Intragenic origins due to short G1 phases underlie oncogene-induced DNA replication stress

Abstract

Oncogene-induced DNA replication stress contributes critically to the genomic instability that is present in cancer. However, elucidating how oncogenes deregulate DNA replication has been impeded by difficulty in mapping replication initiation sites on the human genome. Here, using a sensitive assay to monitor nascent DNA synthesis in early S phase, we identified thousands of replication initiation sites in cells before and after induction of the oncogenes CCNE1 and MYC. Remarkably, both oncogenes induced firing of a novel set of DNA replication origins that mapped within highly transcribed genes. These ectopic origins were normally suppressed by transcription during G1, but precocious entry into S phase, before all genic regions had been transcribed, allowed firing of origins within genes in cells with activated oncogenes. Forks from oncogene-induced origins were prone to collapse, as a result of conflicts between replication and transcription, and were associated with DNA double-stranded break formation and chromosomal rearrangement breakpoints both in our experimental system and in a large cohort of human cancers. Thus, firing of intragenic origins caused by premature S phase entry represents a mechanism of oncogene-induced DNA replication stress that is relevant for genomic instability in human cancer.

Extended Data Figure 1. Experimental setup to study S phase entry and DNA replication initiation.

a, Cyclin E protein levels, as determined by immunoblotting, in cells with normal levels of cyclin E (NE) and cells overexpressing cyclin E (OE) (2.5 days after tetracycline withdrawal). Actinin serves as a loading control. This is a representative example of more than ten independent replicates. b, Experimental outline of the protocol used to monitor S phase entry by flow cytometry and of the EdUseq protocols: EdUseq-HU, EdUseq-noHU and EdUseq-HU/release. *, indicates that EdU was added 30 min before harvesting the cells. c, Flow cytometry profiles of cells with normal levels of cyclin E (NE) and cells overexpressing cyclin E (OE), after mitotic shake-off (0 h) and 14 and 10 h later, respectively, after the cells had been released in media containing HU and EdU. 2C and 4C, DNA content of G1 and G2 cells, respectively. This is a representative example of more than ten independent replicates. d, DNA content versus EdU incorporation flow cytometry plots of NE and OE cells. EdU-positive NE and OE cells were gated blue and red, respectively. 2C and 4C, DNA content of G1 and G2 cells, respectively. The gating strategy for these data is shown in Supplementary Information Fig. S1.

Extended Data Figure 2. Identification of replication timing domains by REPLIseq.

a, Experimental outline of the REPLIseq protocol and FACS profiles of cells with normal levels of cyclin E (NE) and cells overexpressing cyclin E (OE) for 1, 2 or 7 days (1d, 2d and 7d, respectively). Cells were sorted according to DNA content into early (blue), mid (green) and late (yellow) S phase fractions. 2C and 4C, DNA content of G1 and G2 cells, respectively. b, Assignment of replication timing (RT) domains. The fractions of the genome that were replicated in early, mid or late S phase were determined on the basis of the REPLIseq profiles of the NE and 2d OE cells. c, Distribution of early, mid and late replication timing bins in 2d OE cells according to their replication timing in NE cells. d, Comparison of the origin firing profile determined by EdUseq-HU (Fig. 1b) and the early S replication profile determined by REPLIseq in NE cells. Replication timing (RT) domains and genic/intergenic (Ge/iG) regions are as in Fig. 1b. Bin resolution, 10 Kb; ruler scale, 100 Kb. e, REPLIseq profiles of the first 10 Mb of chr7 of cells expressing normal levels of cyclin E (NE) or overexpressing cyclin E (OE) for 1, 2 or 7 days (1d, 2d and 7d, respectively). Profiles are shown separately for the cells in early, mid and late S phase. Replication timing (RT) domains and genic/intergenic (Ge/iG) regions are as in Fig. 1b. Bin resolution, 10 Kb; ruler scale, 100 Kb.

Extended Data Figure 3. Further characterization of replication origins.

a, Adjusted average sigma values at 1 Kb resolution around 1,828 origins refined by performing the EdUseq-HU protocol in the additional presence of mimosine or aphidicolin (light and dark gray, respectively), compared to OE and NE cells treated only with HU (pink and blue, respectively). aσ, adjusted σ. b, Distribution of the subset of origins refined at 1 Kb resolution (1,828 origins) relative to origin type, as determined by the original 10 Kb resolution assignment: CN, constitutive; Oi2, oncogene-induced 2; Oi, oncogene-induced. c, Distribution of constitutive (CN) and oncogene-induced (Oi2 and Oi) origins, refined at 1 Kb resolution (1,828 origins), according to replication timing (RT) domains (E, early; M, mid; L, late S phase). d, Distribution of constitutive (CN) and oncogene-induced (Oi2 and Oi) origins, refined at 1 Kb resolution (1,828 origins), according to gene annotation in all replication timing domains (all-RT) or only in the early S phase replicating domains (E-RT). e, Transcription (EUseq) levels (median) in NE (light gray) and OE (dark gray) cells at sites of constitutive (CN) and oncogene-induced (Oi) origins, refined at 1 Kb resolution, at various time points after mitotic shake-off. Origins mapping to genic (Ge) or intergenic (iG) genomic bins were plotted separately. rσ, relative σ. f, Scatter plots of EdUseq-noHU σ values (log2) at all origins (CN, purple; Oi2, pink; Oi, red) at 10 Kb resolution for NE vs OE cells not treated with HU (Fig. 1h). EdU was present during the indicated time points.

Extended Data Figure 4. Further characterization of transcription profiles and effects of DRB.

a, Genome-wide comparison (all genomic bins) of newly-synthesized transcripts (EUseq) in cells expressing normal levels of cyclin E (NE) vs cells overexpressing cyclin E (OE) cells at the indicated time points after mitotic shake-off. The two genomic bins mapping to the CCNE1 (cyclin E) gene are colored red; rσ, relative sigma. b, List of early and mid S large genes along which replication initiation and transcription profiles were plotted in Fig. 2d. For each gene, the association with a common fragile site (CFS) is indicated. c, Inhibition of transcription by DRB. EU incorporation was monitored by fluorescence microscopy in control, DRB-treated (9 h), and in DRB-treated (9 h) and released (5 h) cells. The nuclei of the cells were counterstained with DAPI. d, DNA content versus EdU incorporation flow cytometry plots of NE cells 14 h after mitotic shake-off. The cells were treated with DRB for the indicated times. EdU-positive NE cells were gated in blue. 2C and 4C, DNA content of G1 and G2 cells, respectively. e, Newly-synthesized transcript profiles (EUseq) at a representative genomic region in NE cells treated with DRB for 9 h after mitotic shake-off and then released for 30 or 120 min (release 30 min, dark gray; release 120 min, light gray; overlap: color; direction of transcription: green, forward; red, reverse; yellow, bidirectional). The red arrow indicates the transcription of the gene harboring oncogene-induced origins at our example locus on chromosome 7 and the green arrow indicates another large gene in this locus. Replication timing (RT) domain and gene (Ge/iG) annotations are as in Fig. 1b; rσ, relative sigma. f, Average transcription (EUseq, log₂rσ) in NE cells treated with DRB for 9 h after mitotic-shake-off and then released for 30, 120 or 240 min, along the length of large genes (>0.35 Mb for early S and >0.65 Mb for mid S genes). The genes are grouped according to replication timing (RT; early, mid S) and level of transcription (high Tx, upper tercile; mid Tx, middle tercile). rσ, relative σ; #Ge, number of genes averaged at each position.

Extended Data Figure 5. Accelerated entry into S phase and firing of novel intragenic origins upon Myc activation.

a, Myc activation (3 days after adding 4-OHT), as determined by immunfluorescence, in cells with non-induced (NM) and induced (OM) Myc activity. Nuclei were counterstained with DAPI. Representative images from two independent experiments are shown. b, Quantification of EdU positive cells at different time points after mitotic shake-off. Means and SDs were calculated from three independent experiments; gray symbols, individual data points. NM, normal Myc activity; OM, induced Myc activity. c, Replication initiation (EdUseq-HU) profiles at a representative genomic region in OM and NM cells, harvested at the indicated times after mitotic shake-off. Peak heights are represented as sigma values (σ). Replication timing (RT) domains and gene annotations (Ge/iG) are as in Fig. 1b. Bin resolution, 10 Kb; ruler scale, 100 Kb. d, Classification of constitutive (CN) and oncogene-induced (Oi2 and Oi) origins based on relative height ratios in OM versus NM cells (OM:NM). e, Scatter plots of EdUseq-HU σ values at origins (CN, purple; Oi2, pink; Oi, red) in NM vs OM cells at the indicated time points after mitotic shake-off. f, Distribution of CN, Oi2 and Oi origins in OM and NM cells according to RT domains (E, early; M, mid; L, late S phase). g, Distribution of CN, Oi2 and Oi origins in OM and NM cells according to gene annotation in all replication timing domains (all-RT) or only in the early S phase replicating domains (E-RT). h, Relative adjusted sigma ratios of replication origins identified in NE (normal cyclin E activity), NM, OE (cyclin E overexpression) or OM cells. Left, number of origins identified in NE or NM cells grouped according to their relative height ratios between these two cell lines. Right, number of Oi origins identified in OE or OM cells grouped according to their level of induction relative to the NE and NM cells, respectively. i, Newly-synthesized transcript profiles (EUseq) at a representative genomic region in OM and NM cells 10 h and 14 h after mitotic shake-off, respectively (NM: light gray; OM: dark gray; overlap: green, forward; red, reverse; yellow, bidirectional direction of transcription). Replication timing (RT) domain and gene (Ge/iG) annotations are as in (c). j, Genome-wide comparison (all genomic bins) of transcription in OM vs NM cells 10 h and 14 h after mitotic shake-off, respectively. rσ, relative σ. k, Median transcription (EUseq) levels in NM (light gray) and OM (dark gray) cells at CN and Oi origins mapping to genic (Ge) or intergenic (iG) genomic bins at 14 and 10 h after mitotic shake-off, respectively.

Extended Data Figure 6. S phase entry and replication initiation profiles of HeLa and RPE1 cells.

a, Percentage of EdU positive HeLa and RPE1 cells at different time points after mitotic shake-off (0 h). Means and individual data points are shown from two independent experiments. b, Replication initiation (EdUseq-HU) profiles at a representative genomic region in HeLa and RPE1 cells at the indicated time points after mitotic shake-off. The profile of U2OS cells expressing normal levels of cyclin E (NE, blue) serves as reference. c, Scatter plots of EdUseq-HU σ values (log₂) at all origins (CN, purple; Oi2, pink; Oi, red) in HeLa and RPE1 cells vs U2OS cells with normal levels of cyclin E (NE) at the indicated time points after mitotic shake-off.

Extended Data Figure 7. Replication initiation and transcription profiles at selected genomic loci.

a, Replication initiation (EdUseq-HU) profiles at three genomic loci in different cells lines, from top to bottom: cells overexpressing cyclin E (OE) vs cells with normal cyclin E (NE) activity, harvested 6 and 14 h after mitotic shake-off, respectively; cells with induced Myc (OM) activity vs normal Myc (NM) activity, harvested 6 and 14 h after mitotic shake-off, respectively; HeLa cells harvested at 6 h vs 14 h after mitotic shake-off and RPE1 cells harvested 14 h after mitotic shake-off. Peak heights are represented as sigma values (σ). Replication timing (RT) domains and gene annotations (Ge/iG) are as in Fig. 1b. Bin resolution, 10 Kb; ruler scale, 100 Kb. The green arrows indicate the direction of transcription of the example gene of each locus harboring oncogene-induced origins. b, Replication initiation (EdUseq-HU) profiles of control (noDRB) and DRB-treated (0-9 h) NE cells harvested 14 h after mitotic shake-off. The same genomic loci as in (a) are shown, focusing on the genes harboring the oncogene-induced origins. Replication timing (RT) domain and gene annotations (Ge/iG) are as in (a). c, Newly-synthesized transcript profiles (EUseq) of NE cells, 2 and 14 h after mitotic shake-off (2h, light green; 14 h, gray; overlap, dark green) shown only for the example genes harboring the oncogene-induced origins shown in (a) (indicated by the green arrows). Replication timing (RT) domain and gene (Ge/iG) annotations are as in (a); rσ, relative sigma.

Extended Data Figure 8. Fork collapse at Oi origins induced in cells with normal levels of cyclin E by inhibiting transcription in early G1.

a, Replication initiation (EdUseq-HU, 14 h HU block) and fork progression (EdUseq-HU/release 60 min) profiles at a representative genomic region in U2OS cells with normal levels of cyclin E (NE) that were treated or not with DRB during the first 7 h of G1. Replication timing (RT) domains and gene annotations (Ge/iG) are as in Fig. 1b. Bin resolution, 10 Kb; ruler scale, 100 Kb. b, Average fork progression (no release and 60 min release) at constitutive (CN) and oncogene-induced (Oi) origins located in highly transcribed regions in control and DRB treated (first 7 h of G1 phase) NE cells. aσ, adjusted average σ.

Extended Data Figure 9. Association of Oi origins with genomic rearrangements and replication timing profiles of cancer rearrangement breakpoints.

a, Mapping of translocations (Transloc; n=27,364) identified by LAM- HTGTS to genomic regions replicated from Oi origins (oncogene-induced replication initiation domains, OiRDs) with the analysis restricted to the early S replicating domains. The fraction of translocations mapping to OiRDs is shown for non-transcribed (0), low (Lo), medium (Me) and highly (Hi) transcribed genomic bins, as well as for all early S replicating bins. Statistical comparisons, using random permutation analysis, are between the NE (normal cyclin E activity, blue) and OE (cyclin E overexpression, pink) samples. The distribution of OiRDs in the genome (gray) is shown as reference. b, Mapping of genomic rearrangement (Rearr; n=136) breakpoints, identified previously in the same cells overexpressing cyclin E (OE) to the oncogene-induced replication initiation domains (OiRDs), according to transcription levels, as in (a), with the analysis restricted to the early S replicating domains. Statistical comparisons, using random permutation analysis, are between observed (red) and genomic (gray) frequencies. NS, not significant. c, Mapping of genomic rearrangement (Rearr; n=490,711) breakpoints from a TCGA pan-cancer dataset to the oncogene-induced replication initiation domains (OiRDs), according to transcription levels, as in (a), with the analysis restricted to the early S replicating domains. Statistical comparisons, using z-scores, are between observed and genomic frequencies. d, Mapping of genomic rearrangement (Rearr; n=490,711) breakpoints in common cancer types from a TCGA pan-cancer dataset to the oncogene-induced replication initiation domains (OiRDs), with the analysis restricted to the early S replicating domains. KIRC, kidney renal cell; COAD, colon adenocarcinoma; HNSC, head and neck squamous cell; UCEC, uterine cervix; GBM, glioblastoma multiformae; LUAD, lung adenocarcinoma; LUSC, lung squamous cell; BRCA, breast; BLCA, bladder; OV, ovary. Statistical comparisons, using z-scores, are between observed (red) and genomic (gray) frequencies. e, Distribution of cancer rearrangement breakpoints according to the replication timing data obtained from the REPLIseq experiment shown in Extended Data Fig. 2.

Extended Data Figure 10. Proposed mechanism for oncogene-induced DNA replication stress.

During the length of a normal G1 phase, transcription progressively inactivates intragenic origins, such that upon S phase entry origin firing is restricted to intergenic domains. Following oncogene activation, cells enter prematurely into S phase, prior to the inactivation of all intragenic origins. This results in bidirectional forks within highly transcribed genes, leading to conflicts between the replication and transcription machineries, fork collapse, DNA DSBs and genomic instability.

Figure 1. Firing of novel origins upon cyclin E overexpression.

a, Percentage of EdU positive cells (mean and SD; n=3 independent experiments; gray symbols, individual data points) at the indicated times after mitotic shake-off. OE, overexpression of cyclin E; NE, normal levels of cyclin E. b, Replication initiation profiles (EdUseq-HU) at a representative genomic region in OE and NE cells harvested 6 and 14 h after mitotic shake-off, respectively. RT, replication timing (blue, early; green, mid; yellow, late S phase); Ge, genes (green, forward direction of transcription; red, reverse; yellow, unspecified; blue, multiple genes within bin); iG, intergenic (gray). Bin resolution, 10 Kb; ruler scale, 100 Kb; σ, sigma. c, Classification of origins based on adjusted σ value ratios in OE over NE cells: CN, constitutive < 2-fold; Oi2, oncogene-induced 2 > 2-fold; Oi, oncogene-induced > 4-fold. d, e, Distribution of CN, Oi2 and Oi origins according to RT (d) (E, early; M, mid; L, late S phase) and gene annotation (e) (All-RT, all RT domains; E-RT, early S RT genomic domains). f, g, Replication initiation profiles (EdUseq-HU) at a representative genomic region (f) and scatter plots of EdUseq-HU values at all origins (g) in NE and OE cells at the indicated times after mitotic shake-off. h, Replication initiation profiles (EdUseq-noHU) at a representative genomic region in OE and NE cells. EdU was present during the indicated times following mitotic shake-off.

Figure 2. Suppression of intragenic origin firing by transcription.

a, Newly-synthesized transcript profiles (EUseq) at a representative genomic region in cells with normal cyclin E (NE) levels or overexpressing cyclin E (OE), 6 h after mitotic shake-off (OE: dark gray; NE: light gray; overlap: color; direction of transcripts: green, forward; red, reverse; yellow, bidirectional). Replication timing (RT) and gene (Ge/iG) annotations are as in Fig. 1b. rσ, relative sigma. b, Median transcript levels (EUseq) at the genomic bins corresponding to genic (Ge) and intergenic (iG), constitutive (CN) and oncogene-induced (Oi) origins in NE (light) and OE (dark gray) cells at the indicated times after mitotic shake-off. c, Transcription (EUseq) and replication initiation (EdUseq-HU) profiles at the indicated times after mitotic shake-off for the gene marked by the arrow in (a). d, Average transcription (EUseq) and replication initiation (EdUseq-HU or EdUseq-noHU) along the length of large (>0.35 Mb for early S; >0.65 Mb for mid S), transcribed (Tx) genes in OE, NE and DRB-treated NE cells at various time points after mitotic shake-off. High Tx, upper tercile; mid Tx, middle tercile; rσ, relative σ; aσ, adjusted σ; #Ge, number of averaged genes. e, f, Replication initiation (EdUseq-HU) profiles at a representative genomic region (e) and scatter plots of EdUseq-HU values at all origins (f) in control (no DRB) and DRB-treated (0-5, 0-7 or 0-9 h) NE cells harvested 14 h after mitotic shake-off.

Figure 3. Collapse of Oi forks due to conflicts with transcription.

a, Fork progression profiles (EdUseq-HU/release) at a representative genomic region in cells overexpressing cyclin E (OE) or with normal cyclin E (NE) levels that were arrested with HU for 10 or 14 h, respectively, after mitotic shake-off and then released for the indicated time periods. EdU was added 30 min before harvesting the cells. The no release (EdUseq-HU) profile is shown as reference. Replication timing (RT) and gene (Ge/iG) annotations are as in Fig. 1b. σ, sigma. b, Genome-wide average fork progression at constitutive (CN) origins in OE and NE cells from the experiment shown in (a). Fork speeds were determined from the distances traveled by the forks between the 90 and 150 min release time points. aσ, adjusted σ. c, Replication initiation (EdUseq-HU) profiles at a representative genomic region in OE and NE cells (upper panel, reference) and fork progression profiles of OE cells released from a 6 h HU block for the indicated times and labeled with EdU for 30 min before harvesting (lower panels, EdUseq-HU/release). d, Genome-wide average fork progression at constitutive (CN) and oncogene-induced (Oi2 and Oi) origins located in highly or lowly transcribed (Tx) regions (upper and lower terciles, respectively) from the experiment shown in (c). e, Rescue of fork progression (EdUseq-HU/release) at Oi origins within a representative genomic region in DRB-treated OE cells. DRB was added 4 h after mitotic shake-off and kept until harvesting. ctl, control (no DRB). f, Genome-wide average fork progression at CN, Oi2 and Oi origins located in highly transcribed (Tx) regions from the experiment shown in (e).

Figure 4. Oi origins are associated with DNA DSB formation and genomic rearrangements.

a, Translocations identified by LAM-HTGTS within a representative genomic region in cells overexpressing cyclin E (OE) and cells with normal cyclin E (NE) activity shown as vertical lines (red and blue, respectively, with color intensity reflecting the number of translocations per genomic bin) in the context of the replication initiation profiles of these cells (EdUseq-HU). b, Number of translocations (#Transloc) per origin in OE and NE cells at oncogene-induced (Oi) origins and surrounding genomic bins. Data are plotted separately for origins mapping to highly (red, blue) and lowly (gray) transcribed (Tx) sites (upper and lower terciles, respectively) and statistical comparisons between OE and NE samples were performed by random permutation tests. c-f, Mapping of translocations (Transloc; n=27,364) identified by LAM-HTGTS (c), genomic rearrangement (Rearr; n=136) breakpoints identified previously in the same cells overexpressing cyclin E (OE) (d) and genomic rearrangement (Rearr; n=490,711) breakpoints from a TCGA pan-cancer dataset (e, f) to genomic regions replicated from Oi origins (oncogene-induced replication initiation domains, OiRDs). The fraction of translocations/breakpoints mapping to OiRDs is shown for non-transcribed (0), low (Lo), medium (Me) and highly (Hi) transcribed genomic bins, and for all genomic bins (c-e) or for all genomic bins in common cancer types (f). The distribution of OiRDs in the genome is shown in gray. Statistical comparisons are between NE (blue) and OE (pink) samples (c) or between observed (red) and genomic (gray) frequencies (d-f) and were performed either by random permutation tests (c, d) or by calculating z-scores (e, f). NS, not significant. KIRC, kidney renal cell; COAD, colon adenocarcinoma; HNSC, head and neck squamous cell; UCEC, uterine cervix; GBM, glioblastoma multiformae; LUAD, lung adenocarcinoma; LUSC, lung squamous cell; BRCA, breast; BLCA, bladder; OV, ovary. g, Proposed mechanism for oncogene-induced DNA replication stress.