|Home | About | Journals | Submit | Contact Us | Français|
Severe levels of hypoxia lead to replication arrest which is independent of the S-phase checkpoint, the DNA damage response and transformation status. DNA fiber analysis demonstrates hypoxia-induced arrest occurs during both the initiation and elongation phases of replication and this correlates with a rapid decrease in available dNTPs. Due to fluctuating tumour oxygen tensions, arrested hypoxic cells can undergo rapid reperfusion and reoxygenation leading to reoxygenation-induced DNA damage. Reoxygenation-induced replication restart is inhibited in chronically exposed cells and is associated with repression of numerous replication factors, including E2F-dependent repression of MCM6 and the loss of phosphorylation and correct localisation of RPA which, in turn leads to a decrease in the hypoxia-induced DNA damage response. In contrast, reoxygenation-induced replication restart can occur after acute exposures to hypoxia and is accompanied by extensive reoxygenation-induced DNA damage and compromised DNA repair. Cells reoxygenated after acute hypoxia exposures undergo rapid p53-dependent apoptosis. These studies indicate that cells, lacking functional p53, which experience reoxygenation after acute, but not chronic, exposure to hypoxia contribute to increased genomic instability and potentially tumourigenesis.
Regions of hypoxia occur in all solid tumours as a result of the abnormally formed tumour vasculature and inadequate perfusion of the tumour mass. The occurrence of regions of low oxygen is an indicator of poor patient prognosis due to increased chemo- and radioresistance, genomic instability and metastatic potential (1, 2). The occurrence of episodes of reoxygenation following hypoxic exposures of various degrees is inherent to the dynamic nature of the tumour vasculature (3, 4). The range of oxygen tensions within a tumour varies between approximately 8% and 0.02% (near anoxia, 100–150 μm from blood vessels) (5). Previous reports demonstrated that severe hypoxia (pO2 <0.02%) induces an S-phase arrest due to inhibition of DNA synthesis in a HIF1-independent manner (6, 7). Hypoxia-induced replication stress was shown to induce activation of the DNA damage response (DDR) apical kinases ATM and ATR and phosphorylation of downstream targets such as p53, Chk1 and Chk2, albeit in the absence of detectable DNA damage (6, 8, 9). In contrast, reoxygenation was shown to rapidly induce DNA damage and, consequently, a more conventional Chk2-dependent DDR response leading to a G2/M arrest (10, 11). Replicative stress and the subsequent induction of the DDR occurs early in tumourigenesis and has been proposed to act as a barrier to tumour progression (12, 13). It has been suggested that hypoxia could contribute to this, as hypoxic regions can form in pre-neoplastic lesions (13, 14).
In this study we have characterized the hypoxia-induced S-phase arrest, showing that, at the molecular level, it is associated with an inhibition of DNA replication in both the initiation and elongation phases. Chronic hypoxia exposure actively induces disassembly of the replisome, preventing replication restart after reoxygenation. However, acutely exposed cells are still able to restart replication despite the presence of an active checkpoint response and reoxygenation-induced DNA damage. In a tumour, whereas regions of chronic hypoxia cells eventually die and undergo necrosis, replication restart in regions acutely subjected to low oxygen could potentially lead to an accumulation of unrepaired lesions and increased genomic instability (15). This is further aggravated by hypoxia-dependent decreased DNA repair (16). Our findings highlight the critical role of hypoxia in tumour progression and indicate that under chronic conditions, hypoxia can limit tumour progression whilst under acute conditions, cycles of hypoxia-reoxygenation increase genomic instability.
RKO, HCT116 and U2OS cells were grown in DMEM with 10% FCS; IBR3 and IBR3 hTert were grown in DMEM with 15% FCS. RKO -neo and -E7 lines were a gift from Dr. Kathleen Cho (University of Michigan). HCT116 spheroids were grown for 15 days by adding single cell suspensions obtained from exponential cultures to spinning flasks (4×106 total cells). Pimonidazole (1μM) was added 24 hours prior to fixation. All cell lines were mycoplasma free.
Hypoxia treatments were carried out in a Bactron II anaerobic chamber (Shell labs) or an In vivo 400 (Ruskinn). Cells were plated on glass dishes. Acute hypoxia was considered up to 12 hours hours whilst chronic exposure was 16–48 hours. Where no period of reoxygenation is indicated, cells were harvested inside the chamber with equilibrated solutions.
Cells were labeled with 10μM BrdU (Sigma) 1h prior to harvesting. Cells were processed for FACS analysis on a FacsSort flow cytometer (BD) after staining with mouse anti-BrdU antibody (BD Biosciences) and Alexa-fluor 488 conjugated goat anti-mouse IgG (Invitrogen). Quantification of cell populations using WinMDI 2.9 software (Scripps Research Institute).
p53 Silencer Select Validated siRNA (Applied Biosystems Ambion) or Stealth RNAi Negative control (Invitrogen) duplexes at a final concentration of 50nM, transfected into RKO cells using DhamaFECT (Thermo Scientific) according to manufacturers instructions. 5′–3′ oligo sequence: GUAAUCUACUGGGACGGAAtt
The DNA fiber technique was performed as described recently (17), with some specific modifications, shown (Fig. S2). Fiber spreads were examined using a Radiance confocal microscope (Biorad). Red (CldU) and green (IdU) tracks were measured and replication structures quantified using ImageJ software (National Institutes of Health). At least 100 replication tracks were measured and 200 replication structures counted/condition/experiment.
Cells were lysed in UTB (9M urea, 75mM Tris-HCl pH 7.5 and 0.15M β-mercaptoethanol) and sonicated briefly. Preparation of chromatin-bound protein extracts was performed as described (18). Primary antibodies were anti-MCM6, MCM7, Chk1, Rad51, p53 and actin (Santa Cruz), anti-RPA32 and MCM5 (Dr. C. Bauerschmidt), anti-PCNA (Calbiochem), anti-HIF1α (BD Biosciences), anti-pATM S1981 (Epitomics), anti-pChk1 S317 (Cell Signaling Technology) and anti-TopBP1 (Abcam). Secondary antibodies were Alexa-fluor 680 conjugated goat anti-mouse and anti-rabbit IgG (Invitrogen). Detection using the Odyssey infrared imaging technology (LI-COR Biosciences).
RPA and 53BP1 staining was as described (8). Single-stranded DNA and BrdU staining was carried out as previously described(19). Cells were visualized using a Nikon 90i microscope or a Radiance confocal microscope (Biorad).
Staining was performed as described, with some modifications (20); antigen retrieval was performed using EDTA pH 8.0 for Hydroxiprobe and EDTA pH 9.0 for CAIX and MCM6. Primary antibodies used were Hydroxiprobe (Chemicon), anti-CAIX (M75, Dr. Postor Ekova) and anti-MCM6 (Santa Cruz).
The extracts for quantification of nucleotide levels were carried in hypoxia with 3% TCA as described (21).
Quantitative real time PCR was performed as previously (22), using the Express SYBR GreenER qRT-PCR supermix kit (Invitrogen) following manufacturers recommendations. Reactions were carried out in a 7500 Fast real time PCR detection system (Applied Biosystems). Expression levels were normalized to 18s rRNA and VEGF was used as a positive control.
Cells were transfected with the reporter constructs (80 ng/well) using Lipofectamine transfection reagent (Invitrogen), alongside pCMV-Renilla (0.2ng/well) for normalization. Firefly and Renilla luciferase activities were measured using the Dual Glo Luciferase assay (Promega).
Statistical significance of differences between data sets was determined assessed by using Student’s t-test. Statistical significance was assumed if p<0.05 or lower and is noted in the figures. Error bars represent +/− SE.
Exposure to low oxygen levels results in a rapid cessation of DNA synthesis in the absence of DNA damage (7). Oxygen levels must be below 0.1% to induce an arrest, shown by BrdU incorporation, all subsequent experiments were carried out at 0–0.02% O2 unless otherwise stated (Fig. 1A and Fig. S1). The oxygen dependency of the arrest confirmed previous findings indicating that the S-phase arrest is independent of HIF-1 (7). The rapid cessation of DNA synthesis observed in response to severe levels of hypoxia, described here and previously, used techniques such as 3H-thymidine labelling or BrdU incorporation followed by FACS analysis (7, 10, 23). These methodologies allow the overall levels of DNA synthesis to be determined in a hypoxic cell but do not give any information about the stage of replication at which the block occurs. In order to fully investigate hypoxia-induced replication arrest we have made use of the DNA fiber technique, which allows visualisation of individual DNA strands and of different replication structures, as well as measurement of the speed of ongoing replication forks (Fig. S2) (24). RKO cells were exposed to hypoxia for 4, 6 or 8 hours and the replication structures observed, quantified and characterised as stalled forks, ongoing forks or new origins (Fig. 1B). It should be noted that a stalled fork could also be a first label termination as these two structures are indistinguishable using this technique. After 4 hours of hypoxia the number of stalled/terminated forks increased significantly and continued to accumulate over the following 4 hours. Concomitant with the increase in stalled forks we saw a decrease in ongoing forks. Importantly, we determined that origin firing was significantly repressed in hypoxic conditions, being non-existent at the 8 hour time point.
We examined the rate of replication at ongoing fork speeds, measuring both average and median rates (Fig. 1C). After 4 hours of hypoxia the number of ongoing forks had fallen by approximately 75% and the average and median rates of the remaining were decreased significantly when compared with normoxic samples, with a shift of the distribution severely towards the left. The fork speeds were reduced further at 6 and 8 hours of hypoxic exposure, being over 15 fold reduced at the later time point when compared to normoxic forks (0.05kb/min versus 0.83kb/min median rates) (Fig. 1C, data not shown). Many of the hypoxic-ongoing forks have a barely detectable level of second label leading us to consider that they are not truly representative of ongoing forks and that replication may be completely abrogated in these conditions. These data demonstrate that DNA replication is blocked at both the initiation as well as elongation phase in response to hypoxia.
The ability of ribonucleotide reductase (RNR) to synthesize the four dNTPs required for DNA replication is dependent on a tyrosyl free radical within the enzymes active site. This free radical is generated in an oxygen-dependent manner and is therefore compromised under hypoxic conditions (7, 23, 25). The ability of RNR to synthesize the four dNTPs required for DNA replication is dependent on a tyrosyl free radical within the enzymes active site. This free radical is generated in an oxygen-dependent manner and is therefore compromised under hypoxic conditions (25). Nucleotide levels were measured by HPLC with spectrophotometric detection in HCT116 cells exposed to hypoxia or as a control, hydroxyurea (HU) (Fig. 1D). Levels of dNTPS decreased significantly (within 1 hour) in response to hypoxia suggesting that a loss of RNR activity under hypoxic conditions could be the underlying cause or contribute to the hypoxia-induced arrest. Significantly, dNTP levels did not decrease during exposure to milder hypoxia (2% O2) which does not induce replication arrest (data not shown). The absolute levels of RNR protein were not found to change throughout the experiment (data not shown). We also noted that energy charge was not altered during the first 24 hours of hypoxic treatment, indicating decreased energy levels did not contribute to the arrest (Fig. S3). This suggests a model whereby stalled forks arise during exposure to hypoxia as a result of reduced levels of dNTPs and trigger signalling pathways which inhibit origin firing which, is also re-enforced by a lack of available nucleotides.
The fate of hypoxia-arrested cells after reoxygenation is significant as reoxygenation induces significant levels of damage and major DNA repair pathways including homologous recombination are inhibited during hypoxia/reoxygenation (9, 16, 26). To investigate replication restart in this scenario we pulse labelled RKO cells with BrdU to follow S-phase cells through hypoxia and reoxygenation, cells were scored as being unlabelled or labelled in the G1, S or G2 phases, (Fig. 2A). Labelled S-phase cells exposed to hypoxia (16 hours) failed to progress into G2 after reoxygenation, suggesting they were replication incompetent. Unlabelled G1 cells were able to progress through to S-phase after reoxygenation, suggesting that the hypoxia-induced G1 arrest is not permanent and that the cells were viable (27). This data indicates that chronic (16 hours) hypoxia exposure compromises the ability of the S-phase arrested cells to restart DNA replication.
Previous micro-array analysis in mouse embryonic fibroblasts suggested that a number of genes involved in DNA replication or repair were repressed in hypoxic conditions, including Asf1b, MCM3, MCM4, MCM5, MCM6, MCM7, Rad51 AP1 and Fen1 (22). In addition, polysomes were fractionated after 24 or 48 hours of hypoxia followed by qRT-PCR to identify gene products actively repressed in hypoxic conditions (data not shown). Of the previously identified genes, MCM3, MCM4, MCM5 and Asf1b, were found to be repressed. We have verified by qRT-PCR the repression of components of the replication machinery under hypoxic conditions in a human cell line (Fig. 2B). Before proceeding further we determined if hypoxia-down-regulation of key replication factors occurred at both the protein level and in vivo. The MCM complex (MCM2–7) is loaded onto the chromatin in the G1 cell cycle phase to form the pre-replicative complexes (pre-RC) and their helicase function is activated upon initiation and required for duplex unwinding during ongoing replication (28, 29). Firstly, the HCT116 cell line was grown as spheroids before staining for MCM6 and the hypoxia marker, pimonidazole (Fig. 2C i-v). The outer layers of the spheroids show a strong nuclear stain of MCM6, whilst the cells surrounding the necrotic core show no or greatly reduced levels of MCM6. As expected the pattern of pimonidazole staining is inverse to that seen for MCM6 i.e. more positive towards the necrotic core indicating increasing levels of hypoxia. U87 cells were grown as xenograft tumours and treated with the anti-angiogenic agent bevacizumab (30). The tumours were sectioned and stained for CAIX and MCM6 (Fig. 2C vi-ix). As expected viable cells within CAIX positive regions showed decreased levels of MCM6. In both models we noticed that MCM6 levels fall before cells become obviously necrotic supporting our hypothesis that this is an active mechanism induced in response to hypoxia and not a result of cell death. It should be noted than in both systems additional micro-environmental stresses not present in in vitro experiments may also play a role in the repression of MCM6, for example nutrient deprivation (glucose and amino acids) and acidosis.
The E2F transcription factors have been shown to be important for the cell-growth regulated expression of the MCM proteins as well as for the repression of DNA repair during hypoxia (31, 32). We investigated whether the hypoxia-induced repression of MCM proteins is E2F dependent firstly using the over-expression of HPV E7. HPV E7 disrupts the E2F/pocket protein interaction and also targets pocket proteins for degradation thus interfering with E2F activity (31). The expression levels of MCM3-7 were compared in two RKO cell lines over expressing E7 with a matched control, (Fig. 3A). In each case the hypoxia-induced repression was significantly alleviated by the presence of E7. To validate the involvement of the E2Fs we made use of a MCM6 reporter construct with all the identified E2F binding sites mutated (32). In response to hypoxia we observed a 10–20 fold repression of the MCM6 promoter in contrast to a 5xHRE luciferase construct which was induced 40 fold, (Fig. 3B). Loss of E2F binding significantly altered the repression of MCM6 in response to hypoxia.
Our next step was to examine what effect the repression of the MCM mRNAs had on the protein levels of MCM5, MCM6 and MCM7 (Fig. 3C). In each case the protein levels decreased and most significantly so after more chronic hypoxia exposures and it should be noted that this was not due to a loss of S-phase cells (Fig. 2A). Since the protein levels for the MCMs were not completely abrogated, potentially due to the long half-life of these proteins (24h) (33), we investigated whether the remaining MCM proteins were functional, by determining their cellular location. Extraction of chromatin-bound proteins after chronic exposure to hypoxia demonstrated no association of MCM5, MCM6 or MCM7 with the chromatin, indicating a complete lack of replisome function (Fig. 3C). The GINS proteins have been demonstrated to interact with the MCMs and are essential for DNA replication (34). We investigated one of the GINS, Psf2 and found that total levels of Psf2 decreased rapidly in hypoxia. We have also shown that, as predicted by the polysome assay, levels of polymerase δ are repressed in hypoxic conditions and show decreased chromatin association. This was not a general effect as PCNA remained chromatin bound during both acute and chronic hypoxia. Taken together these data indicate that replication does not resume after chronic periods of hypoxia due to an active disassembly of the replisome including both the helicases and polymerases. The mechanism behind this appears multi-factorial. This data is supportive of a transcriptional model in which activating E2Fs are replaced on the MCM6 promoter with repressive E2Fs, possibly E2F4 as this is the most sensitive to E7 expression during hypoxia exposure (35). In addition, the MCM complex is not retained at the replication fork potentially due to a loss of interacting factors such as Psf2. The MCM complex has been shown to be phosphorylated by ATM and ATR in response to replication stress (MCM2 and MCM3) whilst, MCM7 has been shown to interact with ATRIP (36). Both the ATM and ATR kinases have been shown to be active during hypoxia and it is therefore plausible that they contribute to MCM/replisome stability.
The accumulation of stalled replication forks or replicative stress has been shown to activate the DDR. Acute hypoxia exposure induced both ATM-S1981 and Chk1-S317 indicating the activity of both the ATM and ATR kinases. However, these signals were lost or decreased with increasing hypoxia exposure time (Fig. 4A). An integral step in the DDR to replication stress is the recognition of regions of ssDNA at stalled forks and binding of RPA. The levels of RPA did not change during hypoxia although there was a robust ATM-dependent phosphorylation during acute exposure that was absent after longer times (Fig. 4B and S4A, S4B). RPA was found to be chromatin associated during acute but not chronic hypoxia. In support of this we found that RPA formed clear nuclear foci in response to hypoxia but that these decreased as exposure time increased (Fig. 4C). We then verified that the hypoxia-induced RPA foci were indeed regions of ssDNA by co-staining for BrdU in labelled but non-denatured cells. The RPA foci and regions of ssDNA completely co-localised (Fig 4D). This prompted us to count the number of cells positive for ssDNA by BrdU staining. We found that this signal also decreases with increasing exposure time indicating that the signal which initiates the DDR in response to hypoxia decreases after extended periods, potentially due to slow/residual polymerase activity (37). These data are supportive of the hypoxia-induced DDR stabilising the replisome during acute exposure, potentially through phosphorylation of the MCM complex although we have not been able to verify this.
In response to chronic exposure to hypoxia essential components of both the replication machinery and the DDR are repressed which together lead to the destabilization of the replication fork and prevent subsequent re-start. However, closer examination of our data indicates that this may not be the case during shorter/acute exposures (less than 12 hours). This raises the possibility that cells that have undergone hypoxia-induced replication arrest for shorter time periods might be capable of replication restart. To address this we repeated the BrdU chase experiment and found that labelled S-phase cells did indeed progress through the cell cycle to G2 (Fig. 5A). To investigate this further we have again made use of the DNA fiber technique. After acute exposure to hypoxia (6 hours), followed by reoxygenation, there was a significant increase in the number of ongoing forks, an increase in the number of new origins and, as expected, a decrease in the number of stalled forks (Fig. 5B). Using the mitotic shake-off technique, we have ruled out any contribution of hypoxic G1 cells entering S-phase during reoxygenation in the fiber analysis (Fig. S5A). Our data suggests that hypoxia-induced replication arrest is dependent on a decrease in dNTPs (Fig. 1D), here we investigated the dNTP levels after reoxygenation. As previously, we saw a rapid and significant drop in all four nucleotides in response to hypoxia however, these levels returned to normal within 2 hours of reoxygenation (Fig. 5C). The same trend was also seen in HCT116 cells (data not shown). Interestingly, we found that this correlated with reoxygenation-restarted replicating fork rates which were slower than normoxic conditions but return to normal within 2 hours after reoxygenation (Fig. 5C). Despite reoxygenation-induced damage and activation of the ATM/ATR signalling pathways, DNA replication resumes during reoxygenation after acute hypoxia.
In addition to its role in regulating initiation and elongation in unperturbed cells, Chk1 has been shown to regulate these processes following replication stress such as UV, camptothecin and HU (18, 38-40). However, the role of Chk1 in hypoxia-mediated arrest and reoxygenation-induced restart is unknown. Pharmacologic inhibition of Chk1 by treatment with Gö6976 did not significantly change the number of ongoing forks, suggesting that Chk1 did not affect the hypoxia-induced arrest (Fig. 6A). Chk1 inhibition led to an overall decrease in average and median speed even in the absence of hypoxic stress, which was not further exacerbated by hypoxia exposure (data not shown). During replication re-start after acute hypoxia however, loss of Chk 1 led to a significant increase in the number of new origins firing. This is supported by previous data indicating a role for Chk1 in mediating the S-phase checkpoint (41, 42). Again, this was verified using the mitotic shake-off technique (Fig. S5A).
We have considered that replication-restart might be restricted to tumour cell lines or possibly cells deficient in mismatch repair as used here and have therefore investigated this after reoxygenation in a non-transformed cell line (43). Primary 1BR3 and immortalised 1BR3 hTert cell lines were exposed to hypoxia and reoxygenation (Fig. 6B). In both cases, cells arrested in response to hypoxia, as determined by a significant increase in the number of stalled/terminated forks. In both cases Chk1 was also phosphorylated during hypoxia exposure (data not shown). Reoxygenation-induced replication-restart was measured in both cell lines and seemed slightly more robust in the 1BR3 line compared to the 1BR3 hTert. These data demonstrate that replication-restart is not restricted to transformed cell lines after reoxygenation but is a more widespread phenomenon.
These findings led us to hypothesise that in order to prevent cells which have undergone replication-restart in these conditions contributing to an increase in genomic instability they might undergo apoptosis. RKO cells were treated with siRNA to p53 or a non-targeting control (ntc) followed by 6 hours of hypoxia and reoxygenation, the levels of apoptosis were then determined (Fig. 6C). The p53 status did not affect the level of hypoxia-induced apoptosis which was approximately 5%. However, after reoxygenation there was a significant increase in apoptosis which was p53-dependent. These data indicate that p53-proficient cells which undergo replication-restart should be eradicated from the tumour population by apoptosis whilst those which have p53-pathway mutations escape and contribute to genomic instability.
We have conclusively demonstrated that replication is abrogated in hypoxic conditions during both the initiation and elongation phases and that this correlates with falling dNTP levels potentially resulting from decreased ribonucleotide reductase activity. Until now it has been unclear whether cells exposed to hypoxia-induced replication arrest are capable of completing S-phase after a reoxygenation event. Our data demonstrate that this is critically dependent on the exposure time to hypoxia.
Even when still viable, chronically-arrested cells are incapable of replication re-start in response to reoxygenation and therefore do not contribute to continued tumourigenesis. We have attributed this to a complete loss of replisome stability as well as hypoxia-mediated down-regulation of the homologous recombination pathway. In contrast, our data demonstrates that after shorter periods of hypoxia-induced replication arrest cells are able to resume replication in response to reoxygenation. Importantly, this occurs in the presence of ROS-induced DNA damage and when essential DNA repair pathways are somewhat inhibited including, non-homologous end joining, mismatch repair and homologous recombination (16, 26). A number of genes, essential to these repair pathways, have been identified which are specifically repressed during hypoxia including; Rad51, Rad52, BRCA1, BRCA2, Ku70, DNA-PKcs, LigIV, MLH1, MSH2 and MSH6 (44-48). Together this indicates that cells that resume DNA synthesis after a reoxygenating event do so in the presence of DNA damage and with impaired DNA repair capabilities. Our data demonstrate that if these re-started cells retain p53 activity they undergo apoptosis. However, since the majority of tumour cells harbour p53 or p53-pathway mutations it is more likely that cycling through periods of acute hypoxia followed by reoxygenation will lead to an accumulation of unrepaired lesions and increased genomic instability (15). Encouragingly, our findings indicate that therapeutic options, such as bevacizumab, may enhance the potential effects of DNA repair/replication inhibitors and therefore suggest a combined approach may be an effective way to target hypoxic regions. In addition, these findings suggest that combining repair/replication inhibitors with therapies with vascular normalisation properties may effectively target cells undergoing reoxygenation-induced replication re-start.
Daria Bochenek and Ioanna Ledaki for technical assistance with qRT-PCR and spheroid growth. Dr Eva Petermann for extensive help with DNA fiber preparation/analysis, Kiyoshi Ohtani for MCM reporter constructs, Penny Jeggo for the 1BR3 cell lines. Amato Giaccia, Denise Chan, Claus Sørensen and Thomas Helleday for critical reading of the manuscript. Supported by the NIHR Biomedical Research Centre, Oxford. EMH, ZB and IMP are funded by the CRUK, grant reference C6515/A9321 awarded to EMH.