Replication Fork Reversal Occurs Spontaneously after Digestion but Is Constrained in Supercoiled Domains*

Replication fork reversal was investigated in undigested and linearized replication intermediates of bacterial DNA plasmids containing a stalled fork. Two-dimensional agarose gel electrophoresis, a branch migration and extrusion assay, electron microscopy, and DNA-psoralen cross-linking were used to show that extensive replication fork reversal and extrusion of the nascent-nascent duplex occurs spontaneously after DNA nicking and restriction enzyme digestion but that fork retreat is severely limited in covalently closed supercoiled domains.

Replication fork reversal was investigated in undigested and linearized replication intermediates of bacterial DNA plasmids containing a stalled fork. Two-dimensional agarose gel electrophoresis, a branch migration and extrusion assay, electron microscopy, and DNA-psoralen cross-linking were used to show that extensive replication fork reversal and extrusion of the nascent-nascent duplex occurs spontaneously after DNA nicking and restriction enzyme digestion but that fork retreat is severely limited in covalently closed supercoiled domains.
Replication fork reversal (RFR) 2 defines a condition that occurs in vivo when a replication fork encounters an obstacle such as a DNA lesion. In these cases, it is thought that the replication fork halts and retreats to impede its collapse (1,2). The nascent strands separate from their corresponding parentals and anneal to each other to form a fourth arm. The resulting structure resembles a Holliday junction that behaves as a substrate for recombination enzymes (3)(4)(5)(6). It was also found that for covalently closed circles of bacterial plasmids exposed to moderate concentrations of intercalating agents in vitro, whereas unreplicated forms acquire positive (ϩ) supercoiling after all their native negative (Ϫ) supercoiling was removed, partially replicated forms appear unable to acquire (ϩ) supercoiling and keep the same electrophoretic mobility as their nicked counterparts (7,8). These results were interpreted as an indication that partially replicated plasmids containing a fork are unable to acquire (ϩ) supercoiling as all of it is adsorbed by RFR. Curiously, in this case, RFR does not affect the electrophoretic mobility of RIs as in the presence of intercalating agents, these CCRIs show the same electrophoretic mobility as their nicked counterparts (7,8).
Two-dimensional agarose gel electrophoresis (9) is increasingly used to analyze linearized replication intermediates (RIs) isolated from cells that have been exposed to different types of DNA-damaging agents (10 -13). In the autoradiograms of these two-dimensional gels, the identification of a diffused pattern named "cone signal" led some authors to claim that the molecules responsible for this pattern contained reversed forks formed in vivo (13). In fact, the detection of such a pattern is currently viewed as a direct evidence for the occurrence of RFR in vivo (10 -15). However, whether or not the cone signal detected in repair-deficient mutants is due to RIs containing reversed forks that formed in vivo is uncertain as an identical pattern previously described as a "triangular smear" was also detected in two-dimensional gels of DNA isolated from undamaged wild-type cells of a number of different species and interpreted as indicative for delocalized termination of DNA replication (16 -26). An important drawback of all these studies is the random location of the event under study, namely a DNA lesion or the site where two forks growing in opposite directions meet. To overcome this potential problem, here we studied RFR in RIs containing a fork stalled at a specific site (7). pBR18-TerE@StyI and pBR18-TerE@AatII ( Fig. 1) contain the Escherichia coli polar replication terminator TerE (27,28) cloned at different distances from the unidirectional ColE1 replication origin in pBR18 (29). Digestion of the RIs of these plasmids with restriction enzymes that cut inside as well as inside and outside the replication bubble generates double-and simple-Ys, respectively (Fig. 1). The RIs of plasmids where the fork stalls at a specific site accumulate, generating a prominent signal in two-dimensional gel autoradiograms that can be readily distinguished on top of the simple-or double-Y patterns (7, 30 -32). Analysis of these plasmids undigested and after restriction enzyme digestion revealed that RFR readily forms in vitro but that fork retreat is severely limited in covalently closed domains (33). Extensive retreat and total extrusion of the nascent-nascent duplex, on the other hand, occurs spontaneously after nicking or restriction enzyme digestion.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Culture Medium-The E. coli strain used in this study was DH5␣FЈ. Competent cells were transformed with monomeric forms of the plasmids as described before (7,29). Cells were grown in LB medium containing 100 g/ml ampicillin at 37°C. Plasmid DNA isolation was performed as described previously (7,29).
Two-dimensional Agarose Gel Electrophoresis and Southern Transfer-The first dimension was in a 0.4% agarose gel in Trisborate-EDTA buffer at 1 V/cm at room temperature for 22 h. The agarose lane containing the DNA/HindIII marker was excised, stained with 0.3 g/ml EthBr, and photographed. During this period, the agarose lane containing the DNA sample was kept in the dark. The second dimension was in 1% agarose gel in Tris-borate-EDTA buffer containing 0.3 g/ml EthBr run perpendicular with respect to the first dimension. The dissolved agarose was poured around the excised agarose lane from the first dimension, and electrophoresis was at 5 V/cm in a 4°C cold chamber for 10 h. Southern transfer was performed as described before (7,29).
Psoralen Cross-linking-To perform psoralen cross-linking, 50 -100 ng of DNA in a total volume of 20 l was incubated with 10 g/ml 4,5Ј,8 trimethylpsoralen (Sigma) for 1 h at room temperature in the dark in a 96-well open plate and subsequently irradiated with a 500-watt high pressure mercury lamp (model TQ 700; Original Hanau) on ice in a open plastic dish for 15 min. The lamp was placed 7 cm above the plastic dish, and the light was filtered through a Pyrex glass to eliminate radiation below 300 nm. Psoralen stock solution was prepared in 100% ethanol. This procedure was performed either before or after DNA digestion. Subsequently, the two-dimensional agarose gel electrophoresis was carried out as described before.
Non-radioactive Hybridization-pBR322 that only hybridizes to the plasmid was used as a probe. DNA was labeled with the random primer fluorescein kit (PerkinElmer Life Sciences). Membranes were prehybridized in a 20-ml prehybridization solution (2ϫ saline/sodium phosphate/EDTA, 0.5% Blotto, 1% SDS, 10% dextran sulfate, and 0.5 mg/ml sonicated and denaturated salmon sperm DNA) at 65°C for 4 -6 h. Labeled DNA was added, and hybridization lasted for 12-16 h. Hybridized membranes were sequentially washed with 2ϫ SSC and 0.1% SDS, 0.5ϫ SSC and 0.1% SDS, and 0.1ϫ SSC and 0.1% SDS for 15 min each at room temperature except for the last wash, which took place at 65°C. Detection was performed with an antifluorescein-AP conjugate and CDP-Star (PerkinElmer Life Sciences) according to the instructions provided by the manufacturer.
Preparation of DNA Samples Enriched for Specific RIs-Specific molecules were isolated from agarose gels following the procedure described by Olavarrieta et al. (7) with minor modifications. After restriction digestion, DNA isolated from exponentially growing cells was analyzed in a one-dimensional agarose gel, the lane was cut and incubated with 0.1 M NaCl in TNE buffer (10 mM Tris-HCl, pH 8.0, 0.1 mM EDTA, 100 mM NaCl) at 65°C for 4 h, and the selected DNA sample was electroeluted out of the agarose gel and resuspended in distilled water.
Electron Microscopy-The purified DNA sample was spread on EM grids under non-denaturating conditions in redistilled water by the benzyldimethyl-alkyl ammonium chloride (BAC) method (34).
Branch Migration and Extrusion Assay-The agarose lane of the first dimension containing the DNA sample was incubated with 0.1 M NaCl in TNE buffer (10 mM Tris-HCl, pH 8.0, 0.1 mM EDTA, 100 mM NaCl) at 65°C for 4 h without 0.3 g/ml EthBr or in the presence of 0.3 g/ml EthBr. Subsequently, the second dimension was performed as described before.

RESULTS AND DISCUSSION
It is generally thought that RFR is repressed in (Ϫ) supercoiled molecules and favored by (ϩ) supercoiling (7,8,35). It was recently shown, however, that formation of Holliday-like junctions at both forks of a replication bubble creates a topological constraint that prevents further regression of the forks (33). To confirm this observation for a different plasmid, we used a modification of two-dimensional gels where the agarose lane containing the DNA that came out from the first dimension was heated at 65°C in the presence of 0.1 M NaCl for 4 h before the second dimension took place (33,36). This condition favors branch migration and extrusion of the fourth arm of Holliday junctions in vitro (37,38). The corresponding autoradiograms are shown in Fig. 2 together with interpretative diagrams. Detection of two novel signals (Fig. 2, marked in the diagrams with black arrows) and their electrophoretic mobility during the second dimension clearly indicated that in some, although not all, DNA molecules, heating caused extrusion of  (7): the ColE1 unidirectional origin, the E. coli terminator sequence there, and the ampicillin resistance gene. B, the terminator utilization substance (TUS) binds Ter sites, and the Ter-TUS complex acts as a polar replication fork barrier (27,48). In consequence, blockage of the replication fork at TerE leads to the accumulation of a specific RI containing an internal bubble and a total mass 1.26 times the mass of unreplicated plasmids. C, digestion of this RI containing a stalled fork with AflIII generates a double-Y of 1.26ϫ. D, map of pBR18-TerE@AatII showing the relative position of its most relevant features. E, in this case, blockage of the replication fork at TerE leads to the accumulation of a specific RI containing a larger internal bubble and a total mass of 1.60ϫ. F, digestion of this RI with AflIII generates a double-Y of 1.60ϫ, whereas the larger fragment resulting from its double digestion with AflIII and PvuI corresponds to a simple-Y of 1.81ϫ (G). Finally, double digestion of the same RI with PstI and EcoRI generates a simple-Y of 1.70ϫ (H).
the two nascent strands (nascent-nascent duplex). During the second dimension, the new molecular species migrated as open circles and linear forms of 2627 bp, precisely the distance between the ColE1 origin and the TerE site in pBR18-TerE@AatII, indicating that the new linear fragment corresponded in fact to the extruded double-stranded fourth arm. In the very same autoradiograms, however, no extrusion occurred for CCRIs. Therefore, we concluded that extensive branch migration and extrusion of the fourth arm was impeded in CCRIs regardless of whether the DNA was (Ϫ) or (ϩ) supercoiled. This was unexpected as it is generally thought that (ϩ) supercoiling actually favors RFR and complete extrusion of the nascent-nascent duplex (8,35). This observation, on the other hand, agrees with the finding that RIs are able to acquire electrophoretic mobility and become (ϩ) supercoiled when exposed to very high (0.3 g/ml and above) concentrations of EthBr (7,33), suggesting that RFR is favored by low and moderate levels of (ϩ) supercoiling but is inhibited when the torsional stress reaches certain threshold. Postow et al. (8) used atomic force microscopy to study the topology of RIs containing stalled forks in the presence of 5 M EthBr (equivalent to 1.97 g/ml). They noticed that under these conditions, RIs become heavily supercoiled but interpreted that this supercoiling was an artifact induced during deposition of the molecules onto mica. It was later shown, however, that RIs recover electrophoretic mobility and are able to acquire (ϩ) supercoiling when exposed to 0.3 g/ml and higher concentrations of EthBr (7) due to the topological locking mechanism activated as soon as RFR forms at both forks of a replication bubble (33).
The observation that extensive RFR and complete extrusion of the nascent-nascent duplex are prevented in (Ϫ) as well as (ϩ) supercoiled RIs prompted us to reinvestigate whether or not the DNA molecules containing reversed forks that are putatively responsible for the cone signal identified in two-dimensional gels indeed form in vivo (10 -13). To this end, we combined two-dimensional gels with psoralen cross-linking, the branch migration and extrusion assay described above, and EM. If RFR occurs in vivo, the signal should be detected in restriction fragments of all sizes regardless of the extent of replication (Fig. 3). Moreover, if RFR forms and retreats unconstrained once a replication fork stalls at a DNA lesion and these lesions occur at random, the mass of the RI with the stalled fork would vary between 1.0 and 2.0ϫ. As clearly depicted in Fig. 3, the signal expected for a mix of molecules where the fork stalls and retreats from different sites is not the cone depicted in Fig. 3B as a gray triangle but rather a smear covering the whole area limited by X-shaped recombinants to the left, and the ascending portion of the simple-Y pattern to the right, as also painted in Fig. 3D in gray. We used two-dimensional gels to examine restriction fragments of different sizes (4.3, 4.4, 3.1, and 3.6 kb, respectively) where the RIs containing stalled forks were double-Ys of 1.26ϫ (1.26 times the mass of unreplicated fragments) for pBR18-TerE@StyI digested with AflIII and 1.60ϫ for pBR18-TerE@AatII also digested with AflIII. We also examined simple-Ys of 1.81ϫ for pBR18-TerE@AatII digested with AflIII and PvuI and 1.70ϫ for the same plasmid digested with PstI and EcoRI (Fig. 1, circular  and linear maps). The spikes emanating from the prominent spot on top of the simple-or double-Y arcs (Fig. 4, represented in red in the diagrams) were easily recognized, although their location, intensity, and extension varied from one gel to the other. Moreover, they did not necessarily fit into the so-called cone signal described elsewhere (10 -13). This spike extended almost exclusively below the accumulated spot for the 1.26ϫ RI, and it also extended below the accumulated spot but showed a small bulge above it for the 1.60ϫ RI and extended both above and below the accumulated spot for the 1.81ϫ RI. Similar signals have been observed for RIs of specific masses in other systems as well (15,39). It is important to note, however, that only a small fraction of the accumulated RIs experienced RFR (Fig.  4). The strength of the spot generated by accumulated RIs indicated that most of them were pretty stable, and a discrete signal for molecules that experienced RFR was detected only for RIs that accumulated due to fork stalling. To confirm that these signals were generated by molecules displaying reversed forks, the agarose lane containing the DNA that came out of the first dimension was heated at 65°C in the presence of 0.1 M NaCl for 4 h before the second dimension took place. The results obtained are shown in Fig. 4, middle vertical panel, with corresponding interpretative diagrams. Note that heating between the first and second dimensions eliminated the original diagonal signals and generated novel ones that ran perpendicular to the first dimension in all cases (this was remarkable for the 1.81ϫ RI) and are represented in Fig. 4 in red in the middle vertical diagrams. It seems likely that the bulk of materials that run at the position of the TerE-stalled RIs contained molecules where the fork retreated to some extent but where the nascent-nascent duplexes were too small to affect their mobility in an appreciable manner. These molecules were nevertheless susceptible to branch FIGURE 3. Two-dimensional gel patterns elicited by X-shaped recombinants and replication intermediates containing reversed forks. Panel I, X-shaped recombination intermediates of two identical molecules where a single recombination event occurred at different sites from one end to the other. One molecule is represented in red, and the other is represented in blue. Panels II-IV, replication intermediates where the fork stalled and retreated to different extents after the molecules have replicated 81% in panel II, 60% in panel III, and 26% in panel IV. Parental strands are depicted in black, and nascent strands are depicted in red. The green bar represents the replication block. Molecules are drawn in two different ways for better comparison with X-shaped recombinants. Below, the "classical" two-dimensional gel patterns (A), the cone signal (B), the signals expected for the populations illustrated in panels I-IV above (C), and the smeared signal expected for a mixture of populations where the replication fork stalled and retreated from randomly located sites (D) are shown. The linear and X-shaped recombinant patterns are depicted in black, the simple-Y pattern is depicted in red, the bubble pattern is depicted in light blue, and the double-Y pattern is depicted in green. migration and extrusion by heat. This interpretation would account for both the new vertical spike, if not all such molecules completely extruded in response to heat, as well as the new spots. We speculated that molecules where the replication fork stalled at TerE and have undergone various extents of RFR, which in turn altered both their first and their second dimension mobilities, generated the original diagonal spikes. As DNA heating was performed after the first dimension was completed, it could not alter first dimension mobilities, but after total extrusion of the nascent-nascent duplexes, the molecules that originally gave rise to the diagonal spikes would now yield a horizontal bulge that should be detected aside of the new signals. In fact, this was clearly the case for both of the two new spots observed, particularly in the case of the 1.81ϫ sample (Fig. 4, indicated with blue arrows in the middle bottom diagram).
For all the fragments studied, the identification of prominent spots that migrated together with the accumulated RI during the first dimension and as linear molecules with the same electrophoretic mobility of unreplicated forms during the second dimension indicated that these molecules resulted from complete extrusion of the fourth arm. Moreover, in the autoradiogram corresponding to the 1.81ϫ RI, a second smaller spot was detected that corresponded to the extruded nascent-nascent linear duplex of 2567 bp. This linear duplex was not detected for the other two fragments examined due to their smaller size. To further confirm that these vertical signals were indeed generated by molecules containing reversed forks, a DNA sample enriched for the 1.60ϫ molecules (Fig. 4, encircled by a black dotted line in the corresponding middle autoradiogram) was prepared and examined at the EM (shown in Fig. 4, at the far right panel). As this sample included molecules of different FIGURE 4. RIs containing a fork stalled at a specific site display a spike signal in two-dimensional gels that corresponds to molecules containing reversed forks. Autoradiograms of two-dimensional gels corresponding to three different fragments where the RI that accumulated was 1.26ϫ, 1.60ϫ, and 1.81ϫ are shown to the far left, with corresponding interpretative diagrams to their right. In these diagrams, the spikes emanating from the accumulated RI are depicted in red and indicated by blue arrows. The signal depicted as a dotted black arc corresponds to broken RIs (49), whereas those depicted in gray correspond to linears and X-shaped recombinants (9,50,51). Autoradiograms of two-dimensional gels corresponding to the same fragments where the DNA was exposed to 65°C in the presence of 0.1 M NaCl between the first dimensions (1st dim) and the second dimensions are shown in the middle, with corresponding interpretative diagrams to their right. A DNA sample enriched for the molecules responsible for the vertical signal encircled with a dotted line in the autoradiogram corresponding to the 1.60ϫ was prepared and examined at the EM. For preparation of these enriched DNA samples, see Ref. 7. This DNA was spread on EM grids under non-denaturing conditions in redistilled water by the BAC method (52). Bar ϭ 0.2 m. A, a simple-Y showing no reversed fork. B-D, RIs containing reversed forks. Interpretative diagrams are shown to the right of each electron micrograph. In these diagrams, numbers indicate the percentage figure represented by each arm. The parental duplex is indicated in blue and green, whereas the nascent strands are depicted in red.
sources (plasmid as well as chromosomal) and the extent of fork retreat was expected to vary significantly from molecule to molecule, the criterion employed to select those molecules that experienced RFR was the following one. Molecules containing a reversed fork are only those where two arms are identical and the increase in length of each of the other two compensate the progressive loss in length of the first pair (Fig. 3). This criterion allowed us to select a number of molecules where the fourth arm was readily identified (Fig. 4, right panel, indicated by black arrows in the interpretative diagrams) and confirmed that these were indeed RIs where the replication fork had retreated to different extents (7,40,53).
Once we confirmed that molecules containing reversed forks generated the vertical signal emanating from the accumulated spots, we investigated whether RFR occurred at the stalled fork in vivo or in vitro. To this end, we used two different and complementary approaches. First, pBR18-TerE@AatII was digested with PstI and EcoRI to generate RIs of 1.70ϫ containing no replication fork stalled at TerE. In this case, the fork of the RI that accumulated corresponded to the ColE1 unidirectional origin (Fig. 1, D and H). Analysis of these RIs by two-dimensional agarose gel electrophoresis revealed a spike and a pattern that were almost identical to the one shown in Fig. 4 at the bottom left (data not shown). This observation suggested that RFR was likely to occur after digestion in vitro.
To test this hypothesis, we used psoralen cross-linking to prevent any further branch migration (41). DNA molecules corresponding to the 1.81ϫ sample (Figs. 1 and 5) were crosslinked with psoralen either after or before DNA digestion and analyzed in two-dimensional gels. Notice that the pattern corresponding to the sample cross-linked with psoralen after DNA digestion (Fig. 5B) was almost identical to the one generated by the untreated sample (Fig. 5A) except that the spike generated by molecules containing reversed forks was now decorated with a number of regularly distributed extra spots. Psoralen intercalation and cross-linking do not occur in a uniform fashion (41). The extra spots might well correspond to sites where psoralen cross-linking occurred in a preferential mode. Branch migration could still shift between cross-links but could not move across them. Surprisingly, no signal for molecules containing reversed forks was detected when psoralen cross-linking took place before DNA digestion (Fig. 5C). In addition, in this case, each of the spots was duplicated. It is well known that the amount of intercalating agents is reduced almost by half in nicked as compared with covalently closed circles (42). The detection of doublets in this autoradiogram was a consequence of this phenomenon. For each doublet, the spot showing slower electrophoretic mobility derived from molecules that were covalently closed at the time of psoralen cross-linking and, in consequence, captured almost double the amount of psoralen (43,44).
It is important to note, however, that our results do not indicate that RFR does not occur in vivo. It has been well established that in E. coli, once a replication fork hits a DNA lesion, RecA (45) and/or RecG (4) promotes regression of the stalled fork, generating a Holliday-like junction (HLJ) that is subsequently processed by the RuvABC complex to allow replication restart by PriA (3,46,47). This retreat of the forks that occurs in vivo, however, is severely constrained by DNA supercoiling and probably cannot extend very long (33). The results we showed here question to what extent the cone signal detected in twodimensional gels (10 -13) reflects the limited retreat of the forks that may occur in vivo as opposed to the extensive branch migration that takes place in vitro after restriction enzyme FIGURE 5. Replication fork reversal occurs spontaneously but only after restriction enzyme digestion. Autoradiograms of two-dimensional gels where the RI that accumulated was 1.81ϫ are shown together with their corresponding interpretative diagrams to the right. A, an untreated sample is shown on top. The DNA sample was cross-linked with psoralen either after (B) or before restriction enzyme digestion (C). Notice that no spike was detected when psoralen cross-linking occurred before restriction enzyme digestion. Arrows at the bottom right indicate the signals expected for linear molecules of 0.80ϫ (2567 bp).
digestion. Furthermore, delocalized termination of DNA replication, which also generates a triangular smear (16 -26), could be enhanced by fork stalling at DNA lesions and might certainly contribute to the so-called cone signal.
In summary, here we showed that extensive RFR and extrusion of the fourth arm occurs spontaneously but only after nicking or DNA restriction enzyme digestion. These results strengthen the observation that the extent of fork retreat is severely constrained in supercoiled domains, probably due to the topological locking that triggers when RFR forms at both forks of a replication bubble (33).