Histone H4K20 tri‐methylation at late‐firing origins ensures timely heterochromatin replication

H4K20 methylation is critical for S‐phase progression and PR‐Set7 replication licensing functions

As shown in human U2OS cells (Appendix Fig S1), loss of mammalian PR‐Set7 results in improper S‐phase progression, which is in part attributed to defects in replication origin activity (Jørgensen et al, 2007; Tardat et al, 2007). To provide evidence that this S‐phase phenotype is directly caused by a decrease in the levels of H4K20 methylation rather than defects in the methylation of other PR‐Set7 targets, we examined the cell cycle of human U2OS cells transduced with a high titer of retroviral vectors encoding a FLAG‐tagged histone H4^K20A mutant carrying a lysine to alanine substitution at position 20 (K20A). Cells transduced with a virus encoding a FLAG‐tagged wild‐type histone H4^WT were used as controls. As shown in Fig 1A and B, immunoblot analysis showed that both FLAG‐tagged H4^WT and H4^K20A proteins were expressed at similar levels and were efficiently incorporated into chromatin 3 days after retroviral infection (Fig 1A, lanes 3 and 6), thereby leading to the replacement of half of the pool of the endogenous histone H4 (Fig 1B, top panel). Incorporation of FLAG‐H4^K20A mutant into chromatin slightly increased PR‐Set7 levels but led to a specific decrease in the global levels of the three methylated H4K20 states, with a stronger reduction for H4K20me1 and H4K20me3 (Fig 1B). Cell‐cycle profiles of FLAG‐H4^WT and FLAG‐H4^K20A cells were then analyzed the same day by measuring the DNA content according to low (gates 1) and high (gates 2) levels of FLAG‐H4 expression (Fig 1C; gates 1 and 2, respectively). We noticed that the amount of cells with the highest levels of H4^K20A (i.e., with the strongest decrease in H4K20me) was low compared to FLAG‐H4^WT cells (Fig 1C, upper panels, compare gates 1 and 2), suggesting that the H4K20A mutation is relatively toxic. Conspicuously, whereas cells with low levels of H4^K20A displayed a cell‐cycle profile similar to FLAG‐H4^WT cells (Fig 1C, gates 1), cells expressing high levels of FLAG‐H4^K20A accumulate in S‐phase, suggesting that they fail to complete DNA replication as previously described in PR‐Set7‐depleted cells (Tardat et al, 2007; Appendix Fig S1).

Our results suggest that appropriate H4K20 methylation levels may be required for PR‐Set7 functions in the licensing of replication origins. To verify this, we measured by FACS the levels of re‐replication induced by the proteolytic‐resistant PR‐Set7^PIPmut mutant in cell populations expressing similar and non‐deleterious levels of FLAG‐tagged H4^WT or FLAG‐tagged H4^K20A, 3 days before infection with retrovirus encoding PR‐Set7^PIPmut mutant (Fig 1C, gate 1). PR‐Set7^PIPmut contains mutations in the PCNA binding domain that inhibit CRL4^cdt2‐mediated degradation and induce PR‐Set7 stabilization (Tardat et al, 2010; Beck et al, 2012b). The resulting re‐replication phenotype is caused by repeated PR‐Set7‐induced replication origin licensing (Tardat et al, 2010), which can be observed by the appearance of cells with DNA content greater than 4N, 2 days after PR‐Set7^PIPmut expression (Fig 1D, panel a). As shown in Fig 1D, expression of the PR‐Set7^PIPmut, but not of the catalytic inactive PR‐Set7^{PIPmut+SETmut} mutant, caused the appearance of ~30% of re‐replicated cells in both parental and FLAG‐tagged H4^WT cells (Fig 1D, compare panels a–d). In contrast, the level of re‐replication upon PR‐Set7^PIPmut expression was greatly reduced in FLAG‐tagged H4^K20A cells (Fig 1D, panels e–f) correlating with the reduced amount of H4K20 residues in these cells. These results provide the evidence that H4K20 methylation is indeed a critical downstream effector of PR‐Set7 replication licensing functions.

H4K20 methylation is an enhancer of origin efficiency

K20 methylation enhances histone H4 interaction with ORC in vitro (Beck et al, 2012b), suggesting that this histone mark might be sufficient to induce DNA replication at specific loci in vivo. To test this possibility, we examined the impact of H4K20 methylation on the replication of Epstein–Barr virus‐derived (EBV) episomes from origin DNA sequences that involve cellular pre‐RC complex for licensing. As illustrated in Fig 2A, the replication and stability of EBV episomes require the binding of the EBV‐encoded protein EBNA1 to two DNA elements: (i) the family of repeats (FR) element that ensures the segregation and stability of episomes during cell division and (ii) the dyad symmetry (DS) element that serves as replication origin in human cells by co‐recruiting ORC to this DNA sequence (Hammerschmidt & Sugden, 2013). To assess whether PR‐Set7‐induced H4K20 methylation on chromatin induces origin activity by its own, a GAL4‐binding UAS sequence was introduced into a DS‐deleted EBV plasmid named FR‐UAS, which is incompetent for replication but not for EBN1A‐mediated segregation. We then examined the rescue of the replication competence of this plasmid in HEK293 cells stably expressing EBNA1 and either GAL4 DNA‐binding domain (GAL4), the wild‐type GAL4‐PR‐Set7^WT, or the methylase‐deficient GAL4‐PR‐Set7^SETmut fusion proteins (Fig EV1A). HEK293 cell line was chosen because it is suitable for transfection and EBV episome replication (Gerhardt et al, 2006). As observed previously (Beck et al, 2012b), quantitative chromatin immunoprecipitation (ChIP‐qPCR) showed that GAL4‐PR‐Set7 binding at the UAS site, but not of the GAL4 or GAL4‐PR‐Set7^SETmut proteins, led to a local increase in H4K20me1 followed by its conversion to higher H4K20me states, as indicated by the appearance of high H4K20me3 enrichment at this locus (Fig EV1B). To measure episome replication efficiency, low molecular weight DNA was harvested 6 days after plasmid transfection and the number of replicated episomes was quantified by bacterial transformation after DpnI digestion to eliminate the bacterial‐derived plasmids initially transfected into GAL4 cell lines. The results are shown in Fig 2B. As controls, similar experiments were performed with the wild‐type FR‐DS plasmid and the FR‐ORI^RDH plasmid (Fig 2A), which contains the 300‐bp ORC‐binding DNA fragment of the human replication origin (ORI^RDH) of the RDH gene instead of the DS element (Gerhardt et al, 2006). The results are shown in Fig 2B. Compared to the FR‐DS plasmid, the FR‐ORI^RDH plasmid displayed a low replication competence in all cell lines (Fig 2B), which is representative of the low efficiency of short mammalian origin DNA sequences to define an origin by their own (Gerhardt et al, 2006). We noticed that the replication of the FR‐UAS plasmid was almost undetectable in GAL4 and GAL4‐PR‐Set7^SETmut cells but not in GAL4‐PR‐Set7‐expressing cells, where this plasmid exhibited a low replication similar to that of the FR‐ORI^RDH plasmid (Fig 2B). These results indicate that a local increase in H4K20 methylation, as observed with mammalian DNA sequences, is not sufficient to define a fully competent origin of replication.

Since origin efficiency might depend on the presence of multiple origin determinants (Méchali et al, 2013; Valton et al, 2014), we next examined the effect of inducing H4K20 methylation in the vicinity of the ORI^RDH origin. For this, we created the FR‐UAS‐ORI^RDH plasmid (Fig 2A) and tested the replication competence of this plasmid as described above. As shown in Fig 2D, the ORI^RDH efficiency in FR‐UAS‐ORI^RDH plasmid was fourfold increased in GAL4‐PR‐Set7 cells, whereas it remained unchanged in GAL4 and GAL4‐PRSet7^SETmut cells (Fig 2C). ChIP‐qPCR analysis showed that this stimulation of ORI^RDH activity GAL4‐PR‐Set7 cells was associated with an accumulation of H4K20me1/me3 and a higher efficiency of MCM2‐7 loading, as measured by the significant increase in the levels of MCM3 at the UAS and ORI^RDH sequences in FR‐UAS‐ORI^RDH plasmid (Fig 2D). Importantly, the same enhancer effect on MCM2‐7 loading and origin activity was also observed for the viral origin sequence oriP in a FR‐UAS‐DS plasmid transfected in GAL4‐PR‐Set7‐expressing cells (Fig EV1B and C), confirming the ability of PR‐Set7‐induced H4K20 methylation to stimulate replication initiation at defined origins. Taken together, these results show that PR‐Set7‐induced H4K20 methylation is likely not sufficient to promote efficient origin function but rather serves to strengthen the activity of defined origins by enhancing the recruitment and/or stability of pre‐RC complexes at these origins.

PR‐Set7‐mediated stimulation of origin activity depends on the further methylation of H4K20me1 by Suv4‐20h1/h2

Since the ORC shows a higher affinity for H4K20me2/3 than for H4K20me1 in vitro (Oda et al, 2010; Vermeulen et al, 2010; Kuo et al, 2012), we asked whether the stimulatory effect of H4K20 methylation on origin activity depends on the switch from H4K20me1 to higher H4K20me states induced by Suv4‐20h1/h2 methyltransferases. To address this question in vivo, we repeated the replication assay with FR‐ORI^RDH and FR‐UAS‐ORI^RDH plasmids in GAL4‐PR‐Set7 cells in the presence of A‐196, a chemical inhibitor of Suv4‐20h1/h2. Immunoblot analysis of whole‐cell extracts confirmed the specificity of the A‐196 treatment, which led to a strong reduction in the levels of H4K20me2/3 and the subsequent increase in H4K20me1 without an impact on the levels of histone H3 tri‐methylation at lysines 4, 9, and 27 (Appendix Fig S2). Consistent with these results, ChIP‐qPCR analysis showed that H4K20me1, but not H4K20me3, was highly enriched at the UAS and ORI^RDH sequences in GAL4‐PR‐Set7‐expressing cells treated with A‐196 (Fig 3A). In this experimental context, the FR‐UAS‐ORI^RDH and FR‐ORI^RDH plasmids displayed a similar replication activity without any significant difference in the levels of MCM3 loading at the ORI^RDH origin (Fig 3B and C). These results confirm that PR‐Set7‐mediated H4K20me1 is only a prerequisite for replication origin stimulation (Beck et al, 2012b), which depends on additional methylation of H4K20me1 by Suv4‐20h enzymes.

Loss of Suv4‐20h activity on H4K20me1 specifically impairs heterochromatin replication timing

We next sought to determine whether the ability of Suv4‐20h and higher H4K20 methylation states to enhance replication origin activity plays an essential role in the replication of all the genome or only at specific chromatin regions. To address this question, we examined at genome‐wide levels the spatiotemporal replication program of mouse embryonic fibroblast (MEF^364.2) line derived from Suv4‐20h2 ^−/− , Suv4‐20h1 ^−/flox, Cre‐ER embryos and in which the remaining knocked‐in floxed SUV4‐20H1 gene can be deleted by addition of 4‐hydroxytamoxifen (4OHT). Four days after 4OHT treatment, immunoblot analyses showed a complete and irreversible disappearance of H4K20me2 and H4K20me3 followed by a concomitant accumulation of H4K20me1 in MEFs^364.2 (Fig 4A, left panels). At this time, untreated and 4OHT cells were pulse‐labeled with BrdU and sorted into early and late S‐phase fractions by flow cytometry. Nascent BrdU‐substituted DNA of each fraction was then isolated and labeled with fluorescence dyes before hybridization on microarrays covering the mouse genome, at the exception of centromeric and telomeric regions, with one probe every 13 kb. The replication timing of these cells was determined using the Agilent CGH algorithm that generates a log ratio value corresponding to early‐replicating regions when the log ratio is positive and to late‐replicating regions when the log ratio is negative. As illustrated in Fig 4A for a fragment of chromosome 11 (right panel), the loss of Suv4‐20h in 4OHT‐treated MEFs^364.2 specifically delays the replication of some mid‐ and late‐replicating chromatin domains, whereas the replication of early domains remained largely unaffected. Consistent with these results, 4OHT‐treated MEFs^364.2 displayed a lower proliferation rate and tended to accumulate in late S‐phase as shown by FACS analysis (Fig EV2A and B). Furthermore, replication‐timing delays of the same chromatin regions were also observed in fibroblast lines derived from Suv4‐20h1/2 knockout embryos (Fig EV2C), thereby indicating that the disrupted replication pattern upon loss of Suv4‐20h was highly reproducible and cell line independent. To verify that the replication‐timing alterations observed in 4OHT‐treated MEFs^364.2 are indeed triggered by the loss of H4K20me2/3 states, we generated as described above the replication‐timing profile of untreated MEF^364.2 expressing similar levels of FLAG‐tagged histone H4^WT or H4^K20A. As shown in Fig 4B, expression of histone H4^K20A led to a decrease in the levels of three H4K20 methylation states, which was followed by the same replication‐timing alterations than those observed in 4OHT‐treated MEFs^364.2 (see arrows in Fig 4A and B). In addition, we noticed that both histone H4^K20A‐ and H4^WT‐expressing MEFs^364.2 displayed some replication‐timing changes in early‐replicating domains compared to untreated cells, which might be caused by the artificial expression of FLAG‐tagged histone H4 proteins in these cells (compare Fig 4A and B, right panels). Conspicuously, although H4^K20A expression caused only a partial decrease in H4K20me2/3 levels compared to the loss of Suv4‐20h, we found that 54% of replication‐timing alterations in 4OHT‐treated MEFs^364.2 were also detected in H4^K20A‐expressing MEFs^364.2, demonstrating that these replication defects are indeed caused by the loss of Suv4‐20h activity on H4K20me1.

Further analysis of replication‐timing delays in 4OHT‐treated MEFs^364.2 revealed that 80% of these delays occurred in late‐replicating domains (Fig 4C). These domains correspond to large genomic regions with a median size of 2.2 megabases (Fig 4D). Quantification of these data shows that 114 distinct late‐replication domains, which cover ~15% of the genome, displayed a significant replication‐timing delay upon loss of Suv4‐20h (P‐value < 1 × 10⁻³). In mammalian cells, late‐replication domains have been reported to be gene poor and often to coincide with heterochromatin regions (Rhind & Gilbert, 2013). In line with these results, the 114 altered replication domains exhibited the lowest gene density and coverage with activating chromatin mark H3K27 acetylation, but the highest coverage with the heterochromatin repressive mark H3K9me2 compared to stochastic or unmodified early/mid/late‐replication domains in Suv4‐20h‐null cells (Fig 4E–G, statistical analysis in Appendix Table S1). Altogether, these results show that Suv4‐20h activity on H4K20me1 is specifically required for the replication timing of late domains corresponding to heterochromatin regions.

High levels of H4K20me3 at origins are required for proper heterochromatin replication

Since H4K20me3 is a hallmark of heterochromatin (Schotta et al, 2004), we next examined by ChIP‐qPCR whether H4K20me3 is particularly enriched at replication origins in late‐replicating heterochromatin domains that were delayed in Suv4‐20h‐null cells. The results of four replication origins randomly selected in a delayed domain and of four origins in a non‐delayed late domain of the chromosome 11 are shown in Fig 5. Moreover, the results of 10 additional origins in different delayed and non‐delayed late domains are shown in Fig EV3A. The localization of all these origins was previously found by nascent DNA strand purification in MEF lines (Cayrou et al, 2011, 2015). Compared to a control origin in an early‐replicating domain and a flanking negative sequence (NEG) close to ORI‐1, the levels of H4K20me3 at origins in non‐delayed late domains were generally weak or close to background levels in all cellular conditions, whereas a peak of H4K20me3 was detected at almost all origins in delayed late domains before but not after 4OHT treatment (Figs 5A and EV3A). Consistent with these results, H4K20me1 was found highly enriched instead of H4K20me3 at these origins in 4OHT‐treated cells (Fig EV3B). Importantly, the decrease in H4K20me3 levels at origins in delayed late domains was also observed in H4^K20A‐expressing MEF^364.2 (Fig EV4A), thereby suggesting that the replication‐timing alterations upon loss of Suv4‐20h‐mediated H4K20 methylation principally occur in heterochromatin regions where origins are normally associated with high levels of H4K20me3.

To determine whether the lack of H4K20me3 has indeed an impact on the licensing and activity of these origins, wild‐type as well as H4^WT‐ and H4^K20A‐expressing MEFs^364.2 were transduced with retroviral vectors encoding FLAG‐tagged MCM2 or MCM5 and ChIP‐qPCR experiments were performed with FLAG antibody to measure pre‐RC formation at different origins. In parallel, we isolated and quantified by qPCR the levels of short nascent DNA strands at these origins (SNS), which is a direct measure of their efficiency (Valton et al, 2014). As shown in Figs 5B and C, and EV4B–D, all origins tested were normally associated with FLAG‐tagged MCM5 and/or MCM2 proteins and displayed high SNS levels in untreated or H4^WT cells. In contrast, whereas the loading of MCM proteins and the activity of origins in non‐delayed late domains largely remained unaffected, the abrogation of H4K20me3 upon 4OHT treatment or its decrease in H4^K20A cells coincided with the inability of H4K20me3‐associated origins to recruit appropriate levels of MCM proteins and to efficiently initiate DNA synthesis (Figs 5B and C, and EV4B–D). These results show that high levels of H4K20me3 are critical for the full activity of an important subset of late‐firing origins, thereby explaining the delay in heterochromatin replication observed in Suv4‐20h and H4K20 methylation‐depleted cells.

ORCA/LRWD1 binding to H4K20me3‐associated late‐firing origins is required for heterochromatin replication

Previous studies have shown that tri‐methylated H4K20 peptides can directly interact in vitro with the WD40 domain of the ORCA‐associated protein ORCA/LRWD1 (Beck et al, 2012b), which is suspected to play an important role in the replication and organization of heterochromatin by facilitating the recruitment of ORC (Wang et al, 2016). We therefore reasoned that H4K20me3 might ensure the proper replication timing of heterochromatin by inducing the recruitment of ORCA at specific late‐firing origins. To test this hypothesis, FLAG‐tagged ORCA was stably expressed in the different MEFs^364.2 lines and the ability of this protein to associate with origins in delayed and non‐delayed late domains of the chromosome 11 was evaluated by ChIP‐qPCR in Suv4‐20h‐null and H4^K20A‐expressing cells. Whereas FLAG‐tagged ORCA was found associated with all origins tested in untreated and H4^WT cells, the abrogation of H4K20me3 upon Suv4‐20h inactivation or H4^K20A expression coincided with the inability of ORCA to associate specifically with the set of origins that are normally highly enriched in this methylation mark (Figs 6A and EV5). Consistent with the absence of ORCA binding to this specific subset of late‐firing origins, which control the replication of ~15% of the mouse genome (Fig 4), the amount of chromatin‐associated pool of ORCA was reduced of 40% in Suv4‐20h‐null cells compared to control cells (Fig 6B). Altogether, these results suggest that ORCA may associate with origins through multiple mechanisms and that the binding of ORCA to H4K20me3 is specifically required for its association with a subset of late‐firing origins.

To determine whether the recruitment of ORCA to H4K20me3‐enriched origins is necessary for the activity of these origins and the proper replication of associated late domains, we induced the silencing of ORCA by two independent shRNA (Fig 6C) and then examined the replication timing and activity of H4K20me3‐associated origins in ORCA‐depleted cells as described above. Although ORCA was not fully depleted (Fig 6C), the replication of mid‐ and late‐replicating domains was delayed in shRNA ORCA‐treated cells, while early‐replicating domains remained unchanged or eventually slightly advanced (Fig 6D and Appendix Fig S3). Clearly, 64% of late‐replicating domains delayed in Suv4‐20h‐null and H4^K20‐expressing cells were also found significantly altered in ORCA‐depleted cells, as shown for the same fragment of chromosome 11 in the three cellular conditions (compare Figs 4A and B, and 6D). In line with these results, we found that ORCA depletion impairs the levels of MCM loading and replication initiation at H4K20me3‐associated origins (Fig 6E and F). Moreover, consistent with the ability of ORCA to interact with Suv4‐20h (Giri & Prasanth, 2015), ORCA depletion also caused a significant reduction of H4K20me3 at these origins, thereby indicating a mutual dependence between the binding of ORCA and the accumulation of high levels of H4K20me3 at these sites (Fig 6G). Taken together, these results suggest a model where the timely replication of a large fraction of heterochromatin regions in mammalian cells depends on the concerted activities of PR‐Set7 and Suv4‐20h in order to generate H4K20me3‐mediated ORCA binding at a subset of late‐firing origins and thus to enhance the licensing and efficiency of these origins.

Plasmids and shRNA constructs

pCDNA‐UAS(5X)‐LUC, pMSCV PR‐SET7^FY/AA, and PMSCV PR‐Set7^{FY/AA‐SETmut} were previously described (Tardat et al, 2010). pQCXIP H4^WT was generated by cloning the histone H4 cDNA and a C‐terminal 3× FLAG sequence as an Age1‐Not1 fragment into the pQCXIP vector (Clontech). pQCXIP H4^K20A substitution mutant was generated by site‐directed mutagenesis (Agilent). pQCXIP GAL4‐PRSet7 and pQCXIH 3xFlag‐ORC2, ORCA, MCM2, and MCM5 were generated as described for pQCXIP H4^WT. pDEST‐ZEO‐GAL4^DBD‐RfA plasmid was first generated by cloning GAL4^DBD and RfA sequences into pCDNA3‐ZEO and subsequently used to shuttle PRSet7^WT and PR‐Set7^SETmut sequence from pDON221 vectors using clonase LR in order to generate pZEO‐GAL4‐PR‐Set7^WT and pZEO‐GAL4‐PR‐Set7^SETmut. Control and ORCA shRNA lentiviral constructs (Sigma) were previously described (Chan & Zhang, 2012), and viral production and infection were performed as described in Tardat et al (2007).

Cell culture and establishment of stable cell lines

Cells were grown in Dulbecco's modified Eagle's medium (Invitrogen) with 10% fetal bovine serum (Sigma). For Suv4‐20h inhibition in human cells, A‐196 (Sigma) was added to a final concentration of 5 μM. Primary mouse embryonic fibroblast cells with inducible deletion of Suv4‐20h1 (E364.2) were established by standard procedures from Suv4‐20h1^flox/−; Suv4‐20h2^−/−; Cre‐ER embryos at E13.5 as described previously (Hayashi & McMahon, 2002; Schotta et al, 2008). As oxidative stress is associated with the loss of Suv4‐20h enzymes (Schotta et al, 2008), MEFs^364.2 were cultured under low oxygen conditions in a tri‐gas incubator containing 2% O₂ and 5% CO₂. FLAG‐tagged pre‐RC protein expression was established by transduction with the appropriate pQCXIP retroviral vectors. Deletion of floxed Suv4‐20h1 allele in MEF364.2 lines was achieved by treatment with 4‐hydroxytamoxifen (1 μM) for 48 h. Establishment of GAL4 and GAL4‐PR‐Set7WT and GAL4‐PRSet7^SETmut‐expressing HEK293 EBNA1⁺ cell lines were established by transfection of pZEO vectors followed by Zeocin selection (20 μg/ml).

Fluorescence‐activated cell sorting (FACS) analysis

Cells were prepared as described previously (Tardat et al, 2010). For cell‐cycle analysis, cells were washed with PBS and resuspended in PBS containing 7‐amino‐actinomycin D (Sigma) and RNase A (100 mg/ml) for 4 h at room temperature. Samples were run on a FACSCalibur (Becton Dickinson) and data analysis was performed using Flowjo software (Tree Star).

ChIP‐qPCR

ChIP experiments with MEFs were performed as described previously (Tardat et al, 2010). ChIP‐qPCR on EBV episomes was performed with transfected HEK293 EBNA1⁺ cell lines. Detailed methods and primer sequences used for qPCR are available in Appendix Supplementary Methods. ChIP was performed with 300 μg of sheared chromatin with the appropriate antibody as follows: anti‐GAL4 (Santa Cruz), anti‐FLAG M2 (Sigma), anti‐H4K20me1 (Diagenode), anti‐H4K20me3 (Abcam and Diagenode), anti‐ORCA (Bethyk), and anti‐MCM (Ritzi et al, 2003). Quantitative PCR analysis was performed using the Roche LightCycler 480 System and the SYBR Green I Master (Roche) according to the manufacturer's instructions. Quantitative PCR values were calculated as fold enrichment relative to isotype IgG control.

Nascent DNA strand purifications and analysis

Nascent strands (NS) were purified as described previously (Cayrou et al, 2011). Briefly, after extraction by DNAsol, NS are crudely separated from genomic DNA by sucrose gradient. Fractions of interest containing NS were then processed by the lambda exonuclease to eliminate contaminating DNA. Three samples were collected from three independent cell cultures. Control was obtained after treatment of the high sucrose fractions (which do not contain NS) by lambda exonuclease digestion, in order to remove lambda exonuclease artefactual digestion. NS levels at specific origins were then evaluated by qPCR using specific sets of primers (Appendix Table S2) with the Roche LightCycler 480 System and the SYBR Green I Master (Roche).

Immunoblot analysis and small‐scale biochemical fractionation

The following antibodies were used: anti‐PRSet7 (1:1,000, Cell Signaling), anti‐H4 and anti‐H4K20me1, me2, me3 (1:1,000, Cell Signaling), anti‐H4 acetylation (Active motif, 1:1,000), anti‐H3K9me3 (1:1,000, Abcam), anti‐FLAG (1:1,000, Sigma), anti‐H3K4me3 (1:1,000; Abcam), anti‐H3K27me3 (1:2,000; Upstate), anti‐GAL4^DBD (1:10,000, Santa Cruz), anti‐MEK1 (1:1,000, Millipore), anti‐HCF‐1 (1:5,000), anti‐ORCA (1:1,000, Bethyl), and anti‐MCM (Cell Signaling). Small‐scale biochemical fractionation to purify chromatin‐enriched fraction was prepared with 10⁷ cells as described previously (Tardat et al, 2010).

Plasmid replication assay

Detailed protocol is available in the Appendix Supplementary Methods. GFP‐positive reporter plasmids (1 μg) were transfected into HEK293 EBNA1⁺ cell line stably expressing the respective GAL4‐fusion protein using Lipofectamine 2000 (Life technologies). Transfections with comparable efficiencies were verified by visualizing GFP‐positive cells. Six days post‐transfection, cells were harvested according to the HIRT protocol (Gerhardt et al, 2006). After centrifugation (2,000 × g) for 1 h at 4°C, DNA was purified by phenol–chloroform extraction and digested with 40 U DpnI (NEB) in the presence of RNase (Roche). Digested DNA (300 ng) was electroporated into Electromax DH10B competent cells (Invitrogen) and ampicillin‐resistant colonies, representing the number of recovered plasmids, were counted. The FR‐DS plasmid was always transfected in parallel, and the number of resulting colonies was used for normalization. Statistical analysis of replication efficiency was performed using paired Student's t‐test.

Replication‐timing analysis

A detailed method is provided as Appendix Supplementary Methods. Briefly, 30 × 10⁶ cells were incubated with BrdU for 1 h before ethanol fixation. Fixed cells were sorted into early and late S‐phase fractions using INFLUX 500 (Cytopeia BD Biosciences). After lysis and anti‐BrdU immunoprecipitation, whole‐genome amplification was conducted (WGA, Sigma) and early and late neo‐synthesized DNAs were labeled with Cy3 and Cy5 ULS molecules (Genomic DNA Labeling Kit, Agilent). The hybridization was performed according to the manufacturer's instructions on 4 × 180K mouse microarrays (SurePrint G3 Mouse CGH Microarray Kit, 4 × 180K, AGILENT Technologies, reference genome: mm9). Microarrays were scanned with an Agilent High‐Resolution C Scanner using a resolution of 2 μm and the autofocus option. Feature extraction was performed with the Feature Extraction 9.1 software (Agilent technologies). Analysis was performed with the Agilent Genomic Workbench 5.0 software. The log₂ ratio timing profiles were smoothed using the Agilent Genomic Workbench 5.0 software with the Triangular Moving Average option (500‐kb windows). The analysis of replication‐timing data was performed using algorithms from CGH applications (Agilent Genomic Workbench 5.0 software), particularly the aberration detection algorithms (Z‐score with a threshold of 1.8) that define the boundaries and magnitudes of the regions of DNA loss or gain corresponding to the late‐ and early‐replicating domains, respectively. A Student test (t‐test) was performed on the average of the log² values of every domain with R program 3.2.3 and significant difference is annotated when P‐value < 10⁻³. The intersection with different data sets was performed with GALAXY tools and t‐test was performed to identify significant differences. The microarray data have been deposited in the Gene Expression Omnibus (GEO) (accession no. GSE69084). Positions of genes used for gene coverage come from RefSeq mm9. Positions of H3K27ac peaks used for coverage come from GSM1631248 (GEO database). Positions of H3K9me2 peaks used for coverage come from GSM887877 (GEO database).