Hypoxic Stress Facilitates Acute Activation and Chronic Downregulation of Fanconi Anemia Proteins Free

This article is featured in Highlights of This Issue, p. 965

Hypoxia is a distinctive feature of solid tumors that contributes to cancer progression, aggressive phenotype, metastasis, treatment resistance, and poor patient prognosis (1). Of the myriad of cellular changes generated by hypoxia, an important contributor to these phenomena is hypoxia-induced genomic instability (2). Many studies have demonstrated that the tumor microenvironment, and hypoxia in particular, can promote the development of diverse genetic lesions, including point mutations, deletions, DNA overreplication, fragile site activation, and chromosomal rearrangements (3–6). Investigation of the genomic instability induced by hypoxia has revealed that hypoxia itself, in the absence of reoxygenation, does not induce direct DNA damage, but rather regulates the activity of multiple DNA repair processes, with several pathways being acutely activated and then chronically repressed (2).

The DNA damage checkpoint kinases, ataxia telangiectasia mutated (ATM) and ATM- and Rad3-related (ATR), are both activated upon treatment with hypoxia. Hypoxia induces ATM autophosphorylation, which is required for the downstream phosphorylation of CHK2 (CHEK2), 53BP1 (TP53BP1), Kap1 (TRIM28), and DNA-PKcs (PRKDC; refs. 7 and 8). ATR forms nuclear foci in hypoxia and is required for phosphorylation of CHK1 (CHEK1), H2AX (H2AFX), and p53 (TP53; refs. 9 and 10). Hypoxia also induces phosphorylation of BRCA1 in a CHK2-dependent manner (11). The activation of these signaling pathways, particularly the ATR–CHK1 pathway, has been proposed to be a consequence of hypoxia-induced replication arrest, which occurs via a decrease in the ribonucleotide pool and results in accumulation of single-stranded DNA (12, 13). Loss of ATR/CHK1 activity does not prevent replication arrest, but does lead to DNA damage detectable by the comet assay and reduced survival during reoxygenation because of apoptosis, suggesting that during hypoxia ATR may protect or stabilize stalled replication forks (12).

Although hypoxia rapidly activates DNA damage signaling pathways, exposure to hypoxia over more extended periods of time results in repression of multiple DNA repair pathways, including nucleotide excision repair, mismatch repair, and DNA double-strand break (DSB) repair (14–17). DSB repair seems to be repressed under hypoxia via the coregulation of 2 important DSB repair proteins, RAD51 and BRCA1 (16, 18). Independently of cell-cycle phase and HIF-1α (HIF1A), RAD51 and BRCA1 expression is reduced after 24 and 48 hours of hypoxia at both the protein and mRNA levels. The transcriptional downregulation of RAD51 and BRCA1 occurs via a shift in transcription factor binding at consensus E2F sites in their proximal promoter regions from the activating E2F1 factor to the repressive E2F4/p130 (RBL2) factor (18, 19). The significance of the downregulation of DNA repair in hypoxic cells is exemplified by their increased sensitivity to mitomycin C, cisplatin, and PARP inhibitors (17, 20). Thus, chronic hypoxia exposure, by reducing DNA repair capacity, promotes genomic instability but also sensitizes cells to DNA-damaging chemotherapeutics.

Fanconi anemia (FA) is a rare genetic disease characterized by congenital abnormalities, bone marrow failure, and predisposition to leukemia and solid cancers (21). At the cellular level, FA is a chromosomal instability disorder marked by hypersensitivity and the formation of chromosomal aberrations upon treatment with DNA interstrand crosslink (ICL)–inducing agents, including mitomycin C, cisplatin, and diepoxybutane (21). Mutations in at least 15 different genes can confer the FA phenotype, and their protein products cooperate in a common pathway required for repairing DNA ICLs (21). Eight FA proteins, FANCA/B/C/E/F/G/L/M, form a core complex that functions as an E3 ubiquitin ligase required for the key activating step of the FA pathway: monoubiquitination of FANCD2 and FANCI (22, 23). FANCM and its associated protein FAAP24 recruit the core complex to replication forks stalled at ICLs, whereas FANCL functions as the catalytic subunit of the E3 ligase in conjunction with the E2 ligase UBE2T (24–26). Upon monoubiquitination, FANCD2 and FANCI form nuclear foci where they colocalize with additional DNA repair proteins and coordinate the removal of ICLs via structure-specific nucleases, translesion synthesis polymerases, and homologous recombination machinery (21).

The FA pathway is also activated during unperturbed S-phase and strongly induced in response to replication stress. Treatment of FA-deficient cells with replication inhibitors results in chromosomal aberrations and fragile site breakage (27). Mechanistically, ubiquitinated FANCD2, along with BRCA1, BRCA2, and RAD51, is required to stabilize stalled replication forks and prevent their degradation (28, 29). In addition, FANCD2 and FANCI foci are visible on metaphase spreads at the extremities of anaphase bridges, possibly depicting sites where replication has failed to complete before mitosis (30). The FA pathway thus seems to play a critical role in protecting cells from the adverse consequences of replication stress.

Based on the impact of hypoxia on DNA repair and the role of the FA pathway in responding to replication stress (a condition induced by hypoxia), we hypothesized that the FA pathway may play a key role in hypoxia. In this study, we have investigated the function of the FA pathway in the cellular response to hypoxia, mimicking one aspect of the tumor microenvironment. We have characterized the initial activation of the FA pathway in hypoxia, demonstrating that FANCD2 and FANCI are ubiquitinated upon acute hypoxic stress in an ATR-dependent but HIF-independent manner. With longer hypoxic exposure, we have found that FANCD2 and FANCI are downregulated at the protein and mRNA levels via a pathway analogous to the downregulation of RAD51 and BRCA1. Finally, we have established the functional significance of an intact FA pathway in avoiding hypoxia-induced genetic instability by demonstrating that FANCD2 protects hypoxic cells from accumulation of γH2AX, DNA damage, and apoptosis.

HeLa, A549, and MCF7 cells were obtained from ATCC. RKO-Neo and RKO-E7 cells were provided by Dr. K. Cho (University of Michigan). PD20+EV, PD20+FD2, and PD20+KR cells were provided by Dr. G. Kupfer (Yale School of Medicine). Growth conditions are described in the Supplementary Methods.

Deferoxamine, hydroxyurea, and mitomycin C (Sigma) were dissolved in H₂O and used at final concentrations of 250 μmol/L, 2 mmol/L, and 1 μmol/L, respectively. N-Ethylmaleimide (Sigma) was dissolved in EtOH and added to lysis buffer at a final concentration of 4 mmol/L. VE-821 (Vertex Pharmaceuticals), KU-55933 (Santa Cruz), and NU-7441 (Tocris Bioscience) were dissolved in DMSO and used at final concentrations of 1 μmol/L, 10 μmol/L, and 1 μmol/L, respectively. Cisplatin (Sigma) was dissolved in dimethylformamide and used at a final concentration of 20 μg/mL.

Hypoxic conditions were established as previously described (16). Details are provided in the Supplementary Methods.

Cells were lysed in AZ lysis buffer (50 mmol/L Tris, 250 mmol/L NaCl, 1% Igepal, 0.1% SDS, 5 mmol/L EDTA, 10 mmol/L Na₂P₂O₇, 10 mmol/L NaF) supplemented with Protease Inhibitor Cocktail (Roche) and 4 mmol/L N-Ethylmaleimide (Sigma). Phosphatase Inhibitor Cocktail (Roche) was added in phospho-protein analysis experiments. For separation of monoubiquitinated and nonubiquitinated FANCD2 and FANCI, 5% SDS-PAGE was used. Where shown, band intensities were quantified using ImageJ64 software. Antibodies are described in the Supplementary Methods.

These assays are described in the Supplementary Methods.

Assays were performed as previously described (18). Briefly, total RNA was isolated using an Absolutely RNA Miniprep Kit (Agilent Technologies) and used to synthesize cDNA using a High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). The resulting cDNA was used in PCRs containing Taqman Universal PCR Master Mix (Applied Biosystems), Taqman Gene Expression Assay Mix containing premixed primers and probes for FANCD2, FANCI, BRCA1, and 18S (Applied Biosystems), and Rox Reference Dye (Invitrogen). An Mx3000P RT-PCR system (Strategene) was used to measure fluorescence intensity in real-time and to calculate cycle thresholds.

Cells were plated, allowed to adhere overnight, and then placed under severe hypoxia or normoxia for 48 hours. Immediately upon removal from hypoxia, cells were trypsinized, washed with PBS, and resuspended in LM Agarose (Trevigen). Neutral single-cell gel electrophoresis was conducted using the CometAssay Electrophoresis System (Trevigen) at 21 V for 1 hour. Data were collected with an EVOS FL microscope (Advance Microscopy Group) and analyzed with CometScore software (TriTek Corporation).

Caspase activity was measured with the Caspase-Glo 3/7 Assay (Promega) according to the kit protocol. Details are provided in the Supplementary Methods.

The proximal promoter regions (1,000 bp upstream and 100 bp downstream of the transcription start site) of FANCD2 and FANCI were analyzed for E2F1 binding sites using the JASPAR CORE database (31). Vertebrate basewise conservation by PhyloP and multispecies alignment of the FANCD2 predicted E2F site were obtained from the UCSC Genome Browser Human Feb. 2009 (GRCh37/hg19) assembly (32).

To study the role of the FA pathway in hypoxia, we began by examining the ubiquitination status of FANCD2 and FANCI. We exposed HeLa cells to severe hypoxia (<0.01% O₂) or normoxia for 24 or 48 hours and visualized monoubiquitinated FANCD2 and FANCI versus the nonubiquitinated forms by 5% SDS-PAGE. Cells treated with mitomycin (MMC) or hydroxyurea (HU), known inducers of FANCD2 ubiquitination, served as positive controls (33). We found that after 24 hours of hypoxic treatment, the fractions of ubiquitinated FANCD2 and FANCI increased, although less dramatically than with MMC or HU treatment (Fig. 1A). After 48 hours of hypoxia, the proportion of ubiquitinated forms remained elevated, but total levels of the proteins decreased (Fig. 1A).

We next tested the impact of the hypoxia-mimetic deferoxamine (DFX) on FANCD2 and FANCI ubiquitination. DFX chelates iron, thereby inactivating proline hydroxylases and stabilizing HIF-1α (34). Similarly to hypoxia, in HeLa cells treated with 250 μmol/L DFX, the fractions of ubiquitinated FANCD2 and FANCI progressively increased after 6 and 24 hours of treatment, whereas total levels of both proteins decreased substantially after 48 hours (Fig. 1B). We also observed initial FANCD2 ubiquitination followed by decreased total FANCD2 expression upon treatment with DFX in A549, MCF7, and RKO cells and upon treatment with severe hypoxia in RKO cells (see below and Supplementary Figs. S1 and S2). More moderate levels of hypoxia (0.1–1% O₂) were not sufficient to induce FANCD2 ubiquitination (Supplementary Fig. S3).

To confirm that the shift of FANCD2 seen upon treatment with hypoxia or DFX was because of monoubiquitination at lysine 561 (the site of ubiquitination in response to MMC), we utilized patient-derived fibroblasts lacking FANCD2 and complemented with an empty vector (PD20+EV), wild-type FANCD2 (PD20+FD2), or a nonubiquitinatable FANCD2 mutant with substitution of lysine 561 with arginine (PD20+KR). We exposed these cells to hypoxia for 24 to 48 hours or to DFX for 6 to 48 hours and performed FANCD2 Western blots. As expected, FANCD2 was ubiquitinated upon treatment with hypoxia or DFX only in PD20+FD2 cells (Fig. 1C and D). With longer treatment, FANCD2 total protein levels were decreased in both PD20+FD2 and PD20+KR cells indicating that the downregulation of FANCD2 is not ubiquitination dependent.

FANCD2 and FANCI form a heterodimer, and their ubiquitinations are codependent (23). We therefore examined whether FANCI ubiquitination is also induced by hypoxia in the PD20 set of cells. We found that FANCI was ubiquitinated upon hypoxia treatment in cells proficient for FANCD2 but not in cells lacking FANCD2 or expressing the ubiquitination-mutant FANCD2 (Fig. 1E).

Activation of the FA pathway in response to DNA damage and replication stress is regulated by the ATR signaling pathway. ATR and CHK1 phosphorylate FANCD2 and are required for FANCD2 monoubiquitination and foci formation (21, 33). Phosphorylation of FANCI at 6 conserved ATR sites additionally plays a pivotal role in FANCD2 monoubiquitination and foci formation (35). In contrast, ATM phosphorylates FANCD2 and is required for activation of the IR-induced S-phase checkpoint but is dispensable for FANCD2 monoubiquitination (36).

To determine whether hypoxia-induced FANCD2 ubiquitination also depends on ATR, we used the small-molecule ATR inhibitor, VE-821 (ATRi), as well as ATM and DNA-PK inhibitors, KU-55933 (ATMi) and NU-7441 (DNA-PKi), respectively. HeLa cells were treated with the kinase inhibitors, exposed to hypoxia or DFX, and analyzed by Western blotting. In this experiment, strong FANCD2 ubiquitination occurred after 48 hours hypoxia, and we observed that only ATRi blocked this ubiquitination (Fig. 2A, lane 18). Quantification of the ratio of monoubiquitinated to nonubiquitinated FANCD2 (the L:S ratio) demonstrated that the ratio increased from 0.27 in normoxia to 1.47 upon treatment with 48 hours hypoxia, was reduced to 0.68 by ATRi, and was not significantly altered by ATMi or DNA-PKi. Densitometric analysis of the FANCD2 L:S ratio in DFX-treated cells demonstrated inhibition of FANCD2 ubiquitination by ATRi at the 24- and 48-hour time points (Fig. 2B, lanes 18 and 28). As expected, MMC- and HU-induced FANCD2 ubiquitination was also blocked specifically by ATRi (Fig. 2A, lanes 23 and 28). In addition, FANCI Western blots revealed that only ATRi reduced hypoxia-induced FANCI ubiquitination, which is most evident at the 24-hour time point (Fig. 2C, lane 6). Interestingly, combined treatment with ATRi and ATMi did not block hypoxia- or DFX-induced FANCD2 ubiquitination and was less efficient at blocking MMC- and HU-induced FANCD2 ubiquitination (Fig. 2A, lanes 20, 25, and 30; Fig. 2B, lanes 20 and 30). This result is consistent with prior findings that ATM deficiency increases FANCD2 ubiquitination in response to IR, HU, and MMC (33) and further suggests that ATM inhibition can potentiate FANCD2 ubiquitination through a pathway independent of ATR.

To verify that the kinase inhibitors were functional in our assay, we performed phospho-protein Western blots to examine their activity in blocking ATM, CHK1, and CHK2 phosphorylation. We found that ATMi blocked the autophosphorylation of ATM at serine 1981, when used both independently and in combination with ATRi (Fig. 2A, lanes 17, 20, 22, 25, and 27; Fig. 2B, lanes 17, 20, 27, and 30). ATRi blocked CHK1 phosphorylation at serine 345, a known downstream target of ATR (Fig. 2A and 2B, lanes 18, 20, 28, and 30), although we noted that CHK1 was phosphorylated after 6 hours DFX treatment even in the presence of ATRi (Fig. 2B, lane 8), suggesting that ATRi was not effective at this early time point and likely explaining the persistence of FANCD2 ubiquitination at this time point. Finally, ATRi induced CHK2 phosphorylation at threonine 68 in normoxic cells (Fig. 2A, lanes 3 and 13; Fig. 2B, lanes 13 and 23), as has been observed in another study (37). CHK2 phosphorylation in treated cells is likely complicated by the downregulation of total CHK2 under hypoxia (8).

To further validate our findings, we utilized genetic inhibition of ATR and ATM by siRNA depletion. These experiments revealed that ATR depletion blocked FANCD2 and FANCI ubiquitination induced by hypoxia and DFX, whereas ATM depletion had no effect (Supplementary Fig. S4), serving to further strengthen our conclusion that hypoxia-induced FANCD2 and FANCI ubiquitination is ATR dependent.

Hypoxia-inducible factors HIF-1 and HIF-2 (EPAS1) are dimeric transcription factors that are stabilized by hypoxia and mediate many of the downstream cellular effects of hypoxia (34). To determine whether hypoxia-induced FANCD2 ubiquitination depends upon HIF signaling, we examined its ubiquitination upon depletion of the HIF-1α or HIF-2α subunits via lentiviral shRNA knockdown. We observed a similar extent of FANCD2 ubiquitination in HIF-proficient and HIF-deficient cells upon treatment with DFX or hypoxia (Supplementary Fig. S5A and S5B). We also analyzed FANCD2 and FANCI ubiquitination in the VHL-mutant renal cell carcinoma 786–0 cell line, which overexpresses HIF-2α and fails to express HIF-1α. At baseline, there is no increased ubiquitination in the mutant cells compared with the VHL-complemented line (in which HIF-2α is suppressed), which indicates that HIF-2α overexpression is not sufficient for inducing FANCD2 or FANCI ubiquitination. After hypoxic exposure, ubiquitination occurs in both mutant and corrected cells, indicating that HIF-1α is not required for the ubiquitination of FANCD2 or FANCI (Supplementary Fig. S5C). Altogether, theses results provide strong evidence that hypoxia-induced ubiquitination of FANCD2 and FANCI is HIF independent.

In response to DNA damage and replication stress, FANCD2 forms ubiquitination-dependent nuclear foci that colocalize with many additional DNA repair proteins, including FANCI, BRCA1, BRCA2, RAD51, and γH2AX (22, 23, 38, 39). To investigate whether hypoxia induces FANCD2 foci formation, we performed immunofluorescence microscopy on PD20+FD2 cells treated with DFX or hypoxia. We found that FANCD2 did indeed form nuclear foci after either treatment and that the foci colocalized with γH2AX (Fig. 3A). Untreated cells had a baseline percentage of foci-positive cells near 10%, which increased to approximately 40%, 50%, and 60% in hypoxia, DFX, and MMC-treated cells, respectively (Fig. 3B). As expected, immunofluorescence microscopy on PD20+EV cells demonstrated the absence of FANCD2 staining, whereas immunofluorescence microscopy on PD20+KR cells treated with hypoxia revealed nuclear FANCD2 without any foci (Fig. 3C).

After the initial ubiquitination of FANCD2 induced by hypoxia and DFX, we noticed that longer treatments with hypoxia or DFX resulted in decreased FANCD2 protein levels. By treating HeLa cells with DFX for 6 to 72 hours and performing Western blots for both FANCD2 and BRCA1, we found that the decrease in FANCD2 paralleled that of BRCA1 (Fig. 4A). We found no difference in the stability of FANCD2 protein after treatment with DFX (Supplementary Fig. S6A), suggesting regulation at the transcriptional or translational level. We thus sought to determine whether the decrease in FANCD2 protein levels is because of reduced mRNA levels using quantitative real-time PCR (qRT-PCR). We found that after extended treatment with DFX, FANCD2 mRNA levels in HeLa cells progressively decreased, closely matching the decrease in BRCA1 mRNA (Fig. 4B and C). DFX-treated A549 and MCF7 cells similarly demonstrated a decrease in FANCD2 protein and mRNA that paralleled the decrease in BRCA1 (Supplementary Figs. S1 and S2).

The hypoxic transcriptional repression of BRCA1, as well as RAD51, is regulated by E2F4/p130 complexes (18, 19). Specifically, hypoxia induces p130 dephosphorylation and nuclear accumulation allowing the formation of repressive E2F4/p130 complexes, which bind to E2F consensus sites in the proximal promoters of BRCA1 and RAD51. We report here that DFX treatment similarly elicits a decrease in the hyperphosphorylated form of p130 and an increase in the hypophosphorylated form (Supplementary Fig. S6B), suggesting that it could induce similar regulatory pathways. The E2F sites mediating the downregulation of BRCA1 and RAD51 are identical sequences contained in a 9-bp region of homology, found in the same orientation, located just upstream of the transcription start sites, and are evolutionarily conserved, suggesting coregulation of these proteins in hypoxia (19). Three putative E2F-binding sites have also been identified in the promoter region of FANCD2 (40). We further analyzed these FANCD2 E2F sites for signs that they may undergo coregulation with BRCA1 and RAD51. Strikingly, one of the E2F sites (the +7 site) is highly homologous to the BRCA1 and RAD51 sites, all are found on the negative strand, and all are within 20 bp of the transcription start site (Fig. 4D). This site is also highly conserved among vertebrates (Fig. 4E).

Given the homology among the FANCD2, BRCA1, and RAD51 E2F binding sites, we hypothesized that hypoxia-induced transcriptional repression of FANCD2 may also be mediated by E2F4/p130. To test this, we utilized RKO cells containing an HPV16-E7-expressing vector (RKO-E7) or a control vector (RKO-Neo). HPV16-E7 protein binds to p130, prevents its interaction with E2F proteins, and targets it for degradation, preventing its ability to transcriptionally regulate genes with E2F4 (41). After treatment with DFX, we found that the decrease in FANCD2 protein levels, like BRCA1, was attenuated by the expression of HPV16-E7 (Fig. 4F). Of note, the expression level of HPV16-E7 also decreased with DFX treatment in our system, likely explaining the eventual suppression in FANCD2 and BRCA1 levels in RKO-E7 cells after prolonged DFX exposure. We next performed qRT-PCR using the RKO-Neo and RKO-E7 cells and found that FANCD2 and BRCA1 mRNA levels displayed a very similar pattern (Fig. 4G). In RKO-Neo cells, both mRNAs decreased slowly over time even in untreated cells (likely because of reduced proliferation as cells become more confluent) but decreased dramatically to nearly undetectable levels with DFX treatment. In contrast, in RKO-E7 cells, there was little to no mRNA decrease in untreated cells and a much less dramatic mRNA decrease in DFX-treated cells. Again, the slow decrease in FANCD2 and BRCA1 mRNA in DFX-treated RKO-E7 cells may be because of the eventual reduction in HPV16-E7 protein itself.

To confirm that the downregulation of FANCD2 upon DFX treatment is representative of downregulation occurring with hypoxic treatment, we repeated the experiments in RKO-Neo and RKO-E7 cells treated with hypoxia for 24 and 48 hours. As with DFX, expression of HPV16-E7 in hypoxia-treated cells blocked the decrease in FANCD2 and BRCA1 protein levels (Fig. 4H). After 48 hours hypoxia, FANCD2 protein levels were reduced to 50% in RKO-Neo cells but remained at 100% in RKO-E7 cells (Supplementary Fig. S7). FANCD2 and BRCA1 mRNA levels decreased to less than 40% after 24 hours hypoxia in RKO-Neo with no decrease in RKO-E7 cells (Fig. 4I). After 48 hours hypoxia, FANCD2 and BRCA1 mRNA levels did decrease in RKO-E7 cells, but still remained above the levels seen in RKO-Neo cells (Fig. 4I) and may reflect the decrease in HPV16-E7 protein expression after 48 hours hypoxia (Fig. 4H). These results suggest that FANCD2, like BRCA1 and RAD51, is transcriptionally repressed by the E2F4/p130 transcription factors upon exposure to hypoxia or DFX.

Finally, we asked whether FANCI is transcriptionally regulated in hypoxia. The promoter region of FANCI does contain several potential E2F binding sites, although none share the high similarity, close proximity to the transcription start site, and strand orientation of the FANCD2/BRCA1/RAD51 sites (data not shown). Using the same RNA samples from HeLa cells treated with DFX, however, we found that FANCI mRNA levels do significantly decrease after 48 to 72 hours of treatment (Supplementary Fig. S8A). FANCI mRNA also decreases in RKO-Neo cells treated with hypoxia or DFX, and the downregulation is abated by overexpression of HPV16-E7 (Supplementary Fig. S8B and S8C).

Histone variant H2AX is phosphorylated at serine 139 by ATM in response to DNA DSB formation. This modified histone, called γH2AX, is required for the accumulation of DNA repair proteins at sites of DNA damage and for activation of cell-cycle checkpoints. In addition, H2AX is phosphorylated and forms nuclear foci in response to replication fork arrest caused by HU or UV in an ATR-dependent manner (42). Severe hypoxia also results in the accumulation and nuclear foci formation of γH2AX and is suspected to be because of replication fork stalling (10).

Given the role of the FA proteins at stalled replication forks, it is not surprising that FA cells have constitutively elevated levels of γH2AX (43) and accumulate excess γH2AX after UV-induced fork stalling (44). We asked whether FANCD2 might also be required to limit the accumulation of γH2AX in cells under hypoxic stress. We performed Western blots for γH2AX in PD20+EV, PD20+FD2, and PD20+KR cells treated with severe hypoxia for 24 to 48 hours or with MMC or HU as controls. We found that hypoxia, MMC, and HU all induced large elevations in γH2AX levels in PD20+EV and PD20+KR cells, but only small increases in the PD20+FD2 cells (Fig. 5A). After 48 hours hypoxia, the results were even more dramatic, with FANCD2-deficient and -mutant cells displaying 18- and 12-fold increases in γH2AX compared with a 4-fold increase in FANCD2-corrected cells (Fig. 5A).

We hypothesized that excess γH2AX in FANCD2-deficient cells signaled either an increased number of stalled replication forks or increased DNA DSBs arising from collapsed replication forks. To determine whether FANCD2-deficient cells develop increased DNA DSBs in response to hypoxia, we turned to the neutral single-cell gel electrophoresis (comet) assay. In this assay, DNA DSBs are detected as an increase in the tail moment of the DNA comets derived from the cells. After treatment with severe hypoxia or normoxia for 48 hours, we found that PD20+EV and PD20+KR cells exposed to hypoxia had significant increases in tail moment compared with PD20+FD2 cells (Fig. 5B). Data averaged from 3 independent experiments demonstrated 4.4- and 3.2-fold increases in the mean tail moment in PD20+EV and PD20+KR cells exposed to hypoxia, respectively, whereas PD20+FD2 cells had less than a 2-fold increase. Prior studies have reported that severe hypoxia in the absence of reoxygenation does not generate DNA damage in normal cells (9, 45). The small increase in tail moment that we observe in the PD20+FD2 cells may result from early damage because of reoxygenation during sample preparation. Regardless, these data indicate that functional FANCD2 is important for protecting cells from hypoxia-induced DNA damage and suggest that at least some of the excess γH2AX that accumulates in hypoxic FANCD2-deficient cells is because of DNA DSB formation. We investigated whether the formation of DSBs manifests as an increase in chromosomal aberrations via cytogenetic analysis of the PD20 set of cells after treatment with normoxia, hypoxia, or MMC. Although MMC induced an increase in aberrations in the absence of functional FANCD2, we were unable to detect an increase in aberrations under hypoxia, suggesting that this assay was not sensitive enough to detect hypoxia-induced damage (Supplementary Fig. S9).

In the DNA damage assays, some of the comets from the hypoxia-treated PD20+EV and PD20+KR cells had the appearance of “hedgehog” or “cloud” comets with long or large tails and very small heads (Fig. 5C). Such comets can potentially represent early apoptotic cells (46). To determine whether loss of FANCD2 results in elevated hypoxia-induced apoptosis, we first assessed for cleavage of PARP in PD20+EV, PD20+FD2, and PD20+KR cells upon treatment with hypoxia or normoxia. We observed slightly elevated levels of cleaved PARP in all hypoxia-treated cells after 24 hours and a more substantial increase after 48 hours (Fig. 6A). Significantly, the elevation of cleaved PARP was greater in PD20+EV and PD20+KR cells compared with PD20+FANCD2 cells.

As a second marker of apoptosis, we measured combined caspase-3 and -7 activity in PD20 cells treated with hypoxia using a luminescence-based assay. We observed that caspase activity was elevated in hypoxic PD20+EV and PD20+KR relative to PD20+FD2 cells (Fig. 6B). The level of caspase induction by hypoxia was comparable to caspase induction with cisplatin treatment in all 3 groups of cells. Together, these experiments indicate that FANCD2, and its ability to undergo ubiquitination, is crucial for limiting hypoxia-induced apoptosis.

In this study we have identified a novel stimulus for activation of the FA DNA repair pathway. Exposure to hypoxia or the hypoxia-mimetic deferoxamine induces the monoubiquitination of FANCD2 and FANCI in an ATR-dependent manner. Upon hypoxia-induced ubiquitination, FANCD2 forms nuclear foci colocalizing with γH2AX. The ubiquitination of FANCD2 is functionally significant as it prevents excess accumulation of γH2AX and DNA damage measured by the comet assay. Furthermore, the presence of wild-type FANCD2, but not a ubiquitination-defective mutant, protects cells from hypoxia-induced apoptosis. Following the acute activation of the FA pathway, longer exposure to hypoxia results in transcriptional downregulation of FANCD2 and FANCI in a manner analogous to the downregulation of RAD51 and BRCA1 by the E2F4/p130 transcription factor. Altogether, our results establish a key role for the FA pathway in the acute response to hypoxia and identify a new mechanism that may contribute to genomic instability induced by prolonged hypoxia.

We believe that the activation of the FA pathway in hypoxia is most likely a response to the replication stress known to occur under hypoxia. The precise role of FANCD2 in replication stress remains incompletely understood, but several recent studies have shed light on potential mechanisms (28, 47). Upon replication stalling, cells deficient for FANCD2 or its ubiquitination, as well as cells lacking BRCA1, BRCA2, or RAD51, demonstrate destabilized replication forks and increased chromosomal aberrations (28). Fork destabilization is dependent upon MRE11 nuclease activity whereas expression of mutant RAD51 that forms hyperstable DNA filaments compensates for FANCD2 deficiency, suggesting that FANCD2 protects stalled forks from nucleolytic degradation potentially through stabilization of RAD51-DNA filaments (28). Additional work has demonstrated that FANCD2 and FANCI directly associate with MCM proteins and that FANCD2 is necessary for initially restraining DNA synthesis and preventing the accumulation of ssDNA upon nucleotide depletion (47). Therefore, FANCD2 and FANCI may have multifunctional roles in protecting cells from replication stress.

Subsequent to the activation of FANCD2 and FANCI upon hypoxic exposure, we found that both proteins are transcriptionally downregulated. Interestingly, hypoxic cells have previously been reported to have increased sensitivity to DNA cross-linking agents, including MMC and cisplatin, which has been attributed to a deficiency in homologous recombination (17). However, because FA-deficiency causes sensitivity to ICLs, the downregulation of FANCD2 and FANCI is likely to contribute. Furthermore, the FA proteins are known to promote homologous recombination and single-strand annealing (48), suggesting that their downregulation may also contribute to the repression of DNA DSB repair observed under hypoxia (16). FANCD2 and FANCI downregulation seems to be coordinated with the downregulation of the homologous recombination proteins BRCA1 and RAD51. For all 4 genes, the downregulation can be prevented by overexpression of HPV16-E7, which inhibits transcriptional repression by p130/E2F4. Hoskins and colleagues demonstrated that the p130/E2F4 complex can bind directly to the FANCD2 promoter (40), supporting a direct role for p130/E2F4 downregulation of FANCD2 in hypoxia. In addition, FANCC and FANCG expression can also be regulated positively by E2F1 and E2F2 and negatively by Rb and p130 (40). The coordinate regulation of numerous DNA repair proteins suggests an evolutionary basis. Under hypoxic stress, where cells have little metabolic reserve, it would be advantageous for cells to decrease expression of nonessential genes in a coordinated manner. Moreover, it is well established that downregulation of DNA repair genes can increase genomic instability, generating “stress-induced mutagenesis” to allow more rapid adaptation to the environment (49).

The downregulation of FANCD2 and FANCI also raises the interesting possibility that functional FA pathway deficiency could be related to the early observations of fragile site activation in hypoxia (6). Hypoxia induces breaks at fragile sites, the fusion of double minutes, the amplification of double minutes to form larger double minutes, and the reintegration of double minutes at other chromosomal fragile sites, processes that may underlie some of the chromosomal aberrations seen in solid tumors (6). Recent work has shown that the chromosomal break-points found in the cells of patients with FA colocalize with aphidicolin-induced fragile sites (50). FANCD2-deficient cells also demonstrate increased breakage at common fragile sites after treatment with replication inhibitors (27). Therefore, it will be interesting to test the hypothesis that the downregulation of the FA pathway under hypoxia can promote fragile site activation.

Our finding that FANCD2 activation occurs in hypoxia and protects cells from hypoxia-associated DNA damage reveals a key mechanism by which cells cope with the stresses of the tumor microenvironment. Inhibiting the FA pathway would therefore be expected to target hypoxic cells and potentially sensitize them to other DNA damaging agents. Indeed, ATR inhibition has been shown to sensitize hypoxic cells to radiotherapy (51), and our work suggests that it may do so, in part, via an impact on FA pathway activation. In addition, the downregulation of the FA pathway upon chronic hypoxia potentially creates a specific vulnerability in hypoxic tumor cells, possibly explaining the increased sensitivity of hypoxic cells to MMC and cisplatin. We anticipate that our findings on the regulation of the FA pathway in hypoxia will enable further investigation of these therapeutic implications.

No potential conflicts of interest were disclosed.

Conception and design: S.E. Scanlon, P.M. Glazer

Development of methodology: P.M. Glazer

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.E. Scanlon, P.M. Glazer

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.E. Scanlon, P.M. Glazer

Writing, review, and or revision of the manuscript: S.E. Scanlon, P.M. Glazer

Study supervision: P.M. Glazer

The authors thank G. Kupfer, P. Sung, S. Longerich, Z. Yun, and K. Cho for providing reagents, and D. Hegan, J. Czochor, N. Balasubramanian, and Y. Lu for technical assistance.

This work was supported by NIH Grant R01ES005775 to P.M. Glazer and NIH Medical Scientist Program Training Grant T32GM007205.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.