Acute hydroxyurea-induced replication blockade results in replisome components disengagement from nascent DNA without causing fork collapse

During S phase, replication forks can encounter several obstacles that lead to fork stalling, which if persistent might result in fork collapse. To avoid this collapse and to preserve the competence to restart, cells have developed mechanisms that maintain fork stability upon replication stress. In this study, we aimed to understand the mechanisms involved in fork stability maintenance in non-transformed human cells by performing an isolation of proteins on nascent DNA-mass spectrometry analysis in hTERT-RPE cells under different replication stress conditions. Our results show that acute hydroxyurea-induced replication blockade causes the accumulation of large amounts of single-stranded DNA at the fork. Remarkably, this results in the disengagement of replisome components from nascent DNA without compromising fork restart. Notably, Cdc45-MCM-GINS helicase maintains its integrity and replisome components remain associated with chromatin upon acute hydroxyurea treatment, whereas replisome stability is lost upon a sustained replication stress that compromises the competence to restart.


Introduction
Several endogenous and exogenous factors can compromise DNA replication dynamics, thereby resulting in the accumulation of replication stress, one of the major sources of genomic instability. Cells respond to this stress by activating several mechanisms to ensure the stabilization of forks, prevent origin firing and cell cycle progression, and, thus, avoid cell division in the presence of under-replicated or damaged DNA [1]. However, extensive replication stress can result in the loss of fork stability and therefore fork collapse [2]. Fork collapse has been used to describe several potentially different processes, including replisome dissociation (loss of replisome stability) and the formation of doublestrand breaks (DSBs). Therefore, fork collapse has usually been understood as a pathological process involving the inactivation of replication forks [2,3]. Moreover, despite the fact that evidence relating to the possible reactivation of collapsed forks has increased in recent years, this reactivation might be mediated by a sub-type of homologous recombination (HR) known as break-induced replication (BIR) which is highly mutagenic [4], thus indicating that the prevention of fork collapse is essential to preserve genome stability.
In response to fork stalling or slowdown, the singlestranded DNA (ssDNA) that accumulates at stalled forks is protected by the heterotrimeric complex RPA. This RPAcoated ssDNA acts as a platform for the recruitment of ATR kinase [5], among other proteins. Once activated, ATR and its downstream kinase Chk1 [6] are believed to maintain the stability of the fork through several mechanisms. For instance, ATR-mediated origin firing inhibition prevents the extensive ssDNA accumulation, which results in RPA exhaustion and global replication fork breakage [7]. Moreover, ATR promotes the association of FANCD2 with MCM helicase to restrain fork progression and prevent ssDNA accumulation [8]. Furthermore, Chk1 protects replication forks from Mus81 endonuclease-mediated fork breakage [9]. In addition, chromatin immunoprecipitation studies on yeast concluded that replisome is disassembled in the absence of checkpoint kinases after exposure to hydroxyurea (HU), thus indicating a possible role for ATR in replisome stabilization [10][11][12]. However, this idea has been challenged by other studies [13,14], so the maintenance of replisome stability during replication stress and the role of ATR in this process remain controversial.
Proteins other than RPA are also involved in ssDNA and replication fork protection. For example, cohesins are thought to participate in fork protection and stability maintenance [15,16]. In addition, Rad51 protects ssDNA from Mre11 nuclease-mediated degradation [17]. Moreover, BRCA2 and FANCD2, among others, contribute to ssDNA protection in an Rad51-dependent manner [18][19][20]. In addition to its role in protecting ssDNA, Rad51 is also involved in HR-independent fork restart [2] and the promotion of fork reversal [21,22].
According to our previous data, the competence to restart is maintained in non-transformed human cells after acute but not sustained HU treatment [23]. To gain a better understanding of the mechanisms involved in fork stability maintenance in non-transformed human cells, we analyzed the HU-induced changes at replication fork level in hTERT-RPE cells under those conditions. Our data show that acute HUinduced replication blockade generates the accumulation of large amounts of ssDNA at parental DNA strand. Strikingly, in contrast to findings in other models [14], our iPOND data show that replisome components are disengaged from nascent DNA under these conditions in non-transformed human hTERT-RPE cells. Interestingly, replisome stability is maintained and forks preserve the ability to restart upon acute HU treatment despite this disengagement. By contrast, sustained replication blockade results in the dissociation of the replisome components and fork collapse.
For S-phase synchronization, cells were cultured in thymidine-containing medium for 22 h, washed with PBS and released into fresh medium for 2 h.

Isolation of proteins on nascent DNA (iPOND)
An adapted version of iPOND that overcomes its efficiency limitation thanks to the use of superparamagnetic beads was performed as described previously [24]. Briefly, EdUlabelled and -treated cells were fixed in 1% PFA for 10 min and quenched with 0.125 mM glycine (pH 7) for 5 min at RT. Cells were harvested, pelleted by centrifugation, and lysed in lysis buffer (ChIP Express kit, Active Motif) for 30 min at 4 °C. Lysates were passed through a 21-gauge needle ten times, and the nuclei were pelleted by centrifugation and then rinsed with PBS supplemented with (Roche) before being subjected to click reaction for 30 min at RT with 0.2 mM biotin azide (Invitrogen). After click reaction, nuclei were pelleted, rinsed with PBS + PIC, resuspended in shearing buffer (ChIP Express kit, Active Motif), and sonicated (Bioruptor, Diagenode) for 15 min at high intensity (30-s/30-s on/off pulses) to obtain 100-300 bp size fragments. Streptavidin-conjugated Dynabeads M-280 (Invitrogen) were washed three times with 1× blocking buffer (1% Triton X-100, 2 mM EDTA (pH 8), 150 mM NaCl, 20 mM Tris-HCl (pH 8), 20 mM beta-glycerol phosphate, and 2 mM sodium orthovanadate, PIC), and then blocked for 1 h at RT with blocking buffer containing 10 mg/mL salmon-sperm DNA (Sigma-Aldrich). Lysates were then incubated with previously blocked Dynabeads (1:10) for 30 min at RT. Finally, beads were washed twice with washing buffer (containing 150 mM NaCl) and twice with highsalt washing buffer (containing 500 mM NaCl) before the proteins were eluted in Laemmli buffer for analysis by either MS or immunoblot.
See also Supplementary Materials and Methods.

Immunoprecipitation and immunoblots
Chromatin fractions were obtained as described in [25] and incubated with 5 μg of anti-MCM3 antibody (homemade) or 5 μg of rabbit immunoglobulin Gs (IgG, Sigma, I8140) for 1 h at 4 °C in rotation. Protein immunocomplexes were pulled down with Pierce ® Protein A Agarose beads (Thermo Scientific, 20333), washed with washing buffer (50 mM Tris pH 7.4, 150 mM NaCl, 1 mM EDTA, and 1% triton), and eluted in Laemmli buffer. Total lysates were harvested in 2% SDS-containing 67 mM Tris-HCl (pH 6.8) buffer. Samples were boiled for 15 min, protein content was measured, and then, samples were normalized and diluted in Laemmli buffer.

Flow cytometry
Cells were synchronized, labelled, and treated as indicated in the corresponding figure legend. After that, cells were collected by trypsinization and fixed in 70% ethanol for at least 2 h at − 20 °C before staining them with BrdU antibody (anti-BrdU, Abcam, ab6326; 1/250). DNA was counterstained with RNAse-containing 1% PI (Sigma-Aldrich).

Chromatin-enriched fraction isolation
Chromatin-enriched fractions were obtained as described in [23], and then, chromatin-enriched fractions were processed and analyzed by immunoblot, as before.

DNA fiber assay
DNA fiber assay was performed in accordance with a protocol described in [2]. Abcam (ab6326; 1/1000) and Becton-Dickinson (347580; 1/200) anti-BrdU antibodies were used for CldU and IdU labelling, respectively. Images were obtained using Leica TCS-SL confocal microscopy with a 63× oil immersion objective, and then analyzed using Fiji software. The number of fibers analyzed in each experiment is indicated in the corresponding figure legend.

Qibc
QIBC experiments were performed as described in [7]. Briefly, cells were pre-extracted with 0.5% Triton X-100-PBS for 1 min at 4 °C and fixed in 4% PFA for 10 min at RT. Cells were then incubated with anti-BrdU (Becton-Dickinson; 1/50), anti-RPA (Homemade; 1/1000), and anti-YH2AX (Millipore; 1/2000) primary antibodies for 1 h at RT and with Alexa Fluor Plus secondary antibodies for a further 1 h at RT. Images were obtained with a motorized Olympus IX-81 wide-field microscope. Automated unbiased image acquisition was carried out with the proprietary Scan Acquisition software.

BrdU immunofluorescence under native conditions
BrdU-labelled and -treated cells were rinsed with PBS, pre-extracted with 0.5% Triton X-100-PBS for 10 min at 4 °C, and then fixed with 3% PFA/2% sucrose-PBS solution for 10 min at RT. After being washed with PBS, cells were blocked for 1 h at RT with 0.05% Tween 20-containing/3% BSA-containing PBS. Cells were then incubated with the anti-BrdU (Becton-Dickinson; 1/50) antibody diluted in blocking solution for 1 h at 37 °C, washed with blocking solution for 15 min at RT, and then incubated with the secondary antibody (anti-mouse 488; 1/500) for 20 min at 37 °C. Finally, cells were washed with blocking solution for 15 min at RT and DNA was counterstained with 0.1 mg/mL RNAse-containing 1% (15 min at 37 °C). BrdU staining was analyzed by confocal microscopy and intensity measured using Image J software. Data from four different experiments were normalized based on the control conditions and statistical analysis (t students) performed with all data grouped.

Agarose gels
DNA was run at 100 V in 1.5% agarose gels to analyze sample sonication. DNA was stained with SYBRTM Safe (Invitrogen) and analyzed with Gel Doc™ EZ Imager (Bio-Rad).

Statistical analysis
All statistical analyses were performed using GraphPad Prism 6 software. Paired or unpaired t test analyses were performed as indicated. Values marked with asterisks are significantly different: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). n.s. was used to indicate an absence of statistical significance.

Analysis of nascent DNA-bound proteins after acute or sustained replication stress in non-transformed human cells
Isolation of proteins on nascent DNA (iPOND) [26] combined with mass spectrometry (MS) [14,27] has been described as a powerful tool to characterize the human replisome and fork-associated proteins. In addition, the development of more efficient versions of iPOND [24] has helped to address its limitation with respect to the need for liberal amounts of starting material and has extended its use to a wide range of cell lines. Thus, to define the changes at replication fork level that compromise fork stability in hTERT-RPE cells, we performed an iPOND-MS experiment that compared acute (2 h) and sustained (14 h) HU treatment (See "Materials and methods" and Supplementary Materials and Methods for details). Importantly, we set up the experimental conditions to ensure that the DNA synthesis was completely inhibited. The following data confirm that DNA synthesis was completely inhibited after 15 min of 10 mM HU treatment: (1) no CldU/IdU incorporation is observed by DNA fiber analysis after this time [23]; (2) no progression through cell cycle is observed by flow cytometry analysis under these conditions (Fig. S1a); (3) no H3 immunoprecipitation is observed by iPOND when EdU is added after 15 min of HU treatment, which indicates that EdU is not incorporated into the DNA (Fig. S1b).
Remarkably, since HU takes 15 min to completely stall the replication forks of hTERT-RPE cells ( Fig. S1b; lanes D and F), EdU was maintained in the media for the first 15 min of HU treatment (Fig. S1c). This prevents the labelled region from being displaced away from replication forks. The EdU incorporated during these first minutes of HU treatment increases the amount of labelled DNA ( Fig. S1b; lanes B and D). Therefore, it was necessary to include a condition in which cells were harvested just after the initial 15 min of EdU + HU treatment (15′ EdU/HU; Fig. S1c). The amount of incorporated EdU will be the same as in 2 h HU and 14 h HU in this case, and thus, this condition is essential to properly analyze the HU-treated samples. The analysis was performed in triplicate in S-phase synchronized cells (to overcome the restriction in the amount of limited starting material) and 3 × 10 7 hTERT-RPE cells were used in each condition.
A total of 716 proteins were identified by MS (Table S1). The pulse/chase enrichment ratios showed that the proteins most enriched at nascent DNA in the pulse condition included DNA polymerase δ, DNA polymerase ε, DNA primase, DNA ligase, RFC1, RFC2, and other replisome components, thereby validating the effectiveness of the proteomic approach taken for replisome component identification (Fig. S1d). To monitor the HU-induced changes, proteomic data were processed (Supplementary Materials and Methods) to generate a heatmap of protein abundance across the experimental conditions and, in addition, proteins were clustered according to their best-known function (Table S2). The 57% of the proteins considered nascent DNA-bound proteins in the pulse (because they present fold enrichments higher than two compared to the chase) had previously identified replication-related functions (Fig. S1e). Furthermore, the 81% of the proteins considered that nascent DNA-bound proteins in the pulse in our proteomic analysis were also classified as nascent DNA-bound proteins in similar studies (Table S3), thus validating the reliability of our approach.

Replisome components are disengaged from nascent DNA upon acute HU treatment in hTERT-RPE cells
As expected, the results of the iPOND-MS experiment showed that most of the replisome components were enriched at nascent DNA in the pulse and 15′EdU/HU conditions (Fig. 1a), indicating that they were directly or indirectly associated to nascent DNA. Strikingly, most of them decreased their association with nascent DNA straight after an acute (2 h) HU treatment that does not compromise competence to restart [23].
Proteomic data were in most cases recapitulated by iPOND combined with western blot (WB) experiments (Fig. 1b). Consistent with the MS results, iPOND-WB experiments showed that MCM6 helicase was displaced away from nascent DNA after acute replication stress. Moreover, the iPOND-WB experiments confirmed that Cdc45, another component of the CMG complex that was not identified in our iPOND-MS experiment, was also displaced away from nascent DNA under this condition (Fig. 1b). According to our iPOND-MS experiments, not only the CMG, but also other replisome components such as those involved in lagging-strand metabolism (DNA Ligase 1, DNA polymerase α, DNA polymerase δ, Fen1, Primase 2) were disengaged from nascent DNA upon 2 h HU treatment (Fig. 1a). Consistently, iPOND-WB experiment showed that the association of Fen1 with nascent DNA was decreased upon HU treatment (Fig. 1b). Notably, in line with our previous data [23], proteins involved in maintaining fork stability and promoting their restart, such as Rad51, FANCD2, and SMC1/3 cohesins [2,[15][16][17]19], were present in nascent DNA after acute but not sustained HU treatment (Fig. 1).

Replisome disengagement from nascent DNA results from HU-induced accumulation of parental ssDNA
The fact that the replisome components were displaced away from nascent DNA prompted us to analyze the way in which forks were remodeled after HU-induced acute replication stress. In human cells, HU-induced replication stress is believed to cause two different types of fork-remodeling events, functional uncoupling of helicases-polymerases and fork reversal, neither of which compromises fork restart [22]. Fork uncoupling would cause the accumulation of large amounts of ssDNA in the parental strand (Fig. S2a); by contrast, fork reversal would generate ssDNA in the nascent DNA strand (Fig. S2a).
To determine which of those remodeling events predominantly occurred upon acute HU treatment in hTERT-RPE cells, we analyzed the accumulation of ssDNA on each DNA strand. The analysis of ssDNA accumulation by native BrdU staining [28] showed a slight increase in nascent ssDNA upon acute HU treatment (Fig. 2a). In line with previous reports [22], this suggests that HU-induced replication stress leads to fork reversal in hTERT-RPE cells. Nonetheless, quantitative image-based cytometry (QIBC) [7] showed a much higher ssDNA accumulation in the parental strand than in the nascent strand under the same conditions (Fig. 2b, c). Moreover, an acute HU treatment was sufficient to induce the accumulation of large amounts of RPA on chromatin (Fig. S2b), which showed a good correlation with the amount of ssDNA detected by native BrdU staining in the parental but not in the nascent strand (Fig.  S2c). Notably, in contrast to parental ssDNA, the amount of nascent ssDNA increased along with the HU treatment and reached its maximum once replication forks showed a clear accumulation of DNA breaks (Figs. 2c, S2d). These results strongly suggest that upon HU treatment, fork reversal is a rare event compared to fork uncoupling in hTERT-RPE cells. However, we cannot discard that the amount of ssDNA generated by each event is different. Hence, to confirm that fork uncoupling is the predominant event upon HU treatment, we analyzed the changes in the amount of parental and nascent ssDNA upon depletion of SMARCAL1 [29,30], FANCM [31], ZRANB3 [32], or FBH1 [33], translocases involved in fork reversal (Figs. 2d, S2e). As suggested by our previous results, this experiment discarded fork reversal as the predominant event upon HU treatment, since the depletion of the different translocases had none or very mild effect on decreasing the amount of nascent and on increasing the amount of parental ssDNA.
Besides fork uncoupling, ssDNA accumulation could also be explained by CtIP-induced fork resection [34]. However, we did not observe CtIP phosphorylation or RPA hyperphosphorylation upon acute HU treatment (Fig. S2F). Likewise, nascent DNA degradation by Mre11 could also result in the accumulation of ssDNA in the parental strand [35]. Nonetheless, fiber analysis of cells treated with mirin (Mre11 inhibitor) [36] demonstrated that nascent DNA was not, at least largely, degraded upon acute HU treatment (Fig. S2G). Taken together, the above results suggest that  and Tables S1, S2, and S3 helicase-polymerase uncoupling is the predominant remodeling event upon acute HU treatment.

hTERT-RPE cells maintain the competence to restart upon acute replication stress in the absence of CDK activity
As replisome components were absent in nascent DNA after acute replication stress (Fig. 1), we wondered whether the previously observed restart [23] was due to the activation of nearby origins, which cannot be distinguished in the fiber analysis due to their proximity to stalled forks [37]. To rule out this possibility, we performed DNA fiber analyses in which we inhibited activation of new origins by adding the CDK inhibitor Roscovitine [38] during the treatment and second labelling period [39][40][41]. The efficacy of Roscovitine was validated by the reduction in the number of new origin firing events (Fig. 3a). Moreover, consistent with a possible role of CDKs in the promotion of normal replication fork progression [42], the length of the IdU (second labelling) tracks was shorter in Roscovitine-treated samples (Fig. S3).
Remarkably, the addition of Roscovitine did not impair competence to restart after acute replication stress (Fig. 3a), thus indicating that restart was not due to newly fired origins. To further confirm these results, we repeated the experiment with Cdc7 inhibitor either alone or in combination with Roscovitine (Fig. 3b). As in the case of Roscovitine, the effect of the Cdc7 inhibitor was validated by the observed decrease in the number of new origin firing events. Notably, consistent with our previous results, none of the drugs impaired the ability to restart after a 2 h HU treatment. PrimPol is a primase/polymerase that is able to re-prime DNA synthesis without the need of a 3′ end, and consequently can synthesize the leading-strand until it reaches the CMG complex [43][44][45]. To test whether this protein was contributing to the restart after a 2 h HU treatment, we tested the effect of its depletion by DNA fiber (Figs. S3b, S3c).
In contrast to what occurs upon other genotoxic stresses, PrimPol does not seem to mediate the restart after a 2 h HU treatment.

CMG helicase maintains its integrity and association with chromatin after acute HU treatment
The fact that replication forks maintain their competence to restart in the absence of CDK activity (Fig. 3) when CMG is not associated with nascent DNA (Figs. 1, S4) suggests that CMG helicase must be displaced away from nascent DNA but still present in chromatin, as the assembly of new CMG complexes is impaired under these conditions [42,46]. Consistent with this, the analysis of chromatin-enriched fractions showed that CMG helicase remains associated with chromatin after acute HU treatment (Fig. 4a). As expected, Roscovitine treatment decreased the association of Cdc45 and GINS with chromatin in untreated cells due to termination events (Fig. 4b). By contrast, CDK inhibition did not reduce their association with chromatin in HU-treated samples (Fig. 4b), which strongly supports the idea that the previously formed CMG helicase complexes remain associated with chromatin under these conditions. Consistent with the ability to restart in the presence of Roscovitine, and the fact that CMG helicase remains bound to chromatin under these conditions, co-IP experiments demonstrated that the integrity of the CMG complex is preserved upon acute HU treatment (Fig. 4c).
Analysis of the chromatin-bound fractions showed that other replisome components, such as those involved in lagging-strand metabolism (DNA polymerase α, DNA polymerase δ, and Fen1), also remained associated with chromatin upon acute HU treatment (Fig. 4a). Consistent with the accumulation of ssDNA observed in Fig. 2, chromatin-bound RPA levels increased upon 2 h HU treatment. In contrast, but in agreement with previous reports [14], association of PCNA with chromatin decreased straight after a 2 h HU treatment. Notably, most of the replisome components, including the CMG complex, decreased their association with chromatin upon a sustained HU treatment that compromises the ability to restart [23] (Fig. 4a). In line with this, Tipin and Claspin, which are involved in fork stability maintenance as part of the replication pausing complex (RPC) [47], maintained their association with chromatin after a 2 h but not that much after a 14 h HU treatment. Likewise, Rad51, FANCD2, and SMC1/3 cohesins also decreased their association with chromatin only after a sustained replication stress (Fig. S4c).
Collectively, our results indicate that acute HU treatment does not challenge CMG integrity or its association with chromatin in hTERT-RPE cells, and suggest that fork restart is mediated by recycling previously formed CMG complexes.

Replication resumption after acute HU treatment does not result in the accumulation of under-replicated regions in hTERT-RPE cells
The above results indicate that, in line with previous reports [22], forks can be remodeled without compromising their ability to restart. Moreover, our data suggest that forks resume replication after acute HU treatment using the same CMG complexes that had been disengaged from nascent DNA. To define the dynamics of replication resumption under these conditions, we analyzed ssDNA disappearance upon release from HU treatment by QIBC (Fig. 5a). The amount of parental ssDNA decreased almost completely after a 30 min release from HU treatment. Consistent with this, chromatin-bound RPA levels were similar to the control  S5a). DNA replication under stress conditions can result in the accumulation of under-replicated regions, which leads to the presence of 53BP1 bodies in the subsequent G1 phase [48][49][50]. Thus, we wondered whether replication resumption after replisome disengagement from nascent DNA upon acute replication stress can also lead to the accumulation of under-replicated regions. Remarkably, the percentage of G1 cells presenting 53BP1 bodies did not significantly increased upon release from acute HU treatment (Fig. 5b). Notably, consistent with the previously observed competence to restart after acute HU treatment, this was not due to changes in the number of cells arriving at G1 (Fig. S5b). Together, these results support the idea that replisome disengagement from nascent DNA upon acute HU treatment does not challenge replication or compromise genome integrity.

Discussion
The maintenance of genome integrity is essential for the existence of all organisms, and DNA replication is one of the events in which the genome is most vulnerable. Replication forks stall frequently as a result of encounters between the replication machinery and different obstacles in the DNA template or a lack of basic elements to replicate the DNA, such as nucleotides. While preventing the restart of stalled forks may negatively influence the proliferative capacity of cells and runs the risk of genomic information being lost, their restart may also lead to the acquisition of mutations, since some restart pathways are error prone [51][52][53]. In this regard, the mechanisms involved in maintaining the ability to restart and their regulation remain unclear. Likewise, the fate of the replisome machinery during replication stress is still controversial [10][11][12][13]. To clarify previous concerns, we decided to perform a comparative iPOND-MS analysis of the proteins associated with replication forks under these conditions, while taking into account our previous data showing that the ability to restart is compromised in non-transformed human cells after sustained but not acute replication stress [23].
Interestingly, using an improved version of iPOND [24], we show here that replisome components of non-transformed human hTERT-RPE cells, including the CMG helicase, are disengaged from nascent DNA but still present in chromatin after acute replication stress (Fig. 5c). Notably, fork stability-related proteins such as Rad51, FANCD2, and SMC1/3 [2,[15][16][17]19] increase their association with nascent DNA under these conditions, thus indicating that replisome components are being specifically displaced. These results, together with the fact that cells maintain the competence to restart under these conditions, indicate that replisome disengagement from nascent DNA does not result in fork collapse. It has previously been reported by Dungrawala et al. [14] that most of the replisome components are stably bound to nascent DNA, and only decrease due to termination events upon 3 mM HU treatment in HEK293T cells. This discrepancy could be related to the different experimental setting or the use of a different cellular model. On one hand, they use a dose of HU that does not fully block EdU incorporation, and thus, EdU is maintained in the media during the treatment, thereby making it possible to label newly fired forks. By contrast, in our experiment, we use a dose of HU that completely blocks EdU incorporation after 15 min of treatment [23] (Fig. S1b). On the other hand, our analysis was performed in non-transformed human cells, while theirs was carried out in HEK293T tumor cells. As already noted by Dungrawala et al. [14], the replication rate in untreated conditions is threefold slower in HEK293T than in hTERT-RPE cells, which would affect the speed of unwinding and, therefore, the helicase-polymerase uncoupling. Thus, we cannot exclude that the observed discrepancy might also be due to differences between cell types.
Remarkably, the association of the replisome components with chromatin is in most cases only lost after a sustained replication stress that compromises the ability to restart. The fact that replisome components are bound to chromatin after acute HU treatment in the presence of Roscovitine, which prevents the activation of new origins [39][40][41], rules out the possibility of new origin firing masking their loss. Our data also show that the integrity of the CMG complex is maintained upon acute HU treatment; this, together with the fact that replication forks are able to restart under these conditions even in the presence of Roscovitine and/or Cdc7 inhibitor, in which the formation of new complexes must be abrogated [42,46], supports the idea that CMG helicase The fact that replisome components are disengaged from nascent DNA but still present in chromatin upon acute replication stress implies that replication forks might have been remodeled. It has previously been reported that HU can cause two different types of fork-remodeling events that do not compromise fork restart: fork reversal and fork uncoupling [22]. In this regard, our results show that there is ssDNA exposed in the nascent strand upon 2 h HU treatment [33], which strongly suggests the presence of reversed forks upon acute replication stress in hTERT-RPE cells. However, the increase in chromatinbound RPA levels upon acute replication stress cannot be explained by reversed forks. In addition, depletion of translocases known to be involved in fork reversal had only a mild effect in the amount of parental and nascent ssDNA exposed upon 2 h HU treatment. Moreover, the presence of high amounts of ssDNA in the parental DNA strand rather suggests that most of the forks are uncoupled. Consistent with this, there is a strong correlation between chromatin-bound RPA levels and the ssDNA in the parental strand. Accumulation of parental ssDNA could also be due to resection. Nonetheless, the amount of ssDNA present in the parental strand is not time-dependent and no clear increase in resection marks such as P-CtIP and P-RPA is observed. Likewise, Mre11-dependent DNA degradation may also cause ssDNA accumulation in the parental strand [35], but, in contrast to sustained HU treatment, we show that acute treatment does not cause Mre11dependent DNA degradation in hTERT-RPE cells. Taken together, these data suggest that fork uncoupling is the predominant remodeling event upon HU-induced acute replication stress in hTERT-RPE cells. Functional helicase-polymerase uncoupling is believed to occur in response to a variety of genotoxic stresses in several species, including humans [22,54]. In fact, the ssDNA that accumulates during this process is thought to be required to induce the activation of the checkpoint response [55]. In this regard, it has generally been considered that uncoupling might cause an extended loop of ssDNA without compromising the interaction between nascent DNA and the replisome. However, our data show that replisome components are disengaged from nascent DNA during this process (Fig. 5c). Notably, this disengagement does not compromise fork restart, thereby suggesting that replication forks are much more dynamic than initially expected. Interestingly, a recent study using an in vitro prokaryotic DNA replication system [56] shows that leading-and lagging-strand machineries work uncoupled from each other. Their data show that, at least in prokaryotes, leading-and lagging-strand DNA polymerases function independently within a single replisome. Notably, the stochastic function of each polymerase results in its transient disengagement from the replicative helicase without blocking replication fork progression, thus demonstrating the high plasticity of replication forks.
Our data clearly show that, after acute replication stress, eukaryotic replication forks can be restarted despite the disengagement of replisome components from nascent DNA. The lagging-strand per se has a discontinuous back-stitching  DNA synthesis mechanism and could, therefore, easily be reinitiated by DNA polymerase α. By contrast, leadingstrand DNA synthesis occurs in a continuous manner, thereby challenging its restart. In this regard, one possibility to restart leading-strand synthesis would be to reinitiated DNA synthesis from the site where the replisome is situated. This would generate long gaps of under-replicated ssDNA behind the fork that could be repaired by Rad51mediated mechanisms in G2 [57,58]. In agreement with this, it has been reported that HU causes the accumulation of The arrows indicate double-positive cells. The average percentage of 53BP1 foci (> 6) and cyclin D1 (CycD1) double-positive cells relative to total G1 (cyclin D1-positive) cells is shown (right panel). Error bars represent standard deviation (paired t test, n = 3). More than 1000 cells were analyzed in each condition. c The current model proposes two types of remodeling events upon 2 h HU treatment: functional helicase/polymerase uncoupling and fork reversal. Our data show that an open replication fork, where replisome components are disengage from nascent DNA, is the predominant remodeling event upon 2 h HU treatment. In addition, our data also show that replication forks maintain the competence to restart after this remodeling. > stands for "more than"; < stands for "minus than". See also Fig. S5 post-replicative ssDNA gaps in hTERT-RPE cells that could be reflecting DNA synthesis restart at a different distance from the original block [22]; however, we have also shown that the bulk of ssDNA disappears rapidly once replication is reinitiated. Another possibility to resume leading-strand DNA synthesis would be to re-engage DNA polymerase ε with the 3′ end of the nascent strand. DNA polymerase ɛ is strongly associated with CMG [59], and thus, in order to engage it with the nascent 3′ terminus, a linker molecule might be needed to bring the replisome close to the DNA end. Alternatively, the soluble store of DNA polymerase δ could also be engaged to the 3′ end of the nascent strand and carry on the DNA synthesis until it reaches the slowmoving CMG complex, when it would be dissociated by collision release and DNA synthesis would be continued by the DNA polymerase ε [60]. Finally, similarly to what occurs upon UV light-induced damage, re-priming of PrimPol or DNA polymerase α could also promote DNA synthesis until the end terminus of the nascent DNA is engaged with the replisome. In this case, the re-priming events could generate small gaps behind the fork that would be filled by other polymerases or repaired by HR [43][44][45]61]. Nonetheless, our results indicate that at least PrimPol does not seem to be involved in the restart after a 2 h HU treatment. Regardless of the mechanism, it is relevant that reinitiation occurs after the disengagement of replisome components from nascent DNA, and that this does not seem to compromise genomic instability, since an increase in 53BP1 foci-containing cells in G1 is not observed. As previously mentioned, during the last years, increasing evidences support the idea of replication forks being much more plastic than what one could initially expect. Recent studies have shown, for instance, that replication forks can reverse without compromising their ability to restart [22] or that, at least in vitro, leading-and lagging-strand DNA polymerases function independently within a single replisome [56]. In line with this, we show here that replication forks of non-transformed hTERT-RPE cells are more plastic and resilient than expected, and that they can restart from a situation involving disengagement of replisome components from nascent DNA, apparently without inducing genomic instability.