An intrinsic S/G2 checkpoint enforced by ATR

See allHide authors and affiliations

Science  24 Aug 2018:
Vol. 361, Issue 6404, pp. 806-810
DOI: 10.1126/science.aap9346

An additional cell cycle checkpoint

Cell division is controlled by checkpoints that regulate the temporal order of the cell cycle phases, including the G1/S, G2/M, and metaphase/anaphase transitions. Yet there are no known control mechanisms for a fourth fundamental transition—the S/G2 transition. Saldivar et al. report a switchlike control mechanism that regulates the S/G2 transition. The checkpoint kinase ATR senses ongoing DNA replication in S phase and represses the mitotic transcriptional network, ensuring that DNA replication in S phase is completed before mitosis.

Science, this issue p. 806


The cell cycle is strictly ordered to ensure faithful genome duplication and chromosome segregation. Control mechanisms establish this order by dictating when a cell transitions from one phase to the next. Much is known about the control of the G1/S, G2/M, and metaphase/anaphase transitions, but thus far, no control mechanism has been identified for the S/G2 transition. Here we show that cells transactivate the mitotic gene network as they exit the S phase through a CDK1 (cyclin-dependent kinase 1)–directed FOXM1 phosphorylation switch. During normal DNA replication, the checkpoint kinase ATR (ataxia-telangiectasia and Rad3-related) is activated by ETAA1 to block this switch until the S phase ends. ATR inhibition prematurely activates FOXM1, deregulating the S/G2 transition and leading to early mitosis, underreplicated DNA, and DNA damage. Thus, ATR couples DNA replication with mitosis and preserves genome integrity by enforcing an S/G2 checkpoint.

To avoid loss of genetic information, cells must replicate their DNA before mitosis (1). Recent data show that replication delays the onset of mitosis (2), yet how cells sense the end of the S phase and coordinate these processes remains a fundamental biological question. A checkpoint monitoring the end of DNA replication has been proposed (3), but effectors of such an S/G2 checkpoint have not been identified. DNA damage can activate the checkpoint kinase ATR (ataxia-telangiectasia and Rad3-related), arresting S or G2 cells by inactivating the CDC25–cyclin-dependent kinase (CDK) pathway (4). Because ATR ensures completion of replication before mitosis during normal proliferation (5, 6), we hypothesized that it may regulate the S/G2 transition.

To test this hypothesis, we used quantitative image-based cytometry (QIBC) (fig. S1) (7) combined with pulse-chase assays (fig. S2) and measured S-to-M and G2-to-M progression. S-phase cells treated with ATR inhibitors (ATRi) underwent accelerated mitotic entry (Fig. 1, A and B). Unexpectedly, ATR inhibition in G2 did not accelerate mitotic entry rates, whereas WEE1 inhibition did (Fig. 1C). This extends previous findings (5, 6) suggesting that ATR acts in the S phase and not in the G2 phase of normal cell cycles to delay mitosis.

Fig. 1 ATR represses a mitotic gene network during the S phase.

(A) S-to-M or G2-to-M progression assay showing the cellular distribution at time = 0 and the gating scheme. Details are shown in fig. S2. 2n and 4n refer to the number of sets of chromosomes. h, hours; DAPI, 4′,6-diamidino-2-phenylindole. (B and C) Fraction of S-gated (B) or G2/M-gated (C) hTERT RPE-1 cells in mitosis as a function of time. Error bars, SEM of n = 3. (D) Experimental design and representative images of hTERT RPE-1 cells in G1, early to late S phase (S1 to S4), G2, or mitosis. m, minutes. (E) QIBC plots of DNA content versus EdU mean intensity, with mean cyclin B cytoplasmic intensity shown by a color scale. Boxes indicate gated populations used for analyses. (F) Mean cyclin B cytoplasmic intensities of gated populations from (E). Error bars, SEM of n = 4. **P < 0.01, ****P < 0.0001. (G) RT-qPCR (quantitative reverse transcription polymerase chain reaction) analysis of cyclin B1 mRNA from synchronized hTERT RPE-1 cells described in fig. S6. Error bars, SEM of n = 4. *P < 0.05, ***P < 0.001. (H) Unsupervised clustering analysis of mock or ATRi and early S, late S, or G2 synchronized cells (selected comparisons shown). Heatmap indicates the fold change of normalized RNA-seq reads between indicated samples (n = 3). (I) Top 4 GO terms for group 5 genes from (H). (J) Volcano plot of log10 false discovery rate–corrected P values and log2 fold change in gene expression for ATRi versus mock treatment of cells in the late S phase.

These observations are consistent with ATR inhibition shortening the S phase. Nevertheless, a pulse-chase-pulse assay (fig. S3) and live-cell imaging with an EYFP-PCNA (enhanced yellow fluorescent protein–proliferating cell nuclear antigen) biosensor (8) (fig. S4) revealed that S-phase shortening alone cannot explain the combined shortening of S and G2 after ATR inhibition (fig. S3, B to D). We hypothesized that ATR controls subsequent G2 duration by delaying accumulation of promitotic factors (9). Inhibiting ATR led to premature cyclin B protein accumulation in the S phase (Fig. 1, D to F, and fig. S5). Moreover, ATR inhibition prematurely increased cyclin B1 mRNA in early-S-phase synchronized cells, whereas in mock-treated synchronized cells, cyclin B mRNA increased concomitantly with the S/G2 transition (Fig. 1G and fig. S6). To determine whether ATR similarly controls the transcription of additional genes, we performed RNA sequencing (RNA-seq). Unsupervised clustering analysis of sequenced reads revealed a group of genes prematurely up-regulated with ATRi in the S phase (Fig. 1H, group 5; fig. S6E; and table S1). Consistent with their normal G2 expression, Gene Ontology (GO)–term analysis of group 5 showed enrichment for promitotic factors (Fig. 1, I and J). These findings suggest a G2/M gene network poised for transcription in the S phase but repressed by ATR until G2.

Next, we analyzed publicly available ChIP (chromatin immunoprecipitation) sequencing, also known as ChIP-seq, data for transcription factor enrichment at the promoters of group 5 genes. B-MYB and FOXM1, transactivators of a mitotic transcription program (10, 11), were highly enriched at these sites (fig. S7A). We asked whether knockdown of either would prevent premature cyclin B accumulation. B-MYB knockdown caused a G1 arrest, precluding further study (fig. S7, B to D). FOXM1 knockdown, which did not block S-phase entry, prevented premature cyclin B accumulation in ATR-inhibited cells (fig. S7, B and E), confirming that FOXM1 drives premature cyclin B expression.

To determine whether ATR regulates B-MYB and FOXM1, we monitored their phosphorylation, a requirement for their activation (11, 12). Whereas both proteins normally exhibit a profound phosphorylation shift at the S/G2 transition, ATR inhibition in the early S phase triggered immediate hyperphosphorylation (Fig. 2, A and B, and fig. S7F). In asynchronous cells, single-cell QIBC analysis revealed that FOXM1 threonine-600 (T600) phosphorylation (fig. S8A), a CDK-dependent activating modification (12), rises in the early S phase, then again in G2. ATR inhibition caused rapid and premature FOXM1 phosphorylation throughout the S phase, raising pFOXM1 to G2 levels (Fig. 2, C to F, and fig. S8, B to D). This premature phosphorylation occurred even after blocking replication with thymidine (fig. S9A). Moreover, mRNA FISH (fluorescence in situ hybridization) analysis of the FOXM1 target PLK1 revealed that ATRi induced its premature expression with or without thymidine (fig. S9, B and C). Thus, ATR suppresses FOXM1 phosphorylation and downstream transcription in S-phase cells, irrespective of S-phase progression.

Fig. 2 ATR controls an S/G2 FOXM1 phosphorylation switch.

(A) Western blots of hTERT RPE-1 cells synchronized as in fig S6A. Async, asynchronous. (B) Quantification of FOXM1 band intensities from (A) plotted as red lines (right y axis). Black lines (left y axis) represent the fraction of S-phase synchronized cells, calculated as described in fig. S6, B to D. (C) Experimental design and representative images of S-phase hTERT RPE-1 cells. Scale bar, 15 μm. (D) QIBC plots of DNA content versus EdU mean intensity, with mean pFOXM1 T600 intensity shown by a color scale. Boxes indicate gated populations used for analyses. (E) Median pFOXM1 intensities from (D). Error bars, SEM of n = 4. (F) pFOXM1 T600 mean intensity in cells treated as described in (C). (G) QIBC plot of EdU-labeled cells with 4n DNA content. Boxes indicate gated populations used for analysis in (H). Numbers indicate cells in each gated population. The S/G2 population was determined as described in fig. S10. (H) pFOXM1 T600 mean intensity in indicated populations. In (F) and (H), whiskers indicate the 10th and 90th percentiles, boxes span the 25th to 75th percentiles, and lines inside boxes represent medians.

The marked difference in S- and G2-phase pFOXM1 levels suggests that a specific regulatory event defines the S/G2 transition. Thus, we measured pFOXM1 levels during the late S phase, the S/G2 transition (fig. S10), and the early G2 phase. pFOXM1 levels rose abruptly in the S/G2 population (Fig. 2, G and H), a finding confirmed using EYFP-PCNA and pFOXM1 imaging (fig. S11, A and B). Given that this population makes up less than 4% of the late S and early G2 cells, phosphorylation is switchlike, in that it happens rapidly at the S/G2 transition.

Paradoxically, cyclin A–CDK phosphorylates FOXM1 (12), but total cyclin A–CDK activity gradually increases throughout the S phase (13), which is inconsistent with switchlike behavior. To test whether a specific cyclin A–CDK complex mediates this behavior, we measured pFOXM1 levels after either CDK2 or CDK1 inhibition. CDK2 inhibitors had little effect on FOXM1 phosphorylation (Fig. 3, A and B, and fig. S12A), although they decreased EdU (5-ethynyl-2′-deoxyuridine) incorporation (fig. S12B). In contrast, CDK1 inhibition prevented FOXM1 phosphorylation at the S/G2 transition and after ATR inhibition (Fig. 3, A and C, and fig. S12). Furthermore, analog-sensitive CDK1 inhibition (14) reduced FOXM1 phosphorylation, and CDK2 inhibition had no additional effect (fig. S13). These results show that CDK1 phosphorylates FOXM1 at the S/G2 transition.

Fig. 3 ATR is active until G2, preventing CDK1-dependent FOXM1 phosphorylation.

(A) Experimental design and QIBC plots of DNA content versus EdU mean intensity, with mean pFOXM1 T600 intensity shown by a color scale, in hTERT RPE-1 cells. Boxes indicate gated populations used for analyses. (B and C) Median pFOXM1 intensities of populations shown in (A). Error bars, SEM of n = 3. (D) Representative images of indicated cell stage in hTERT RPE-1 cells. (E) ATR activity during an unperturbed cell cycle as described in fig. S14. Error bars, SEM of n = 3.

Given that ATR suppresses FOXM1 phosphorylation until G2, we hypothesized that ATR activity may abruptly decline as the S phase ends, relieving an intrinsic checkpoint. Consistent with this hypothesis, ATR-dependent H2AX phosphorylation (γH2AX) increased as cells entered the S phase, peaked in the mid–S phase, and rapidly decreased as cells completed DNA replication (Fig. 3, D and E, and fig. S14, A to D). Thus, ATR is active throughout the normal S phase, but its activity drops at the S/G2 transition, allowing rapid FOXM1 phosphorylation.

In mammalian cells, either a RAD9A-TOPBP1 or ETAA1-dependent pathway activates ATR (4). ETAA1 knockdown reduced ATR activity in an unperturbed S phase, whereas RAD9A knockdown did not, even though it reduced γH2AX after replication stress (Fig. 4A and fig. S14E). Moreover, deleting the ETAA1 ATR-activation domain greatly reduced ATR activity in an unperturbed S phase, but not in response to replication stress, whereas auxin-mediated degradation of a TOPBP1-mAID fusion had the opposite effect (Fig. 4, B and C, and fig. S15). Thus, ETAA1 activates ATR during an unperturbed S phase, whereas TOPBP1-RAD9A activates ATR to enforce the replication stress response, providing a rationale for the existence of multiple ATR activators.

Fig. 4 The ETAA1-ATR pathway couples the S phase and mitosis.

(A) ATR activity in small interfering RNA (siRNA)–transfected hTERT RPE-1 cells during an unperturbed cell cycle as described in fig. S14. Error bars, SEM of n = 3. (B) Illustration of the mAID degron system fused to TOPBP1. Ub, ubiquitin. (C) ATR activity in HCT116 cell lines mock-treated or auxin-treated (2 hours) as described in fig. S15A. Error bars, SEM of n = 3. AAD, ATR-activation domain. (D and E) S-to-M progression assay (see Fig. 1, A and B) in hTERT RPE-1 cells 40 hours after siRNA transfection. Representative images are 6 hours after EdU wash-off. (E) Fraction of S phase–gated cells in mitosis as a function of time after EdU wash-off. Error bars, SEM of n = 3. (F) Representative images of anaphase cells mock- or ATRi-treated for 8 hours, 40 hours after siRNA transfection. (G) Fraction of cells with UFBs. Error bars, SEM of n = 3. **P < 0.01; ns, not significant.

ATR enforces its checkpoint functions in part by activating CHK1 (checkpoint kinase 1), an effector kinase that inhibits CDKs (4). CHK1 inhibition also triggered premature FOXM1 phosphorylation and cyclin B accumulation (fig. S16, A to C). Thus, the repressive activity of an ETAA1-ATR-CHK1 pathway on the cyclin A–CDK1–FOXM1 axis controls the S/G2 transition (fig. S16D).

FOXM1 overexpression is sufficient to accelerate mitotic entry (10); therefore, we asked whether ATR slows the S-to-M progression by controlling FOXM1. Consistent with this hypothesis, reducing FOXM1 to levels that permit mitotic entry, while suppressing premature cyclin B expression in ATR-inhibited cells (fig. S7E), prevented ATRi-induced premature mitosis (Fig. 4, D and E). Furthermore, partial FOXM1 knockdown reduced ultrafine anaphase bridges (UFBs) and 53BP1 bodies after ATR inhibition (Fig. 4, F and G, and fig. S17). Because these both indicate a failure to complete DNA replication before mitosis (5, 15, 16), we conclude that proper control of the S/G2 phosphorylation switch promotes the completion of DNA replication and prevents genome instability.

Our data unveil an ATR-controlled S/G2 phosphorylation switch that initiates the G2/M transcription program. By repressing CDK1, ATR blocks this switch until the S phase ends to properly time G2-specific events. Although the molecular signal activating ATR in an unperturbed S phase is unclear, ETAA1’s role suggests that single-stranded DNA (ssDNA) generated by ongoing replication, rather than unreplicated DNA, is a critical component. Transiently formed RPA-coated ssDNA could serve as a platform for colocalizing ETAA1 and ATR, sustaining ATR activity throughout the S phase. The decline in ATR activity that occurs once replication is complete then signals the S/G2 transition, executed via the CDK1-pFOXM1 switch (fig. S18).

Because ATR ensures that G2 depends on S-phase completion, we refer to this as an intrinsic S/G2 checkpoint and propose that this checkpoint monitors replication completion (3). This checkpoint prevents a cellular identity crisis in which the S and G2 phases overlap, which would cause underreplication, early mitosis, and subsequent DNA damage. Given the frequent overexpression of FOXM1 in cancer (17), deregulation of this fundamental cell cycle transition could be a common event contributing to cancer genome instability.

Supplementary Materials

Materials and Methods

Figs. S1 to S18

Table S1

References (1822)

References and Notes

Acknowledgments: We thank F. Ochs, J. Ferrell, and the Cimprich laboratory for comments on the manuscript. Funding: This work was supported by grants from the NIH to K.A.C. (ES016486), D.C. (CA102729), and T.M. (GM127026); the Wellcome Fund (107022 and 203149) to W.C.E.; the American Cancer Society (PF-15-165-01–DMC) and the Burroughs Wellcome Fund Postdoctoral Enrichment Program to J.C.S.; the German Research Foundation DFG (HA 6996/1-1) to S.H.; and by a Mexican Government CONACYT fellowship to F.C.-S. Author contributions: J.C.S., W.C.E., D.C., K.S., and K.A.C. designed experiments. J.C.S., S.H., F.C.-S., K.S., J.R.P., L.X., T.E.B., and M.C. performed experiments. M.J.B. and M.C. performed the bioinformatics analyses. J.C.S., M.J.B., M.C., T.M., and K.A.C. analyzed the data. J.C.S. and K.A.C. wrote the manuscript. Competing interests: None declared. Data and materials availability: RNA-seq data are available in the GEO repository (accession number GSE116131).

Stay Connected to Science

Navigate This Article