Chronic hypoxia promotes hypoxia-inducible factor-1α–dependent resistance to etoposide and vincristine in neuroblastoma cells Free

Hypoxia is widespread in solid tumors as a consequence of poorly structured tumor-derived neovasculature. Direct measurement of low oxygen levels in a range of adult tumor types has correlated tumor hypoxia with advanced stage, poor response to chemotherapy and radiotherapy, and poor prognosis. Little is known about the importance of hypoxia in pediatric tumors; therefore, we evaluated the effects of hypoxia on the response of the neuroblastoma cell lines SH-EP1 and SH-SY5Y to the clinically relevant drugs, vincristine, etoposide, and cisplatin. Short periods of hypoxia (1% O₂) of up to 16 hours had no effect on drug-induced apoptosis or clonogenic survival. Prolonged hypoxia of 1 to 7 days leads to reduction in vincristine- and etoposide-induced apoptosis in SH-SY5Y and SH-EP1 cells, and this was reflected in increased clonogenic survival under these conditions. Neither short-term nor prolonged hypoxia had any effect on the clonogenic response to cisplatin in SH-SY5Y cells. Hypoxia-inducible factor-1 (HIF-1) α was stabilized in these cell lines within 2 hours of hypoxia but was no longer detectable beyond 48 hours of hypoxia. Up-regulation of carbonic anhydrase IX showed HIF-1α to be transcriptionally active. Down-regulation of HIF-1α by short hairpin RNA interference and the small-molecule 3-(5′-hydroxymethyl-2′-furyl)-1-benzylindazole reduced hypoxia-induced drug resistance. These results suggest that prolonged hypoxia leads to resistance to clinically relevant drugs in neuroblastoma and that therapies aimed at inhibiting HIF-1α function may be useful in overcoming drug resistance in this tumor. [Mol Cancer Ther 2006;5(9):2241–50]

Neuroblastoma is the most common extracranial solid tumor of childhood. Many patients present over the age of 1 year, with disseminated disease and well-characterized adverse molecular signatures, including amplification of MYCN. Even with intensive multiagent induction chemotherapy, high-dose therapy with autologous stem cell rescue, radiotherapy, surgery, and maintenance chemotherapy with 13-cis-retinoic acid, the survival for this group of patients is poor (1).

Hypoxia, a reduction in the level of tissue oxygen to 1%, is commonly found in tumors as a result of decreased blood flow in the tumor-derived neovasculature (2). Hypoxia correlates with decreased survival and advanced stage in a range of adult tumors, including squamous cell carcinomas of the head and neck and carcinoma of the cervix (3). Hypoxia has long been known to decrease the effectiveness of radiotherapy, and the effectiveness of several chemotherapeutic drugs may also be modified by hypoxia. One of the principle modulators of tumor cell response to hypoxia is the transcription factor hypoxia-inducible factor-1 (HIF-1). HIF-1 exists as a heterodimer of HIF-1α and HIF-1β (also known as aryl hydrocarbon receptor nuclear translocator; ref. 4). Whereas HIF-1β is constitutively expressed, HIF-1α levels are normally kept low through proteasomal degradation. The ubiquitinylation and targeting for degradation of HIF-1α is mediated through its binding to the protein product of the von Hippel-Lindau tumor suppressor gene under conditions of normal oxygen tension (5, 6). Under hypoxic conditions, this interaction is disrupted, allowing dimerization with HIF-1β and transactivation of target genes (7).

HIF-1 transcriptionally up-regulates the levels of two proapoptotic members of the Bcl-2 family of proteins, NIX and BNIP3, in several different tumor cell types (8, 9). BNIP3-mediated cell death is thought to be independent of cytochrome c release and caspase activation. BNIP3 however induces mitochondrial permeability pore opening and many of the morphologic features of necrosis. This mode of cell death has been referred to as aponecrosis (10). HIF-1 is also able to induce p53-dependent apoptosis (11) and hypoxia has been shown to induce apoptosis in neuroblastoma cell lines in vitro, acting at least in part by up-regulation of the cell surface expression of the CD95/Fas death receptor. In this setting, overexpression of MYCN resulted in an increase in the sensitivity of neuroblastoma cells to hypoxia-induced apoptosis (12). It is thus likely that chronic hypoxia actively selects for tumor cells with nonfunctional hypoxia-induced apoptotic pathways. These tumor cells are predicted to fail to engage apoptosis after drug or radiation treatment and this pleiotropic resistance may account for the correlation observed between hypoxia and poor prognosis. Previous studies in our laboratory have shown that hypoxia leads to HIF-1α-dependent down-regulation of the proapoptotic Bcl-2 family protein Bid in colon carcinoma cell lines, and together with hypoxia-dependent, but HIF-1α-independent, down-regulation of Bad, Bim, and BNIP3, this leads a decrease in etoposide-induced apoptosis (13).

Neuroblastoma cell lines are able to stabilize HIF-1α in response to hypoxia and up-regulate HIF-1 target genes, including vascular endothelial growth factor and tyrosine hydroxylase, both in vitro and as xenografts (14). Multiple angiogenic factors (including vascular endothelial growth factor) are expressed in neuroblastomas in vivo, and higher level expression correlates with advanced disease and poor outcome (15). Thus, we investigated the effects of hypoxia and the role of HIF-1α in the response of neuroblastoma cell lines to etoposide, vincristine, and cisplatin, drugs that are widely used in the treatment of this disease.

SH-EP1 and SH-SY5Y were the kind gift of Dr. Robert Ross (Fordham University, Bronx, NY) and were maintained in DMEM-F12 with 10% FCS (Gibco, Invitrogen, Paisley, United Kingdom).

For hypoxia experiments, cells were cultured and treated in a sealed modular incubator (InVivo2 Hypoxia workstation 400) flushed with 1% O₂, 5% CO₂, and 94% N₂ (hereafter referred to as hypoxia).

Cells were plated at 500 per well in six-well tissue culture plates (Costar, Corning, NY). Twenty-four hours after plating, cells were incubated in hypoxia or left in normoxia. Twenty-four hours later, cells were treated with etoposide (0–100 μmol/L; Sigma, Gillingham, United Kingdom), vincristine (0–0.8 μmol/L; Sigma), or cisplatin (0–50 μmol/L; Sigma) for 1 hour under the same oxygen levels as the preceding 24 hours. Medium was then changed and cells were either replaced in hypoxia or placed back in normoxia for 7 days. Colonies were fixed with 70% methanol and stained with methylene blue and colonies of >50 cells were counted. Survival fraction is calculated as number of colonies in the test condition/number of colonies in the control/untreated well and plotted logarithmically against drug dose. Absolute clonogenicity (number of colonies/number of cells originally seeded) was not significantly different between normoxic and hypoxic conditions.

Cells were treated as above with clonogenic IC₉₀ doses of each drug and harvested by trypsinization at defined time points. Cells were fixed by dropwise addition of ice-cold 95% (v/v) ethanol. Cells were stained with propidium iodide (10 μg/mL; Molecular Probes, Invitrogen, Paisley, United Kingdom) and analyzed on a FACScalibur flow cytometer (Becton Dickinson, Oxford, United Kingdom) using the argon laser tuned to 488 nm and where red fluorescence (propidium-bound DNA) was detected at 630 ± 22 nm. Cells (30,000) were analyzed at a flow rate of 300 to 500 per second. WinMDI and Modfit softwares (Becton Dickinson) were used to evaluate the data. The fold difference between the percentage of sub-G₁ cells in normoxia and hypoxia is shown; thus, if under hypoxia, the sub-G₁ percentage was 25%, and if under hypoxia, this was 12.5%; this is a 2-fold decrease. Each experiment was repeated thrice, and mean fold differences in sub-G₁ percentage between hypoxia and normoxia were calculated.

Cells were treated in exponential growth phase with either etoposide (50 μmol/L for 1 hour) or vincristine (0.2 μmol/L for 1 hour) and harvested at defined time points following treatment by mechanical detachment. Cells were suspended in lysis buffer [50 mmol/L Tris (pH 7.5), 50 mmol/L sodium fluoride, 10 mmol/L β-glycerophosphate, 10 mmol/L sodium pyrophosphate, 1 mmol/L EDTA, 1 mmol/L EGTA, 0.2% (v/v) Triton X-100, 0.1 mmol/L phenylmethylsulfonyl fluoride, 10 μL/mL protease inhibitor cocktail (Sigma)] and sonicated at 10 Hz for 10 seconds. Protein content was estimated using Bio-Rad (Hemel Hempstead, United Kingdom) protein assay reagent. Lysate was boiled for 15 minutes in sample buffer [21% (v/v) glycerol, 4% (w/v) SDS, 0.1% (v/v) β-mercaptoethanol, bromothenol blue in 0.5mol/L Tris (pH 6.8)], run on appropriate percentage polyacrylamide gels, and transferred to polyvinylidene difluoride membranes (Immobilon, Millipore, Watford, United Kingdom). Membranes were blocked with 5% (w/v) nonfat dried milk/0.05% (v/v) TBS-Tween 20 and probed with primary antibody in 0.05% (v/v) TBS-Tween 20 at 4°C overnight and then with secondary antibody conjugated to horseradish peroxidase in 5% (w/v) milk/0.05% (v/v) TBS-Tween 20 for 1 hour at room temperature. Blots were visualized with the enhanced chemiluminescence system (Amersham, Chalfont St. Giles, United Kingdom) and analyzed on a Fuji LAS-1000 Plus imaging system with AIDA software (Fuji, Bedford, United Kingdom). Primary antibodies used were carbonic anhydrase IX (Santa Cruz Biotechnology, Santa Cruz, CA), HIF-1α (BD Transduction Laboratories, Oxford, United Kingdom), Bid N-19 (Santa Cruz Biotechnology), actin (act40, Sigma), caspase-3 (Cell Signalling, Danvers, MA), and poly(ADP-ribose) polymerase (PARP; Cell Signalling). Secondary antibodies were either goat anti-mouse horseradish peroxidase or goat anti-rabbit horseradish peroxidase (both from DAKO, Ely, United Kingdom).

Cells were plated in 75-cm² tissue culture flasks and left in either normoxia or hypoxia. After 48, 72, 120, or 168 hours of hypoxia, the proteasome inhibitor MG132 (50 μmol; Sigma) was added and incubated with cells for 8 hours in hypoxia before harvesting. Cells were then subjected to immunoblotting for HIF-1α as described above. As a control for the effectiveness of this dose and duration of treatment, normoxic cells were incubated with MG132 for 8 hours before harvesting for immunoblotting for HIF-1α.

The pSilencer1 vector (Ambion, Huntingdon, United Kingdom) containing the previously published HIF-1α target sequence (13) was modified by cloning in green fluorescent protein and a puromycin resistance gene. SH-EP1 and SH-SY5Y cells were transfected with the resulting vector by electroporation at 260 V and capacitance of 1,050 μF. Forty-eight hours after transfection, cells were placed into medium containing puromycin (0.6 μg/mL) for 10 days. Cells expressing green fluorescent protein were isolated with a FACSVantage cell sorter (Becton Dickinson) set to excite at 488 nm and cells that exhibited green fluorescence at 520 ± 30 nm were collected. Cells were reseeded in six-well plates at 500 per well and single green fluorescent protein–positive clones were picked after 7 days. After a further 10 days in puromycin, cells were screened for HIF-1α expression by immunoblotting after 8 hours of hypoxia as described previously. Stable clones with and without short hairpin RNA (shRNA)–mediated repression of HIF-1α expression were chosen for each cell line. These were then used for clonogenic assay, apoptosis measurement, and immunoblotting for cleaved caspase-3 and PARP after treatment with vincristine and etoposide in prolonged hypoxia as described previously.

3-(5′-Hydroxymethyl-2′-furyl)-1-benzylindazole (YC-1) is a small molecule that inhibits HIF-1α protein expression post-transcriptionally by an unknown mechanism (16). Cells were plated for clonogenic assays as described above. Cells were then treated with vincristine as described previously but additionally treated with 10 μmol/L YC-1 (Calbiochem, Merck Biosciences, Nottingham, United Kingdom) with varying schedules as described in Results.

Quantitative reverse transcription-PCR assays were designed by inputting sequences into the Probe finder software.⁴

Assays were done using amplification primers HIF-1α TTTTTCAAGCAGTAGGAATTGGA (left) and TTCCAAGAAAGTGATGTAGTAGCTG (right; Invitrogen, Paisley, United Kingdom) and labeled probe 71 from the Universal Probe Library (Roche, Applied Science, Lewes, United Kingdom). Tubulin and actin were selected as housekeeper genes from a panel of five. Amplification primers for actin were GGATGCCACAGGACTCCAT (left) and ATTGGAATGAGCGGTTC (right) and labeled probe 11; amplification primers for tubulin were TTAACCATGAGGGAAATCGTG (left) and CTGATCACCTCCCAGAACTTG (right) and labeled probe 64. RNA was harvested at defined time points after drug treatment using the RNeasy kit (Qiagen, Crawley, United Kingdom). Reverse transcription of 1 μg RNA was done with Taqman reverse transcription reagents (Roche Applied Science) following the manufacturer's instructions. Quantitative PCR was done with 5 ng cDNA as template using Taqman master mix (Roche Applied Science) and an ABI prism 7900HT sequence detection system (Roche Applied Science).

Statistical significance was tested using standard ANOVA methods (weighted least squares regression-weighted by weight) and set at P < 0.05.

To investigate the effects of hypoxia on the drug sensitivity of neuroblastoma cells, the clonogenic response of cells were compared using three regimes: (a) cells were maintained and treated in normoxia (hereafter termed normoxia), (b) cells were incubated under hypoxia for 23 hours before drug treatment and then released into normoxia (termed acute hypoxia), and (c) cells were incubated under hypoxia for 23 hours before drug treatment and then maintained in hypoxia (termed chronic hypoxia). In contrast to our previous findings for etoposide-treated colorectal cancer cells (13), acute hypoxia had no effect on the clonogenic survival of SH-EP1 and SH-SY5Y cells after exposure to vincristine or etoposide (Fig. 1A). However, chronic hypoxia led to a significant increase in the clonogenic survival of SH-EP1 and SH-SY5Y cells after exposure to vincristine and etoposide (Fig. 1A). The effect of chronic hypoxia on clonogenic response to cytotoxics was drug specific as, although chronic hypoxia led to a significant increase in clonogenic survival of SH-EP1 cells after exposure to cisplatin, it had no effect on the survival of SH-SY5Y cells (data not shown). Prolonged hypoxia had no significant effect on the cell cycle profile of either cell line with or without exposure to either vincristine or etoposide.

Where an increase in clonogenic survival under prolonged hypoxia was observed, it correlated with a reduction in drug-induced apoptosis (Fig. 1C). In both SH-SY5Y and SH-EP1 cells exposed to vincristine, apoptosis was 6- to 8-fold greater under normoxia than under chronic hypoxia by 4 days after drug treatment, and in SH-SY5Y cells, etoposide-induced apoptosis was 5-fold greater under normoxia than under prolonged hypoxia 4 days after drug treatment. Chronic hypoxia had no effect on cisplatin-induced apoptosis in SH-SY5Y cells (data not shown). Differences in drug-induced apoptosis in Fig. 1C (measured by the appearance of a sub-G₁ population) were confirmed biochemically by lower levels of the 17- and 19-kDa cleavage products of caspase-3 and the 85-kDa cleavage product of the caspase-3 substrate, PARP (note the differences in both full-length and cleaved forms between hypoxia and normoxia; Fig. 1B). Acute hypoxia had no effect on apoptosis induced by any of these three drugs, again contrasting with the results we had obtained with colon cancer cell lines previously (13) but correlating with the lack of effect of acute hypoxia on clonogenic survival after drug treatment. Thus, chronic exposure to hypoxia leads to increased clonogenic survival in neuroblastoma cells after vincristine and etoposide treatment to which decreased drug-induced apoptosis contributes. However, this is not a universal effect and depends on the type of cellular damage inflicted as chronic hypoxia had no effect on the clonogenic response of SH-SY5Y cells to cisplatin.

HIF-1α protein was undetectable in either cell line under normoxic conditions but became detectable within 2 hours of hypoxia (Fig. 2A). HIF-1α protein remained detectable for up to 48 hours in hypoxic SH-EP1 and SH-SY5Y cells but was undetectable again beyond 48 hours of hypoxia. HIF-1α activity was assessed in these experiments by the up-regulation of protein levels of carbonic anhydrase IX, a known transcriptional target of HIF-1α (17). Like HIF-1α, carbonic anhydrase IX levels were undetectable in normoxia and were increased by 48 hours of hypoxia (Fig. 2B). Interestingly, carbonic anhydrase IX expression was still detectable after 168 hours of hypoxia, long after HIF-1α levels were undetectable, presumably reflecting its far longer half-life as reported previously (18).

To investigate the molecular mechanism for the loss of HIF-1α beyond 48 hours of hypoxia, cells were exposed daily to the proteasome inhibitor MG132 for 8 hours before harvesting for immunoblotting. Under these conditions, HIF-1α was detectable in normoxia and for up to 72 hours in hypoxia but not beyond this point (Fig. 2C). Next, the level of HIF-1α mRNA was measured by quantitative real-time PCR. Tubulin and actin were established as housekeeper genes for this experiment, and no changes in HIF-1α mRNA levels relative to these two genes were seen in either cell line over 168 hours of hypoxia. The mean Pearson correlation coefficient from three independent experiments for changes in C_T in SH-SY5Y cells was 0.978614 (HIF-1α against tubulin) and 0.955119 (HIF-1α against actin); the mean Pearson correlation coefficient from two independent experiments for changes in C_T in SH-EP1 cells was 0.929021 (HIF-1α against tubulin) and 0.930396 (HIF-1α against actin). Thus, as has been reported for other cell lines, HIF-1α levels in neuroblastoma cell lines are regulated by proteasomal degradation and not by changes at the mRNA level up to 48 to 72 hours. However, the lack of HIF-1α protein stabilization beyond 48 to 72 hours despite maintained mRNA levels suggests a nonproteasomal mechanism of HIF-1α loss.

To assess the functional role of HIF-1α in hypoxia-induced drug resistance, stable clones of SH-SY5Y and SH-EP1 cells were generated, in which HIF-1α expression was repressed by shRNA interference (Fig. 3A). The clonogenic response to vincristine and/or etoposide of these clones (SH-EP 1, clone 2; SH-SY5Y, clone 3) was then compared with those of clones without knockdown of HIF-1α (SH-EP1, clone 1; SH-SY5Y, clone 4). Down-regulation of HIF-1α significantly decreased the clonogenic survival of both SH-EP1 and SH-SY5Y cells after vincristine treatment and of SH-SY5Y cells after etoposide treatment in prolonged hypoxia (Fig. 3B). This reduction in clonogenic survival after drug treatment in HIF-1α knockdown neuroblastoma cells was only seen under hypoxic conditions (data not shown). The differences in clonogenic survival after drug treatment were mirrored by increased drug-induced apoptosis in HIF-1α-repressed cells after both drug treatments in chronic hypoxia reported by an increase in the sub-G₁ population (Fig. 3D) and greater cleavage of PARP after drug treatment of low HIF-1α expressors compared with high HIF-1α expressors (Fig. 3C). Taken together, these data suggest that the reduction in drug-induced apoptosis and increase in clonogenic survival in SH-EP1 cells after vincristine treatment and SH-SY5Y cells after vincristine and etoposide treatment was dependent on HIF-1α function in these cells.

We have reported previously acute changes in several proapoptotic Bcl-2 family proteins after exposure of colon cancer cell lines to hypoxia both in vitro and as xenografts (13). In these experiments, Bid was identified as a HIF-1 transcriptional target. Therefore, the response of proapoptotic Bcl-2 family proteins to hypoxia in SH-EP1 and SH-SY5Y cells, and Bid, in particular, was studied. In clear contrast to the colon carcinoma cell lines, no acute changes in Bid, Bim, Bad, or Bax were observed (Fig. 4; data not shown). The down-regulation of Bid in SH-EP1 cells after 72 hours of hypoxia was noted (Fig. 4), although >50% reduction in drug-induced apoptosis had already occurred at this time point (Fig. 1B), making it unlikely that this change contributes to the drug resistance of SH-EP1 cells under hypoxia.

The data presented thus far suggest that HIF-1α function is required for the resistance of SH-EP1 and SH-SY5Y cells to vincristine and SH-SY5Y cells to etoposide under chronic hypoxia. However, the duration of HIF-1α stabilization by hypoxia was limited to a 48-hour period. It was therefore of interest to determine first whether vincristine and etoposide treatment affected the duration of HIF-1α up-regulation under hypoxia and, second, whether the period of HIF-1α stabilization modulated the degree of drug resistance. Treatment with vincristine and etoposide prolonged the time during which HIF-1α was detectable under hypoxic conditions at the protein level. For both cell lines, drug treatment for 1 hour after 23 hours of hypoxia leads to detectable expression of HIF-1α up to 168 hours of hypoxia (Fig. 5A), whereas without drug treatment, this expression could not be seen beyond 48 hours (Fig. 2). Although vincristine and etoposide treatment was able to stabilize HIF-1α protein, real-time quantitative PCR did not show any changes in HIF-1α mRNA levels after drug treatment in prolonged hypoxia compared with changes in the two housekeeper genes tubulin and actin (Fig. 5B). The next set of experiments were designed to determine whether this prolonged expression of HIF-1α after vincristine treatment of both SH-EP1 and SH-SY5Y cells was important to their resistance to vincristine under prolonged hypoxia. YC-1 blocks HIF-1α expression post-transcriptionally, both in vitro and in vivo (16). SH-EP1 and SH-SY5Y cells were treated with 10 μmol/L YC-1 in two schedules, either (a) at the same time as they were placed in hypoxia followed 23 hours later by vincristine treatment and then removal of YC-1 (short treatment; Fig. 5C, schedule 4) or (b) with continuous treatment with YC-1 from the time that they were treated with vincristine (prolonged treatment; Fig. 5C, schedule 3). HIF-1α expression was significantly up-regulated after 24 hours of hypoxia, and this up-regulation was inhibited in the presence of YC-1 (Fig. 5D, compare normoxia with schedules 1 and 2). As before vincristine treatment leads to prolonged expression of HIF-1α, which remained detectable at 168 hours, however, this expression was much reduced in the continuous presence of YC-1 (Fig. 5D, compare schedules 3 and 4). Removal of YC-1 leads to a rapid return to normal levels of HIF-1α, so that 24 hours after removal, HIF-1α levels were comparable with those without YC-1 treatment (Fig. 5D, right, compare schedules 1, 2, and 5). Thus, HIF-1α expression level was reduced before vincristine treatment and allowed to return to normal thereafter (Fig. 5D, schedule 4), or cells with normal (high) HIF-1α levels were treated with vincristine and then HIF-1α levels were reduced (Fig. 5D, schedule 3). In this way, the effects of these two perturbations of HIF-1α levels on the clonogenic response of SH-EP1 and SH-SY5Y cells to vincristine in prolonged hypoxia were compared. Prolonged treatment with YC-1, which attenuates the duration of HIF-1α stabilization, significantly decreased the clonogenic survival of both cell lines to vincristine treatment under prolonged hypoxia. On the other hand, short treatment with YC-1, which reduced HIF-1α levels acutely without effecting HIF-1α between 48 to 168 hours, had no effect on the clonogenic response of SH-SY5Y cells to vincristine and significantly increased the clonogenic survival of SH-EP1 cells under these conditions (Fig. 5E). Thus, the timing of HIF-1α stabilization relative to drug treatment SH-EP1 and SH-SY5Y was important for resistance to chemotherapy-induced apoptosis in hypoxia.

Hypoxia induces a multitude of adaptive changes in cells. Not least among these is the up-regulation of several proapoptotic members of the Bcl-2 family of proteins, including BNIP3 (19) and PUMA (20), and the induction of apoptosis. The ability of tumor cells to evade apoptosis is fundamental to their survival in the hypoxic microenvironment. Transformed cells, in which Bcl-2 is overexpressed or Bax and Bak are lost, become resistant to hypoxia-induced apoptosis and develop genomic instability (20). Moreover, hypoxia contributes further to genetic instability by HIF-1α-dependent transcriptional repression of the mismatch repair genes MSH2 and MSH6 (21). We have shown previously in colon carcinoma cell lines that this tumor cell resistance to apoptosis can contribute to drug resistance. These cells respond to hypoxia by down-regulating Bax, Bid, Bim_EL, Bad, and BNIP3 in vitro (and Bax and Bid in vivo) and this, at least in part, makes them more resistant to etoposide-induced apoptosis in hypoxia (13). Hypoxia is also able to prevent nerve growth factor withdrawal–induced cell death in sympathetic neurons by down-regulating Bim_EL (22). Neuroblastoma cell lines have been shown to stabilize HIF-1α and HIF-2α in hypoxia and to up-regulate hypoxia-responsive genes, including vascular endothelial growth factor in these conditions. In addition, these cell lines have been observed to down-regulate neuronal genes and up-regulate genes associated with neural crest tissue, and this has been interpreted as dedifferentiation toward an immature phenotype (14). Gene expression analysis of neuroblastoma cell lines exposed to hypoxia for 72 hours showed up-regulation of several genes associated with cell survival and drug resistance (23).

Neuroblastoma cell lines are often biphenotypic in culture, and for this reason, it is often easier to work on phenotypically stable subclones than with patient-derived parental lines. SH-EP1 and SH-SY5Y are S-type and N-type subclones of SK-N-SH, derived in the 1980s and extensively characterized (24). Neuroblastoma is a clinically heterogenous tumor, and a significant minority of patients have amplification of the oncogene MYCN, which correlates with poor outcome (1). It is impossible for any cell line model to be fully representative of the broad clinical range of this tumor. Neither SH-EP1 nor SH-SY5Y has amplification of MYCN, like most clinical tumors. Within these subclone pairs, there are significant differences in chemosensitivity as we have described previously (25) and it is thus important to include both in this type of study. Our study shows that chronic, but not acute hypoxia, leads to resistance to drug-induced apoptosis in these neuroblastoma cell lines in vitro. In agreement with previous data, no induction of apoptosis by hypoxia in the two neuroblastoma cell lines was observed (12). In contrast to our previous work with colon carcinoma cell lines, no effect was seen on drug-induced apoptosis with short periods of hypoxia (24 hours or less); in addition, the acute down-regulation of Bid and Bax observed previously in colon cell lines and in the HT1080 fibrosarcoma cell line (26) did not occur in the neuroblastoma cell lines (see Fig. 4). These differences suggest that different tumor cell types use different mechanisms to evade hypoxia-induced apoptosis, although the functional consequence of resistance to chemotherapy-induced apoptosis may be the same. The effects of hypoxia on drug-induced apoptosis were not universal and were drug dependent as has been described previously in other cell lines (27). A small but significant decrease in the sensitivity of (the already resistant; ref. 28) SH-EP1 to etoposide-induced apoptosis in hypoxia was observed. In these neuroblastoma cell lines, cisplatin causes cell cycle arrest in S and G₂ phases, up-regulation of p53 at the protein level, and the induction of apoptosis (25). However, neither acute nor chronic hypoxia had any effect on cisplatin-induced apoptosis in SH-SY5Y cells. Interestingly, anoxia (<0.1% O₂) has been reported recently to reduce etoposide-induced apoptosis in the neuroblastoma cell line SK-N-BE (2), and in this study, drug-specific effects were also seen because the same 24-hour exposure to anoxia had no effect on cisplatin- and melphalan-induced cell death (29). The anoxic treatment in this study was of the same length as our acute exposure to hypoxia, which had no effect on vincristine- or etoposide-induced apoptosis in SH-EP1 and SH-SY5Y cells, but this difference may simply reflect the different responses of tumor cells to anoxia compared with hypoxia; indeed, the authors found that most neuroblastoma cell lines were unable to tolerate 24 hours of anoxia, although we have observed no toxic effects of chronic hypoxia on any neuroblastoma cell lines. Reported mean oxygen levels in several different adult tumors range from 0.34% in pancreatic carcinoma up to 2.1% in head and neck carcinoma (30); thus, there will be areas of both hypoxia and anoxia within any tumor. Of note, HIF-1α stabilization is far greater in hypoxia than in anoxia and this may also contribute to the different observations (31). In the study by Das et al. (29), the effect of anoxia was reduced by interfering with the vascular endothelial growth factor/Flt1 pathway with antagonistic Flt1 antibodies and by down-regulating the Flt1 receptor with shRNA interference and the authors implicated an autocrine loop between vascular endothelial growth factor/Flt1 and HIF-1α. Such a pathway would be affected by shRNA interference– and YC-1-mediated down-regulation of HIF-1α and may be one potential mechanism by which HIF-1α is able to mediate cytotoxic drug resistance, although as these authors report that SH-EP1 cells do not express the Flt1 receptor, it cannot be the only pathway involved in hypoxia-induced drug resistance.

Our data suggest that HIF-1α function(s) are required for resistance to chemotherapy-induced apoptosis in chronic hypoxia in neuroblastoma cells (Fig. 3). Reduction of HIF-1α by shRNA interference or by treatment with YC-1 had no effect on the drug responses of these cells in normoxia, although HIF-1α-null mouse embryonic fibroblasts have been reported to be significantly more sensitive to carboplatin and etoposide in both normoxic and hypoxic conditions (32). This discrepancy may simply reflect inherent differences between neuroblastoma cells and mouse embryonic fibroblast and/or the differences in HIF-1α levels between our cells with repressed HIF-1α and HIF-1α-null mouse embryonic fibroblasts. Down-regulation of HIF-1α protein in cells maintained in hypoxia has been reported previously in HeLa (33), endothelial, and Hep3B cells (34) and A549 lung adenocarcinoma cells (35). However, the time course in all of these cell lines was shorter than in our neuroblastoma cells, with HIF-1α protein levels observed to decline after 4 to 18 hours of hypoxia in contrast to our observation of maintained expression over 48 hours of hypoxia. A large number of pathways have been described that may be important in the negative regulation of HIF-1α (for review, see ref. 36). However, most of these act by impairing the transcriptional activity of HIF-1α. Down-regulation of HIF-1α at the protein level beyond 48 hours of hypoxia was observed in the neuroblastoma cells, which could be inhibited to a limited extent by the proteasome inhibitor MG132, but no decrease in the level of HIF-1α mRNA, suggesting that the decrease in HIF-1α between 48 and 72 hours of hypoxia is due to proteasomal degradation. In endothelial cells, this degradation of HIF-1α in prolonged hypoxia has been suggested to be dependent on reactive oxygen species and is inhibited by prostaglandin I₂ (37), although such a mechanism may be specific to endothelial cells. Endogenous antisense to HIF-1α has been proposed to be important in the down-regulation of HIF-1α in A549 lung adenocarcinoma cells (35), but such a mechanism involves the destabilization and down-regulation of HIF-1α mRNA, which we did not observe in neuroblastoma cells. A transcription-dependent feedback loop, independent of the proline hydroxylase–modulated von Hippel-Lindau pathway, yet still dependent on proteasomal degradation, and which may depend on acetylation of HIF-1α, has been described recently in A549 cells (38). Such a pathway could explain the down-regulation of HIF-1α between 48 and 72 hours of hypoxia in these neuroblastoma cells, although it does not explain the loss of HIF-1α beyond 72 hours, which could not be stabilized by exposure to a dose of MG132 that was able to stabilize HIF-1α in normoxia. Beyond 72 hours of hypoxia, inhibition of a nonproteasomal degradation of HIF-1α may become an important factor that modulates drug response.

Our data using YC-1 to down-regulate HIF-1α suggest that it is not only the function of HIF-1α per se that is important in mediating drug resistance in chronic hypoxia but also the timing of HIF-1α expression (Fig. 5). The differential effect of HIF-1α inhibition in relation to the timing of cytotoxic drug dose is mirrored by the relationship between HIF-1α inhibition and radiosensitivity (39). Inhibiting HIF-1α activity after radiation treatment with YC-1 led to a significant enhancement of its efficacy against tumor xenografts, and this effect was maximal if HIF-1α activity was inhibited after the radiation had been delivered (40). The need to consider the appropriate scheduling of HIF-1α inhibition with the administration of conventional cytotoxic drug/radiation therapies has clear clinical implications. Several agents are now being developed that target HIF-1α by modifying protein levels, such as YC-1 (41, 42) or by inhibiting the transcriptional activity of HIF-1α (43) or by interfering with the binding of HIF-1α to the critical cofactor p300 (44). Our data suggest that these agents can be effective as sensitizers to cytotoxics in hypoxic cells, but that to do so, they must maintain HIF-1α repression at least in these neuroblastoma cells. Given the often transient nature of hypoxia within tumor tissue, it may also be important to maintain HIF-1α repression to prevent drug resistance in tumor cells that have only just become hypoxic, in which HIF-1α levels may be increasing after cellular exposure to cytotoxic agent.

We thank David Ryder (Department of Medical Statistics, Christie Hospital, Manchester, United Kingdom) for statistical advice and Emma Saunders and Stuart Pepper (Molecular Biology Core Facility, Paterson Institute for Cancer Research, Manchester, United Kingdom) for assistance with quantitative PCR.