Histone deposition promotes recombination-dependent replication at arrested forks

Replication stress poses a serious threat to genome stability. Recombination-Dependent-Replication (RDR) promotes DNA synthesis resumption from arrested forks. Despite the identification of chromatin restoration pathways after DNA repair, crosstalk coupling RDR and chromatin assembly is largely unexplored. The fission yeast Chromatin Assembly Factor-1, CAF-1, is known to promote RDR. Here, we addressed the contribution of histone deposition to RDR. We expressed a mutated histone, H3-H113D, to genetically alter replication-dependent chromatin assembly by destabilizing (H3-H4)2 tetramer. We established that DNA synthesis-dependent histone deposition, by CAF-1 and Asf1, promotes RDR by preventing Rqh1-mediated disassembly of joint-molecules. The recombination factor Rad52 promotes CAF-1 binding to sites of recombination-dependent DNA synthesis, indicating that histone deposition occurs downstream Rad52. Histone deposition and Rqh1 activity act synergistically to promote cell resistance to camptothecin, a topoisomerase I inhibitor that induces replication stress. Moreover, histone deposition favors non conservative recombination events occurring spontaneously in the absence of Rqh1, indicating that the stabilization of joint-molecules by histone deposition also occurs independently of Rqh1 activity. These results indicate that histone deposition plays an active role in promoting RDR, a benefit counterbalanced by stabilizing at-risk joint-molecules for genome stability.

Flaws in the DNA replication process are a source of genome and epi-genome instability. Numerous Replication Fork Barriers (RFBs) and replication-blocking agents interrupt fork elongation, causing recurrent temporary pauses to a single replisome and occasional terminal fork arrest. Stressed forks are fragile DNA structures prone to chromosomal aberrations which may result from faulty replication-based DNA repair events [19]. Chromatin establishment and maturation take place during DNA replication, a critical step for the inheritance of the epigenome [20]. Histone supply and chromatin assembly regulate fork stability and elongation [21,22]. Fork obstacles interfere with histone dynamics, including histone recycling and inheritance of histone marks, resulting in adjacent loci liable to epigenetic changes [2,23]. Thus, stressed forks are instrumental in triggering chromosomal aberrations and chromatin changes by mechanisms that remain to be fully understood.
A variety of DNA repair factors are engaged in the timely resumption of fork elongation. Homologous recombination (HR) is a key DNA repair pathway that preserves fork integrity and replication competence through a process called Recombination-Dependent Replication (RDR) [24]. At the pre-synaptic step, the recombinase Rad51 binds single stranded DNA (ssDNA) exposed at arrested forks, to form a filament with the assistance of mediators such as yeast Rad52 and mammalian BRCA2. After homology search, the Rad51 filament promotes strand invasion into an intact homologous DNA template, usually the sister chromatid or the parental DNA ahead of the fork, to form a displacement loop (D-loop). Then, the invading 3' end allows DNA synthesis to be primed and the reassembly of replication factors to restart the fork (reviewed in [25]). D-loops can be disassembled by DNA helicases such as the human RecQ helicase BLM and its fission yeast orthologue Rqh1 [26]. Because eukaryotic genomes contain numerous dispersed and repeated sequences, RDR can occasionally generate chromosomal rearrangements. In these circumstances, RecQ helicases are instrumental to limit the likelihood of faulty RDR creating chromosomal aberrations [27,28].
Chromatin factors handle histone dynamics at DNA lesions to provide access to DNA repair machineries and to prime DNA repair [1]. Subsequent to DNA repair, chromatin restoration is a necessary step to engage physiological processes such as transcription restart and turning-off the checkpoint response [29,30]. The crosstalk to couple DNA repair and chromatin restoration are poorly understood and it is unknown if RDR, that allows the resumption of DNA synthesis at dysfunctional forks, is coupled to histone deposition. We and others have reported that HR-mediated DNA synthesis is liable to homology-dependent template switches (TS) during both the initiation of RDR and the progression of restarted forks; those TS being promoted by CAF-1 [27,31,32,33]. We have proposed that CAF-1 prevents the disassembly of the D-loop by Rqh1, in a PCNA-dependent manner [34]. Whether this role of CAF-1 in stabilizing joint-molecules requires its histone deposition activity is unknown.
Here, we advance from our previous work by showing that HR-mediated DNA synthesis is coupled to Asf1 and CAF-1-mediated histone deposition, a step necessary to promote RDR. At a site-specific fork arrest, RDR-coupled histone deposition prevents the disassembly of D-loop intermediates by Rqh1. Consistent with chromatin assembly occurring at specific joint-molecules, CAF-1 recruitment to sites of HR-dependent DNA synthesis requires Rad52. DNA synthesis-coupled histone deposition and Rqh1 act synergistically to promote cell survival to replication stress, but not to DSBs. In the absence of Rqh1, histone deposition favors spontaneous deletion-type recombinants, indicating that the role of chromatin assembly in promoting D-loop stabilization occurs also independently of Rqh1. Our data indicate that nucleosome assembly stabilizes D-loop intermediates to promote replication fork recovery. Therefore, we reveal a novel replication-dependent crosstalk between DNA repair factors and chromatin assembly to ensure repair-synthesis and balance genome stability at sites of fork-arrest.

Asf1 promotes TS during initiation of RDR
We previously reported that CAF-1 promotes TS during initiation of RDR, in a way that the D-loop is protected from disassembly by Rqh1. We asked if other histone chaperones are involved in RDR. To this end, we employed a previously described site-specific fork arrest assay in which replication of a specific genomic locus is strictly dependent on HR [27]. The assay consists of two polar Replication Fork Barrier (RFB), called RTS1, integrated as inverted repeats at both sides of the ura4 + gene, abbreviated as the t>ura4<ori locus ( Fig 1A). Upon activation of the RFB, the binding of Rad51 and its Rad52 loader allows blocked forks to be restarted to overcome the RFB. Occasionally, Rad51 promotes newly replicated strands to switch template and invades the opposite RTS1 sequence to form a D-loop which primes the restarted DNA synthesis on a non-contiguous DNA template. This TS pathway results in the formation of stable joint-molecules (JMs), referred to as D-loop and Holliday junction-like intermediates, whose resolution generates at least two RDR products: an acentric and a dicentric chromosome [34]. The formation of acentric and dicentric chromosome is strictly dependent on Rad52 and serves as a marker for RDR completion [27].
We applied the RDR assay to a temperature-sensitive allele of the essential asf1 gene, asf1-33, which was reported to be defective for H3 deposition at restrictive temperature [35]. Cells were cultured at semi-permissive temperature (32˚C) at which asf1-33 mutated cells exhibited sensitivity to 0.02% methyl methane sulfonate (MMS, an alkylated DNA agent that impedes replication fork progression) (S1A Fig). Chromosome analysis by Pulsed Field Gel Electrophoresis (PFGE) coupled with Southern-blot hybridization showed that the amount of acentric fragment, a RDR product, was reduced by 2 fold in asf1-33 cells, as well as in a mutant lacking CAF-1 (i.e. in pcf1 deleted cells), indicating that Asf1 promotes RDR (Fig 1B). The analysis of replication intermediates by two-dimensional gel electrophoresis (2DGE) showed that signals corresponding to arrested forks were similar in all analyzed strains (see bottom left numbers on Fig 1C), indicating that the defect in RDR is not a consequence of a less efficient RFB activity in strains lacking Asf1 (Fig 1C). JMs intensity was reduced in the asf1-33 mutant (Fig 1C  and 1D), suggesting that Asf1 preserves JMs during RDR. We concluded that Asf1 promotes TS during initiation of RDR, as reported for CAF-1.
We investigated the role of other histone chaperones in RDR by analyzing the level of acentric chromosome upon RFB induction. We found no decrease on acentric level in cells lacking the HIRA complex (involved in replication-independent H3-H4 deposition), the FACT complex, Rtt106, the two orthologues of the S. cerevisiae H2A-H2B histone chaperone Nap1 and Nap2, and Chz1, a histone variant H2AZ chaperone (S1B and S1C Fig). Of note, we observed a slight increase in the level of the acentric chromosome in the absence of Nap1, but this was not observed in the absence of the other H2A-H2B histone chaperone Nap2. Thus, the two key histone chaperones Asf1 and CAF-1, known to mediate (H3-H4) 2 tetramer deposition in a DNAsynthesis dependent manner, promote RDR.

tetramer formation
To address the role of histone deposition during RDR, we decided to genetically disrupt replication-dependent chromatin assembly by altering the stability of (H3-H4) 2 tetramer to inhibit their stable deposition onto DNA. To this end, we employed a mutated form of H3 containing a single substitution of histidine to aspartic acid (H3-H113D). This mutation was reported to inhibit CAF-1-mediated nucleosome deposition in vitro [36]. Diagram of the t>ura4<ori locus, in which t refers to the telomere-proximal side (gray lines), ura4 refers to the wt gene (blue line), > and < refers to the polarity of the two RTS1-RFBs (orange bars) and ori refers to replication origins (opened black circle, the largest one being the most efficient origin). Green bar indicates the rng3 probe. The RDR assay consists of a polar Replication Fork Barrier (RFB), called RTS1, integrated at the ura4 + gene, 5Kb away from a strong replication origin at the centromere-proximal side. An inverted repeated RTS1 sequence is integrated at the telomere-proximal side of ura4 + to generate the t>ura4<ori locus. Due to the main replication direction, the barrier activity of the centromere-proximal RTS1-RFB is predominant over the activity of the telomere-proximal RFB. The RFB activity is mediated by the RTS1-bound protein Rtf1, the expression of which is regulated by the nmt41 promoter repressed in the presence of thiamine. Upon Rtf1 expression, > 90% of forks are blocked at the centromere-proximal RFB. The binding of the Rad51 recombinase and its Rad52 loader allow the blocked fork to be restarted to overcome the RFB. Faulty restart events occur in~2-3% of cells/replication: Rad51 promotes newly replicated strands to switch template and to invade the opposite inverted RTS1 sequence. DNA synthesis is then initiated from the 3' invading strand on a non-contiguous DNA template, resulting in a stable early JM, referred to as a D-loop intermediate. Upon arrival of the converging fork, a second template switch event results in the formation of a later JM, referred to as Holliday junction (HJ)-like intermediate whose resolution generates at least two recombination products: acentric and dicentric which levels are a marker of RDR completion. The interface between the two H3-H4 dimers involves the C-terminal region (residues 106 to 131) of the two histones H3, called here H3 and H3' [4,37] (Fig 2A and 2B and S2A Fig). We examined X-ray structures of nucleosomes, chosen among the numerous available experimental models on the basis of homology with S. pombe histone H3 (see details in Materials and methods). The H113 emerges as a key residue in the H3:H3' interface with H113 of one histone H3 being anchored to the second histone H3' by a dense network of contacts involving six residues (S2B Fig). Each H113 forms two intermolecular hydrogen bonds with C110 and D123, reinforced by Van der Waals contacts with four residues: A114, R116, K122 and L126. By replacing a neutral or positively charged histidine residue by a negatively charged aspartate that is positioned in front to another aspartate, D123, the H113D mutation generates a prohibitive electrostatic repulsion in the intact H3:H3' organization. Previous works reported that several mutations (C110E, L12R-I130R, H113A, L126A and A114Y) prevent the H3:H3' interface formation because of an overwhelming energetic penalty; such mutations are lethal in budding yeast [12,38,39]. By analogy with these cases, the H113D mutation likely drastically destabilizes the H3:H3' interface, precluding the (H3-H4) 2 tetramer formation.
In S. pombe cells, 3 genes (hht1, hht2 and hht3) encode a single histone H3 protein. The H113D mutation was introduced in the hht2 gene, which is expressed throughout the cell cycle [40,41]. Cells expressing H3-H113D were viable with no apparent growth defect ( Fig  2C). The deletion of hht1 and hht3 is viable (hht1-d hht3-d) and cells maintain histone H3 protein levels similar to those in wild type (wt) strain [42,43]. To obtain a strain expressing only H3-H113D, the hht2-H113D mutated strain was crossed with an hht1-d hht3-d strain. Combining the hht2-H113D mutation with both deletions resulted in a synthetic lethality ( Fig 2C). Thus, when H3-H113D is the sole histone H3 expressed, cells are not viable, in agreement with this mutation inhibiting (H3-H4) 2 tetramer formation. Still, we observed that combining hht2-H113D with either single htt1 or hht3 deletion preserves cell viability but causes a severe synthetic growth defect ( Fig 2C). This strongly suggests that wt H3 is abundant enough to allow wt (H3-H4) 2 tetramer to be formed and to preserve cell viability but mixed (H3-H113D-H4-H3-H4) tetramer are also unstable. In further experiments, we have employed hht2-H113D mutation in hht1 + hht3 + background, unless stated otherwise.
We asked if H3-H113D is incorporated into chromatin using chromatin fraction assays. As expected, H3-HA was found chromatin-bound whereas H3-H113D-HA was not detected in the chromatin fraction, indicating that this mutated histone is very poorly incorporated into nucleosomes assembled onto DNA (Fig 2F). The wt H3 (expressed from hht1 and hht3 gene) was equally detected in the chromatin fraction of hht2-HA and hht2-H113D-HA cells, indicating that wt H3 is in fine incorporated within wt nucleosomes in H3-H113D expressing cells. The interface between the two histones H3, colored here in light (H3) and dark blue (H3'), maintains H3 and H4 in the tetramer form (H3-H4) 2 . In the H3:H3' interface, H113 of each histone (orange for H113 in H3 and red for H113 in H3') is deeply MMS-induced DNA damage resulted in a slightly more abundance of PCNA bound to the chromatin whereas H3-H113D-HA was still not detected in the chromatin fraction. Our data indicate that in hht2-H113D mutated cells, H3-H113D inhibits stable tetramer formation and is therefore not deposited onto chromatin, while wt (H3-H4) 2 tetramers are formed and assembled onto chromatin.

H3-H113D alters the process of replication-coupled nucleosome assembly
We investigated the consequences of inhibiting tetramer formation on the chromatin structure in vivo. We asked whether nucleosome organization was perturbed in cells expressing H3-H113D by examining the nucleosomal status of total DNA and newly replicated DNA by micrococcal nuclease (MNase) digestion. We employed strains able to incorporate BrdU, a thymidine analogue [44]. Cells were blocked in early S-phase by hydroxyurea treatment and released in BrdU-containing media for 20 minutes to label newly replicated DNA.  [45,46,47], the absence of CAF-1 resulted in an alteration of the nucleosomal organization of newly replicated DNA ( Fig 3A, bottom panel, BrdU staining). Compared to wt strain, the proportion of BrdU-positive mono and di-nucleosome population was significantly increased in pcf1-d cells, revealing a higher sensitivity to MNase as an indication of a decreased nucleosomal density of the newly replicated chromatin ( Fig 3B, right panel, Replicated chromatin, p<0.005). This is consistent with the loss of the chromatin assembly function of CAF-1. A nucleosomal ladder was still detected in pcf1-d cells, indicating that, in vivo, nucleosome assembly on newly replicated DNA can occur to some extent in the absence of CAF-1. In human cells, the histone chaperone HIRA acts as a salvage pathway to maintain chromatin integrity during DNA replication [48]. Consistent with this, we found that the deletion of pcf1 is synthetic lethal with the deletion of hip1, the gene encoding one subunit of the fission yeast HIRA complex (S4A Fig). Therefore, the highest MNase sensitivity of the newly replicated DNA observed in pcf1-d cells is consistent with a transient alteration of the replication-coupled chromatin assembly process. Remarkably, we observed a similar increased in MNase sensitivity of the newly replicated chromatin in cells expressing H3-H113D (Fig 3A and 3B, right panel, p<0.005), indicating that replication-coupled chromatin assembly is altered in these cells. The nucleosomal pattern of the newly replicated DNA was similarly affected in the double mutant pcf1-d hht2-H113D, compared to each single mutant, suggesting that the absence of CAF-1 and unstable (H3-H4) 2 tetramers alter the same replication-coupled chromatin assembly pathway (Fig 3A and 3B, right panel, p<0.005). These data are consistent with the report that H3-H113D inhibits in vitro the chromatin assembly activity of CAF-1 [36]. Together with the observation that H3-H113D is poorly incorporated into assembled nucleosome onto DNA without affecting the MNase sensitivity of total DNA, our data are consistent with a transient defect in replication fork-associated chromatin assembly in cells expressing H3-H113D, as observed in the absence of CAF-1. Although other histone deposition pathways might be affected, our data support that the CAF-1-mediated chromatin assembly pathway is impaired in vivo in H3-H113D expressing cells.
We probed for interactions between H3-H113D and histone chaperones involved in RDR. We found that H3-H113D-HA was poorly bound to Asf1-MYC, compared to H3-HA ( Fig  3C). However, the binding of Asf1-MYC to wt H3 was not affected in cells expressing H3-H113D, arguing that the histone chaperone function of Asf1 is preserved in hht2-H113D mutated cells (S4B Fig). Then, we employed a strain expressing Pcf1-YFP that we previously showed targeted to replication foci and able to interact with PCNA [34]. Unexpectedly, we found that pcf1-YFP was synthetic lethal with the deletion of hip1, as observed for the deletion of pcf1 (S4A Fig). Nonetheless, Pcf1-YFP was able to interact with H3, H4, PCNA and Pcf2-MYC (S4C- S4E Fig). Thus, Pcf1-YFP is able to promote protein-protein interactions essential to its chromatin assembly function, but it is not functional enough to ensure cell viability in the absence of HIRA. Then, we probed for interactions with H3-H113D and found that Pcf1-YFP bound H3-H113D-HA to the same extent than H3-HA (Fig 3C and S4E Fig). Also, in cells expressing H3-H113D, the ability of Pcf1-YFP to bind Pcf2-MYC, PCNA and H3 and H4 was not affected (S4C- S4E Fig). Altogether, these data indicate that in hht2-H113D mutated cells, CAF-1 forms complexes with H3-H4, H3-H113D, and PCNA. Since H3-H113D inhibits tetramer formation, a step necessary to CAF-1 promoting histone deposition, our data are consistent with H3-H113D altering CAF-1 function in replication-coupled chromatin assembly in vivo, as observed in vitro [36].

Histone deposition promotes TS during initiation of RDR by preventing Dloop disassembly
The H3-H113D mutation offers the possibility to question the role of replication-coupled chromatin assembly in promoting TS during the initiation of RDR. We thus applied the RDR assay to hht2-H113D hht1 + hht3 + mutated cells and found that JMs intensity was reduced as well as the subsequent accumulation of acentric chromosome, one RDR product (Fig 4A-4D). Of note, the efficiency of the RFB was found similar in all strains analyzed (see bottom left numbers on Fig 4C). Rad52 was able to bind the active RFB in cells expressing H3-H113D, indicating that H3-H113D impairs RDR downstream Rad52 recruitment to arrested forks, as reported for CAF-1 ( Fig 4E) [34]. Importantly, similar defects in RDR were observed in the double pcf1-d hht2-H113D mutant compared to each single mutant, showing that H3-H113D impairs RDR by altering CAF-1-mediated histone deposition. We reported that in the absence of CAF-1, JMs are faster disassembled by Rqh1 [34]. Remarkably, similar interactions were observed in hht2-H113D cells in which the deletion of rqh1 restored the intensity of JMs signal, whereas the deletion of rqh1 alone did not affect the intensity of JMs (Fig 4C and 4D). No defect in RDR were observed in strains in which genes encoded H3 are deleted (either hht2-d or hht1-d hht3-d cells (Fig 4A-4D and S5 Fig), further supporting that H3-H113D favors Dloop disassembly as a consequence of impairing replication-coupled histone deposition, not as a consequence of its lower abundance. Altogether, these data establish that histone deposition promotes TS during RDR by preventing D-loop disassembly by Rqh1.

H3K56Ac is dispensable to TS during initiation of RDR
The H3K56Ac modification marks newly synthetized histone and was proposed to regulate nucleosome assembly to contribute to the DNA damage response [49]. Hence, the importance of acetylation in addition to histone deposition has to be considered. We found that in contrast to wt H3 that was efficiently acetylated by Rtt109 in hht2-H113D cells, we could no detect the  (Fig 3C). We thus tested the role of H3K56Ac in RDR. We analyzed two strains: one expressing a single H3 protein which cannot be acetylated on K56 (hht1-d hht3-d hht2-K56R) and a rtt109 deleted strain. Consistent with previous reports [17], both strains were defective for H3K56Ac (S5B Fig). We observed no changes on the level of acentric chromosome (S5C and S5D Fig). Thus, H3K56Ac is dispensable to TS during the initiation of RDR.

CAF-1 recruitment to sites of HR-dependent DNA synthesis
Our data indicate that CAF-1 and histone deposition promote RDR by stabilizing D-loop intermediates. Since DNA synthesis is primed from the 3'end of the invading strand within the D-loop, one could expect CAF-1 to associate with such sites of DNA synthesis in vivo. D-loops are transient DNA structures that form into the genome upon DNA damage. Therefore, we took advantage of the RDR assay in which the replication of the ura4 gene is dependent on HR to ask if CAF-1 associates with this site of HR-dependent DNA synthesis. Pcf1-YFP was found enriched at the two RFB sites in ON condition, compared to OFF condition (Fig 5, primer  pairs 2 and 4, left panel). Remarkably, the binding of Pcf1-YFP to ura4 was stimulated upon activation of the RFB (Fig 5, primer pair 3, left panel). To test if such binding to RFBs depends on the formation of recombination intermediates, we repeated these analyses in rad52-d cells in which JMs do not occur [27]. Pcf1 binding to ura4 and two RFB sites was no longer stimulated by the activation of the RFB in the absence of Rad52 (Fig 5, middle panel). We concluded that CAF-1 recruitment to HR-mediated DNA synthesis requires the formation of JMs. We asked if inhibiting (H3-H4) 2 tetramer formation prevents CAF-1 binding to sites of HR-mediated DNA synthesis. Upon activation of the RFB, Pcf1-YFP was found enriched at the two RFBs and at ura4 in cells expressing H3-H113D, showing that RDR defect in hht2-H113D strain is not a consequence of CAF-1 inability to bind the site of HR-mediated DNA synthesis. Altogether, our data demonstrate that CAF-1 associates to sites of HR-dependent DNA synthesis, downstream of Rad52 and JMs formation.

RDR-coupled histone deposition promotes cell resistance to camptothecin in the absence of Rqh1
We have addressed the role of fork-associated histone deposition to the cell response to replication stress. The H3-H113D mutation or the pcf1 deletion decreased cell survival of rqh1 deleted cells upon camptothecin (CPT, a topoisomerase I inhibitor) treatment, and resulted in a moderate synthetic growth defect upon MMS treatment (Fig 6A and 6B). This was not a consequence of a loss of Rqh1-Pcf1 interaction in cells expressing H3-H113D as Rqh1-MYC associated with Pcf1-YFP in hht2-H113D mutated cells (S4D Fig). Also, DNA damage did not stimulate chromatin-bound H3-H113D-HA, suggesting that unstable tetramers are unlikely to be assembled during repair synthesis (Fig 2F). Interestingly, such synthetic interactions were not observed in response to bleomycin (Fig 6C), a DSB-inducing agent. These data reveal that in the absence of Rqh1 activity, cell resistance to topoisomerase I inhibitor relies on CAF-1-mediated histone deposition.

(H3-H4) 2 tetramer formation favors deletion type recombinant in the absence of Rqh1
We asked if the role of fork-associated histone deposition in promoting HR event might also occur independently of Rqh1. The lack of Rqh1 results in a spontaneous hyper-recombination phenotype caused by its multiple roles in processing HR intermediates [50]. We asked if histone deposition contributes to the anti-recombinase activity of Rqh1 by monitoring spontaneous HR events using an assay for intra-allelic recombination between direct ade6 repeats ( Fig  7A) [51]. Gene conversion (GC) and synthesis dependent strand annealing (SDSA) result in conversion type, whereas crossover (CO) and single strand annealing (SSA, a DNA repair pathway independent of Rad51-mediated strand exchange) give rise to deletion type. In rqh1-d cells, conversion and deletion rate were increased by 2-and 3-fold, respectively, which is consistent with known Rqh1 anti-recombinase activities (Fig 7B) [50]. The hht2-H113D mutation had minor impact on HR outcomes but when combined with rqh1-d, it decreased by 35% the rate of deletion type without affecting the rate of conversion type (hht2-H113D and rqh1-d hht2-H113D compared to wt and rqh1-d respectively). These data establish that a significant part of the deletion events occurring in the absence of Rqh1 are caused by (H3-H4) 2 tetramer formation. Thus, histone deposition also promotes HR events by other mechanisms than counteracting Rqh1 activity. These data are consistent with antagonistic activities of Rqh1 and histone deposition in controlling HR outcomes: Rqh1 promotes D-loop disassembly to favor a conservative pathway whereas histone deposition promotes D-loop stability to favor deletion events such as CO.

Discussion
Fork-obstacles open up the risk of genome instability. HR ensures the timely completion of genome duplication by restarting dysfunctional forks, but this comes at the expense of TS events that can generate genome rearrangements. Here, we established that HR-mediated DNA synthesis is coupled to histone deposition, mediated by CAF-1 and Asf1, and that this process is necessary to promote RDR (Fig 8). We established that inhibiting (H3-H4) 2 tetramer formation impacts RDR and D-loop resolution, in a manner dependent on CAF-1 function. Our data support a mechanism during which RDR is coupled to CAF-1-mediated histone deposition to stabilize D-loop intermediates to promote RDR. We reveal a novel interplay between chromatin restoration and DNA repair factors; a crosstalk necessary to cell survival to replication stress. Our finding highlight that chromatin restoration is an integral part of the RDR process. We propose that histone deposition plays an active role in RDR to avoid discontinuity in chromatin assembly upon replication stress, a benefit counterbalanced by stabilizing at-risk joint-molecules for genome stability.  A complex network of histone chaperones handles histone dynamics at replication forks to promote chromatin assembly. We establish that only a subset of this network is pivotal to promote RDR. Despite recent advances in understating the mechanism of CAF-1-mediated histone deposition in vitro, it remains challenging to define point mutations to generate mutated forms of CAF-1 unable to interact with H3-H4 in vivo [12]. To overcome this, we took advantage of the H3-H113D mutated form, reported to inhibit CAF-1-mediated histone deposition in vitro [36]. We show that H3-H113D precludes (H3-H4) 2 tetramer formation, is poorly incorporated into nucleosomes assembled onto DNA and alters the organization of the newly replicated chromatin. Remarkably, the H3-H113D mutation mimics the absence of CAF-1 in impairing RDR. We cannot exclude that H3-H113D has potential CAF-1-independent effects, but we have observed no additive effect between the lack of CAF-1 and H3-H113D indicating that the inhibition of replication-coupled histone deposition is responsible for the defect in RDR. RDR requires Asf1, the 3 CAF-1 subunits and the ability of CAF-1 to interact with PCNA [34], supporting the view that Asf1 and CAF-1 promote RDR by coupling histone deposition to the step of DNA synthesis. Consistent with this, CAF-1 binds to site of HR-mediated DNA synthesis in a Rad52-dependent manner, indicating that the early steps of RDR must be engaged for the subsequent recruitment of CAF-1. The binding of Rad52 to arrested forks is not affected when inhibiting histone deposition or in the absence of CAF-1 [34], consistent with CAF-1-mediated histone deposition acting downstream of Rad52.
We proposed that CAF-1-mediated histone deposition acts as a chromatin restoration pathway during the DNA synthesis step of RDR. In this scenario, histone deposition would occur onto the DNA duplex of the extended D-loop (Fig 8). We cannot exclude that histone deposition occurs onto the displaced strand of the D-loop, as it has been proposed that nucleosome can be deposited onto ssDNA [52]. RDR requires two histone chaperones that mediate chromatin assembly in a DNA synthesis manner and requires the ability of CAF-1 to interact with PCNA. Therefore, we favor the first hypothesis in which histone deposition is coupled to DNA synthesis during D-loop extension. Extensive works have address the role of chromatin factors in regulating DNA lesions accessibility to DNA repair machineries but how chromatin restoration is coupled to the DNA repair event is poorly understood [3]. Our data put forward a crosstalk between the DNA repair machinery and the step of chromatin restoration that plays an active role in promoting RDR.
Histone deposition coupled to RDR impacts D-loop resolution. One mechanism involves nucleosome assembly onto D-loop intermediates to counteract Rqh1 activity. Possibly, histone deposition onto extended D-loops creates a substrate less favorable to the Rqh1 activity. RecQ helicases act as motor proteins/helicases to migrate DNA junctions which may be prevented by assembled nucleosomes. It remains challenging to address whether histones are deposited onto D-loops as these structures are very transient. JMs visualization by 2DGE requires a step of enrichment in replication intermediates, technically incompatible with chromatin immuno-precipitation approaches to address histone binding to JMs. Nonetheless, we establish that inhibiting replication-coupled histone deposition favors D-loop disassembly. We provide insights into the nature of the chromatin required to protect D-loops. Asf1 acts as an histone chaperone to present H3-H4 to Rtt109 and generates the H3K56Ac mark and then transfer H3-H4 to CAF-1 [7]. We found H3K56Ac being dispensable to promote TS during the initiation of RDR. Possibly, Asf1 acts as a donor histone to CAF-1. However, H3-H113D binds to CAF-1 likely as a H3-H4 dimer but not to Asf1, indicating that H3-H4 can be handed off to CAF-1 independently of Asf1, a pathway that remains to be clarified.
Nucleosome deposition onto extended D-loop may impose topological constraints. D-loop disassembly is a topoisomerase-mediated mechanism [53]. Thus, topological constraints resulted from DNA wrapped around nucleosome may be easily relieved by topoisomerase 3. When seeking for additional chromatin factors required for RDR, we found Nap1 and Nap2, two histone H2A-H2B chaperones, to be dispensable to protect D-loop. This suggests that the deposition of (H3-H4) 2 tetramer, but not the formation of a nucleosome, is sufficient by itself to counteract Rqh1 activity and to limit topological constraints.
Fork-associated histone deposition affects HR outcomes by mechanisms that are also independent of Rqh1 activity. During HR-mediated DSB repair, disassembly of D-loops extended by DNA polymerase ensures a non-crossover outcome [26]. We report that in the absence of Rqh1, a significant part of the deletion type events is favored by histone deposition. Chromatin assembly at D-loop intermediates may also counteract the activity of other helicases in disrupting JMs [26]. Interestingly, spontaneous rates of gene conversion were unaffected by inhibiting histone deposition. This suggests that DNA synthesis associated to GC and SDSA is too short in length to favor histone deposition. Nonetheless, our data reveal that the antagonistic activities of RDR-coupled histone deposition and Rqh1 in D-loop resolution are pivotal to balance genome stability at arrested forks and to promote cell resistance upon replication stress.
During unchallenged replication, the concerted action of multiple histone chaperones coordinates the assembly of chromatin behind the fork to achieve recycling of parental histones and deposition of newly synthetized histones [20]. RDR results in the progression of noncanonical forks in which both strands are synthetized by the DNA polymerase delta, which contrasts with the division of labor between DNA polymerase delta and epsilon at origin-born replication forks [54]. Such restarted forks are liable to replication errors such as multiple template switches, replication slippage and U-turn [32,33,55]. Despite these unusual features, our data suggest that restarted forks remain coupled to histone deposition and thus may help to ensure continuous chromatin assembly upon replication stress.
Fork obstacle and replication stress interfere with the inheritance of epigenetic marks [23,56,57]. On the one hand, the post-replicative repair of DNA lesions/structures, left un-replicated behind the fork, is uncoupled from chromatin assembly and recycling of parental histones. On the other hand, it was proposed that the bypass of DNA secondary structures, such as G quadruplexes, by PrimPol allows the repriming of DNA synthesis very closely to the fork obstacle, allowing for the maintenance of replication-coupled chromatin maturation [58]. The choice of the pathway employed to overcome fork obstacles may influence the maintenance of histone deposition coupled to restarted forks. Our data indicate that RDR, a main pathway to bypass fork obstacles, is coupled to histone deposition. In the view that replisomes are often interrupted by numerous obstacles, RDR-coupled histone deposition contributes to cell resistance to replication stress and may ensure continuity in assembling chromatin upon replication stress. Nonetheless, this comes at the expense of stabilizing JMs which can be detrimental to genome stability.

Standards yeast genetics
Yeast strains used in this work are listed in S1 Table. Gene deletion and gene tagging were performed by classical and molecular genetics techniques [59]. Strains containing the RTS1-RFB were grown in supplemented EMM-glutamate media containing 60μM of thiamine. To induce the RTS1-RFB, cells were washed twice in water and grown in supplemented EMM-glutamate media containing thiamine (Rtf1 repressed, RFB OFF condition) or not (Rtf1 expressed, RFB ON condition) for 24 hours or 48 hours.

Analysis of replication intermediates by 2DGE
Replication Intermediates (RIs) were analyzed by 2DGE as previously described [61]. RIs were migrated in 0.35% agarose gel in 1X TBE for the first dimension. The second dimension was migrated in 0.9% agarose gel 1X TBE supplemented with EtBr. DNA was transferred onto a nylon membrane (Perkin Elmer, NEF988001PK) in 10X SSC. Membranes were incubated with a 32 P radiolabeled ura4 probe, and RIs were detected using phosphor-imager software (Typhoon-trio) and quantified with ImageQuantTL.

Spontaneous recombination assay
Spontaneous recombination rate was assayed using strains containing a direct repeat of two nonfunctional ade6 alleles flanking a functional ura4 gene [51]. Strains were kept on low adenine EMM plates lacking uracil to prevent selection for Ade + and Urarecombinants. Dark pink colonies were streaked on supplemented YE plates and 23 to 26 independent single colonies for each strain were used to calculate Ade+ recombinant rate. Appropriate dilutions were plated on supplemented YE plates (to determine cell survival), EMM plates lacking adenine (to score spontaneous recombination rate, Ade+ recombinants) and EMM plates lacking adenine and uracil (to score gene conversion rate, Ade+ Ura+ recombinants). Colonies were counted after 5-7 days of incubation at 30˚C. The rates of Ade+ and Ade+ Ura+ recombinant were calculated as described in [65].

Analyzed X-ray structures of nucleosome
There is no experimental model of nucleosome containing histones from S. pombe. Among the numerous available structures of nucleosome, only three of them include yeast histones, all from S. cerevisiae (PDB codes 1ID3, 4JJN and 4KUD), with resolutions of~3 Å. Indeed, the highest resolution was obtained for a nucleosome containing histones from Xenopus laevis (PDB code 1KX5, resolution of 1.9 Å). Since the histone sequences are extremely conserved, S. cerevisiae and Xenopus laevis H3 histones share 92% of residues with H3 of S. pombe. More specifically, the region surrounding H113 is well preserved across S. pombe, S. cerevisiae and Xenopus laevis, with a very good score being observed for the couple S. pombe / Xenopus laevis (S2A Fig). Given the reasonable sequence agreement, 1ID3, 4JJN, 4KUD and 1KX5 were analyzed using PDBsum to describe the interface between two H3-H4 dimers [66].

Data availability
Numerical data used for graphs are provided in the supplementary information S1 Data.  2).

(A)
Amino acid sequences of the region containing H113 and its interacting partners in the H3: H3' nucleosomal interface. This table gives the amino acid sequences of the relevant part of H3 in Schizosaccharomyces pombe and the considered X-ray structures. The X-ray structures are referenced by their PDB codes. 1KX5 contains histones from Xenopus laevis; 1ID3, 4JJN and 4KUD include histones from Saccharomyces cerevisiae. The residues in red differ from those of Schizosaccharomyces pombe. The residues on yellow background form a network of contact with H113 (green background) in the H3:H3' interface. (B) Schematic representation of the contact network involving H113 in the H3:H3' interface. H3'-H113 interacts with 6 residues of H3. Two hydrogen bonds (red arrows) are reinforced by Van der Waals contacts (grey arrows). Identical, symmetric interaction pattern is observed with H3-H113 and C110, H3'-D123, A114, R116, K122 and L126. The interface analysis was carried out with PDBsum (44).

(C)
Top panels: effect of the H113D mutation on histone H3-HA level transcribed from hht1, hht2 or hht3. Tubulin was used as loading control. Bottom panels: Quantification of top panels: the level of H3-H113D-HA and H3-HA were normalized to tubulin. Individual values are plotted ± the range. (TIF) S3 Fig. BrdU incorporation during and after HU treatment (related to Fig 3). Logarithmic growing cells from SL1077 strain were arrested in early S-phase with HU treatment. A pulse of 20min BrdU (400μM) incorporation was done after 4 hours of HU block or after releasing cells in a fresh media. Top panel: BrdU-incorporated genomic DNA was digested with increasing amount of MNase and migrated on ethidium bromide-containing agarose gel. Bottom panel: After transfer onto nitrocellulose membrane, incorporated BrdU was revealed using anti-BrdU antibody.