Human DDK rescues stalled forks and counteracts checkpoint inhibition at unfired origins to complete DNA replication

Summary

Eukaryotic genomes replicate via spatially and temporally regulated origin firing. Cyclin-dependent kinase (CDK) and Dbf4-dependent kinase (DDK) promote origin firing, whereas the S phase checkpoint limits firing to prevent nucleotide and RPA exhaustion. We used chemical genetics to interrogate human DDK with maximum precision, dissect its relationship with the S phase checkpoint, and identify DDK substrates. We show that DDK inhibition (DDKi) leads to graded suppression of origin firing and fork arrest. S phase checkpoint inhibition rescued origin firing in DDKi cells and DDK-depleted Xenopus egg extracts. DDK inhibition also impairs RPA loading, nascent strand protection, and fork restart. Via quantitative phosphoproteomics, we identify the BRCA1-A complex subunit MERIT40 and the cohesin accessory subunit PDS5B as DDK effectors in fork protection and restart. Phosphorylation neutralizes auto-inhibition mediated by intrinsically disordered regions in both substrates. Our results reveal mechanisms through which DDK controls the duplication of large vertebrate genomes.

eTOC blurb

Eukaryote genomes duplicate via spatially and temporally regulated firing of replication origins. Using chemical genetics and phosphoproteomics, Jones et al. show that human DDK promotes origin firing by counteracting the S phase checkpoint, and also enables the uncoupling, protection, and restart of stalled forks.

Graphical Abstract

Introduction

Eukaryotes duplicate their genomes by initiating DNA synthesis at thousands of replication origins during S phase. A detailed picture of this process has been developed in Saccharomyces cerevisiae, where DNA replication has been not only dissected genetically but also reconstituted in vitro (Georgescu et al., 2015; Heller et al., 2011; Yeeles et al., 2015). The origin recognition complex (ORC) initially loads MCM complexes onto DNA as inactive double hexamers (Evrin et al., 2009; Remus et al., 2009). Thereafter CDK (cyclin-dependent kinase) and DDK (Dbf4-dependent kinase) trigger assembly of the active CMG (Cdc45 MCM GINS) helicase and recruitment of additional replisome components (Heller et al., 2011; Yeeles et al., 2015), resulting in hexamer separation and mutual bypass via lagging-strand exclusion (Abid Ali et al., 2016; Douglas et al., 2018; Georgescu et al., 2017; Yuan et al., 2016). Genetic and biochemical epistasis experiments indicate that intrinsically disordered regions (IDRs) at the N-terminus of MCM4 and MCM6 are critical DDK targets in CMG assembly (Deegan et al., 2016; Randell et al., 2010; Sheu and Stillman, 2010), while Sld2 and Sld3 are critical CDK targets (Tanaka et al., 2007; Zegerman and Diffley, 2007). Yeast DDK is also a major target of the S phase checkpoint kinase Rad53, which inhibits new origin firing in DNA-damaged cells by binding to and phosphorylating DDK’s Dbf4 subunit (Chen et al., 2013; Zegerman and Diffley, 2010).

Less is known about how DDKs and S phase checkpoint kinases control origin firing in vertebrates. While DDK is required for CMG assembly and DNA replication in Xenopus egg extracts (Silva et al., 2006; Takahashi and Walter, 2005), conditions that bypass this requirement have not been identified. Moreover, DDK activity stays high in Xenopus egg extracts and human somatic cells treated with etoposide or other inducers of replication fork arrest (Lee et al., 2012; Tenca et al., 2007; Tsuji et al., 2008). These observations imply that, in vertebrates, the S phase checkpoint uses mechanisms other than DDK inactivation to block new origin firing, potentially because these organisms also rely on DDK to stabilize and restart stalled replication forks.

To date two small-molecule inhibitors (PHA-767491 and XL413) have been used to investigate DDK function in human cells and Xenopus egg extracts (Chen et al., 2015; Dungrawala et al., 2015; Toledo et al., 2013; Vannier et al., 2013). However PHA-767491 potently inhibits Cdk9 (also called P-TEFb) (Montagnoli et al., 2008), which phosphorylates and activates RNA polymerase II for transcriptional elongation (Larochelle et al., 2012; Parua et al., 2018), and promiscuously inhibits other kinases (Hughes et al., 2012). As a result, the effects of this compound are difficult to interpret with high confidence. While XL413 is a more potent and selective inhibitor of DDK in vitro, it has been reported to have poor bioavailability in some cancer lines (Sasi et al., 2014) and inhibits DNA replication and cell proliferation less effectively than PHA-767491 (Alver et al., 2017; Rainey et al., 2017). For this reason PHA-767491 is used more frequently than XL413 in DDK-related experiments (Hiraga et al., 2017; Moiseeva et al., 2019; Sasi et al., 2018; Sparks et al., 2019).

To overcome these limitations, we created a chemical-genetic system for human DDK. In brief, isogenic CDC7-null cells were reconstituted with alleles that accept or reject bulky purine analogs as ATP-competitive inhibitors. As shown below, this approach affords vastly improved penetrance and empirically validates on-target specificity into each experiment. We report that DDK inhibition (DDKi) disproportionately affects origin firing, fork elongation, replication foci formation, and DNA synthesis in late-replicating regions of the genome, whereas early regions are much less sensitive. Asymmetric sensitivity to DDKi was not an intrinsic property of late origins, but rather imposed on them by the S phase checkpoint, as inhibiting either ATR or Chk1 restored CMG assembly, origin firing, and DNA replication in DDKi cells and DDK-depleted Xenopus egg extracts. Our studies also implicate DDK in multiple aspects of replication fork metabolism, including suppression of spontaneous fork arrest, formation of RPA-coated single-stranded DNA (ssDNA) tracts and protection of nascent strands during fork stalling, and efficient restart after nucleotide repletion. Using quantitative phosphoproteomics, we identify cohesin and the BRCA1-A complex as DDK effectors in these events and show that DDK phosphorylates and neutralizes unstructured auto-inhibitory modules in each complex, analogous to DDK activation of the MCM helicase (Sheu and Stillman, 2010). Taken together, our studies provide a comprehensive and substantially revised framework for DDK’s roles in vertebrate DNA replication.

Results

Human cells exhibit escalating sensitivity to DDKi across S phase

To enable tight chemical-genetic control over human DDK, we used adeno-associated virus (AAV)-mediated gene editing (Berdougo et al., 2009) to delete CDC7 conditionally in hTERT-immortalized retinal pigment epithelial (RPE1) cells, which are non-transformed and have a female diploid (46, XX) karyotype (Figure S1A-B). CDC7flox/Δ cells were transduced with retroviral constructs encoding wildtype (Cdc7^wt) or analog-sensitive (M134G, hereafter Cdc7^as) versions of the kinase. Both versions enabled recovery of CDC7-null clones after transient expression of Cre recombinase (Figure 1A). While Cdc7^wt and Cdc7^as cells proliferated at the same rate, the latter were sensitive to growth inhibition by 1-NM-PP1, a bulky purine analog (Figure 1B). To monitor DDK activity in vivo we used a monoclonal antibody that recognizes MCM2 (S40) phosphorylation. Cdc7^as phosphorylated this site with moderately (2- to 3-fold) lower efficiency than Cdc7^wt, similar to other analog-sensitive kinases (Burkard et al., 2009; Holland et al., 2010; Maciejowski et al., 2010). 1-NM-PP1 decreased this modification to undetectable levels in Cdc7^as cells but did not affect Cdc7^wt cells (Figure 1C). In this regard our chemical-genetic system is not only allele-specific but also more penetrant than currently available DDK inhibitors (Figure S1C). Crucially our system preserves Cdk9-cyclin T1 phosphorylation of RNA polymerase II, a known off-target and confounding effect of the PHA compound (Figure S1D).

Figure 1. Human cells exhibit an escalating requirement for DDK activity across S phase.

(A) CDC7 conditional–null and analog-sensitive RPE1 cells were generated as described in STAR Methods and Figure S1. Endogenous and transgene-encoded Cdc7 were detected by immunoblotting. FLAP designates a composite localization and purification tag (FLAG-EGFP-TEV-S-peptide). (B) Cell growth in the presence or absence of 1NM-PP1 (mean ± SD; data are from three experiments). (C) Kinetics of DDK inhibition (DDKi). Lysates from Cdc7^wt and Cdc7^as cells treated with 1NM-PP1 were blotted to detect total or serine 40-phosphorylated MCM2. Gel loading was varied to enable comparison over a wider dynamic range. Results are representative of three experiments. (D) DDKi inhibits DNA synthesis in late S phase cells. Cdc7^wt and Cdc7^as cells were treated with or without 1NM-PP1 for the indicated time periods and pulse labeled with BrdU for 30 min prior to fixation. Flow cytometry was used to monitor DNA content (PI staining) and rate of synthesis (BrdU). Overlay histograms display DNA synthesis in near-4N cells (late S/G2/M phase). Results are representative of three experiments. (E) DDKi blocks progression from early to late S phase replication foci. Cdc7^wt and Cdc7^as cells expressing EGFP-PCNA and mKO2–hCdt1 (30–120) were synchronized in G0 via serum starvation, then refed in the presence of 1NM-PP1 and followed by spinning disk confocal microscopy. Time 0 denotes S phase entry as judged by 50% loss of mKO2-hCdt1. (F) DDKi delays formation of early replication factories. Individual traces from Cdc7^wt (black) or Cdc7^as (red) cells treated with 1NM-PP1 were used to compare the kinetics of mKO-hCdt1 degradation and EGFP-PCNA foci formation. (G) The number and intensity of PCNA foci in early S phase cells (t=100 to 300 min) and late S phase cells (t=400 to 1000 min) were compared using one-way ANOVA and Holm-Sidak tests. Error bars indicate SEM. Data were quantified from timelapse recordings of 8 cells per condition.

BrdU labeling and flow cytometry were used to examine transit through S phase (Figure 1D). Untreated Cdc7^as cells and Cdc7^wt cells with or without 1-NM-PP1 had the expected “flat” profile, in which the extent of BrdU incorporation was constant across S phase. In contrast Cdc7^as cells treated with 1-NM-PP1 exhibited a “sharkfin” profile, in which BrdU incorporation rose at the G1/S transition but then declined as DNA content (i.e., the fraction of the genome already replicated) increased (Figure 1D). Over time, these DDK-inhibited (DDKi) cells approached ~4N DNA content but did not complete DNA synthesis or enter mitosis. As a complementary approach we expressed and imaged the FUCCI sensor mKO2-Cdt1 (30–120) (Bainor et al., 2018; Sakaue-Sawano et al., 2008) and EGFP-PCNA in cells that had been synchronized by serum starvation and refed in the presence of 1-NM-PP1. The midpoint of Cdt1 degradation was used to identify the G1/S boundary, while PCNA foci were used to monitor DNA replication. Control cells initiated DNA replication in concert with Cdt1 destruction and finished it 650 ± 30 min later (n=14; Figure 1E). PCNA initially formed small foci throughout the nucleus but then redistributed into large foci at the nuclear and nucleolar peripheries, reflecting a transition from early to late replication (Leonhardt et al., 2000). DDKi delayed (but did not prevent) the formation of small PCNA foci (Δt = 130 mins) (Figure 1F). However large peripheral foci were not seen (n=12), even in cells tracked for ≥1000 min after Cdt1 destruction (n=7) (Figure 1E and 1G).

These findings prompted us to examine the impact of DDKi on replication across the genome. We first generated an RPE1-specific replication profile by flow sorting G1 and S phase cells and sequencing their genomes to measure locus-specific changes in copy number (Figure 2A and S2A-B) (Koren et al., 2012). Separately we sequenced the genomes of Cdc7^wt and Cdc7^as cells synchronized at the G1/S boundary via double-thymidine block and released in the presence or absence of 1-NM-PP1 (Figure 2A and S2B). Using a stringent significance threshold (p=0.01) we identified 230 regions whose extent of duplication changed by at least one standard deviation after DDKi (Table S1 and Figure S2C). Ninety percent (n=206) were under-represented regions, spanned one or several Mb, and correlated with late replication in unsynchronized RPE1 cells. In contrast over-represented regions (n=24) were much shorter and early replicating (Figure 2B-C, Figure S2C, and Table S1). The late-replicating X chromosome was especially sensitive to DDKi, as 78 Mb (50% of the entire chromosome) was under-replicated (Table S1). Our studies indicate that DNA replication becomes increasingly DDK-dependent as human cells transit through S phase. In contrast budding yeast DDK activates origins throughout S phase but becomes dispensable for DNA replication and cell division once early origins have fired (Bousset and Diffley, 1998; Donaldson et al., 1998).

Figure 2. DDKi causes under-replication of late regions and accumulation of stalled forks.

(A) Replication of chromosome 4 in flow-sorted RPE1 cells (black) and Cdc7^as cells (blue) released from double-thymidine block in the presence of 1-NM-PP1. Significantly under-replicated regions are shaded. See also Figure S2 and Table S1. Data are from two biological replicates. (B) Distribution of under-replicated regions in DDKi cells versus genomewide replication timing (hg19). (C) Cdc7^wt and Cdc7^as cells were pulse labeled with EdU and chased through S phase in the presence of 1NM-PP1. At various times cells were fixed and stained with antibodies to FANCD2 (red) and 53BP1 (green). Images are maximum-intensity projections of deconvolved z stacks and are representative of three experiments. (D) Quantification of FANCD2 and 53BP1 foci in DDKi and control cells (N=3 experiments). Error bars denote SEM.

DDKi causes replication forks to stall without recruiting RPA

To understand DDK’s graded impact on DNA synthesis, we first asked if DDKi cells accumulate stalled forks. Cdc7^wt and Cdc7^as cells were pulsed with EdU, chased through S phase in the presence of 1-NM-PP1, and then stained with antibodies that detect FANCD2 and 53BP1, both early responders at stalled forks (Dungrawala et al., 2015; Lossaint et al., 2013). About 20 FANCD2 foci were present in both Cdc7^wt and Cdc7^as cells at the time of EdU labeling, reflecting sites of spontaneous fork stalling during unperturbed replication (Figure 2C-D). Cdc7^wt cells resolved these foci during the 1-NM-PP1 chase, demonstrating that fork stalling was reversible (Figure 2C-D). By contrast, Cdc7^as cells accumulated FANCD2 foci and 53BP1 foci during the 1-NM-PP1 chase, indicating that DDKi led to protracted and likely irreversible fork stalling (Figure 2C-D).

Forks stall because of impediments to duplex DNA unwinding or nascent strand synthesis. Due to helicase-polymerase uncoupling, the latter mode of fork stalling generates long tracts of ssDNA that become coated by RPA (Byun et al., 2005). To assess this response we co-stained for FANCD2 and RPA and performed quantitative image analysis. As a positive control for uncoupling, we depleted dNTPs with hydroxyurea (HU). HU triggered widespread recruitment of RPA to FANCD2 foci, as reflected in their congruence on linescans and positive correlation over the entire nucleus (Pearson coefficient R=0.76 ± 0.05 (mean ± SEM, n=9 cells) versus R=0.49 ± 0.07 in untreated cells (n=10); Figure 3A). In contrast, treatment with the interstrand crosslinking agent mitomycin C (MMC) did not induce RPA recruitment to FANCD2 foci (R=0.49 ± 0.04, n=10; Figure 3A). This finding is consistent with the fact that interstrand crosslinks stall forks without inducing helicase-polymerase uncoupling (Huang et al., 2010) and shows that our RPA-FANCD2 colocalization assay can discriminate between different modes of fork stalling.

Figure 3. Replication forks require DDK to efficiently recruit RPA and activate ATR in response to nucleotide depletion.

(A) Cdc7^as cells were pulse labeled with EdU (30 min), treated as specified, then fixed and stained with antibodies to FANCD2 and RPA2. Maximum-intensity projections of deconvolved z stacks and linescans are shown. Images are representative of three experiments. See also Figure S3. (B) Cdc7^as cells were treated with 1NM-PP1 or 2 mM hydroxyurea (HU). ATR activation was assessed with pS345-Chk1 and pS33-RPA2 antibodies. Where indicated, gel loading was varied to enable comparison over a wider dynamic range. (C-D) HU was added to Cdc7^as cells after 2 h in 1NM-PP1. (E) Cdc7^wt and Cdc7^as cells were treated with 1NM-PP1 with or without ATRi for the final hour of treatment. DNA-PK activation was assessed using pS4/8-RPA2 antibodies. (F) Cdc7^as cells were treated with or without 1NM-PP1 for 24 h. Where indicated ATRi, RO-3306 (CDK1i), or roscovitine (CDK2i) was added for the last hour. Results are representative of three experiments.

RPA was not recruited to FANCD2 foci in DDKi cells (R =0.44 ± 0.07, n=10), even in response to HU (R =0.46 ± 0.09, n=15) (Figure 3A). Using XL413 as an orthogonal DDK inhibitor, we observed identical epistasis in HU-treated HCT116 cells (Figure S3A, upper panels). We conclude that human DDK is required after origin firing for template unwinding and buildup of RPA-coated ssDNA at stalled forks.

ATRi enables RPA recruitment to stalled forks in DDKi cells

In line with decreased RPA recruitment at stalled forks, DDKi induced little if any increase in ATR-dependent phosphorylation on its own and suppressed ATR activation in response to HU, as evidenced by lower levels of RPA2 (S33) and Chk1 (S345) phosphorylation (Figure 3B-D). In DDK-proficient cells, ATR limits RPA loading by suppressing fork collapse and DSB end resection, as well as new origin firing (Toledo et al., 2013). To determine if ATR acts similarly in DDK-inhibited cells we used VE-821, a validated ATR-selective inhibitor (Reaper et al., 2011). ATR inhibition (ATRi) restored RPA colocalization with FANCD2 (Figure 3A) and induced DNA-PK phosphorylation on RPA2 (S4/S8), an indicator of DSB end resection (Figure 3E). Treatment with R0–3306 (CDK1 inhibitor) or roscovitine (a CDK2 inhibitor) suppressed this modification, consistent with CDK-dependent activation of fork-cleaving (GEN1, MUS81) and end-resecting (MRE11-RAD50-NBS1, CTIP, EXO1, DNA2) nucleases (Figure 3F) (Pasero and Vindigni, 2017; Pfander and Matos, 2017; Stracker and Petrini, 2011). As a complementary approach we utilized CDK2AF/AF cells, in which the S phase checkpoint cannot inhibit CDK2 via T14 and Y15 phosphorylation but otherwise remains intact (Hughes et al., 2013). CDK2AF/AF cells were highly proficient at cleaving and resecting HU-stalled forks as judged by RPA2 (S4/S8) phosphorylation and DDK-independent colocalization of RPA and FANCD2 (Figure S3A-B). We conclude that DDK promotes template unwinding, RPA recruitment, and ATR activation at stalled forks. In contrast, DDK is not required for fork collapse, end resection, and RPA loading when ATR is inhibited or unable to downregulate CDK2.

ATRi induces origin firing in DDKi cells and DDK-depleted Xenopus egg extracts

As the apical kinase of the S phase checkpoint, ATR not only stabilizes stalled forks but also inhibits new origin firing elsewhere in the genome (Saldivar et al., 2017). Several groups have reported that origin firing in ATR-inhibited cells requires DDK, based on its sensitivity to PHA-767491 (Dungrawala et al., 2015; Moiseeva et al., 2017; Moiseeva and Bakkenist, 2019; Toledo et al., 2013). On the other hand, deleting mrc1+ (claspin) or cds1+ (Chk2) bypasses the requirement for hsk1+ (DDK) in fission yeast, implying that the S phase checkpoint is epistatic (Matsumoto et al., 2011). In light of the PHA compound’s poor selectivity we examined the relationship between DDK and ATR using other approaches. We first used Xenopus egg extracts, which require DDK to assemble the CMG helicase and load RPA onto unwound origin DNA, thus initiating DNA replication (Silva et al., 2006; Takahashi and Walter, 2005). Cdc7-Drf1 and Cdc7-Dbf4 were immunodepleted from low-speed supernatant (LSS), after which sperm chromatin and [α−32P]dATP were added to monitor DNA replication (Figure 4A-C). DDK-depleted extracts did not hyperphosphorylate MCM4 or load Cdc45 and RPA onto chromatin (Figure 4B) and replicated DNA poorly compared to mock-depleted extracts (Figure 4C). However, caffeine (an inhibitor of xATR and xATM) rescued all three defects (Figure 4B-C).

Figure 4. S phase checkpoint inhibition induces CMG assembly, origin firing, and DNA replication in DDK-depleted Xenopus egg extracts and rescues late replication in DDKi cells.

(A) Immunodepletion of major (xDrf1) and minor (xDbf4) DDK activators from Xenopus egg extract low speed supernatants (LSS). Where indicated, lambda phosphatase was added to collapse heterogeneously phosphorylated proteins into discrete bands. (B) Sperm DNA was incubated in mock- or xDrf1/xDbf4-depleted extracts. Where indicated, caffeine was added to inhibit xATR and xATM. Chromatin pellets were analyzed by SDS-PAGE and immunoblotting. (C) Replication of sperm chromatin was monitored by [α-³²P] dATP incorporation as described (Takahashi et al., 2008; Takahashi and Walter, 2005). Results from five experiments (mean ± SD) were compared using two-way ANOVA and Dunnett’s tests. (D) ATRi restores late replication in DDKi cells. Cdc7^as cells were treated with 1-NM-PP1 in the presence or absence of ATRi (VE-821) or ATMi (KU-60019) for 6 h and pulse labeled with BrdU for the final 30 min. DNA content and rate of synthesis were assessed by flow cytometry. Overlay histograms compare BrdU incorporation in near-4N cells. Results are representative of three experiments. See also Figure S4. (E) Cdc7^as cells were treated with or without 1NM-PP1 and ATRi for 6 h. DDK phosphorylation was assessed by immunoblotting for total or serine 40-phosphorylated MCM2. Gel loading and blot exposures were varied to facilitate comparison over a wider dynamic range.

To confirm and extend these results, we compared the effects of selective ATRi or ATMi (using KU-60019) on Cdc7^as cells treated with 1-NM-PP1. Only ATRi stimulated BrdU incorporation in late S phase (Figure 4D). Crucially ATRi did not restore MCM2 (S40) phosphorylation, implying that rescue was not mediated by DDK reactivation (Figure 4E). Experiments with chemically unrelated DDK (XL413), ATR (AZ20), and Chk1 (AZD7762) inhibitors yielded identical results (Figure S4A-C). However, DNA synthesis in CDK2AF/AF cells remained sensitive to DDKi, indicating that ATR and Chk1 can block late replication through CDK2-independent mechanisms (Figure S3C).

Next we analyzed origin firing and fork dynamics by labeling replication tracts on extended DNA fibers. Cells were sequentially pulsed with two halogenated nucleosides (IdU and CldU), after which genomic DNA was stretched on glass slides and probed with IdU- and CldU-specific antibodies (Figure 5A). Through microscopic examination of hundreds of fibers, we quantified fork stalling/termination (IdU-only tracts), origin firing (CldU-only and IdU-flanked CldU tracts), and fork velocity (IdU-CldU tracts). DDKi did not reduce fork velocity (Figure 5B), but instead caused spontaneous fork stalling/termination and decreased origin firing as cells reached late S phase (Figure 5C-D). Inhibiting ATR after 9 hours of DDKi triggered massive origin firing and a reciprocal decrease in fork velocity, but did not affect stall/termination rates (Figure 5B-D). Collectively these results suggest that vertebrate DDKs promote DNA replication by counteracting the ATR-dependent block to origin firing, rather than by triggering initiation directly.

Figure 5. Stalled forks require DDK for nascent strand protection and direct restart.

(A) Cells were pulse-labeled with IdU and CldU for 20 min each, after which extended DNA fibers were prepared and immunostained to detect individual replication tracts. Examples of initiation, elongation, and stall/termination events are shown. (B) DDKi does not affect fork velocity. Data points (n≥280 per condition) and means ± SD are from two experiments. (C-D) ATRi restores origin firing but not fork stalling in DDKi cells. Stalled/terminated forks (IdU-only tracts) and newly fired origins (CldU-only tracts) were quantified as a percentage of all replication tracts (n≥300 tracts per condition from two experiments). Means ± SEM are plotted. P-values were computed using the chi-square test. (E) Fork protection and restart assay. Cells were pulsed with IdU, treated with 1NM-PP1 and/or HU for 5 h, then washed and pulsed with CldU. Examples of fork restart, nascent strand degradation, and irreversible fork arrest are shown. (F) DDK is required for nascent strand protection. IdU tract lengths in HU-arrested Cdc7^as and Cdc7^wt cells (n≥300 per condition) and RPE1 cells (n≥100 per condition) were compared using a Kruskal-Wallis test. Data points and means ± SD are plotted. (G) The frequency of irreversible fork arrest/collapse after HU washout (IdU-only tracts) was determined from at least 300 tracts per condition from two experiments. Means ± SD are plotted.

DDK protects nascent strands and promotes restart at stalled forks

During protracted fork stalling, nascent strands form nuclease-resistant structures that are important intermediates in fork restart (Bhat and Cortez, 2018; Quinet et al., 2017; Rickman and Smogorzewska, 2019). To investigate DDK’s roles in these events, we treated Cdc7^wt and Cdc7^as cells with HU for 5 hours between the IdU and CldU pulses (Figure 5E). DDKi resulted in loss of ~8 kb of nascent DNA during the HU arrest and a 5-fold increase in restart failures (IdU-only tracts) after HU washout (Figure 5F-G). We conclude that DDK protects nascent strands during fork stalling and promotes their extension via direct restart.

Phosphoproteomics reveals DDK effectors in fork protection and restart

To date few DDK substrates other than the MCM helicase have been identified, and none has been implicated in fork protection or restart. To identify such factors, we used quantitative mass spectrometry to compare the phosphoproteomes of control and DDKi cells during fork stalling. In brief, Cdc7^wt and Cdc7^as cells were labeled with heavy or light SILAC amino acids, arrested in mid-S phase via thymidine block, and treated with or without 1-NM-PP1 for 2 hours (Figure 6A). Over 13,000 phosphosites were identified and quantitated by nanoflow liquid chromatography and tandem mass spectrometry (nLC-MS/MS) in two biological replicates. We identified 97 sites on 54 proteins that were significantly (≥ 2-fold change and P ≤ 0.01) and reproducibly downregulated by 1-NM-PP1 in Cdc7^as cells, but not in Cdc7^wt cells (Supplementary Data File 1). Validating our approach, known DDK sites on the N-terminus of MCM2 and MCM4 were downregulated 40- to 500-fold. In addition, we identified DDK-regulated sites on factors involved in DNA replication and repair (e.g., CTF4, UBR5, APLF, TNKS1BP1, MERIT40), chromatin organization and gene expression (PDS5B, SIRT1, ARID3A, SET, CEBPZ, MED14, ZBTB11) and centriole/centrosome function (NEURL4, AKAP12) (Figure 6B).

Figure 6. Quantitative phosphoproteomics identifies DDK targets, including cohesin and BRCA1-A complex subunits involved in fork protection and restart.

(A) SILAC-based screen for unbiased discovery of DDK substrates and effectors. Cdc7^wt and Cdc7^as cells were grown in medium containing light (Arg⁰ and Lys⁰) or heavy (Arg¹⁰Lys⁸) amino acids, arrested in S phase with thymidine for 16 h, and treated with 1NM-PP1 for 2 h. Two biological replicates were analyzed. (B) Phosphopeptides are plotted as fold change with/without DDKi versus signal intensity. Red dots indicate phosphopeptides that were significantly and selectively downregulated by 1-NM-PP1 in Cdc7^as cells in both replicates. (C) Motif analysis of high-confidence DDK-regulated phosphorylation sites (n=97). (D) The BRCA1-associated deubiquitinase (DUB) complex detects and dismantles K63-linked ubiquitin chains at DNA damage sites. DDK phosphorylation sites at the N-terminus of MERIT40 are highlighted. (E) The cohesin complex entraps chromatin fibers to mediate sister chromatid cohesion and chromatin looping. PDS5 proteins regulate the dynamics of entrapment and release via interactions with the ATPase head of Smc3, the α-kleisin Rad21/Scc1, and SA2/Scc3. PDS5B’s isoform-specific C-terminal AT hook and DDK phosphorylation sites are highlighted. (F-H) Nascent strand protection (F and G) and fork restart assays (H) were performed in wildtype, MERIT40^−/−, and PDS5B^−/− RPE1 cells. Where indicated, wildtype, nonphosphorylatable (4A), and IDR-deleted (ΔN, ΔC) versions of MERIT40 and PDS5B were expressed as transgenes. Data points and means ± SD are plotted and come from two experiments. See also Figure S6.

DDK typically modifies serines next to acidic or already-phosphorylated serine residues (Cho et al., 2006; Montagnoli et al., 2006; Wan et al., 2008). This pattern was observed at 44% of DDK-regulated phosphosites, suggesting they were direct DDK targets, while 31% were proline-directed sites and thus likely CDK targets indirectly affected by DDKi (Figure 6C). Using recombinant DDK, we confirmed direct and site-specific phosphorylation of MERIT40, PDS5B, CTF4, UBR5, SET, and NEURL4 in vitro (Figure S5B).

DDK promotes fork protection and restart by phosphorylating intrinsically disordered auto-inhibitory modules in cohesin and the BRCA1-A complex

The N-terminal DDK sites in MCM2, MCM4, and MCM6 are embedded within IDRs that also contain acidic repeats (Parker et al., 2019; Sheu et al., 2014). Similar IDRs and acidic repeats were present in many of the DDK targets detected in our screen (Figure S5A). In light of the fork protection and restart defects in DDKi cells, we focused on the N-terminal IDR of MERIT40 (a subunit of the BRCA1-associated (BRCA1-A) deubiquitinase complex) and the C-terminal IDR of PDS5B (an accessory subunit of cohesin), as both are important for repairing DNA damage caused by replication stress (Brough et al., 2012; Couturier et al., 2016; Jiang et al., 2015; Kyrieleis et al., 2016). PDS5 orthologs were recently implicated in fork progression and nascent strand protection, but their regulation and potential non-redundancy remain obscure (Carvajal-Maldonado et al., 2019; Morales et al., 2020). We used CRISPR/Cas9 to target MERIT40 and PDS5B in RPE1 cells, and thereafter reconstituted each knockout with wildtype (WT) or non-phosphorylatable (4A) transgenes (Figure S5G-H). Basal fork speeds were normal under all conditions (Figure S5I-J). However both MERIT40 and PDS5B required DDK phosphorylation in order to protect (and in the case of MERIT40, to restart) HU-stalled forks (Figure 6F-H). On the other hand PDS5A, which lacks PDS5B’s C-terminal AT hooks and IDR, was dispensable for fork protection (Fig S5E-F). MERIT40 and PDS5B transgenes lacking their respective IDRs efficiently rescued fork protection and restart (Figure 6F-H). These findings suggest that human DDK activates these factors for fork protection and restart by alleviating IDR-mediated auto-inhibition.

Discussion

DDK is required for DNA replication in metazoans, but an accurate understanding of its roles, substrates, and relationships with other S phase kinases has been hampered by the promiscuity and modest potency of existing DDK inhibitors. To overcome this limitation, we generated a chemical genetic system for inhibiting DDK in an allele-specific manner. The high penetrance and selectivity of this system revealed DDK’s critical roles in counteracting the S phase checkpoint at unfired origins and in remodeling stalled forks into platforms for checkpoint signaling, nascent strand protection, and direct restart. Using quantitative phosphoproteomics, we identified a vastly expanded set of DDK substrates in human cells, including subunits of cohesin and the BRCA1-A complex that protect and restart stalled forks in a phosphorylation-dependent manner. Taken together, these results lead to a revised understanding of how DDK ensures the complete duplication of vertebrate genomes.

Like other eukaryotes, vertebrates duplicate their genomes with recurrent timing and spatial patterning during S phase (Fu et al., 2018; Marchal et al., 2019; Rhind and Gilbert, 2013). Current models for replication timing emphasize the role of higher-order chromatin organization and nuclear architecture (e.g. topologically associated domains versus lamin-associated domains) in establishing early- and late-replicating nuclear compartments after mitotic exit (Pope et al., 2014; Ryba et al., 2010; Sima et al., 2019). However, compartment-specific differences in the requirements (or thresholds) for initiating replication remain elusive. Our findings suggest that late-replicating regions require high DDK activity to overcome an ATR- and Chk1-dependent block to origin firing, whereas origins in early-replicating domains are not checkpoint-inhibited and thus require little if any DDK activity for origin firing. This interpretation is consistent with the genetics of DDK bypass in fission yeast (Matsumoto et al., 2011) and can also explain our observations in Xenopus egg extracts, which normally require DDK to initiate DNA replication (Silva et al., 2006; Takahashi and Walter, 2005). In this system xATR inhibition not only reinstated CMG assembly in DDK-depleted extracts but also induced a different MCM4 phosphoshift than seen in mock-depleted extracts (Figure 4B), suggesting MCM4 had been activated by a checkpoint-sensitive kinase. One candidate is CDK2, which phosphorylates MCM4 and other subunits in parallel with DDK. However checkpoint-uncoupling mutations in CDK2 (T14A Y15F) were not sufficient to rescue DDK inhibition, consistent with recent reports that ATR and Chk1 block origin firing through multiple substrates, including FANCI and Treslin (Chen et al., 2015; Guo et al., 2015).

We also show that human cells require DDK after initiation to remodel, stabilize, and restart stalled forks, thus explaining why vertebrate DDKs remain active during replication stress, unlike the yeast kinase (Lee et al., 2012; Tenca et al., 2007; Tsuji et al., 2008). DDKi caused elongating forks to stall spontaneously without recruiting RPA and also blocked RPA recruitment at hydroxyurea-stalled forks, implying a primary defect in template unwinding. We speculate that DDK continuously phosphorylates the CMG helicase after initiation, thereby maintaining its helicase activity throughout elongation, and that this mechanism is especially important during uncoupling, when the helicase switches to a much slower mode of dsDNA unwinding (Sparks et al., 2019).

Consistent with RPA’s central role in checkpoint signaling and fork protection (Bhat and Cortez, 2018; Saldivar et al., 2017), DDKi failed to activate ATR strongly on its own, antagonized ATR activation in response to hydroxyurea, and provoked nascent strand degradation at hydroxyurea-stalled forks. DDKi cells nonetheless had sufficient ATR activity to block fork collapse and new origin firing, consistent with the proposal that ATR is basally activated by RPA on the lagging strand (Saldivar et al., 2017) and inhibits origin firing during normal replication (Moiseeva et al., 2017; Moiseeva et al., 2019; Ruiz et al., 2016). Our results do not exclude the possibility that DDK and ATR cooperate in the context of replisome stabilization (Alver et al., 2017).

Current models of fork protection propose that RPA initiates fork protection via recruitment and regulation of fork remodelers, leading to fork reversal (Berti et al., 2020; Quinet et al., 2017; Tye et al., 2020). Thereafter BRCA1 and BRCA2 recruit Rad51 and promote nucleofilament assembly (Bhat and Cortez, 2018; Rickman and Smogorzewska, 2019). In addition to protecting remodeled forks, Rad51 enables fork restart via homologous recombination (Bhat and Cortez, 2018). Defects in fork protection make nascent strands vulnerable to degradation by various nucleases, but the specific requirements vary across genetic backgrounds (Berti et al., 2020; Tye et al., 2020). We did not detect a positive role for DDK in resecting stalled or collapsed forks in RPE1 cells, which are wildtype for BRCA1 and BRCA2. However, recent reports suggest that loss of BRCA1 or BRCA2 makes fork resection sensitive to DDKi due to synthetic effects on nuclease stability and localization (Rainey et al., 2020; Sasi et al., 2018). While additional work is needed to address these issues, our findings establish DDK as a central regulator of fork protection and identify MERIT40 and PDS5B as relevant substrates.

IDRs have long been recognized to play critical roles in signal transduction, as the same polypeptide can engage in different protein-protein interactions in response to upstream inputs, including phosphorylation (Oldfield and Dunker, 2014; Wright and Dyson, 2015). At the mesoscale, IDR-ligand interactions drive the assembly of membraneless organelles via phase separation (Boeynaems et al., 2018; Shin and Brangwynne, 2017). In these contexts, deleting IDRs or mutating phosphorylation sites therein results in loss of protein-protein interactions, organellar integrity, and biological function. However, the IDRs in PDS5B and MERIT40 were dispensable for fork protection and restart, and mutation of their DDK phosphoacceptor sites did not affect cohesin or BRCA1-A complex assembly or subcellular targeting to stalled replication forks (Figure S6). Taken together, these results imply that DDK activates latent fork-protection and restart activities in cohesin and the BRCA1-A complex via a mechanism that is conceptually similar to MCM helicase activation. IDRs and acidic repeats were also found in other identified DDK substrates that function outside of DNA replication (e.g., gene expression, chromatin organization, and centriole assembly), suggesting DDK may use the same principle to activate other cellular programs during S phase.

STAR Methods

RESOURCE AVAILABlLITY

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Prasad Jallepalli (jallepap@mskcc.org).

Materials Availability

Unique/stable reagents generated in this study are available from the lead Contact, Prasad Jallepalli (jallepap@mskcc.org), with a completed material transfer agreement.

Date and Code Availability

Mass spectrometry data are available via the ProteomeXchange Consortium (http://www.proteomexchange.org) using the dataset identifier PXD014399.

Replication timing data are available via the Sequence Read Archive (https://www.ncbi.nlm.nih.gov/sra) using the BioProject identifier PRJNA684004.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Cell lines

Cell lines used in this study are listed in the Key Resources Table. RPE1 cells were grown at 37°C in a 1:1 mixture of DMEM and Ham’s F-12 medium with 10% fetal bovine serum, 100 U/ml penicillin, and 100 U/ml streptomycin, and 2.5 mM L-glutamine. RPE1-derived CDC7L1flox/Δ cells were generated via two rounds of AAV-mediated gene targeting. Cdc7^wt and Cdc7^as cells were generated via retroviral expression of EGFP-fused Cdc7, infection with AdCre, and limiting dilution. PDS5A−/−, PDS5B−/−, and MERIT40−/− cells were generated via CRISPR/Cas9 gene editing and limiting dilution. Subsequently PDS5B−/− and MERIT40−/− cells were transduced with retroviral vectors that express wildtype (WT), non-phosphorylatable (4A), or truncated (ΔC, ΔN) versions of PDS5B or MERIT40. Retroviral transduction was also used to express these proteins in HeLa cells. Wildtype and CDK2AF/AF HCT116 cells have been described (Hughes et al., 2013). All non-RPE cell lines were cultured in DMEM with 10% fetal bovine serum, 100 U/ml penicillin, and 100 U/ml streptomycin, and 2.5 mM L-glutamine.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
α-tubulin	Santa Cruz	Cat# sc-5286
RPA2	Abcam	Cat# ab2175
RPA2 pS4/S8	Abcam	Cat# Ab87277
RPA2 pS33	Bethyl	Cat# A300-246A
FANCD2	Novus	Cat# NB100-183
MCM2 pS40	Abcam	Cat# Ab133243
MCM2	Bethyl	Cat# A300-191A
CHK1	Santa Cruz	Cat# sc-8408
CHK1 pS345	Cell Signaling	Cat# 13303
PDS5B	Bethyl	Cat# A300-537A
PDS5A	Bethyl	Cat# A300-088A
MERIT40/NBA1	Bethyl	Cat# A302-516A
BRCC36	Abcam	Cat# Ab108295
CDC7	MBL	Cat# K0070-3
53BP1	BD Biosciences	Cat# 612522
BrdU (B44)	BD Biosciences	Cat# 347580
BrdU/CldU (BUI/75)	Abcam	Cat# Ab6326
Abraxas/CCDC98	Abcam	Cat# Ab139191
BRE/BRCC45	Abcam	Cat# Ab95985
MCM7	Santa Cruz	Cat# G966
RNA polymerase II C-terminal repeat domain (CTD) pS2	Abcam	Cat# Ab5095
SMC3	Bethyl	Cat# A300-060
Rad21	Bethyl	Cat# A700-052
BRCA1	Santa Cruz	Cat# sc-6954
Anti-FLAG (M2) antibody	Sigma	Cat# F1804
Anti-FLAG (M2) antibody-coupled magnetic beads	Sigma	Cat# M8823
Goat anti-mouse-Alexa 488 conjugate	Invitrogen	Cat# A11029
Goat anti-mouse-Alexa 568 conjugate	Invitrogen	Cat# A11004
Goat anti-rabbit-Alexa 488 conjugate	Invitrogen	Cat# A11008
Goat anti-rabbit-Alexa 568 conjugate	Invitrogen	Cat# A11011
Bacterial and Virus Strains
BL21 (DE3) competent E. coli	NEB	Cat# C2527
DH10B competent E. coli	Invitrogen	Cat# 18297010
AdCre	Vector Development Lab, Baylor College of Medicine	N/A
Chemicals, Peptides, and Recombinant Proteins
BrdU	Sigma	Cat# B5002 CAS: 59-14-3
hydroxyurea	Sigma	Cat# H8627 CAS: 127-07-1
PHA-767491	Selleckchem	Cat# S2742 CAS: 942425-68-5
1-NM-PP1	Kevan Shokat	N/A
VE-821	Cayman Chemicals	Cat# 17587 CAS: 1232410-49-9
AZ20	Sigma-Aldrich	Cat# SML-1328 CAS: 233339-22-4
AZD-7762	Sigma-Aldrich	Cat# SML-0350 CAS: 1246094-78-9
roscovitine	Sigma-Aldrich	Cat# R7772 CAS: 186692-46-6
RO-3306	Millipore	Cat# 217699 CAS: 872573-93-8
KU-60019	Apex Bio Technology	Cat# 501014294 CAS: 925701-49-1
caffeine	Sigma-Aldrich	Cat# C0750 CAS: 58-08-2
Fugene 6	Roche	E2691
3x FLAG Peptide	Sigma-Aldrich	Cat# F4799
Cdc7-Dbf4	Mark Frattini	N/A
S-protein agarose	Millipore	Cat# 69704
Mitomycin C	Sigma-Aldrich	Cat# M4287 CAS: 50-07-7
Critical Commercial Assays
Click-iT EdU Alexa Fluor 647 Imaging Kit	ThermoFisher	Cat# C10340
Gibson Assembly master mix	NEB	Cat# E2611L
Gateway LR Clonase II enzyme mix	ThermoFisher	Cat# 11791020
Gateway BP Clonase II enzyme mix	ThermoFisher	Cat# 11789100
Pierce GST spin purification kit	ThermoFisher	Cat# 16106
Deposited Data
Phosphoproteomics data	ProteomeXchange Consortium	dataset PXD014399
Replication timing data	Sequence Read Archive	BioProject PRJNA684004
Experimental Models: Cell Lines
RPE1 (hTERT-immortalized human retinal pigmented epithelial cell line)	Clontech	now ATCC CRL-4000
RPE1 Cdc7wt	This paper	N/A
RPE1 Cdc7as	This paper	N/A
RPE1 Cdc7wt and Cdc7as + FLAP-PCNA & mKO-Cdt1 (30-120)	This paper	N/A
RPE1 PDS5B−/− clone #19	This paper	N/A
RPE1 PDS5B−/− clone #26	This paper	N/A
RPE1 PDS5B−/− #19 + pJC1-mCherry-HA-PDS5BWT	This paper	N/A
RPE1 PDS5B−/− #19 + pJC1-mCherry-HA-PDS5B4A	This paper	N/A
RPE1 PDS5B−/− #19 + pJC1-mCherry-HA-PDS5BΔC	This paper	N/A
RPE1 MERIT40−/− clone #22	This paper	N/A
RPE1 MERIT40−/− #22 + pQCXIN-HA-MERIT40WT	This paper	N/A
RPE1 MERIT40−/− #22 + pQCXIN-HA-MERIT404A	This paper	N/A
RPE1 MERIT40−/− #22+pQCXIN-HA-MERIT40ΔC	This paper	N/A
RPE1 PDS5A−/− clone #20	This paper	N/A
HCT116 (human colorectal cancer cell line)	Bruce Clurman (Hughes et al., 2013)	N/A
HCT116 CDK2AF/AF	Bruce Clurman (Hughes et al., 2013)	N/A
HEK293 (human embryonic kidney cell line)	ATCC	CRL-1573
Phoenix-ECO (HEK293-derived retroviral packaging line)	ATCC	CRL-3214
HeLa (human cervical cancer cell line)	ATCC	CCL-2
HeLa + pQCXIN-FLAP-PDS5BWT	This paper	N/A
HeLa + pQCXIN-FLAP-PDS5B4A	This paper	N/A
HeLa + pQCXIN-FLAG-MERIT40WT	This paper	N/A
HeLa + pQCXIN-FLAG-MERIT404A	This paper	N/A
Experimental Models: Organisms/Strains
Xenopus laevis	N/A	RRID:NXR 0.031
Recombinant DNA
pAAV-lacZ	Agilent	Cat# 240071
pRC	Agilent	Cat# 240071
pHelper	Agilent	Cat# 240071
pAAV-CDC7flox	This paper	N/A
pAAV-CDC7Δ	This paper	N/A
pVSV-G	Takara	Cat# PT3343-5
pcDNA5-FRT-TO-FLAP-DEST	Jallepalli Lab (Maciejowski et al., 2017)	N/A
pQCXIN	Takara	Cat# 631514
pQC7N	This paper	N/A
pQC7N-FLAP-Cdc7wt	This paper	N/A
pQC7N-FLAP-Cdc7as	This paper	N/A
MaRX-hygro-mKO2-Cdt1 (30-120)	Gregory David (Bainor et al., 2018)	N/A
pQCXIN	Takara	N/A
hCas9 expression vector	George Church	Addgene 41815
Empty gRNA vector	George Church	Addgene 41824
MERIT40 gRNA vector; target CCACCACCAGTGCAAACTCG	This paper	N/A
PDS5A gRNA vector; target GTGAGATCCTTCGCTAAATC	This paper	N/A
PDS5B gRNA vector; target GCTGCCTTGCTGATATTTTC	This paper	N/A
pQCXIN-HA-MERIT40WT	This paper	N/A
pQCXIN-HA-MERIT404A (S7A S8A T10A S20A)	This paper	N/A
pQCXIN-HA-MERIT40 ΔC (Δ1-20)	This paper	N/A
pQCXIN-FLAP-PDS5BWT	This paper	N/A
pQCXIN-FLAP-PDS5B4A	This paper	N/A
pJC1 (“floxed” bicistronic retroviral vector with mCherry, HA epitope tag, and blasticidin resistance)	This paper	N/A
pJC1-mCherry-HA-PDS5BWT	This paper	N/A
pJC1-mCherry-HA-PDS5B4A (S1407A S1408A S1417A S1418A)	This paper	N/A
pJC1-mCherry-HA-PDS5B ΔC (Δ1394-1447)	This paper	N/A
pDONR221	ThermoFisher	12536017
pDEST15-MERIT40WT (1-60)	This paper	N/A
pDEST15-MERIT404A (1-60 S7A S8A T10A S20A)	This paper	N/A
pDEST15-PDS5BWT (1391-1447)	This paper	N/A
pDEST15-PDS5B4A (1391-1447 S1407A S1408A S1417A S1418A)	This paper	N/A
pDEST15-CTF4WT (351-410)	This paper	N/A
pDEST15-CTF43A (351-410 S393A S394A S407A)	This paper	N/A
pDEST15-SETWT (1-49)	This paper	N/A
pDEST15-SET3A (1-49 S15A T23A S24A)	This paper	N/A
pDEST15-UBR5WT (1971-2000)	This paper	N/A
pDEST15-UBR52A (1971-2000 S1990A T1998A)	This paper	N/A
pDEST15-NEURL4WT (1071-1113)	This paper	N/A
pDEST15-NEURL45A (1071-1113 S1083A S1085A S1086A T1088A S1089A)	This paper	N/A
Software and Algorithms
BWA	(Li and Durbin, 2010)	https://github.com/lh3/bwa
Samtools	(Li et al., 2009)	http://www.htslib.org
MATLAB	MathWorks	N/A
MaxQuant (v1.0.14.7)	(Cox and Mann, 2008)	https://www.maxquant.org
MASCOT (v2.3.02)	Matrix Science	http://www.matrixscience.com/search_form_select.html
Perseus	(Tyanova et al., 2016)	https://maxquant.net/perseus/
Fiji (v2.0)
NIS Elements-AR (v5.41)	Nikon Instruments	https://www.microscope.healthcare.nikon.com/products/software/nis-elements/nis-elements-advanced-research
Prism (v7)	GraphPad	https://www.graphpad.com
Lasergene (v.14)	DNASTAR	https://www.dnastar.com

METHOD DETAILS

AAV and CRISPR/Cas9 mediated genome editing

AAV-mediated editing was performed as described (Berdougo et al., 2009; Rodriguez-Bravo et al., 2014; Rodriguez-Rodriguez et al., 2018). In brief, pAAV-CDC7^flox was constructed by amplifying 5’ and 3’ homology arms from a human genomic BAC containing the CDC7L1 locus. Fragments were assembled around a central FRT-neo^R-FRT-loxP cassette. Thereafter a PstI-marked loxP site was inserted into the 3’ homology arm. The pAAV-CDC7^Δ vector was similarly assembled around a central loxP-neo^R-loxP cassette. These and all other plasmids used in this study were sequenced and aligned to an in silico reference using DNASTAR Lasergene (v.14) to confirm their integrity. Each AAV targeting construct was cotransfected with pRC and pHelper (Helper-Free AAV System, Agilent) into HEK293 cells. Infectious particles were released by freeze thawing and applied to RPE1 cells. G418-resistant clones were isolated in 96-well plates and assayed for gene targeting via genomic PCR and Southern blotting.

For CRISPR/Cas9-based genome editing, overlapping oligonucleotides (60-mers) targeting PDS5A (GTGAGATCCTTCGCTAAATC), PDS5B (GCTGCCTTGCTGATATTTTC), and MERIT40 (CCACCACCAGTGCAAACTCG) were annealed, extended with Phusion polymerase, and cloned into a guide RNA expression plasmid (Addgene 41824) using Gibson Assembly master mix (NEB). RPE1 cells were cotransfected with each construct and a human codon-optimized Cas9 expression plasmid (Addgene 41815) using a Nucleofector 2b device (Lonza) and program T23. After two rounds of cotransfection, cells were plated into 96 well plates at limiting dilution and screened for KO clones by Western blotting.

Chemicals

Where indicated, 4 mM hydroxyurea, 10 μM VE-821, 10 μM KU-60019, 2.5 mM thymidine, 1 μM mitomycin C, 20 μM 1NM-PP1, 20 μM XL413, 20 μM PHA-767491, 20 μM roscovitine, 10 μM R0–3306, 1 μM AZ20 or 0.5 μM AZD7762 were used.

Proliferation assays

Proliferation assays were initiated by seeding T-25 flasks with 105 cells, treating with or without 1NM-PP1 for three days, and counting cells in triplicate on a hemocytometer. Population doublings were calculated as log₂(mean cell count/105). Subsequent passages followed the same procedure.

Retroviral vector cloning and expression

To express Cdc7 from its own promoter, we replaced the CMV promoter and IRES-neoR cassette in pQCXIN (Takara) with a 700-bp fragment upstream of the CDC7L1 transcription start site and an SV40 promoter-driven neoR cassette. The resulting vector (pQC7N) was cut with BamHI and NotI, blunted, and ligated to the PmeI fragment of pcDNA5-FRT-TO-FLAP-DEST (Maciejowski et al., 2017) to generate a Gateway-compatible destination vector. Wildtype and analog-sensitive Cdc7 were transferred from donor clones using LR Clonase II enzyme mix (ThermoFisher).

The vector pJC1 was constructed by inserting a loxP-mCherry-HA-loxP cassette and a blasticidin-resistance cassette into the first and second multiple cloning sites of pQCXIX (Takara). Wildtype and mutant versions of PDS5B and MERIT40 were cloned into pJC1 using Gibson assembly master mix (NEB) and into Gateway-compatible versions of pQCXIN using LR Clonase II (ThermoFisher) and donor clones.

Retroviral constructs and pVSV-G were transfected into Phoenix-ECO cells using Fugene 6 (Roche). Viral supernatants were supplemented with 5 μg/ml polybrene and applied to cells for 18 h. Stably expressing cell lines were selected with 0.5 mg/ml G418, 0.1 mg/ml hygromycin, or 10 μg/ml blasticidin.

GST fusion vector cloning, induction, and purification

Synthetic gBlocks (IDT) encoding wildtype or phosphosite-mutant fragments of MERIT40, PDS5B, CTF4, SET, and UBR5 and flanked by attB1 and attB2 sites were recombined into pDONR221 using Gateway BP Clonase II enzyme mix (ThermoFisher). After sequence verification, inserts were transferred into pDEST15 (ThermoFisher) using Gateway LR Clonase II enzyme mix.

E. coli BL21(DE3) strains were transformed with these plasmids, inoculated into LB medium supplemented with 50 μg/ml ampicillin, and grown in shake flasks at 37°C. When cultures reached an OD₆₀₀=0.8, IPTG (0.3 mM) was added. After 4 h of growth, cells were harvested and snap-frozen at −80°C. Frozen pellets were thawed, resuspended, and sonicated in ice-cold buffer A (300 mM NaCl, 10% glycerol, 50 mM NaH₂PO₄, pH 7.5 and 0.025% Nonidet P-40, 0.1 mM PMSF, 1 mM NaF, 1 mM β-glycerophosphate, 20 mM imidazole, 1 mg/ml lysozyme, and additional 0.5% Nonidet P-40). After clearing centrifugation, lysates were loaded onto glutathione spin columns (Pierce GST spin purification kit (ThermoFisher). After washing, GST fusions were eluted in buffer B (50 mM Tris, pH 8.0, 1 mM EDTA, 100 mM NaCl, 10 mM reduced glutathione) and dialyzed in 50 mM Tris, pH 8.0, 10% glycerol, 10 mM NaCl, 2 mM EDTA, 1 mM DTT for storage at −80°C.

In vitro kinase assays

Kinase assays were performed using 20 ng of Cdc7-Dbf4 and 0.5 μg of substrate protein. Reactions (25 μl) contained 5 μCi [γ−32P]ATP (3,000 Ci/mmol), 50 mM Hepes-NaOH, pH 7.4, 10 mM MgSO₄, 1 mM DTT, 1 mM β-glycerophosphate, 1 mM NaF, 0.1% Nonidet P-40, and 100 μM cold ATP. After incubating at 30°C for 20 minutes, reactions were terminated by adding 2X SDS sample buffer and boiling.

DNA fiber analysis

RPE1 cells were serially labeled with IdU and CldU (50 μM) for 20 min each. Extended DNA fibers were prepared as previously described (Jackson and Pombo, 1998). Briefly, cells were trypsinized and resuspended at 1 × 106 cells/ml in PBS. 2 μl of cell suspension was placed onto a glass slide and lysed in 10 μl of lysis buffer (200 mM Tris-HCl (pH 7.4), 0.5% SDS, 50 mM EDTA). After 6 min, the slides were tilted at 15° to allow the DNA to spread. Slides were air-dried for 30 minutes, fixed in methanol and acetic acid (3:1) for 2 min, and refrigerated overnight before immunolabeling. DNA was denatured with 4 M HCl for 20 min at room temperature. Slides were incubated in blocking buffer (PBS + 0.1% Triton X-100 + 10% goat serum) for 1 h. IdU and CldU were detected using rat anti-BrdU (Abcam ab6326, 1:100) and mouse anti-BrdU (BD 347580, 1:100). Secondary antibody staining was performed using Alexa Fluor 488-labelled goat anti-mouse IgG antibody and Alexa Fluor 568-labeled goat anti-rat antibody (1:200). Slides were mounted in Prolong Gold and imaged on a DeltaVision Elite (GE Life Sciences). Replication track lengths were measured using SoftWoRx.

BrdU incorporation and flow cytometry

RPE1 cells were treated with 10 μM BrdU for 30 min, fixed in 70% ethanol and treated with 2 M HCl for 20 min. Cells were incubated in blocking/dilution buffer (PBS + 0.1% Triton X-100 + 10% goat serum) for 20 min, then sequentially stained with BrdU antibody (BD Biosciences 347580, 1:100) and Alexa 488 anti-mouse antibody (ThermoFisher) for 1 h each, followed by 1 μg/ml RNAse A and 2 μg/ml propidium iodide for 2 h. Flow cytometry data were acquired on FACS Calibur and Fortessa instruments (BD Biosciences) and analyzed using FlowJo.

Cell lysis, immunoprecipitation, and Western blotting

For SDS-PAGE, cell pellets were resuspended in SDS lysis buffer (0.1 M Tris pH 6.8, 2% (w/v) SDS and 12% (v/v) β-mercaptoethanol), heated to 95°C for 13 min, and centrifuged at 20,000 × g for 20 min. For native immunoprecipitation, cell pellets were resuspended in NETN lysis buffer (100 mM NaCl, 20 mM Tris-Cl pH 8.0, 0.5 mM EDTA, and 0.5 % Nonidet P-40) supplemented with protease inhibitor cocktail (Sigma), incubated on ice for 20 min, and centrifuged at 20,000 × g for 20 min. Extracts were immunoprecipitated with FLAG antibody (M2)-coupled magnetic beads (Sigma) or S protein-agarose beads (Millipore). FLAG pulldowns were eluted with 100 μg/ml 3X FLAG peptide in Tris-buffered saline (TBS; 50 mM Tris-Cl pH 7.6, 150 mM NaCl). Samples were separated by SDS-PAGE (8%, 10%, and 15% Tris-glycine gels) or SDS-NuPAGE (3–8% Tris-acetate and 4–12% Bis-Tris gels, Invitrogen) and transferred to PVDF membranes. 5% nonfat dry milk in TBS-T (TBS with 0.05% Tween-20) was used for blocking and dilution of primary antibodies and secondary antibody-HRP conjugates. Signals were detected via enhanced chemiluminescence (Western Lightning Plus, PerkinElmer).

Immunofluorescence and quantitative image analysis

Cells on coverslips were fixed with 4% paraformaldehyde (15 min) and permeabilized with 0.2% Triton X-100 (10 min). Where indicated, cells were pulse-labeled with EdU for 30 min before fixation. EdU was detected using a Click-It Alexa Fluor 647 imaging kit (ThermoFisher) as described in manufacturer’s instructions. 3% BSA was used as the blocking and antibody dilution buffer. After mounting in ProLong Plus (Life Technologies), specimens were imaged on a DeltaVision microscope and deconvolved in SoftWoRx.

Quantitative image analysis was performed using Fiji 2.0. To count replication/repair foci, individual nuclei were identified by thresholding the DAPI channel and locating objects larger than 15 px2 using Analyze Particles tool, then transposed to the other channels. The Find Maxima tool (single point output) was used to identify foci within each nucleus. Linescans and colocalization analysis (calculation of Pearson correlation coefficients) were performed using the RGB Profiler tool and JaCoP plugin. BRCA1 recruitment to stalled forks was quantified by measuring BRCA1 and FANCD2 signal intensities within FANCD2 foci. Nuclear regions without foci were used for background subtraction.

Live-cell imaging

Cells stably expressing GFP-PCNA and mKO-Cdt1 (30–120) were grown in 35 mm glass-bottom dishes (MatTek), serum starved for 48 hours, and refed with serum-containing media plus 1-NM-PP1 for 12 hours. Cells were imaged at 5-min intervals on a Nikon Eclipse Ti microscope equipped with 60x and 100x oil objectives, OKO stage-top incubator and CO₂ delivery system, Yokogawa CSU-X1 spinning disk confocal imager, Andor Xyla 5.5 and Photometrics Evolve 512 cameras, and NIS Elements-AR software (v5.41). Individual cells were manually tracked and cropped before export as TIFF stacks. The fluorescence intensity of mKO2-Cdt1 was quantified in Fiji 2.0. The appearance of at least five PCNA foci was used to define the onset of DNA replication.

Replication timing assay

DNA replication timing profiles were generated based on the same principles as previously described (Koren et al., 2014; Koren et al., 2012). Cdc7^as cells were synchronized via double-thymidine block and released into S phase for 6 hours. Where indicated, 1-NM-PP1 was added two hours before release and maintained thereafter. Genomic sequencing libraries (Illumina) were run on a HiSeq using 100bp paired end reads. Reads were mapped to the hg19 reference genome using BWA (Li and Durbin, 2010) and filtered and indexed using samtools (Li et al., 2009). Subsequent analysis was performed in MATLAB. In brief, genomic “windows” of 200 consecutive reads were defined in the control samples and used to count the number of reads per window (normalized to total coverage of the sequencing library) in 1-NM-PP1 treated samples. This mode of normalization removes any influence of genome structure, GC content, and other technical effects on DNA copy number measurements. Replication profiles were then smoothed using a qubic spline and further normalized to an autosomal mean of zero and standard deviation of one (i.e. Z-score normalized). A similar procedure was performed with flow-sorted G1 and S phase RPE1 cells as described (Koren et al., 2014; Koren et al., 2012). Each condition was analyzed in two biological replicates. We also sequenced Cdc7^wt cells with and without 1-NM-PP1 treatment; the resulting copy number profiles were extremely similar (Pearson R>0.97) to those obtained for Cdc7^as cells without 1-NM-PP1.

Identification of over- and under-represented regions

We identified contiguous chromosomal regions whose mean replication timing deviated by more than one standard deviation (>1 SD) after 1-NM-PP1 treatment. In order to identify regions of significant difference from expectation given this threshold, we permuted the replication timing data 100 times and calculated the length of regions that are expected to deviate from the mean by >1 SD in no more than one permutation (i.e. regions with P < 0.01). Regions longer than 660 kb complied with this requirement. We then identified regions of this length or longer that were >1 SD above the mean (over-represented) or > 1 SD below the mean (under-represented). Finally, these regions were compared to the RPE1 replication timing profile.

SILAC labeling and cell culture treatments

Cdc7^wt cells were grown in DMEM:F12 medium with 10% dialyzed fetal bovine serum and either 175 μM unlabeled L-arginine (Arg0) and 250 μM unlabeled L-lysine (Lys0). Cdc7^as cells were grown under the same conditions with heavy isotopes of L-[U-13C₆, 15N₄]-arginine (Arg10) and L-[U-13C₆, 15N₂]-lysine (Lys8). After six cell doublings, 1.6 × 106 cells were seeded per 15-cm dish (6 per experiment). 18 h later, 2.5 mM thymidine was added for 24 h to arrest cells in S phase. Thereafter both Cdc7^wt and Cdc7^as cells were treated with 20 μM 1NM-PP1 during the thymidine arrest. Cells were harvested 2 h later. This labelling scheme was used for two replicate experiments.

Sample preparation and mass spectrometry

Cell pellets were lysed in RIPA buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 1% NP-40, 0.1% sodium deoxycholate, 1 mM EDTA, 5 mM β-glycerolphosphate, 5 mM sodium fluoride, 1 mM sodium orthovanadate, inhibitor cocktail (Roche)) and centrifuged at 7000 xg for 30 min. Supernatant (Sample A) was precipitated immediately. The pellet was lysed in extraction buffer (600mM NaCl, 1% N-octylglucoside in PBS, 5 mM β-glycerolphosphate, 5 mM sodium fluoride, 1 mM sodium orthovanadate, inhibitor cocktail (Roche), 1uL/mL benzonase (Sigma-Aldrich)), incubated for 30 minutes in a water bath sonicator, centrifuged at 13,000 xg for 15 min and the supernatant collected (Sample B).

Samples were precipitated with 4-fold excess acetone, solubilized in 8 M urea (6 M urea, 2 M thiourea in 10 mM HEPES pH 8.0). Equal quantities of protein from each SILAC state were mixed within each replicate. Buffer exchange was performed on the mixed samples using 10kDa Amicon Ultra-0.5 centrifugal filter units (Millipore). Samples were reduced with 1mM DTT for 1 h and alkylated with 5.5 mM chloroacetamide (CAA) for 1 h. Proteins were digested with Lys-C endoproteinase (1:100 m/m) (Wako) for 3 h, followed by dilution to 2 mM urea and digested overnight with Trypsin (1:100 m/m) (Sigma). Trypsin and Lys-C enzymatic activity were quenched by acidification with trifluoroacetic acid (TFA) (2% v/v). Peptides were desalted and concentrated on a C₁₈-SepPak cartridge (Waters), eluted with 40% acetonitrile (ACN) 1% TFA, then 60% ACN 1% TFA. Samples were fractionated by Small Cation Exchange (SCX) chromatography on an S column (GE Healthcare) using an ÄKTA FPLC system (GE Healthcare). The peptides were separated with a linear gradient ranging from 100% to 30% SCX buffer A (5 mM KH₂PO₄, pH 2.7, 30% ACN) for 30 min at a flow rate of 1.0 mL/min. 31 fractions of 2 mL were collected and pooled into 10 samples.

Phosphopeptides were enriched by Titansphere chromatography using titanium dioxide (TiO₂) beads (Macek et al., 2009). Beads were pre-coated for 20 min in 20 mg/mL 2,5-dihydroxybenzoic acid (2,5-DHB) in 80% ACN, 1% TFA. Each sample was incubated with 1 mg of coated beads for 30 min. Beads were transferred to a C₈ filter tip, washed once with 100 μL SCX buffer, once with 100 μL 40% ACN in 0.5% TFA and lastly in 50 μL 80% ACN in 0.5% acetic acid. The phosphopeptides were eluted with 2 × 10 μL 5% NH₄OH followed by 2 × 10 μL 10% NH₄OH in 25% ACN, pH > 11. Samples were concentrated in a speedvac at 60°C to a final volume of 10 μL and thereafter acidified with 20 μL 5% ACN in 1% TFA. The phosphopeptides were desalted and concentrated on a C₁₈ STAGE-tip. Peptides were eluted with 2 × 10 μL 40% ACN in 0.5% acetic acid, concentrated to 5 μL using a speedvac and supplemented with 3 μL 5% ACN 0.1% TFA.

All phosphopeptide samples were analyzed with an EASY n-LC system (Proxeon) interfaced with a Q-Exactive mass spectrometer (ThermoFisher) through a nanoelectrospray ion source. Peptides were loaded on a 15-cm analytical column (75 μm inner diameter), packed in-house with 3-μm C18 beads. Samples were separated with a gradient from 8% to 63% Buffer B (80% ACN and 0.1% acetic acid) over 113 minutes, ramped to 80% Buffer B in one minute and continued for 15 min, then decreased to 5% over 5 minutes, at a flow rate of 200 nL/min. The eluting peptides were electrosprayed directly into the Q-Exactive mass spectrometer.

Data processing and analysis

All raw MS and MS/MS data were analyzed together in MaxQuant v1.0.14.7 (Cox and Mann, 2008) and using the MASCOT search engine (v2.3.02). For protein identification all fragment scans were searched against a concatenated forward and reversed version of the International Protein Index (IPI) database for humans (v. 3.68) supplemented with common contaminants. False discovery rate was set to 1% for peptides, proteins and modifications sites. To determine the phosphorylated amino acid within peptides, MaxQuant calculated the localization probabilities of all putative serine, threonine and tyrosine phosphorylation sites using the PTM score algorithm (Olsen et al., 2010). Only phosphorylation sites with a localization probability higher than 0.75 (class I) were considered. The Perseus software suite (Tyanova et al., 2016) was used to identify significantly regulated phosphorylation sites using Significance B testing (Cox and Mann, 2008). All raw mass spectrometric data files have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository (Vizcaino et al., 2014) with the dataset identifier PXD014399.

Preparation of Xenopus egg extracts, chromatin pull-down, and DNA replication assay.

Low-speed supernatants (LSS) and immunodepletion of LSS were prepared as described previously (Takahashi et al., 2008). Briefly, eggs were collected, dejellied with a 2.2% cysteine–HCl (pH 7.8) solution containing 0.1 mM ethylene glycol-bis (2aminoethylether)-N,N,N′,N′-tetraacetic acid, washed with Barth solution (15 mM Tris–HCl (pH 7.4), 88 mM NaCl, 2 mM KCl, 1 mM MgCl₂, and 0.5 mM CaCl₂), activated with 100 ng/ml A23187 (Sigma Aldrich), washed with Barth solution three times, and then washed with extraction buffer (50 mM Hepes–KOH (pH 7.6), 50 mM KCl, 5 mM MgCl₂, 2 mM dithiothreitol, and 50 mM sucrose). Eggs were packed by spinning at 200 ×g, crushed with centrifugation at 20,000 ×g for 20 min, and the cytoplasmic fraction containing nuclear membranes was collected and clarified by centrifugation at 120,000 ×g for 40 min at 4°C. Two rounds of immunodepletion were performed using 50 μl of LSS and 10 μl of recombinant protein A sepharose (GE Healthcare) coupled with 5 μl of Drf1 and 20 μl of Dbf4 antisera (Takahashi and Walter, 2005) or 25 μl of non-immune serum for 2 h. 10 mM Caffeine was added to LSS. For chromatin spin-down, sperm nuclei were incubated in LSS at 3,000 /μl concentration, and 10 μl of mixtures were diluted with 90 μl of ELB salts (10 mM Hepes–KOH (pH 7.7), 50 mM KCl, and 2.5 mM MgCl₂) containing 0.6% Triton X-100, layered over 150 μl of ELB salts containing 0.5 M sucrose, and centrifuged at 20,000 ×g for 1 min. 200 μl of ELB salts containing 250 mM sucrose was added to chromatin, and spun at 20,000 ×g for 1 min. Chromatin pellets were then resuspended in Laemmli’s sample buffer. DNA synthesis was quantified by supplementing reaction mixes with trace amounts of [α−32P] dATP, 2 μl aliquots were taken at each timepoint and stopped with 8 μl of replication stop buffer (80 mM Tris–HCl (pH 8.0), 5% SDS, 10% Ficoll 400, 8 mM EDTA, 0.13% phosphoric acid, and 0.2% bromophenol blue) containing 2 mg/ml Proteinase K, and incubated for 1 h at 37°C. DNA was separated by agarose gel electrophoresis, the gel was dried on a filter paper, and incorporation of 32P into sperm DNA was quantified by phosphorimaging.

The amount of DNA synthesis was calculated based on the incorporation efficiency of α−32P-dATP into synthesized DNA as previously described (Takahashi et al., 2008; Takahashi and Walter, 2005). Briefly, the theoretical amount of DNA synthesized from a single-round semi-conservative DNA replication was calculated based on the amount of the template DNA used, and the theoretical capacity of DNA synthesis of an extract was calculated based on its dNTP concentration. The ratio of these two parameters provides the incorporation efficiency of radioactive dATP for 100% replication. The amount of DNA synthesis was calculated by dividing the experimental values by the theoretical value. The concentration of dNTP was measured for each preparation of extracts by adding known concentrations of exogenous dNTPs and for depletion assays, calibrated by multiplying 0.8, which is the experimentally-determined dilution factor.

QUANTIFICATION AND STATISTICAL ANALYSIS

MATLAB was used for quantitative and statistical analysis of replication timing data, while MaxQuant, MASCOT, and Perseus were used for mass spectrometry data. Microscopy data were quantified and analyzed using Fiji 2.0 and Prism 7. Sample sizes, experimental replicates, and statistical tests are described in the Figure Legends and Results. Sample sizes were not pre-determined. A multiplicity adjusted significance threshold of P ≤ 0.05 was used throughout the study. Unless stated otherwise, error bars indicate SEM.

Supplementary Material

Data file 1. Supplemental Data File 1, Related to Figure 6. DDK-regulated phosphorylation sites in human cells.

SILAC-encoded Cdc7^as cells were synchronized in mid-S phase with thymidine, treated ± 1-NM-PP1 for 2 hours, and subjected to phosphoproteome analysis as described in the STAR Methods. High-confidence phosphosites and those exhibiting DDK regulation at P≤0.05 or P≤0.01 cutoffs are from two independent experiments.

Table S1. Table S1, Related to Figure 2. Chromosome regions with altered replication timing in DDKi cells.

Chromosomal locations (hg19 coordinates) and lengths of regions identified as under-represented (delayed replication) or over-represented (earlier replication) in DDKi cells, and the percentage of each chromosome that is under-represented in DDKi cells. Under- and over-represented regions were identified based on a P = 0.01 cutoff (see STAR Methods).

Figure 7. Vertebrate DDKs alleviate checkpoint inhibition at unfired origins and promote remodeling, protection, and restart of stalled forks.

(A) DDK promotes origin firing by counteracting the S phase checkpoint. The S phase checkpoint inhibits origin firing in the presence of replication intermediates (RIs) to avoid dNTP and RPA exhaustion. Early origins fire before checkpoint-activating RIs have accumulated and thus require minimal DDK activity. In contrast, origin firing in late S phase cells and Xenopus egg extracts is stringently repressed by the checkpoint and requires high DDK activity. (B) DDK promotes stalled fork uncoupling and RPA recruitment, which are key events in checkpoint amplification and fork protection. DDK also activates the fork protection and restart activities of cohesin and the BRCA1-A complex by phosphorylating auto-inhibitory modules in PDS5B and MERIT40.

Highlights.

• DDKi causes spontaneous fork stalling and inhibits late origin firing

• ATRi and Chk1i rescue origin firing in DDKi cells and frog egg extracts

• DDKi blocks RPA loading, nascent strand protection, and restart at stalled forks

• MERIT40 and PDS5B are key DDK substrates in nascent strand protection and restart