate the HJs out of 549 the region being probed or to resolve the HJ. As a result of this finding, we concluded that neither RecG nor 550 RuvABC are absolutely required for HJs to be processed. RusA, the only other known endogenous HJ resolving 551 enzyme in E. coli, is absent from the strains used in this study. Spontaneous branch migration of the HJ out of 552 the region being examined could explain these results, but subsequent recovery of replication would then also 553 depend upon the branch migration occurring to regenerate the Y-shaped structure to allow replisome re-554 loading. An alternative pathway that has been proposed is the action of RecBCD (Fig. 8)

569
The proposed model (Fig. 8) has some testable predictions. The establishment and all gene knockout strains were created by P1 transduction and confirmed by PCR (Table S1 for list of strains).

600
The sulAp-mCherry construct, created by fusing the sulA promotor to mCherry, was inserted into the