Oncogenic c-Myc induces replication stress by increasing cohesins chromatin occupancy

Oncogene-induced replication stress is a major driver of genomic instability in cancer cells, with a central role in both cancer initiation and progression (1). Despite its critical role in cancer development, the mechanisms that lay at the basis of oncogene-induced replication stress remains poorly understood. Here, we investigate the mechanism of c-Myc-induced replication stress. Our data shows that c-Myc induces replication stress by increasing the amount of cohesins bound to chromatin in the G1 phase of the cell cycle. This is independent of previously suggested mechanisms involving deregulation of replication initiation and transcriptional interference. Restoring the amount of chromatin-bound cohesins to control levels, or preventing the accumulation of cohesins at CTCF sites, in cells experiencing oncogenic c-Myc activity prevents replication stress. Increased cohesins chromatin occupancy correlates with a c-Myc-dependent increase in the levels of the cohesion loader Mau2. Preventing c-Myc-induced increase in Mau2 reduces oncogene-induced replication stress. Together our data support a novel mechanism for oncogene-induced replication stress. Since c-Myc activation is a crucial event in many human cancers (2), identifying the mechanisms through which this oncogene promotes replication stress provides critical insights into cancer biology.

Replication stress (RS) is defined as the slowing-down and/or stalling of DNA replication forks and is a major source of genomic instability in cancer cells, being often caused by activation of oncogenes or loss of tumour suppressors (3), (1,4,5). Despite the critical role of oncogene-induced RS in cancer, the mechanisms that generate it remain unclear (Figure 1a). Different oncogenes have been studied in various systems, and the general consensus is that RS is likely the result of deregulation of replication initiation events (6). Other proposed mechanisms involve the interference between the replication and transcription machineries (7-9). In the case of oncogenic overexpression of Cyclin E, different mechanisms have been reported (7,10,11). The reduced length of G1 following Cyclin E overexpression has been associated with a decrease in licensing events, which results in fewer replication complexes available for replication. This is thought to culminate in genomic instability due to under-replication (12). More recently, Cyclin E overexpression has also been shown to cause RS by increased transcription-replication collisions in transcribed genes (7, 11). In contrast, the onco-genic activity of the transcription factor c-Myc has been reported to increase replication initiation events, thus causing over-replication (4,5). Surprisingly, while c-Myc is thought to induce a large transcriptional program to promote proliferation and growth (13), it has not been linked to increased transcription-replication interference. Besides transcription machineries other large protein complexes bound to DNA could interfere with replisome progression. The cohesin complex is probably one of the most abundant protein complexes interacting with the DNA. Cohesins are ring-shaped multiprotein complexes comprising two major subunits, Structural Maintenance of Chromosome (Smc) 1 and Smc3, along with the kleisin subunit Rad21 and Stag1 and Stag2 in mammalian cells (14). The loading of cohesins onto DNA is highly regulated, and in mammalian cells depends on the activity of the loaders Mau2 and Nipped-B-like protein (NIPBL). The activity of the loaders is antagonised by the release factor Wapl (15). While loading and release occur throughout the entire cell cycle, during S-phase and G2 cohesins interaction with the DNA is more stable (16). This is due to the establishment factors, Esco1 and Esco2 in mammalian cells, which acetylate the Smc3 subunit of the cohesin complex (17), and thus prevent the release activity of Wapl (18). Recent evidence confirmed that in mammalian cells, as previously reported in budding yeast (19), cohesin rings are able to move along the DNA in a transcription-dependent manner (20). Binding of the CCCTC-Binding Factor (CTCF) to CTCF sites is involved in the organisation of spatially interacting regions of chromatin (21), known as topologically associated domains (TADs) (21,22). It has been shown that CTCF sites can act as a road-block for cohesins, with accumulation of cohesins often detected at CTCF sites in mammalian cells (23,24). While the majority of cohesins interact dynamically with the DNA, some are associated more stably, and reside at CTCF sites, where they participate in the organization of chromatin loops (25). To establish which of the different mechanisms contribute to c-Myc-induced RS, we exploited a c-Myc-ER inducible RPE1 hTERT cell line (26). In this system oncogenic c-Myc activation depends on the translocation of the c-Myc-ER protein into the nucleus after 4OH tamoxifen (4OH-T) addition (Figure S1a, S1b), which is independent of c-Myc-ER protein levels ( Figure S1c). 24 hr to 48 hr after 4OH-T addi-tion oncogenic c-Myc activity can be observed via a decrease in colony formation (S1d) and increased gene expression of well-known c-Myc target genes at the mRNA (Figure S1e, S1f) and protein level ( Figure S1g). As shown previously (26), activation of c-Myc-ER by 4OH-T induces RS-induced DNA damage response within 24 hours (Figure S1i, S1j) and shortening of DNA fibre length (Figure S1k, S1l), indicative of slowing down or replication forks, supporting the use of this system to study c-Myc-induced RS.
To gain some initial insights into how c-Myc induces RS, we arrested cells in different cell cycle phases, released them with or without c-Myc activation and tested the levels of RS and DNA damage in the following S phase. First, we used confluency to arrest cells in G0/G1. After release into the cell cycle, the oncogene c-Myc was activated by adding 4OH-T or left untreated as control (Figure 1b, S1m), allowing us to study the first G1 and S phases after the oncogene activation. In order to measure RS, we analysed the length of DNA fibres. The activation of c-Myc reduces the average DNA fibre length compared to control, suggestive of slowing down of replication forks (Figure 1c, 1d and Figure S1n). We then measured DNA damage by monitoring the phosphorylation of the histone H2AX. We observed increased levels of the DNA damage marker γH2AX in S phase cells with both Western blot and immunofluorescence (Figure 1e-h). We confirmed these findings by synchronising the cells via nocodazole shake-off (Figure S1o-s). These data suggest that the activation of c-Myc during G1 and S phase can induce RS and DNA damage.
To test if G1 phase is required for c-Myc to induce RS, we synchronised cells in early S phase by adding hydroxyurea (HU). Subsequently we washed out HU to allow S phase progression with or without c-Myc activation and analysed the levels of RS and DNA damage ( Figure S2a, S2b). Surprisingly, we did not observe any decrease in DNA fibre length in c-Myc activated cells compared to control ( Figure S2c). On the contrary DNA fibres appear longer upon c-Myc induction, indicating that c-Myc activity during S phase does not cause RS, but might even increase the replication capacity of the cell as reported in Pennycook et al (27). Correspondingly, we did not observe any increase in DNA damage in c-Myc cells ( Figure S2d). As expected, c-Myc activation increased replication initiation events (origin firing, Figure S2c) and increased Cyclin E levels, a transcriptional target of c-Myc, as well as Mau2 levels (Figure S2e,S2f). This data suggests that c-Myc-induced increase in origin firing does not cause RS per se. Since HU treatment by itself causes RS, which might mask c-Myc-induced RS, we also released cells synchronously into S phase after a G1 arrest via CDK4/6 inhibition by Palbociclib, as reported by Trotter et al (28). c-Myc was activated, via addition of 4OH-T, either immediately after release, therefore throughout G1 phase, or immediately before entering S phase, and DNA damage and DNA fibre length were analysed as above ( Figure S2g-j). These data confirm that c-Myc activity during G1 is required to generate RS in S phase.
A possible cause could be a decrease in G1 length, which has been associated with reduced origin licensing and consequent RS upon Cyclin E overexpression (12). We evaluated replication origin licensing in pre-extracted samples by measuring chromatin bound Mcm7, a component of the helicase complex that is loaded on the DNA during G1 as in . We did not observe a decrease in origin licensing, ( Figure S2k), indicating that reduced origin licensing is an unlikely causative mechanism for c-Myc-induced RS.
It has previously been reported that RS can result from increased transcription-replication collisions (7,29). To test if interference between replication and c-Myc-induced transcription contributes to RS we decreased global transcription levels after c-Myc activation via treatment with the RNA polymerase inhibitor 5,6-Dichloro-1-β-d-Ribofuranosyl Benzimidazole (DRB) for 2 hr, , like in Kotsantis et al (8), during the first S phase after release from G1 ( Figure S2l-m). We confirmed that c-Myc activation increases global transcription levels, by 5-Ethynyl Uridine (EU) incorporation, and that DRB reduces it, ( Figure S2l). However, DNA fibre analysis shows that this does not rescue the c-Myc-dependent decrease in DNA fibre length ( Figure S2m). This suggests that in this system transcription-replication interference does not significantly contributes to c-Myc-induced RS levels, though we cannot exclude a role for this mechanism in other settings.
We next focused on alternative mechanisms that could contribute to c-Myc-induced RS. A potential source of RS could be protein complexes interacting with the DNA during S phase that could slow down fork progression. A potential candidate for this would be the cohesin complexes. Whilst cohesin has been shown to have a role in preventing RSinduced DNA damage (30), we hypothesised that a c-Mycinduced increase in cohesin occupancy during G1 could slow down replisome progression during S phase as suggested in Kanke et al (31) and Morales et al (32). We therefore analysed the fraction of chromatin bound cohesin subunits Smc1 and Smc3 by Immunofluorescence (IF) of pre-extracted samples, in cells released from G0/G1 with and without c-Myc. These data show that higher levels of Smc1 and Smc3 are detected on chromatin in c-Myc-activated cells both in G1 and in S phase (Figure 2a, S3a). This was confirmed in asynchronous cell populations ( Figure S3b) and by chromatin preparations followed by Western blot analysis (Figure 2b).
Our IF analysis on pre-extracted nuclei together with the chromatin preparation and Western blot analysis establish an overall increase in chromatin occupancy of the cohesin subunits Smc1 and Smc3, but do not reveal where these accumulate in the genome. To investigate if this c-Myc-dependent increase in chromatin bound cohesin accumulates at specific sites, we established genome-wide binding of the cohesin component Smc1 by ChIPseq in cells with activated c-Myc or control (Figure 2c-e). We performed these experiments at 48 hr after c-Myc activation, when we observe consistent levels of RS (Figure S1k), and no obvious differences in cell cycle distribution between control and c-Myc-activated cells ( Figure S1h). These experiments show that cohesin accumulates at 50,493 sites in c-Myc cells compared to 43,906 in control (Figure 2d, (Figure 2e). Motif analysis of the peaks in control shows enrichment of CTCF sites, which is in line with published data (23). Enrichment of CTCF sites can be detected in the peaks in common and in the c-Mycspecific peaks. These data suggest that the c-Myc-dependent increase in chromatin bound cohesin primarily accumulates at CTCF sites, in agreement with the reported location of cohesins on DNA (23, 24) ( Figure 2d). Next, we wanted to investigate if the increase in cohesin chromatin occupancy contributes to c-Myc-induced RS. We tested this by reducing the levels of cohesins on DNA in cells experiencing oncogenic c-Myc activity by targeting the cohesin component Rad21. We used a non-efficient siRNA to ensure that the reduction of Rad21 would not cause cell cycle defects during the first S phase ( Figure S3c). Analysis of Smc1 binding to chromatin confirms that Rad21 knockdown reduces the levels of chromatin-bound cohesins both in untreated and in c-Myc cells (Figure 3a and b). Importantly, the levels of Smc1 on chromatin in Rad21-depleted c-Myc cells are similar to the untreated control, allowing us to test if the increased chromatin binding of cohesins is at the basis of generating RS. Reducing the levels of cohesin chromatin occupancy in cells experiencing oncogenic c-Myc increases DNA fibre length in   the first S-phase in both synchronous and asynchronous cell populations ( Figure 3c and 3d, Figure S3d and S3e). Rad21 depletion did not affect the extent of replication origin firing and the cell cycle profiles ( Figure S3f, S3g), nor the expression level of well-established c-Myc target genes ( Figure  S3h) supporting the idea that it is the increase of cohesins on DNA that contributes to RS. Next, we tested if the c-Mycdependent accumulation of cohesin specifically at CTCF sites was important for c-Myc-induced RS. Depleting CTCF (Figure S3i), completely rescued RS induced by c-Myc, without reducing the global amount of cohesins on DNA or affecting the cell cycle profiles (Figure 3e-h, S3j, S3k). We finally analysed the RS and DNA damage response activation in these cells, following Rad21 and CTCF depletion. We observed a decrease in both Chk1 phosphorylation and γH2AX levels compared to control silencing upon c-Myc activation, suggesting that in both cases RS-induced DNA damage was reduced ( Figure S3l, S3m).
Together these data indicate that the c-Myc-dependent increase in cohesins chromatin occupancy, likely at CTCF sites, causes a slowdown of replication forks to generate RS. To investigate how c-Myc could increase cohesin chromatin occupancy, we tested if its activation affects the expression levels of cohesin subunits and regulators. Smc1, Smc3 and Rad21 protein levels did not change significantly upon c-Myc activation ( Figure S4a). Interestingly, protein levels of the cohesin loader Mau2 increased upon c-Myc activation in both synchronised ( Figure 4a) and asynchronous cells (Figure 4b,S4b), which corresponds to an increase in mRNA levels ( Figure S4c, S4d). In vertebrates, cohesin loading requires NIPBL and Mau2 (33, 34). While NIPBL is the proper cohesin loader, Mau2 stabilises the protein levels of NIPBL (23, 35, 36), therefore we analysed NIPBL levels in c-Myc-activated cells. Both Mau2 and NIPBL protein levels increase upon c-Myc activation ( Figure S4b), while NIPBL mRNA does not change significantly ( Figure S4d), suggesting that c-Myc increases Mau2 expression that in turn stabilises NIPBL.
To verify whether c-Myc-induced increase in cohesin chromatin occupancy could be partially due to increased loading via upregulating Mau2, we reduced Mau2 levels in c-Myc activated cells to control levels using a non-efficient siRNA ( Figure 4c) and tested the presence of cohesins on chromatin and RS. Similar to Rad21 reduction, preventing the increase in Mau2 prevented excess cohesin loading onto chromatin (Figure 4d, e) and RS upon c-Myc activation in both synchronised ( Figure 4f, g and S4e) and asynchronous cells ( Figure  S4f and S3l, m). As for Rad21 knock down experiments, the extent of origin firing and cell cycle distribution is not affected by Mau2 depletion ( Figure S4g). These data indicate that c-Myc-dependent increase in Mau2 levels could be at the basis of the increased loading of cohesins on chromatin, which causes RS during S phase.
To investigate if the c-Myc-dependent increase in Mau2 levels could contribute to the generation of RS we transiently expressed Mau2-GFP, or GFP as control, in RPE1 cells and analysed RPA phosphorylation, marker of RS, and γH2AX, marker of DNA damage, by quantitative immunofluorescence and western blot ( Figure S4h-m). Whilst both markers increase in IF in transfected cells, an increase in phosphorylation of RPA is particularly pronounced in IF and in total lysates. These data suggest that overexpression of Mau2 alone can cause some RS and RS-induced DNA damage. We finally tested whether this mechanism of c-Myc-induced RS contributes to RS in cancer cells (Figure 4h-k). We selected a couple of lung cancer cell lines expressing different levels of c-Myc and measured the length of DNA fibres to evaluate the presence of RS. We observed reduced fibre length in the H1299 cell line compare to the A549 cells, which correlates with higher levels of c-Myc in H1299 cells (Figure 4h, 4i). To establish if a reduction in cohesins can rescue RS levels we depleted Rad21 ( Figure S4n) and measured DNA fibre length in both cell lines (Figure 4j, k and S4o). Rad21 depletion increased fibre length only in the H1299 cells, where c-Myc is highly expressed, but not in A549 cells, supporting a causative role for cohesins in RS, which is in the agreement with our findings. Interestingly depletion of Rad21 in the A549 cells, which do not experience RS, partially reduces fibre length. This suggests a protective role for cohesins in preventing RS, which is in agreement with reported work, supporting a double-edge involvement of cohesins in RS.
Our data show that a c-Myc-induced increase in cohesins on the DNA contributes to the induction of RS. This is different from previously reported mechanisms of oncogeneinduced RS, which are linked to deregulation of replication origin usage and/or transcription/replication interference. We show that the c-Myc-dependent accumulation of cohesins on chromatin, most likely at CTCF sites, can cause RS. Our data indicates that the levels of the cohesin loader Mau2 are upregulated by c-Myc activity, and that overexpression of Mau2 alone can cause RS, providing a potential mechanism through which c-Myc affects cohesin regulation. Overall, our data show that excessive cohesins on chromatin can interfere with the progression of replication forks, thus contributing to oncogene-induced RS. Our findings are surprising in light of previous work which indicates an important role for cohesins in preventing RS and DNA damage (37, 38). Based on our data we speculate that whilst the presence of cohesins during S phase is required to protect stalled forks and repair damaged DNA, excessive cohesin accumulation in an oncogenic context can interfere with the progression of the replisome. This is in agreement with recently published work in mammalian cells which shows that increased presence of cohesins on DNA slows down fork progression (32) and with work in yeast, which shows that DNA damage accumulates in SMCrich genomic regions during replication (39). With c-Myc activation being a crucial event in many human cancers, identifying the mechanisms through which this oncogene promotes RS provides critical insights into cancer biology and therapy.
EdU and EU incorporation. Cells on coverslips were incubated with EdU (final concentration 10 µM) for 30 min and fixed in 4% formaldehyde solution. Cells were permeabilised in 0.2% triton for 5 min and incubated with Click-it reaction cocktail (Click-it Alexa Fluor 647 C-10424 Invitrogen) for 30 min. Nuclei were stained with Hoechst (Invitrogen) solution 1:10000. The coverslips were mounted on slides with mounting medium Fluoroshied (Sigma). Images were obtained with Leica SPE2 40x objective lens and processed with Fiji. For detection of global RNA synthesis levels by 5-Ethynyl Uridine (EU) staining, 1 mM EU was added to cells for 1hr prior to collection. Cells were fixed in 4% formaldehyde. EU detection was performed using the Click-iT RNA Alexa Fluor 488 Imaging kit (Ther-moFisher, C10329) following manufacturer's instructions. Coverslips were rinsed for 2 min in Click-iT reaction rinse buffer and stained in Hoechst solution (1:10,000, Invitrogen H3570) for 5 min at room temperature. Fluoroshield (Sigma, F6182) was used for mounting on slides. Once dry, coverslips were sealed with nail varnish. Leica SPE2 using 40x objective lens and processed with Fiji.
Fibre analysis. Cells were labelled with 25µM CIdU for 15 min at 37 • C and then with 250µM CO2-equilibrated IdU (final concentration 250 µM) for 15 min at 37 • C. Fibre spreading and labelling was performed as in (40). The fibres were stained with primary antibodies (Rat anti-BrdU Abcam ab6326 1:250, Mouse anti-BrdU BD Biosciences 347580 1:100) overnight and with secondary antibodies (Alexafluor 555 goat anti-rat 1:500, Alexafluor 488 goat anti-mouse1:500) for 1.5 hr. Images were obtained with Leica SPE2 63x objective lens and processed with Fiji. 100-200 fibres were measured for each experiment. Composite images were constructed to visualise red and green channels simultaneously. The 'line' tool in Fiji was used to measure length of DNA replicating fibres, which are characterised by the presence of consecutive red and green signals. Total amount of fibres (ongoing fibres, replication origins, replication terminations, stalled forks) and replication origins (characterised by a red track between two green tracks) were counted to quantify the percentage of origin firing.
Chromatin preparation. RPE-1 c-Myc ER cells were seeded in 10 cm dishes and c-Myc expression was activated upon tamoxifen treatment. Cells were harvested after 16 hr of c-Myc activation and the chromatin was isolated with Chromatin Extraction Kit (ab117152, Abcam) according to the manufacture's protocol. The sonication was performed in a Diagenode Bioruptor® sonicator using the program 10 min: 30 s on, 30 s off.
Flow cytometry. For analysis of DNA content by propidium iodide (PI) staining, cells were collected by trypsinisation and fixed in 70% ethanol at -20 • C overnight. After centrifugation, the cell pellet was washed with PBS and resuspended in 100 mg/ml RNaseA and 50 mg/ml propidium iodide in PBS, and incubated at 4 • C overnight. Samples were measured on a BD LSRII flow cytometer using DIVA software (BD) and analysed using FlowJo software.
Survival assay. Cells were treated for 48 hr with 4OH-T or left untreated. The same volume of cell suspension was replated in 5 cm dishes and colonies were left to grow for one week. Cells were finally fixed and stained in 70% EtOH and 0.5% Methylene blue.
CHiPseq. Cells were cultured in 15 cm dishes for 48 hr with or without the addition of 4OH-T. Cells were then washed with 1x PBS and crosslinked in 10ml of 1% formaldehyde at RT for 10 min. Quenching was carried out adding 1 ml of 1.25 M glycine for 10 min at RT. Cells were then scraped in PBS, spin down, resuspended in cold buffer A (100 mM Hepes pH8, 100 mM EDTA, 5 mM EGTA, 2.5% triton) and rocked for 10 min at 4 • C. The same step was repeated using cold buffer B (100mM Hepes pH8, 2M NaCl, 100mM EDTA, 5 mM EGTA, 0.1% triton). Cells were resuspended in cold ChIP buffer (25 mM tris/HCl pH8, 2 mM EDTA, 150 mM NaCl, 1% triton, 0.1% SDS) plus protease inhibitor cocktail (Sigma P8340) and sonicated at maximum output on a Bioruptor for 30 s on/30 s off for 30 min using Diagenode tubes. Sonication was checked on 1% agarose gel. After sonication, lysates where centrifuged for 15 min at maximum speed at 4 • C. Protein A solution was prepared by resuspending beads in ChIP buffer (about 50%) 1 µg/µl BSA and rocking at 4 • C for 15 min. The supernatant (soluble chromatin) was transferred in new tubes and pre-cleared adding blocked protein A solution and rocking for 2 hr at 4 • C. Cleared soluble chromatin was centrifuged for 4 min at 4000rpm at 4 • C. The supernatant was transferred in a new tube and 10 µl was saved as input. The soluble chromatin was incubated overnight with 10 µg of anti Smc1 (Bethyl laboratories A300-055A rabbit). The following day 20 µl protein A beads prepared as above, were added to chromatin, which was then rocked at 4 • C for 2 hr. Beads were spin down for 2 min at 2000 rpm and washed with ChIP buffer, wash solution 1 (25 mM Tris/HCl pH8, 2 mM EDTA, 500 mM NaCl, 1% triton, 0.1% SDS), Wash solution 2 (250 mM LiCl, 1% NP40, 1% NaDOC, 1 mM EDTA, 10 mM Tris/Hcl pH 8) and twice with TE. TE was then removed and elution buffer (1% SDS, 100 mM NaHCO3) was added. All samples were incubated at 65 • C overnight to reverse crosslinking. The day after, samples were purified using QIAquick PCR purification kit (QUIAGEN). The DNA was then diluted in ddH2O. QC was performed using the ThermoFisher Qubit and either the Agillent BioAnalyser or TapeStation. The DNA samples were normalised and prepared into Illumina compatible libraries using the KAPA HyperPrep kit according to the manufacturer's instructions. The libraries were pooled to 4 nM and sequencing was performed on the HiSeq 4000 with at least 75 bp reads.
RTqPCR. RNA was extracted using RNeasy plus mini kit Qiagen. Before column purification cell pellets were vortexed for 30 s in RLT buffer + 1% β-mercaptoethanol. RT-qPCR was carried out using Mesa Blue mastermix (Eurogentec). All reactions were normalised to Gapdh as a control.