CDT1 inhibits CMG helicase in early S phase to separate origin licensing from DNA synthesis

Summary Human cells license tens of thousands of origins of replication in G1 and then must stop all licensing before DNA synthesis in S phase to prevent re-replication and genome instability that ensue when an origin is licensed on replicated DNA. However, the E3 ubiquitin ligase CRL4Cdt2 only starts to degrade the licensing factor CDT1 after origin firing, raising the question of how cells prevent re-replication before CDT1 is fully degraded. Here, using quantitative microscopy and in-vitro-reconstituted human DNA replication, we show that CDT1 inhibits DNA synthesis during an overlap period when CDT1 is still present after origin firing. CDT1 inhibits DNA synthesis by suppressing CMG helicase at replication forks, and DNA synthesis commences once CDT1 is degraded. Thus, in contrast to the prevailing model that human cells prevent re-replication by strictly separating licensing from firing, licensing and firing overlap, and cells instead separate licensing from DNA synthesis.


INTRODUCTION
To duplicate their genome precisely once, eukaryotic cells are thought to strictly separate DNA replication into a period of replication origin licensing and a period of origin firing. [1][2][3] During licensing in G1 phase, cells demarcate future sites of DNA synthesis by loading inactive minichromosome maintenance (MCM)2-7 helicases onto origins of replication. At the start of S phase, cells begin origin firing, whereby replication factors are recruited to the inactive helicases to form active CDC45-MCM2-7-GINS (CMG) helicases and replication forks that duplicate DNA. Critically, it is thought that origin licensing must be strictly separated in time from origin firing to avoid re-replication, which occurs when newly synthesized DNA is re-licensed and replicated again within the same cell cycle. 1,3,4 Avoiding re-replication is crucial for maintaining genome stability, and failure to do so results in gene amplification, DNA damage, oncogenesis, and cell death. 1,5 The G1/S transition is a particularly vulnerable period in the cell cycle when cells must simultaneously inactivate licensing and initiate origin firing. In humans and other vertebrates, avoiding re-replication is critically dependent on the repression of the essential licensing factor CDT1 from the start of S phase through anaphase. 5 Activation of CDT1 during this period is sufficient to trigger re-replication. 6-9 CDT1 activity can be repressed by CDT1 degradation mediated by cullin-RING E3 ubiquitin ligases CRL4 Cdt2 and SCF Skp2 (also known as CRL1 Skp2 ), as well as by GMNN (geminin) binding to CDT1 and hyperphosphorylation of CDT1 by cyclin A-CDK1. 5,10 Both geminin and cyclin A are degraded during G1 by E3 ubiquitin ligase APC/C Cdh1 and only begin to accumulate after APC/C Cdh1 inactivation at the start of S phase, 11-13 while SCF Skp2 -mediated CDT1 degradation is thought to only begin in mid-S phase. 14,15 These findings suggest that degradation of CDT1 by CRL4 Cdt2 alone is responsible for preventing re-replication in early S phase.
However, the exclusive role of CRL4 Cdt2 in inactivating CDT1 at the start of S phase poses a conundrum: for CRL4 Cdt2 to ubiquitinate and degrade CDT1 in S phase, CDT1 must first bind to the replication fork component proliferating cell nuclear antigen (PCNA), 16,17 and therefore, CDT1 degradation can only start after origins have already fired. This regulation would result in an overlap period in early S phase when cells fire origins and could still license DNA before CDT1 is fully degraded. 1,4,17 Since it is expected that fired origins immediately synthesize DNA, this overlap period would be susceptible to re-licensing and re-replication. Even a small overlap could cause re-replication, since human diploid cells contain approximately 6 Gb pairs of DNA and replicate DNA at thousands of sites simultaneously, each of which provides an opportunity for re-replication.
Here, using a single-cell-microscopy-based analysis of human cells, we show that there is an overlap period in early S phase that lasts approximately 30 min, during which active CDT1 is present together with fired origins. Strikingly, using single-cell microscopy and in-vitro-reconstituted human DNA synthesis, we show that in addition to licensing origins in G1, CDT1 has an unexpected second role of inhibiting CMG helicase at replication forks during this overlap. Thus, cells can fire origins while inhibiting DNA synthesis, and this inhibition is only relieved once CDT1 is fully degraded. In this way, cells restrict the amount of synthesized DNA produced in the presence of CDT1 to deter re-replication. Conceptually, our study suggests that instead of temporally separating licensing and firing of origins in early S phase, human cells safeguard genome integrity by using CDT1-mediated CMG helicase inhibition to separate licensing and DNA synthesis.

RESULTS
Active CDT1 is present together with fired origins in early S phase To determine whether CDT1 protein is present together with fired origins of replication ( Figure 1A), we monitored the degradation of a doxycycline (Dox)-inducible CDT1-mCherry fusion protein in live MCF10A cells (a non-transformed human epithelial cell line). We simultaneously imaged EYFP-tagged PCNA, which forms foci at sites of origin firing and DNA synthesis. 18,19 In line with previous studies, 14,20 CDT1-mCherry degradation at S-phase start was coupled to the formation of PCNA foci (Figures 1B and S1A-S1C). Mutant analysis of CDT1 degrons confirmed that CDT1-mCherry degradation at S-phase start is mediated by CRL4 Cdt2 , while SCF Skp2 does not contribute until mid-S phase (Figures S1D-S1F). There is approximately 30 min between the start and completion of CDT1-mCherry degradation (Figures S1A and S1B), suggesting that there is an extended overlap period in early S phase when CDT1 is present together with fired origins.
To determine whether endogenous CDT1 similarly overlaps with fired origins in early S phase, we combined live-cell microscopy of S-phase entry reporters with fixed-cell immunofluorescence (IF) microscopy of endogenous CDT1. To precisely measure S-phase entry in live cells, we imaged a component of the FUCCI(CA) cell-cycle reporter system, human CDT1 (1-100)DCy , which is rapidly degraded by CRL4 Cdt2 in response to origin firing at S-phase start 15 ( Figure 1C). We used this reporter in its original N-terminal mCherry-tagged orientation (referred to here as N-CRL4 Cdt2 reporter) and also created a C-terminal tagged reporter (C-CRL4 Cdt2 reporter), which is degraded with slightly faster kinetics and further facilitates precise measurement of S-phase entry (see STAR Methods for discussion of reporters and Figures 1C, S1G, and S1H). We use the term S-phase entry to refer to the start of origin firing, which is marked by the degradation of the CRL4 Cdt2 reporters ( Figure 1C).
We live-imaged thousands of asynchronously proliferating cells and detected S-phase entry using automated time-lapse analysis of the C-CRL4 Cdt2 reporter. Cells were immediately fixed after live-cell imaging and stained for endogenous CDT1. Single-cell CDT1 levels were then measured by quantitative image-based cytometry (QIBC). 21 Finally, we computationally matched each cell from the fixed-cell QIBC analysis to its S-phase entry time from live-cell measurements. [22][23][24][25] This allowed us to analyze endogenous CDT1 levels as a function of time after S-phase entry ( Figure 1D). We refer to this combined live and fixed-cell technique as retrospective time-lapse synchronized QIBC (RT-QIBC).
Based on RT-QIBC of endogenous CDT1 IF staining in asynchronously cycling cells, endogenous CDT1 takes approximately 30 min to degrade following the start of S phase ( Figure 1E). Control experiments showed that the start of S phase, as measured by the C-CRL4 Cdt2 reporter, coincided with the appearance of chromatin-bound PCNA ( Figure S1I), confirming that origins had fired. Thus, there is an overlap period after the start of S phase where endogenous CDT1 is present together with fired origins.
During this overlap period, CDT1 could be active or could, at least in principle, be inhibited through binding by geminin or hyperphosphorylation by cyclin A-CDK1. However, analysis of an APC/C Cdh1 reporter showed that APC/C Cdh1 , which degrades geminin and cyclin A, is only inactivated near the start of S phase and can in some cells be inactivated after the start of S phase (Figures S1G and S1H), in line with previous findings. 14,15 Accordingly, we observed low levels of geminin and cyclin A in the first 30 min of S phase using RT-QIBC analysis (Figures 1F, 1G, and S2A-S2D). This result is consistent with previous studies showing that geminin and cyclin A contribute to CDT1 inhibition only later in S and G2 after they accumulate to high enough levels. 8,10 Since cyclin E-CDK2 does not hyperphosphorylate CDT1, 10 and since cyclin A and geminin are low in early S phase, we conclude that the CDT1 present in early S phase is active.
Our analysis revealed that unsynchronized cells start S phase with high CDT1 and low geminin and cyclin A ( Figures 1E-1G). In contrast, we found that cells synchronized in early S phase using aphidicolin or thymidine 26 start S phase in a dramatically different state with low CDT1 and high levels of geminin and cyclin A (Figures S2E and S2F). This is consistent with the mechanisms of aphidicolin and thymidine, which block DNA synthesis after origin firing but allow geminin and cyclin A accumulation, as well as Cdt1 degradation in response to chromatin-bound PCNA ( Figure S2E and S2F). These findings argue that singlecell analysis of asynchronous cells is needed to study regulatory events in early S phase.
We conclude that early S phase is characterized by an approximately 30-min-long overlap period, during which replication origins have fired and CDT1 is still present and active. This presents a problem in the maintenance of genome integrity, as synthesized DNA at these fired origins would be susceptible to re-licensing by CDT1 and re-replication.
DNA synthesis is inhibited in the presence of CDT1 We next determined how much DNA is synthesized during the overlap period when origins have fired and CDT1 is still present. We measured the levels of CDT1 together with DNA synthesis, measured by the incorporation of 5-ethynyl-2 0 -deoxyuridine (EdU) into synthesized DNA in an 8 min period just before cell fixation. Strikingly, as cells transitioned from G1 to S phase, CDT1 and EdU staining were mutually exclusive (Figure 2A), arguing that while origins fire in the presence of CDT1, there is no detectable DNA synthesis occurring during the overlap period.
One possible explanation for the lack of EdU incorporation is that CDT1 itself suppresses DNA synthesis. To explore this possibility, we examined EdU incorporation by RT-QIBC cells expressing an APC/C Cdh1 reporter together with high levels of Dox-inducible CDT1-mCherry that remained present in early S phase for a longer period ( Figures 2B and S3A-S3C). Markedly, these cells also exhibited a period of inhibited EdU incorporation after APC/C Cdh1 inactivation, which closely cor-responded to the prolonged time during which CDT1-mCherry was still being degraded ( Figure 2B, shaded area). In line with this interpretation, we identified a prominent population of cells with chromatin-bound PCNA but low EdU incorporation, corresponding to cells that had fired origins but had not yet  (legend continued on next page) ll Article fully degraded CDT1-mCherry ( Figure 2C, lower-right quadrant).
Since CDT1-mCherry in the above experiments was still ultimately degraded in S phase, we more directly tested for a suppressive role of CDT1 by engineering a non-degradable mutant of CDT1 (ND-CDT1). Degradation of ND-CDT1 was prevented by both removing the PCNA-interacting protein (PIP) degron, which is required for PCNA binding and CRL4 Cdt2 -mediated degradation, and mutating the Cy motif, which is required for degradation by SCF Skp2 (Figures 2D, S3D, and S3E). 15,16,20 Markedly, similar to full-length CDT1, ND-CDT1 inhibited EdU incorporation ( Figures 2E and 2F). Critically, inhibition of DNA synthesis did not prevent the firing of origins since the CRL4 Cdt2 reporter was degraded similarly to control cells ( Figures S3F and S3G). Furthermore, continued expression of ND-CDT1 persistently inhibited EdU incorporation and prevented progression through S phase as measured by DNA content ( Figures 2G, S3H, and S3I). To ensure that ND-CDT1 did not interfere with G1 origin licensing, we measured chromatinbound MCM2 to quantify origin licensing 27,28 and found no change ( Figures S3J and S3K). ND-CDT1 also inhibited DNA synthesis in U2OS and HeLa cells, arguing that this inhibition is not cell-type specific and occurs in both non-transformed and transformed cells (Figures S3L and S3M).
As an additional control, we confirmed that endogenous CDT1, not just overexpressed CDT1, can inhibit DNA synthesis when it fails to be degraded in S phase. To prevent the degradation of CDT1 in S phase, we acutely treated cells with MLN-4924, which blocks the activity of cullin-RING E3 ubiquitin ligases, including CRL4 Cdt2 (Figures S4A and S4B). 29 Cells treated with MLN-4924 had suppressed EdU incorporation following S-phase entry ( Figure 3A, siCtrl conditions). Similar to overexpressed CDT1-mCherry, MLN-4924 increased the population of cells with chromatin-bound PCNA and low EdU incorporation ( Figures 3C and 3D, siCtrl conditions). Knockdown of CDT1 partially rescued EdU incorporation, while knockdown of CDKN1A (p21) or CDKN1B (p27), other proteins stabilized by MLN-4924, 30 did not ( Figures S4B and S4C). We conclude that both overexpressed and endogenous CDT1 can suppress DNA synthesis during S phase. These findings provide a potential explanation of how cells avoid re-replication during the overlap period when CDT1 is present together with fired origins in early S phase, as the amount of synthesized DNA, the substrate of re-replication, at these fired origins would be reduced until CDT1 is fully degraded ( Figure 3E).
Geminin counteracts the inhibition of DNA synthesis by CDT1 When we prevented CDT1 degradation using MLN-4924, cells still started to increase EdU incorporation in S phase after a delay ( Figure 3A). While geminin is initially very low in early S phase, it gradually accumulates throughout S phase ( Figure 1F). We considered whether geminin binding to CDT1 could ultimately abrogate the inhibitory action of CDT1 on DNA synthesis. Consistent with this hypothesis, geminin knockdown prolonged the suppression of EdU incorporation in MLN-4924-treated cells ( Figures 3A-3D, S4B, and S4C).
To more directly determine whether geminin antagonizes inhibition of DNA synthesis by CDT1, we measured the suppression of DNA synthesis at different levels of ND-CDT1 in the presence or absence of geminin. Cell populations with inducible ND-CDT1 showed variable expression between cells, and after computationally stratifying cells by their ND-CDT1 expression, we found that EdU incorporation was inhibited in a dose-dependent manner ( Figure 3F). When geminin was allowed to accumulate, only cells with moderate-to-high ND-CDT1 could suppress DNA synthesis 2-3 h post S-phase entry. However, when we knocked down geminin, DNA synthesis stayed suppressed at much lower levels of ND-CDT1 ( Figures 3F and S4D). By analyzing dose responses at a range of geminin levels (Figure S4E), we observed that DNA synthesis was increasingly inhibited by ND-CDT1 as geminin decreased, with the IC 50 for ND-CDT1 decreasing linearly with geminin levels (Figures 3G, 3H, S4F, and S4G).
We conclude that geminin not only has a role in inhibiting CDT1 licensing 5 but also prevents CDT1 from inhibiting DNA synthesis. In this way, DNA synthesis can proceed either after CDT1 is degraded or after geminin levels have sufficiently increased to inhibit CDT1. Nevertheless, during the overlap period of CDT1 and fired origins in early S phase, geminin is low and cannot inhibit CDT1, arguing that suppression of DNA synthesis by CDT1 alone deters re-replication during the overlap period.
CDT1 suppresses DNA synthesis during the overlap period of licensing and firing If endogenous CDT1 is indeed responsible for inhibiting DNA synthesis during the first 30 min of S phase, prematurely inactivating CDT1 should accelerate the start of DNA synthesis. Since CDT1 is required for origin licensing, we could not use long-term CDT1 knockdown. Instead, we made use of the licensing kinetics in MCF10A cells, which complete the majority of origin licensing shortly after anaphase and then further boost licensing during G1 (Figures S5A and S5B). In this way, acutely inactivating CDT1 during G1 would reduce but not prevent origin licensing, which cells can tolerate. 31 We generated a cell line with Dox-inducible geminin to prematurely inactivate CDT1 during G1 ( Figures 4A and S5C). To prevent geminin degradation in G1 by APC/C Cdh1 , we mutated the geminin D-box motif, 13,32 and the resulting cell line induced Geminin DDbox to levels higher than normal G2 geminin levels in $50% of G1 cells ( Figure 4B).
In initial control experiments, where we induced Geminin DDbox in asynchronous cells, cells that received Dox well before mitosis and thus expressed Geminin DDbox by the time they reached (G) RT-QIBC in mitogen-released MCF10A cells. Cells were treated with control siRNA (same experiment as Figure 3F) and induced with Dox. ND-CDT1(+) cells selected based on gating in Figure S3H     (C) Cells were identified in two groups. Pre-licensing: Dox added R5 h before mitosis, and fixed %1 h in G1. Post-licensing: Dox was added %1 h before mitosis and fixed %1 h in S. (D) Pre-extracted cells. MCM2 from unlicensed cells was estimated from G2 MCM2 signal (4N DNA and PCNA negative). Shaded area is lower 50% of cells, corresponding to cells in shaded bar in Figure 4B. n R 434 per condition.
(legend continued on next page) ll Article anaphase had completely inhibited origin licensing, indicating that Geminin DDbox had fully inhibited CDT1 (''pre-licensing,'' Figures 4C and 4D). In contrast, cells that divided shortly after Dox addition, thus going through early G1 without Geminin DDbox , only exhibited the expected moderate reduction in origin licensing resulting from inhibited licensing late in G1 (''post-licensing,'' Figures 4C and 4D). We could therefore examine the latter group of cells to determine whether premature CDT1 inactivation accelerates DNA synthesis at S-phase start. Markedly, in the first 30 min of S phase, cells with CDT1 neutralized by Geminin DDbox exhibited approximately 5-to 10-fold higher EdU incorporation than control cells ( Figures 4E, 4F, and S5D-S5G). We thus conclude that CDT1 is present in early S phase and suppresses DNA synthesis, providing a protective mechanism against re-replication during the overlap period where CDT1 is present together with fired origins.
CDT1 inhibits DNA synthesis independently of the intra-S-phase checkpoint and re-replication Next, we sought to determine the mechanism by which CDT1 inhibits DNA synthesis. The intra-S-phase checkpoint, which limits the rate of DNA synthesis, can be activated in response to rereplication and DNA damage caused by CDT1 dysregulation. 9,33-35 Alternatively, the addition of high levels of CDT1 to replicating Xenopus egg extracts not only triggers checkpoint activation but also directly inhibits replication fork elongation, 36-38 suggesting other plausible mechanisms. Of note, CDT1-mediated inhibition of DNA synthesis in Xenopus egg extract was interpreted to be a safety response to aberrant CDT1 expression during S phase, rather than a protective mechanism routinely utilized during an overlap period in early S phase.
We first focused on whether the intra-S-phase checkpoint mediates the inhibition of DNA synthesis by CDT1. We overexpressed ND-CDT1 in cells with geminin knocked down to maximize the possibility of producing DNA damage and observed no increases in gH2AX, phospho-Chk1(S317) and phospho-Chk2(T68), markers of DNA damage and the intra-S-phase checkpoint ( Figure 5A). As an independent measure of checkpoint activation, we turned to a live-cell reporter of the activity of cyclin E/A complexed with CDK2/1 (cyclin E/A-CDK) (Figure S5H). 39,40 Cyclin E/A-CDK activity increases throughout the cell cycle starting in G1. Cyclin A knockdown confirmed that cyclin A activity does not contribute to the G1 increase in reporter activity ( Figures S5I and S5J), consistent with the low cyclin A IF staining during G1-and S-phase start ( Figure 1G). Since cyclin E/A-CDK activity is partially inhibited by the intra-S-phase checkpoint, it can be used as a proxy for checkpoint activation. 41 Accordingly, hydroxyurea (HU) reduced cyclin E/A-CDK activity in S phase, while inhibitors of checkpoint mediators ATR or WEE1 increased cyclin E/A-CDK activity ( Figure 5B). However, ND-CDT1 expression did not decrease cyclin E/A-CDK activity ( Figure 5C). Furthermore, ND-CDT1 suppressed EdU incorporation with the same IC 50 in the presence of ATR and WEE1 inhibitors ( Figure 5D). Together, these results show that the intra-S-phase checkpoint does not mediate the suppression of DNA synthesis by CDT1.
It has also been suggested that re-replication can inhibit DNA synthesis independently of the intra-S-phase checkpoint. 33,42 If re-replication is necessary for the suppression of DNA synthesis by CDT1, blocking licensing activity in S phase, which is necessary for re-replication, should rescue DNA synthesis. To inhibit licensing, we used a previously developed RPE-1 TP53 À/À cell line with mAID and SMASh-tag inducible degrons knocked into both copies of CDC6 (referred to here as CDC6 d/d ), another essential licensing factor ( Figure 5E). 43 In this cell line, CDC6 can be rapidly degraded to very low levels within 4 h, and control experiments confirmed that degrading CDC6 inhibited origin licensing ( Figure S5K).
We introduced Dox-inducible constructs for ND-CDT1-mCherry and control NLS-mCherry into the CDC6 d/d cell line and synchronized cells in late G1 by releasing cells from G0 into a mimosine arrest, which blocks cells after origin licensing ( Figure S5L). 44 In the final 4 h before releasing cells from mimosine arrest into S phase, we degraded CDC6 to prevent potential further licensing during S phase ( Figures 5F and 5G). ND-CDT1-mCherry inhibited EdU incorporation following mimosine release, regardless of CDC6 degradation, arguing that re-replication is not necessary for CDT1 to inhibit DNA synthesis ( Figures 5H and S5M). Furthermore, this cell line does not have functional TP53 (p53), which has also been implicated in the DNA damage response to re-replication. 9 We conclude that rereplication, p53, and intra-S-phase checkpoint activation are not required for CDT1-mediated inhibition of DNA synthesis in early S phase, arguing that CDT1 directly suppresses DNA synthesis.
CDT1 inhibits replication fork elongation while permitting origin firing We next determined whether CDT1 suppresses DNA synthesis by inhibiting origin firing, or by inhibiting replication fork elongation after origin firing. To quantify origin firing and recruitment of replication factors to the replication fork in the presence of ND-CDT1, we measured chromatin-bound CDC45, TIMELESS, DNA polymerases epsilon, alpha, and delta (Pol ε, a, and d), and PCNA ( Figures 6A and S6A). Replication factors that are part of or bind to the CMG helicase (CDC45, TIMELESS, Pol ε, and Pol a) did not have impaired chromatin association in the presence of ND-CDT1, while Pol d, which synthesizes lagging (E and F) RT-QIBC in post-licensing Dox addition MCF10A cells. EdU added 30 min before fixation. Cells that were in S phase for 12-30 min and had not fully degraded CDT1 were analyzed. Geminin DDbox (+) cells selected based on Geminin stain. (E) For each of 3 independent experiments, the median of cells was taken (n R 31 cells per replicate per condition; Geminin DDbox (À): 120 cells total; Geminin DDbox (+): 213 cells total) and normalized to Geminin DDbox (+) condition. Error bars are mean ± 2 3 SEM (Geminin DDbox (À) cells are 13.6% ± 7.4% of Geminin DDbox (+) cells). See Figure S5G for estimated absolute quantification. (F) Representative EdU images and matching C-CRL4 Cdt2 traces (magenta trace represents time following S entry). Geminin DDbox (À) and Geminin DDbox (+) cells 17.2 and 16.9 min in S phase. Scale bars, 5 mm. See also Figure S5. (legend continued on next page) ll Article strands, and PCNA were present but approximately 50% reduced ( Figures 6A-6C and S6A-S6C). 45 These findings are consistent with our observation that CRL4 Cdt2 , which depends on chromatin-bound PCNA, is still activated in the presence of ND-CDT1 ( Figures S3F and S3G).
To determine whether reduced chromatin-bound Pol d and PCNA are responsible for inhibited DNA synthesis, we simultaneously measured chromatin-bound replication factors and EdU incorporation. In control conditions there was a linear relationship between each chromatin-bound protein (CDC45, TIMELESS, Pol ε, Pol a, Pol d, and PCNA) and EdU incorporation, indicating that each chromatin-bound protein signal is proportional to the number of active replication forks ( Figures 6D and  S6D). However, EdU incorporation was greatly reduced at matching levels of chromatin-bound protein in the presence of ND-CDT1 ( Figures 6D, S6D, and S6E). This was the case even for Pol d and PCNA, where EdU incorporation was much lower than expected given a 2-fold reduction in chromatin-bound levels. These results are consistent with CDT1 inhibiting replication fork elongation at fired origins.
Since PCNA and Pol d are found at lagging-strand Okazaki fragments, 45,46 we hypothesized that the approximately 2-fold reduction in chromatin-bound PCNA and Pol d in the presence of ND-CDT1 was a consequence of reduced fork elongation. Accordingly, chromatin-bound PCNA was lowered by HU, which blocks fork elongation, but was not further lowered in cells also expressing ND-CDT1 ( Figure 6E). Such a reduction of PCNA at stalled forks has previously been reported. 47,48 In aggregate, our results indicate that CDT1 does not act by inhibiting origin firing but rather inhibits replication fork elongation.
To confirm that CDT1 can inhibit replication fork elongation, we used a recently developed in vitro assay for reconstituting CMG-dependent DNA synthesis using purified human proteins. 49 In this assay, replisome components (including CMG helicase) are loaded onto forked DNA templates and can perform leading-strand DNA synthesis alone or leading-and laggingstrand DNA synthesis ( Figure 6F). Importantly, this assay is independent of origin firing. In this system, addition of CDT1 strongly inhibited DNA synthesis in a dose-dependent manner ( Figures 6G and S6F). CDT1 inhibited both leading-and lagging-strand DNA synthesis ( Figures 6G and S6G) and inhibited DNA synthesis not only when added to reactions from the start but also when added to already elongating forks ( Figure 6H). Preincubation of CDT1 together with geminin neutralized its inhibitory effect ( Figures 6G and S6G). Finally, CDT1 inhibited DNA synthesis in a replisome composed only of CMG, Pol ε, and repli-cation protein A (RPA) ( Figure S6H). These in vitro reconstitution experiments demonstrate that CDT1 inhibits replication fork elongation and DNA synthesis.

CDT1 inhibits CMG helicase through its MCM-binding domains
We next focused on how CDT1 suppresses replication fork elongation in early S phase. Replication fork elongation could be suppressed by inhibition of the replicative DNA polymerases or by inhibition of CMG helicase, which is responsible for unwinding double-stranded DNA. CDT1 contains a high-affinity interaction with PCNA through its PIP degron, and it has been suggested that PIP-degron-containing proteins can interfere with the binding of polymerases to PCNA, thereby inhibiting DNA synthesis. 50,51 However, ND-CDT1 has its PIP degron removed (Figure 2D), and CDT1 inhibits replication reactions in vitro independently of PCNA ( Figure S6H), arguing against CDT1 inhibiting polymerases.
To determine whether CDT1 inhibits polymerases or CMG helicase, we examined the production of single-stranded DNA (ssDNA) at the replication fork. If CDT1 inhibited polymerases, ssDNA would accumulate as CMG helicase unwound DNA that the polymerases could not fill ( Figure 7A). 21,36,52 As a measure of ssDNA, we examined the amounts of chromatin-bound ssDNA-binding protein RPA. In control cells without ND-CDT1 expression, there was measurable chromatin-bound RPA in S phase due to the normal production of ssDNA at replication forks ( Figure 7B). Markedly, S-phase cells expressing ND-CDT1 had diminished RPA binding compared with control cells despite having similar amounts of origin firing, as measured by chromatin-bound CDC45. ND-CDT1 also prevented the increase in chromatin-bound RPA in response to HU and ATR inhibitor cotreatment ( Figure 7B), which generates a large increase in ssDNA. 21 These results suggest that CMG helicase-mediated unwinding of DNA is suppressed in the presence of CDT1. These findings are supported by findings in Xenopus egg extract, where the addition of CDT1 also reduced chromatin-bound RPA. 36 Furthermore, CDT1 did not inhibit DNA synthesis in a Pol ε primer extension assay, which uses an ssDNA template and does not contain or rely on CMG helicase ( Figure 7C). 49 In aggregate, our data are consistent with CDT1 inhibiting CMG helicase, rather than polymerase in early S phase.
As part of its role in origin licensing, CDT1 directly binds to soluble MCM helicases through two MCM-binding regions, which results in a conformational change in the MCM helicases that allows their loading onto origins. 5,20,53 This interaction is blocked (C) Mitogen-released MCF10A cells. 2 mM HU added to HU condition 1-2 h before S entry. Shaded area is mean ± 2 3 SEM. n R 336 cells per condition.   (legend continued on next page) ll Article when CDT1 is bound by geminin. 5 In Xenopus egg extracts, truncations in CDT1 that overlapped with MCM-binding regions interfered with the inhibition of DNA synthesis by CDT1, as did geminin addition. 37 Given that a minimal replisome consisting of CMG, Pol ε, and RPA is inhibited by CDT1 in vitro ( Figure S6H), and that Pol ε is not intrinsically inhibited by CDT1 ( Figure 7C), we hypothesized that the same regions in CDT1 that mediate licensing by binding MCM might also interact with and inhibit the activated CMG helicase complex (which contains MCMs). Consistent with this hypothesis, chromatin-bound ND-CDT1 colocalized with chromatin-bound CDC45, a component of CMG helicase (Figures 7D, S7A, and S7B). Since ND-CDT1 cannot bind PCNA due to its lack of PIP degron, 16 the co-localization was not the result of binding to PCNA at the replication fork.

Chromatin-bound
Finally, we tested whether the MCM-binding regions of CDT1 are necessary for CDT1 to inhibit DNA synthesis in human cells. The first MCM-binding region is found at its C terminus, while a second MCM interaction interface was identified near R210 of human CDT1. 5,20 We overexpressed ND-CDT1 with either a truncation at residue 498 in the C-terminal MCM-binding domain (ND-CDT1 D499-546 ), which abolishes licensing and MCM-binding, 54 or point mutations R198A/R210A in the other interface (ND-CDT1 R198A/R210A ), which severely diminishes licensing activity. 55 We examined their inhibitory effect on DNA synthesis with geminin knocked down to account for potential differential geminin regulation of the mutants. ND-CDT1 D499-546 could not inhibit EdU incorporation at all, while ND-CDT1 R198A/R210A was impaired in its ability to inhibit EdU, with an IC 50 approximately double that of normal ND-CDT1 ( Figures 7E and S7C). Consistent with this result, geminin, which prevents binding of CDT1 to MCM helicases, 5 also prevented CDT1-mediated inhibition of DNA synthesis ( Figure 3F). Together, our different lines of evidence support a model in which CDT1 suppresses DNA synthesis by inhibiting CMG helicase activity.

DISCUSSION
Our study focused on the fundamental problem in eukaryotic DNA replication of how cells duplicate the genome precisely once every cell cycle. It is generally thought that the solution to this problem is the strict temporal separation of origin licensing from origin firing to prevent re-replication. Vertebrate licensing regulation is centered around the inhibition of licensing factor CDT1 from S-phase entry to anaphase through CDT1 degradation by CRL4 Cdt2 and SCF Skp2 , and inhibition by geminin and cyclin A. 5 However, only CRL4 Cdt2 was proposed to prevent re-licensing in early S phase, and, given the dependence of CRL4 Cdt2 activity on PCNA bound to replication forks, it has been noted this mechanism cannot fully separate licensing and firing in early S phase. 1,4,17 Considering the large number of replication origins, even a short overlap of CDT1 and fired origins while CDT1 is being degraded could allow for the re-licensing of DNA, raising the question of how cells might prevent re-replication during this vulnerable period.
In this work, we identified an overlap period of CDT1 protein with fired origins in early S phase that lasts approximately 30 min in human cells, during which geminin and cyclin A levels are still low. Strikingly, we found that CDT1 inhibits DNA synthesis during this overlap period, and this inhibition is only relieved once CDT1 is fully degraded. We found that CDT1 suppresses CMG helicase, and thus replication fork elongation, at fired origins through its MCM-binding domains. By suppressing replication fork elongation after origin firing, CDT1 allows its own degradation by CRL4 Cdt2 to be completed before DNA synthesis commences ( Figure 7F). Importantly, this mechanism is robust toward changes in CDT1 expression levels, as cells with higher amounts of CDT1 that take longer to degrade would suppress DNA synthesis longer.
Mechanistically, the ability of CDT1 to suppress CMG helicase during fork elongation while allowing origin firing implies that CDT1 does not prevent the initial origin unwinding performed by CMG during origin firing, and only acts afterward on CMG at elongating forks. One possible explanation could be that origin firing factors, which bind and activate MCM helicases and then are released, can prevent CDT1 from inhibiting CMG specifically during origin firing.
Previous studies have identified responses to re-replication and DNA damage in human cells that reduce DNA synthesis in response to aberrant CDT1 regulation. 7,35,54,56 Critically, most characterized mechanisms require that cells first re-replicate DNA, and the resulting reduced DNA synthesis serves to minimize further damage. In contrast, the CMG helicase inhibition by CDT1 identified in our study can act before re-replication is produced.
Nevertheless, the finding that CDT1 can directly inhibit DNA synthesis still raises the question of why dysregulated CDT1 can result in re-replication and DNA damage at all. 6-9,57 A likely explanation is that the small amount of DNA synthesis in the presence of CDT1 can, over long periods, allow for enough residual DNA synthesis to produce re-replication. Furthermore, since overexpressed or dysregulated CDT1 might be incompletely degraded, CDT1 could be reduced to levels too low for effective (B) G1 mode intensities were subtracted off signals, and values were normalized to ND-CDT1(À) median. n = 2,000 cells per condition. (F) CMG is loaded onto forked DNA template, replisome is assembled, and replication is initiated in the presence or absence of CDT1. Synthesized DNA is radiolabeled for visualization (green strand). (G) Denaturing agarose gel analysis of 6 min replication reactions (15.8 kbp DNA template) with indicated proteins. 10 nM geminin was pre-incubated with equimolar CDT1 where indicated. CMG-independent template labeling is marked. (H) Denaturing agarose gel analysis of pulse-chase experiment (15.8 kbp DNA template) with the indicated proteins. Synthesized DNA was labeled in pulse, and chase was added after 50 s. 10 nM CDT1 was added with pulse or chase. Dashed and solid lines in violin plots are IQR and median. Data representative of 3 independent experiments. See also Figure S6.  M13mp18 Single-stranded DNA

Late-S phase (zoom)
(legend on next page) ll Article suppression of DNA synthesis, but high enough for some relicensing to occur over time. In support of this, non-degradable mutants of CDT1 paradoxically produce less re-replication than wild-type CDT1. 54,56 Overall, our study shows that licensing and firing have an overlap period, arguing for a revision of the concept that origin licensing must be separated from origin firing to prevent re-replication. Instead, cells separate origin licensing from DNA synthesis in early S phase. Importantly, we identify that this separation is enforced by CDT1 inhibiting the CMG helicase after origin firing until CDT1 is fully degraded. All previously identified re-replication prevention mechanisms center around the inhibition of licensing factors as cells enter S phase. 1 In contrast, we identified a new class of licensing regulation whereby a licensing factor itself inhibits S-phase progression. We propose that both classes of regulation are critical for safeguarding genome integrity.
Limitations of the study Based on (1) CDT1 reducing ssDNA produced by CMG, (2) in vitro assays showing CDT1 inhibits CMG-dependent DNA synthesis but does not inhibit Pol ε itself, and (3) the dependence of inhibition on the MCM-binding domains of CDT1, we argue that CDT1 directly inhibits CMG helicase. However, the data cannot exclude the possibility that CDT1 impacts another aspect of replisome function, such as helicase-polymerase coupling, and thereby reduces helicase activity. Additional in vitro assays examining the CMG activity in the absence of polymerase will be needed to distinguish these possibilities. Furthermore, while we found that inhibition of CMG helicase by CDT1 is dependent on the MCMbinding domain of CDT1, we have yet to elucidate the structural and biophysical basis of this inhibition. Finally, while our work in human cells and work in Xenopus egg extract suggests that CDT1 inhibits DNA synthesis in vertebrates, future work is needed to examine whether this mechanism is conserved other organisms such as yeast and Drosophila, which many studies have used as model systems to study DNA replication.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:       have been deposited at Github (https://github.com/MeyerLab/image-analysis-ratnayeke-2022) and Zenodo, and are publicly available as of the date of publication. Zenodo DOI is listed in the key resources table. d Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Cell line generation
Cell cycle reporter cell lines were generated using third-generation lentiviral transduction. 60,61 In short, lentivirus was produced in HEK-293T cells co-transfected with packaging plasmids pMDLg/pRRE (Addgene # 12251, RRID:Addgene_12251), pRSV-rev(Addgene # 1225, RRID:Addgene_12253), and pCMV-VSV-G (Addgene # 8454, RRID:Addgene_8454) together with the lentiviral plasmid with Lipofectamine 2000 (Thermo Cat# 11668019). 72 h after transfection, virus was collected from the supernatant, filtered with a .22 mm filter (Millipore Cat# SCGP00525), and concentrated using 100 kDa centrifugal filters (Millipore Cat# UFC910024). Virus was then transduced into cells in growth media. For constitutively expressed fluorescent constructs, positive fluorescent cells were sorted using a BD Influx cell sorter (performed in Stanford Shared FACS Facility), while Dox-inducible constructs (TetOn in pCW backbone with puromycin selection marker) were selected with 1 mg/mL puromycin until control cells died unless otherwise stated. TetOn cells were grown in the absence of Dox until the time of the experiment unless otherwise stated. See Table S1 for a list of cell lines with reporter combinations.
pCW constructs (TetOn Dox-inducible) expressing HA or mCherry-tagged CDT1 mutants or Geminin DDbox were introduced into these fluorescent reporter cell lines in combinations found in the Table S1. For cells with the APC/C Cdh1 reporter and TetOn-CDT1-mCherry, cells were transduced with CSII-pEF1a-mVenus-hGeminin(1-110), followed by pLV-rtTA3-IRES-Puro, and then pLV-TetOn-CDT1-mCherry. Cell lines transduced with CDT1-mCherry constructs were induced with 500 ng/mL Dox while serumstarved to sort for expressing cells. mCherry positive cells were sorted, and media was then switched to growth media without Dox. MCF10A cells containing a CDK4/6 reporter (not analyzed), Cyclin E/A-CDK reporter, and APC/C Cdh1 reporter used in Figures S4B and S4C were described previously. 58 RPE-1 TP53 -/-CDC6 d/d cells were transduced with CSII-pEF1a-H2B-mTurquoise and CSII-pEF1a-mVenus-hGeminin(1-110), and then pLV-TetOn-ND-CDT1-mCherry or pLV-TetOn-NLS-mCherry. U2OS CDC45-GFP cells were transduced with pCW-ND-CDT1-mCherry-Puro or pCW-ND-CDT1-HA-Puro. HeLa cells were transduced with CSII-pEF1a-H2B-mTurquoise and pLV-mCherry-hGeminin(1-110)-IRES-Puro. ll Article METHOD DETAILS Cell cycle reporters Cell cycle reporters of CRL4 Cdt2 and APC/C Cdh1 activity were used in this study. These reporters were originally developed as the two components of the FUCCI(CA) reporter system. 15 The CRL4 Cdt2 activity reporter corresponds to CDT1(1-100) DCy in the FUCCI(CA) system and is based on a fragment of human CDT1 corresponding to the amino acid 1-100, which is inactive with respect to origin licensing. CDT1(1-100) DCy contains a PCNA-interacting protein (PIP) degron, which mediates CDT1 degradation by CRL4 Cdt2 in response to PCNA at fired origins, and has a removed Cy motif to prevent degradation by SCF Skp2 . 15 The CRL4 Cdt2 reporter is rapidly degraded to low levels at S phase start and reaccumulates at the start of G2. Conversely, the APC/C Cdh1 reporter corresponds to Geminin  from the FUCCI(CA) system and is based on amino acids 1-110 of human Geminin fused to a fluorescent protein (either mVenus or mCherry). Geminin  is degraded at anaphase by APC/C Cdc20 and then by APC/C Cdh1 throughout G1, and reaccumulates at the time of APC/C Cdh1 inactivation at the G1/S transition. Thus, Geminin  reports APC/C Cdh1 activity at the G1/S transition. APC/C Cdh1 inactivation represents a commitment point in the cell cycle and typically occurs near the time of S phase entry and DNA replication, though CRL4 Cdt2 activation in response to origin firing is an explicit measure of S phase entry. 14,15 While these reporters are typically quantified by their presence or absence in single-timepoint measurements, when reporter fluorescence kinetics are measured in single cells using time-lapse microscopy, the precise time of CRL4 Cdt2 activation (the start of degradation of the CRL4 Cdt2 reporter) and APC/C Cdh1 inactivation (the stabilization of the APC/C Cdh1 reporter) can be identified.
In this study, we used two versions of the CRL4 Cdt2 reporter, one of which is an N-terminal mCherry-tagged CDT1(1-100) DCy (referred to as the N-CRL4 Cdt2 reporter), which is identical to the construct used in the FUCCI(CA) reporter system. Since the N-CRL4 Cdt2 reporter is fused to a fluorescent protein on the N-terminus, the PIP degron is in the middle of the construct. Since CDT1 naturally has an N-terminal PIP degron, we hypothesized that reversing the order of the fluorescent protein fusion in a reporter could confer a faster response to the initial origins that are fired in early S phase. As a result, we created a C-terminally tagged CDT1(1-100) DCy (referred to as the C-CRL4 Cdt2 reporter). We found that C-CRL4 Cdt2 responds with slightly faster kinetics at S phase start than N-CRL4 Cdt2 (Figure S1H), which was necessary for looking within the first 15-30 min of S phase by RT-QIBC in Figures 1  and 4. However, both reporters are well suited for RT-QIBC looking at times after the first 15-30 min of S phase, and we use both reporters in this study. C-CRL4 Cdt2 has a similar orientation to the PIP-FUCCI cell cycle reporter, 14 which is based on CDT1(1-17) and is also degraded throughout S phase. We consider the N-CRL4 Cdt2 (originally used in the FUCCI(CA)) system, C-CRL4 Cdt2 , and PIP-FUCCI reporters to all be reporters of CRL4 Cdt2 activity and should all be suitable for use with RT-QIBC.
The cyclin E/A-CDK reporter is a translocation-based reporter that is phosphorylated by cyclin E or A complexed with CDK2 or CDK1 (referred to collectively as cyclin E/A-CDK). 40 It is based on a fragment of human DNA helicase B (amino acids 994-1087), which is phosphorylated by cyclin E/A-CDK. When unphosphorylated in G0 and early G1, this reporter is localized in the nucleus, and as cyclin E/A-CDK activity increases throughout the cell cycle, the reporter becomes progressively localized to the cytoplasm due to increased phosphorylation. Thus, the cytoplasm to nuclear ratio of intensity is a readout of cyclin E/A-CDK activity. Critically, the increase in cyclin E/A-CDK activity in G1 is the result of increasing Cyclin E activity, rather than Cyclin A, which is only present after APC/C Cdh1 inactivation when it is stabilized ( Figure S5I).

Plasmid generation
Plasmids generated in this study were assembled using Gibson assembly of PCR amplified inserts and restriction enzyme digested plasmid backbones. Human full-length CDT1 was amplified out of MCF10A cDNA for CDT1 overexpression, and mutations and tags were introduced through primers or gene synthesis (IDT). ND-CDT1 constructs were created through a truncation of wild-type CDT1 (aa20-546), which removes the CDT1 PIP degron. The CDT1 Cy motif (aa68-70 of full-length CDT1) was mutated to alanine (DCy) to prevent degradation by SCF Skp2 . This sequence was fused at the C-terminus to a flexible linker, SV40 NLS, and either an mCherry or HA tag. For the CDT1 DPIP -mCherry construct, CDT1 (aa20-546) was fused to a flexible linker and mCherry at the C-terminus. For CDT1 DCy -mCherry construct, full-length CDT1 with the Cy motif mutated to alanine was fused to a flexible linker and mCherry at the C-terminus. ND-CDT1 D499-546 was generated from ND-CDT1 and had a truncating mutation which removed the C-terminal amino acids corresponding to residues 499-546 of CDT1. ND-CDT1 R198A/R210A was generated from ND-CDT1 and had alanine mutations introduced at amino acids corresponding to residues 198 and 210 in CDT1. The Geminin DDbox (human Geminin with R23A and L26A mutations) sequence was generated using gene synthesis and HA-tagged. For Dox-inducible TetOn constructs, PCR products were inserted into the pCW backbone (derived from pCW-Cas9, a gift from Eric Lander & David Sabatini, Addgene plasmid # 50661, RRID:Addgene_50661), a bicistronic vector with a TetOn promoter driving gene expression in addition to a constitutive PGK promoter-driven PuroR-T2A-rtTA. pC1-ND-CDT1-mCitrine was created by cloning ND-CDT1 into the pC1 backbone, derived from C1-F-tractin-mCitrine. 63 pLV-hCDT1(1-100)DCy-mCherry-P2A-mVenus-hGeminin(1-110) was generated from full-length CDT1 and Geminin (Human ORFeome V5.1). The N-CRL4 Cdt2 reporter was amplified from tFucci(CA)2/pCSII-EF, and inserted into the pLV backbone to generate pLV-mCherry-hCDT1(1-100)DCy. pLV, CSII and pCW are lentiviral expression plasmids, while pC1 is a mammalian expression plasmid. ll Article siRNA and plasmid transfection MCF10A cells were transfected with siRNA using DharmaFECT 1 (Dharmacon Cat# T-2001-03) according to the manufacturer's protocol using 20 nM siRNA and 1:500 diluted DharmaFECT 1 final concentration unless otherwise stated. Cells were incubated in transfection mixture for 6-24 h in either growth or serum-starvation media, followed by a media change. Pools of 3-4 siRNA oligos (ON-TARGETplus, Dharmacon) were used for siCtrl, siCDT1, siGeminin (siGMNN), siCyclin A (siCCNA2) and si-p21(siCDKN1A). For siCDT1 and siGeminin, siRNAs that do not target hCDT1(1-100) and hGeminin(1-110) were selected to avoid knockdown of the CRL4 Cdt2 and APC/C Cdh1 reporters, respectively. CCNA2 was knocked down since CCNA1 is not expressed in our cell lines. For siGeminin and siCyclin A in cycling cells (Figures 3A, 3B, S5I, and S5J), cells were transfected for 4-6 h and then immediately live imaged. These proteins can be suppressed even with short siRNA treatment due to their rapid protein degradation starting at anaphase. A list of siRNA oligos is in Table S1. HeLa cells were transiently transfected with pC1-ND-CDT1-mCitrine plasmid using In vitro replication assays Protein expression Proteins used in in vitro replication and primer extension assays were expressed and purified as previously described. 49,64 Recombinant Geminin was purchased from Abcam (Cat# ab86447). The sequence for the expression of N-terminal 2X FLAG-tagged CDT1 (FLAG-tag: MDYKDDDGDYKDDD) was codon optimized for expression in insect cells and synthesized by Epoch Life Science Gene Synthesis. To prepare baculovirus for CDT1 expression, the 2X FLAG-CDT1-pACEBac1 construct was transformed into EMBacY E. coli competent cells for bacmid generation. Isolated bacmid was transfected into Sf9 cells using FuGENEÒ HD (Promega Cat# E2311). Baculovirus was amplified, and the resulting virus was used to infect 1 L of Hi5 cells at a density of 1x 10 6 / mL. Cells were harvested on day 3 after infection.

CDT1 purification
The cell pellet obtained from 1 L of baculovirus-infected culture was resuspended in lysis buffer (50 mM Tris-HCl pH 7.5, 10% glycerol, 0.05% NP-40, 1 mM EDTA, 1 mM DTT) + protease inhibitors (cOmplete, EDTA-free, Roche; one tablet per 50 mL buffer, Sigma-Aldrich Cat# 11873580001). Cells were lysed by dounce homogenization, and insoluble material was removed by centrifugation (235,000g at 4 C for 45 min). 2 mL Anti-FLAG M2 affinity gel (Sigma-Aldrich Cat#A2220) was added to the supernatant and incubated for 3 h at 4 C. Resin was collected in a disposable gravity flow column (Bio-Rad) and was washed with 100 mL lysis buffer. CDT1 was eluted in 2 mL lysis buffer + 0.4 mg/mL 33 FLAG peptide and 4 mL buffer + 0.2 mg/mL 33 FLAG peptide. Eluates were pooled and applied to 1 mL HiTrap heparin column equilibrated in PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , and 1.8 mM KH 2 PO 4 ) pH 7.5, 0.5 mM TCEP, 10% glycerol. CDT1 was eluted using a 40 column volume gradient to 1 M NaCl. Peak fractions were pooled, frozen in liquid nitrogen, and stored at -80 C. DNA replication assays Experiments were performed as previously described. 49 Reactions were conducted in replication buffer (25 mM HEPES-KOH (pH 7.6), 0.01% NP-40, 100 mM potassium glutamate, 1 mM DTT, 10 mM Mg(OAc) 2  Primer extension assay Experiments were conducted as previously described. 49 Where indicated in the figure, 10 nM CDT1 was pre-incubated with 10 nM Geminin, and the protein mix was added to the reaction before initiating primer extension by the addition of Pol ε. Pulse-chase experiments Experiments were performed as previously described. 49 Replication reaction conditions were applied, except that the concentration of dCTP in the pulse was reduced to 3 mM. The concentrations of dCTP, dGTP, dATP, and dTTP were increased to 600 mM in the chase. The chase was added after 50 s.

Fixed-cell sample preparation
General protocol Staining and imaging were performed in 96-well glass-bottomed plates (Cellvis Cat# P96-1.5H-N). Cells were fixed in 4% paraformaldehyde in PBS (diluted from Fisher Cat# NC1537886) for 10 min at room temperature, followed by PBS wash. If cells expressed fluorescent proteins which spectrally overlapped with the fluorophores used in later steps, the fluorescent proteins were chemically bleached 65 in 3% H 2 O 2 + 20 mM HCl in PBS for 1 h, washed in PBS, and checked under a microscope to ensure there was negligible residual signal. If fluorescent proteins needed to be quantified in fixed cells prior to immunofluorescence, cells were initially imaged before bleaching and reimaged after staining. For PCNA and CDC45 staining, cells were incubated in ice-cold methanol for 15 min after fixation and then washed in PBS. Cells were permeabilized in 0.2% Triton X-100 in PBS for 10 min and then blocked in blocking buffer A (10% FBS, 1% BSA, 0.1% Triton X-100, 0.01% NaN 3 in PBS) for 1 h. Cells were then incubated with primary antibodies overnight in blocking buffer A at 4 C, washed in PBS, and then incubated with secondary antibodies in blocking buffer A for 1 h at RT. Cells were washed with PBS and then incubated in 1 mg/mL Hoechst 33342 (Invitrogen Cat# H3570) or in 1 mg/mL DAPI (Thermo Cat#62248) in PBS for 10 min, followed by a final PBS wash prior to imaging. Unless otherwise stated, all washes were done with an automated plate washer (aspirate to 50 mL, dispense 250 mL, repeated 9 times, BioTek 405 LS) or by hand (for pre-extracted samples, 3 washes aspirating all liquid). Iterative immunofluorescence If simultaneously staining for targets with antibodies of the same species, the iterative indirect immunofluorescence imaging (4i) technique was used to sequentially image multiple antibodies. 66 In short, the first round of imaging was identical to the general immunofluorescence protocol, with the exception that cells after the post-Hoechst PBS wash were washed in ddH 2 O and then placed in imaging buffer (700 mM N-acetyl cysteine in ddH 2 O, pH 7.4, Sigma-Aldrich A7250). Cells were imaged and then washed in ddH 2 O. The prior-round antibodies were eluted by 3310-min incubations in elution buffer, which consists of 0.5M glycine (Sigma-Aldrich), 3M urea (Sigma-Aldrich), 3M guanidinium chloride (Sigma-Aldrich) and 70 mM TCEP-HCl (Goldbio Cat#TCEP50) in ddH 2 O, pH 2.5, followed by a PBS wash. Cells were then checked under a fluorescence microscope to ensure proper elution. Cells were then blocked with blocking buffer B, consisting of 1% BSA in PBS supplemented with 150 mM maleimide (dissolved just prior to use, Sigma-Aldrich Cat# 129585) for 1 h and then washed in PBS. Cells were then blocked with blocking buffer A for 30 min, followed by primary antibody incubation, and the subsequent steps were the same as in the first round, repeated as needed. Control wells leaving out primary antibodies were always included to ensure there was no residual signal from prior rounds of imaging. Pre-extraction for chromatin-bound protein If chromatin-bound proteins were being stained for, soluble proteins were extracted from cells. Just prior to fixation, media was aspirated off cells, and the plate was placed on ice. Cells were incubated in ice-cold pre-extraction buffer, consisting of 0.2% Triton X-100 (Sigma-Aldrich Cat# x100) + 1x Halt Protease Inhibitor Cocktail (Thermo Cat# 78439) in the selected aqueous buffer. For all proteins pre-extracted except RPA1, pre-extraction buffer was made with PBS, while CSK buffer was used for RPA1, consisting of 10 mM ll Article PIPES (Sigma-Aldrich), 100 mM NaCl (Sigma-Aldrich), 300 mM sucrose (Sigma-Aldrich), 3 mM MgCl 2 (Sigma-Aldrich) at pH 7.0. After a set extraction time, 8% PFA in H 2 O was directly added to wells 1:1 with wide-orifice tips to minimize cell detachment, and cells were fixed for 25 min at room temperature, after which the sample was treated with the general staining protocol. Extraction times were: 4-5 min (PCNA, MCM2, TIMELESS), 3 min (POLA2, POLD2, POLE2), and 2 min (ND-CDT1 and CDC45-GFP in U2OS SoRa imaging and MCM2 in RPE-1).

EdU incorporation and labeling
If measuring 5-ethynyl-2 0 -deoxyuridine (EdU) incorporation, cells were pulsed with 50 mM EdU (Cayman Chemical Cat# 20518) in growth media for 8 min prior to fixation and pre-extraction, unless otherwise stated. EdU is incorporated throughout the EdU pulse, such that incorporated EdU reflects the average rate of DNA synthesis over the length of the pulse. Thus an 8 min short EdU pulse is more reflective of the instantaneous DNA synthesis rate compared to a longer pulse such as 1 h. After blocking cells (prior to primary antibodies), cells were washed once with PBS, and then a click reaction 67 was performed in 2 mM CuSO 4 , 20 mg/mL sodium ascorbate in TBS (Tris 50 mM, NaCl 150 mM pH 8.

Microscopy
Time-lapse imaging, RT-QIBC, and QIBC For automated epifluorescence microscopy, cells were imaged using a Ti2-E inverted microscope (Nikon) or ImageXpress Micro XLS microscope (Molecular Devices). For imaging on the Ti2-E, multichannel fluorescent images were acquired using triple-band (ECFP/ EYFP/mCherry, Chroma: 89006) or quad-band (DAPI/FITC/TRITC/Cy5, Chroma: 89402) Sedat filter sets using an LED light source (Lumencor Spectra X) and Hamamatsu ORCA-Flash4.0 V3 sCMOS camera. 10x (Nikon CFI Plan Apo Lambda, NA 0.45) or 20x (Nikon CFI Plan Apo Lambda, 0.75 NA) objectives were used to acquire images. For imaging on the ImageXpress, images were taken with appropriate single-band filter sets with a white-light source, using a 10x (Nikon CFI Plan Fluor, NA 0.3) or 20x (Nikon CFI Plan Apo Lambda, 0.75 NA) and Andor Zyla 4.2 sCMOS camera. All images were acquired in 16-bit mode, and acquisition settings were chosen to not saturate the signal. Fluorophores and imaging channels were chosen to minimize bleedthrough, and in the case of detectable bleedthrough, it was corrected using bleedthrough coefficients estimated from single fluorophore controls.
For live-cell time-lapse imaging, 96-well plates were imaged within an enclosed 37 C, 5% CO2 environmental chamber in 200 mL of growth media. 4-9 sites were imaged in each well (with the number of wells imaged varying depending on experiment and imaging interval) every 3-12 min (3 min for measurements of CDT1 in early S phase as in Figure 1E, or in Figure S1E or S1I. Longer intervals for measuring longer-term dynamics). Light exposure to cells was limited by using the minimum exposure necessary to maintain an acceptable signal-to-noise ratio on a per-channel basis, and total light exposure was always limited to below 300 ms per site each timepoint. Images were taken with the 10x objective for all live-cell imaging except for experiments shown in Figures 3A,  S1E, S5H, and S1I to maximize the number of cells in the field of view. When performing the live-cell imaging for RT-QIBC, cells were immediately taken off the microscope following the final time point and fixed.
For fixed-cell imaging for RT-QIBC and QIBC, tiled images of the majority of each well (16-36 sites per well) were taken using the 20x objective. When reimaging fixed cells (matching back to either live-cell imaging for RT-QIBC or previous rounds of fixed-cell imaging), the plate position (which can shift slightly when replacing the plate on the microscope) was aligned to approximately the same location and further aligned computationally during image analysis. ll Article Spinning-disk confocal microscopy For high-resolution live-cell imaging of EYFP-PCNA and CDT1-mCherry ( Figures 1B and S1A-S1C), cells were imaged on an automated spinning-disk confocal microscope (Intelligent Imaging Innovations, 3i). This system used a Nikon Ti-E stand, motorized XY stage with piezoelectric Z movement (3i), Andor Zyla 4.2 sCMOS camera, CSU-W1 confocal scanner unit (Yokogawa), and 405/445/ 488/514/561/640 nm LaserStack (3i), controlled using SlideBook 6 (3i). Cells were imaged in a 37 C environmental chamber (growth media was HEPES buffered) using a 60x/1.35NA oil objective (Nikon) with 2x camera binning. Images at the nucleus midplane were taken every 2-3 min in a 5x5 montage which was stitched together after acquisition. H2B-mTurquoise, EYFP-PCNA, and CDT1-mCherry were imaged using a triple-band 445/515/561 excitation filter set.
For fixed-cell imaging of chromatin-bound ND-CDT1 localization, pre-extracted U2OS CDC45-GFP cells were imaged on a SoRa spinning-disk confocal microscope (Marianas system, 3i). This system was similar to the previously described 3i microscope, except it used a Zeiss Axio Observer 7 stand, ORCA-Fusion BT sCMOS Camera (Hamamatsu), CSU-W1 SoRa confocal scanner unit (Yokogawa), and 405/445/488/514/561/637 nm LaserStack (3i). Cells were stained using rabbit anti-HA and anti-rabbit Alexa Fluor 568 (for detecting ND-CDT1) together with mouse anti-GFP and anti-mouse Alexa Fluor 488 to image endogenously tagged CDC45-GFP, together with Hoechst. Images were taken using the 405, 488, and 561 channels using a quad-band 405/488/561/640 nm excitation filter set (3i), with a 63x/1.4NA Plan-Apochromat Oil M27 objective (Zeiss) and 4x magnification changer and no camera binning. The field of view was manually searched without 4x magnification and low exposure in the 488 channel to identify cells that were in S phase. Cells were then imaged in both channels at 5 Z-positions around the midplane of the nucleus (1 mm spaced, only the midplane is shown). No deconvolution was performed, and controls were tested to ensure there was no spectral bleedthrough or cross-binding of secondary antibodies.

Experimental details MCF10A mitogen-release experiments
For mitogen-release experiments, MCF10A cells were serum-starved the day following seeding onto plates (Day 1). For experiments with siRNA knockdowns, cells were transfected with siRNA the next day (Day 2) in serum-free media. Media was changed the following day (Day 3, approximately 48 h after serum-starvation). If inducing ND-CDT1, Dox was added during the media change 6 h before mitogen release. On Day 3, cells were mitogen-released with full growth media and live-imaged. For RT-QIBC and QIBC, cells were fixed 18 h after mitogen-release. For drug treatments (hydroxyurea, AZ-20, MK-1775, MLN-4924), drugs were added 14 h after mitogen-release during live-cell imaging. CDC6 degradation experiments RPE-1 TP53 -/-CDC6 d/d cells 43 had an auxin-inducible degron (mAID), and SMASh-tag knocked into both endogenous CDC6 loci. Cells contained a Dox-inducible OsTIR1 E3 ubiquitin ligase component which is required for mAID degradation. The addition of auxin to cells triggers the degradation of mAID-containing proteins. The SMASh-tag contains degron that is auto-cleaved after protein translation by a protease domain. Addition of BMS-650032 (BMS) inhibits this auto-cleavage, resulting in protein degradation. Thus, addition of Dox/Auxin/BMS triggers a robust degradation of endogenous CDC6.

QUANTIFICATION AND STATISTICAL ANALYSIS
Image analysis Automated analysis of time-lapse imaging of cell cycle reporters, quantitative image-based cytometry (QIBC), and Retrospective Time-lapse Synchronized QIBC (RT-QIBC) were performed using a custom MATLAB (R2020a, MathWorks) pipeline based on previous work. 22 QIBC here is considered to be the high-throughput single-cell quantification of fixed-cell signals (fluorescent proteins, immunofluorescence, EdU staining, DNA stain), while RT-QIBC involves the assignment of QIBC measurements to an explicit time in the cell cycle based on prior time-lapse imaging of cell cycle reporters (N-and C-CRL4 Cdt2 reporters, APC/C Cdh1 reporter, and EYFP-PCNA). In principle, RT-QIBC can be used to quantify any fixed-cell signal (immunofluorescence used in this study, as well as mRNA or DNA FISH, for example) and retrospectively analyze any live-cell reporter or imaging measurements. In this way, the dynamics of fixed cell measurements can be reconstructed with temporal resolution that is only limited by imaging frequency (in this study 3-12 min). The image processing pipeline and code used to generate all figures in this study have been deposited on Github and Zenodo (https://github.com/MeyerLab/image-analysis-ratnayeke-2022, https://doi.org/10.5281/zenodo.7183750), and data can be downloaded at Dryad (https://doi.org/10.5061/dryad.4xgxd2599). Segmentation and signal quantification Raw images were flat-field corrected (also known as shading corrected) to correct for uneven sample illumination. Since images output by the sCMOS camera are the sum of a camera offset value together with the actual detected signal (which is proportional to the sample illumination), we subtracted off the camera offset value and then divided the image by an empirically determined illumination profile. This profile was calculated either from the background autofluorescence in areas without cells in live-cell images (aggregated over a large number of sites) or from sample-free wells filled with autofluorescent blocking buffer A for fixed cell imaging. Confocal movies were not flat-field corrected due to a lack of uneven illumination.
For live-cell imaging, nuclei were automatically segmented from H2B signal using a Laplacian of Gaussian blob detector, which in the case of movies with low contrast, was further refined with active contours. For fixed-cell imaging, nuclei were segmented from the Hoechst or DAPI stain using a threshold determined from histogram curvature. Detected nuclei larger than the median object size were checked using a curvature-based object splitting algorithm that splits cells along two points of high perimeter curvature. If there are more than two putative split points, pairs of points are chosen based on pairs with the highest distance along the perimeter between points divided by the Euclidean distance of the points. For multi-round fixed cell imaging, each imaging round was segmented and aligned to a common round of images. A segmentation mask from a single round (typically the first round) was designated the primary mask and used for quantification of all rounds.
To quantify nuclear cell cycle reporters and fixed-cell signals, the background signal was estimated by taking the 25 th percentile of pixels outside of a dilated nuclear mask (dilated 7.8 mm for predominantly nuclear signals, 15.6 mm for signals with cytoplasmic components) and subtracted off of images. For chromatin-bound CDC45, POLA2, POLD2, POLE2, and TIMELESS signals, the background was not subtracted during image processing but accounted for later during analysis. The mean and median signal within the nucleus were then calculated, and for signals with a cytoplasmic component, the median signal within a ring outside of the nucleus was calculated (region 0.65 mm to 3.25 mm outside the edge of the nucleus). To quantify the puncta area of EYFP-PCNA, a tophat filter (3 pixel radius for confocal imaging, 2 pixel radius for wide-field) was applied to the image, and a series of thresholds of different stringencies were manually chosen and applied to minimize false positives and negatives. The total area of pixels above the thresholds was quantified. Time-lapse tracking For time-lapse imaging, nuclei were tracked using a nearest-neighbor algorithm between each frame and its previous frame. To increase tracking fidelity, the total nuclear signal (the sum of nuclear intensity) was used as an invariant quantity which does not change significantly between frames. Using this, putative aberrant merging and splitting of nuclei during segmentation could be detected and corrected. Mitotic events are detected when two daughter nuclei are detected within the vicinity of a previous nucleus and have a total nuclear signal which is approximately equal to the previous nucleus.
To match fixed-cells to live cells, fixed-cell images were computationally aligned to live-cell images using 2D cross-correlation, and cells with their associated measurements were assigned to their nearest live-cell neighbor. When matching the 20x fixed-cell images back to 10x live-cell images, live images were resized using bicubic interpolation (for alignment and tracking purposes only), or fixed images were mean-value binned. Cell cycle annotation of live-cell data Mitosis was annotated during the process of tracking cells, defined as the separation of the two sets of chromosomes at anaphase. CRL4 Cdt2 activation (defined as the start of CRL4 Cdt2 reporter degradation) and CDT1-mCherry degradation start were annotated by subjecting traces following mitosis or mitogen-release to a drop detection algorithm. This algorithm detects degradation of the reporter at a given time based on a set number of frames following it (the number of frames after a corresponds to the minimum detectable time since degradation start, typically $3 frames). By detecting points using only a set number of frames beyond the degradation point, we avoid biases in the accuracy of detecting cells that just recently degraded compared to cells that degraded much earlier.
Points were checked based on the slope and curvature of the trace within the window being low and high enough, respectively, and having a set decrease in the reporter signal (normalized to reporter expression). APC/C Cdh1 inactivation (defined as the start of APC/C Cdh1 reporter accumulation) was detected in a complementary way, identifying the first point where the slope increases to ll Article a threshold level and the reporter increases from a low level to a threshold-value of persistent increase. All threshold values were empirically determined and validated by eye on at least 200 traces. For both CRL4 Cdt2 and APC/C Cdh1 reporters, the integrated intensity within the nucleus was quantified for trace analysis.
For identification of the start of S phase from PCNA foci, foci were segmented, and the total area of foci was quantified. The transition from G1 to S phase is characterized by an increase of low foci signal to high foci signal. For RT-QIBC analysis, the same algorithm was used for the APC/C Cdh1 reporter (as the increase in puncta area mirrors the rise in APC/C Cdh1 reporter levels). For confocal imaging, a dual threshold algorithm was used. A high foci signal threshold that robustly identified S phase cells was manually determined (50 pixels), and the first point at which the foci area increased above this threshold and was higher than the previous 4 frames was identified. Since the high foci signal threshold typically identified a point well after S phase entry, the final frame before this high identified point which was below a low threshold (3 pixels) was identified as the true S phase entry point.

RT-QIBC
After automated tracking and quantification of live-and fixed-cell imaging, each cell was associated with its corresponding annotated cell cycle reporter traces, as well as multidimensional fixed-cell measurements from QIBC. Based on this, the time elapsed from a point of interest (such as CRL4 Cdt2 activation or mitosis) was used to arrange fixed-cell measurements based on time. Conversely, live-cell traces can be selected based on QIBC measurements (such as the expression of ND-CDT1). For analyses with high time resolution (e.g. Figures 1E-1G and S1I), time offsets for each imaging site and well were accounted for based on the order of well acquisition.

Chromatin-bound ND-CDT1 localization analysis
We measured the Pearson correlation coefficient of chromatin-bound ND-CDT1 relative to chromatin-bound CDC45 (which marks CMG at sites of DNA synthesis) as a measure of colocalization. First, nuclei were segmented using Hoechst DNA stain to confine the analysis to signals within the nucleus. ND-CDT1 and CDC45 signals were masked by the nuclear mask, and then the Pearson correlation coefficient was calculated between these signals. To gauge the significance of these correlations, the ND-CDT1 and CDC45 were randomized by shifting the two channels 40 pixels (1.038 mm) relative to each other. Pearson correlation coefficients were calculated between shifted images (only analyzing pixels within the nucleus) in each of 4 shifting directions (up, down, left, and right), and the mean correlation coefficient of these 4 measurements was considered the randomized data for a given cell. To visualize the ND-CDT1 and CDC45 segmented signals in Figure 7D, each signal was processed with a top-hat filter with a 10-pixel radius and then a Gaussian filter with a 1-pixel radius. Signals were segmented by applying a uniform threshold to the filtered images.

Cell gating
To estimate thresholds in an automated manner for determining cells positive and negative for QIBC staining (for example, chromatin-bound PCNA positive cells for S phase cells) or 2N DNA content, cells known to be in either G1 or S phase based on livecell imaging were identified, and then the 99 th percentile (95 th for Figure S1I) was chosen as the threshold.
For Dox-inducible cell lines, cells within populations expressed different amounts of protein in response to uniform Dox, with a small subset of cells not inducing protein to detectable levels. To correct for this, cells were selected for analysis based on either immunofluorescence staining of induced-protein at the end of the experiment (e.g. HA-tag, CDT1 or Geminin staining), or in the case of fluorescent protein-tagged protein induction (e.g. ND-CDT1-mCherry), cells were selected based on live-cell imaging of the induced protein. Thresholds were based on uninduced cells. G1 and S phase cells were identified through a combination of live-cell reporter quantification and fixed-cell stains. In Figure 2A, late G1 to early S cells were defined by 2N DNA and intermediated Cyclin E/A-CDK activity (0.7 -1.2). In Figure

Quantification corrections
For experiments with chromatin-bound proteins measured after pre-extraction, there were rare sections of the samples that were incompletely extracted of soluble proteins. As a proxy for extraction efficiency, in experiments with APC/C Cdh1 or Cyclin E/A-CDK reporters (which are soluble), the residual fluorescent protein signal was imaged in addition to immunofluorescence. Cells that had high fluorescent protein signal for the reporters were considered incompletely extracted and removed from the analysis.
For the staining of replication factors in Figures 6 and S6, staining was performed in two rounds. In the first round, chromatin-bound replication proteins (CDC45, TIMELESS, POLE2, POLA2, POLD2, PCNA) were stained using Alexa Fluor 647 secondary antibodies simultaneously with EdU staining with AFDye 488. Fluorophores were then bleached for 1 h and re-stained for PCNA using an Alexa Fluor 647 secondary antibody to identify S phase cells. However, there was a low-intensity residual signal from the first round of staining, which was corrected by using an empirically determined residual signal scaling factor.
For Figures S4B and S4C, H2B-iRFP670 was expressed in a bicistronic vector together with the APC/C Cdh1 reporter (P2A sequence). As a result, the APC/C Cdh1 reporter signal could be normalized by the H2B-iRFP670 signal to control for differential expression of the construct between cells. ll Article For pre-extraction experiments of replication factors (Figures 6 and S6), outliers resulting from incompletely extracted cells and imaging artifacts were removed by removing the top 1 percentile of data. In Figure S3B, the outer 1 percentile of CRL4 Cdt2 activation delays from APC/C Cdh1 inactivation was removed to account for misidentified cell cycle transitions.
When equal numbers of cells were plotted from different conditions, cells were randomly subsampled to equalize conditions without replacement.

Data normalization
For normalized stain quantification, a baseline was calculated from G1 levels of the stain and subtracted off of all values, followed by division by the group to be normalized. For EdU quantification shown on a linear scale (for dose-responses and chromatin-bound stain linear fits), the G1 background signal was subtracted off values for a true zero. For Figures 1F and 1G, measurements were normalized to the median G2 signal of each protein. For Figures 4E, S5F, and S5G, the EdU signal was background subtracted and divided by the Dox(-) EdU signal 1 -1.2 h after S phase entry to standardize values between replicates. For Figure 5C, the Cyclin E/A-CDK activity at S phase start was subtracted off values to plot the change of Cyclin E/A-CDK activity from S phase start over time.

Statistical analysis
Details of statistical tests can be found in the figure legends. Comparisons were made with either paired t-tests (for tests between multiple independent replicate experiments) or two-sample t-tests for within-experiment comparisons of measurements with an a of.05.
For linear fits of chromatin-bound stains ( Figure 6), a linear model with a fixed zero-intercept was fit using robust fitting with a bisquare weight function (tuning constant of 2). For fitting dose-response curves cells could be stratified based on their ND-CDT1 expression. Single-cell measurements of EdU and the single-cell expression of ND-CDT1 were fit to a Hill equation of the form EdUð½NDCdt1Þ = EdU max À EdU max À EdU min 1 + IC 50 ½NDCdt1 n using nonlinear regression. EdU max is the EdU incorporation of cells without ND-CDT1, which was set using cells not expressing ND-CDT1. EdU min represents the minimum EdU incorporation, IC 50 is the 50% inhibitory concentration of ND-CDT1 concentration ½NDCdt1, and n is the Hill coefficient. For Figures 3F and 5D, EdU min , IC 50 and n were all fit parameters, while for Figures 3G, 3H, S4F, and S4G, EdU min was set based on high levels of ND-CDT1 expression. Nonlinear regression was performed using the Levenberg-Marquardt algorithm (nlinfit() in MATLAB). Initialization parameters for EdU inhib were estimated from the 5 th percentile of EdU signal, while EdU inhib was initialized based on the median ½NDCdt1 in the cell population, and was n initialized as 1.
For bootstrapped estimators, samples were resampled at least 1000 times, and confidence intervals were calculated using the percentile method. For raincloud plots, 68 which are combined violin and jitter plots, the data distribution was estimated using a kernel smoothing density estimate. The solid and dashed lines in the violin plot correspond to the median and inter-quartile range (IQR).

Visualization
All example cells were extracted from full-sized images through MATLAB scripts and selected based on RT-QIBC or time-lapse analysis. For Figure 7D, cells were selected for imaging based on being in early or late S phase. An example cell was chosen and visualized using ImageJ (v1.53, Fiji distribution) 62 Figure S1. Active CDT1 is present together with fired origins in early S phase, related to Figure 1 (A-C) MCF10A cells expressing EYFP-PCNA and doxycycline (Dox)-inducible CDT1-mCherry (induced 6 h prior to imaging) were imaged using confocal microscopy.
(A) Quantification of PCNA foci detection relative to CDT1 degradation in cells from Figure 1B. The start of CDT1 degradation and first detectable PCNA foci (TPCNA offset), the total time it takes for CDT1 to be degraded (TCDT1 deg), and the overlap duration when CDT1 and PCNA foci are simultaneously visible (TCDT1&PCNA) were determined. TPCNA offset represents the amount of time it takes for PCNA foci to grow large enough to be detectable over soluble pool of PCNA following the start of origin firing and CDT1 degradation. Black × corresponds to frames shown in example cell in Figure 1B.        Figure 2B that inactivated APC/C Cdh1 2, 4, 6, and 8 h prior to fixation.
(D) Western blot validation of ND-CDT1 overexpression (CDT1 antibody detects endogenous CDT1 and ND-CDT1) and Geminin knockdown. Geminin ΔDbox was induced with Dox for 6 h and quantified with anti-Geminin antibody (detects endogenous Geminin and Geminin ΔDbox ). n = 147,420 (Dox(-)), 150,706 (Dox(+)) cells. Figure 4F, co-stained for EdU incorporation, CDT1, and Geminin (detecting both endogenous Geminin and Geminin ΔDbox ). Scale bar = 5 µm. Note, CDT1 levels are likely increased in response to Geminin ΔDbox due to co-stabilization. However, CDT1 is inactivated by Geminin, and CDT1 is still degraded over the same 30 min period ( Figure S5E). (F, G) Calibration of EdU incorporation to absolute DNA synthesis from Figure 4E using RT-QIBC (C-CRL4 Cdt2 reporter). Absolute DNA synthesis (in base pairs) can be inferred by integrating EdU incorporation measurements (area under the curve), which estimates the total DNA EdU signal. The fraction of total DNA synthesis during a given period (a period 1.5-2 h after S phase entry was chosen as a calibration point. Denoted as FDNA 1.5-2h) can be estimated by taking the ratio of the area from 1.5 -2 h after S phase entry (Area1.5-2h) to the total area (AreaTotal). In the experiment from Figure 4E, a 30 min EdU pulse was used, and thus the EdU intensity in cells 2 h after S phase entry in cells without Dox added would be equivalent to FDNA 1.5-2h. (G) The EdU signal in Figure 4E was calibrated based on Figure S5F. Multiplying this by 6×10 9 base pairs (approximate human diploid DNA) gives the equivalent amount of DNA synthesis in base pairs. For each of 3 independent experiments, the median of cells was taken (n ≥ 31 cells per replicate per condition, Geminin ΔDbox (-) 120 cells total, Geminin ΔDbox (+) 213 cells total). Error bars are mean ± 2×SEM.  (B) Comparison of chromatin-bound replisome components in G1 with ND-CDT1, in same experiment as Figure 6B. G1 mode intensities from ND-CDT1(-) were subtracted off signals, and values were normalized to the ND-CDT1(-) S phase condition from Figure 6B. n = 2,000 cells per condition. Representative of 3 individual experiments.

(D) Same example cells as shown in
(C) Comparison of chromatin-bound TIMELESS in S phase (left) and G1 phase (right), analyzed in same way as Figure 6B and Figure S6B. n = 2,000 cells, pooled from 3 wells for ND-CDT1(-), or 7 wells for ND-CDT1(+).
(D) Analysis of EdU incorporation as a function of chromatin-bound protein levels for TIMELESS, POLE2, POLA2 in S phase. G1 mode intensities were subtracted off EdU and chromatin-bound intensity. Line is fit line of linear regression (n = 2000 cells). Representative of 3 independent experiments. Other stains in Figure 6D. TIMELESS staining was done in separate experiment as other stains. Cells were pooled from 3 wells for ND-CDT1(-) or 7 wells for ND-CDT1(+) from a single experiment.
(E) Summary of slopes from fit lines from Figures 6D and S6D. EdU suppression ratio is defined as the ratio of the fit line in the control condition to the ND-CDT1 condition (>1 indicates there is lower EdU incorporation for a given amount of chromatin-bound protein). Bar is mean of 3 independent experiments.
(F) Coomassie-stained SDS-PAGE analysis of purified human CDT1 protein.
(G) Denaturing agarose gel analysis of leading/lagging strand replication reactions (9.7 kbp DNA template) for 20 min with the indicated proteins. 10 nM CDT1 was pre-incubated with equimolar Geminin for 5 min on ice before adding to the reactions where indicated. In all reactions, the concentration of potassium glutamate was 250 mM. Experiment was repeated 3 times.
(H) Denaturing agarose gel analysis of a time course experiment performed on the 15.8 kbp forked DNA template with the indicated proteins. CTF18-RFC, PCNA, and CDT1 were included where indicated. CMG-independent template labeling products are indicated. Experiment was repeated two times.
Dashed and solid lines in violin plots are IQR and median, respectively.