HIF-1-dependent Regulation of Hypoxic Induction of the Cell Death Factors BNIP3 and NIX in Human Tumors

Introduction

Solid tumors are poorly oxygenated compared with normal tissues and contain regions of hypoxia. Induction of apoptosis by hypoxia is one mechanism by which stress-damaged cells can be destroyed, and most tumor cells retain the ability to undergo apoptosis in response to hypoxic stress (1) . However, hypoxia increases the mutation rate of cells (2) , resulting in the selection of mutations that make cells more resistant to apoptosis and less responsive to cancer therapy (3 , 4) . HIF-1 3 is a heterodimeric transcription factor consisting of an oxygen-regulated α subunit (Hif-1α) and a stable nuclear factor, Hif-1β/ARNT, and has been well characterized as a mediator of hypoxic response (reviewed in Ref. 5 ). Under normoxic conditions, Hif-1α is rapidly degraded by the proteosome after being targeted for ubiquitination, a process that is dependent on the pVHL (6) . Under hypoxic conditions, degradation of Hif-1α is suppressed, and transcription of mRNAs encoding hypoxically responsive genes can occur. HIF-1 has been shown to be a factor mediating hypoxia-induced apoptosis; hypoxia increases apoptosis in Hif-1α++ embryonic stem cells and CHO cells, but this process is strikingly reduced in the same cells in which the gene has been disrupted (7) . We screened for genes induced by hypoxia in a breast carcinoma cell line (T47D) using gene expression arrays and detected an up-regulation of BNIP3. Also, using the SAGE map website, 4 virtual subtraction of genes expressed by a glioblastoma cell line (H247) under normoxic and hypoxic conditions revealed up-regulation of NIX in hypoxia (8) . BNIP3 is a proapoptotic mitochondrial protein that was isolated through its interaction with E1B 19K and Bcl-2 (9) . Overexpression of BNIP3 and its homologue NIX (10, 11, 12) in Rat-1 fibroblasts and MCF-7 breast carcinoma cells induces cell death within 12 h. BNIP3 and NIX are expressed ubiquitously in most human tissues as assessed by Northern blotting (12) , although it is not known which cell types express BNIP3 and NIX or if their pattern of expression differs in malignant tissue. A recent study has shown that BNIP3 mRNA levels increase in response to hypoxia in a CHO cell line and that this effect is mediated via Hif-1α (13) . We have characterized the response of BNIP3 and NIX to hypoxia in human cell lines and shown that BNIP3 and NIX are overexpressed in human tumors compared with normal tissue.

Materials and Methods

Cell Culture.

Human cell lines were obtained from the Imperial Cancer Research Fund cell service and grown in DMEM, RPMI 1640, or Hams F-12 supplemented with 10% FCS (Globepharm), l-glutamine (2 μm), penicillin (50 IU/ml), and streptomycin sulfate (50 μg/ml). The human cell lines investigated were: SKBr, T47D, MDA468, MCF7, and MDA231 (breast cancer); SKOV3 (ovarian cancer); HT1080 (fibrosarcoma), MKN45 (stomach cancer); C32 and LS174T (colon cancer); EJ (bladder cancer); DU145 (prostate cancer); U937 and THP-1 (macrophage); RZM (Epstein-Barr virus-transformed normal human lymphocytes); HUVEC (endothelial). RCC4 (renal cancer) cell lines expressing pVHL or empty vector have been described previously (6) . The CHO cell lines used were KA13 (mutated to be defective for Hif-1α) and C4.5 (the parent cell line) and have been described previously (14) . Parallel incubations were performed on flasks of cells approaching confluence in normoxia (humidified air with 5% CO₂) or hypoxia [hypoxic conditions were generated in a Napco 7001 incubator (Precision Scientific) with 0.1% O₂, 5% CO₂, and balance N₂].

Western Blotting.

Cells were homogenized in a lysis buffer containing 8 m urea, 10% SDS, 1 m DTT, and protease inhibitors. Proteins were electrophoresed on a 10% SDS-PAGE gel and transferred onto a polyvinylidene difluoride membrane (Millipore). BNIP3 protein was detected using a mouse antihuman BNIP3 monoclonal antibody (10) followed by goat antimouse horseradish peroxidase (Dako) and enhanced chemiluminescence developing reagents (Amersham). Blots were exposed to film from 30 s to 2 min.

Immunohistochemistry.

Formalin-fixed paraffin-embedded tissue (John Radcliffe Hospital pathology archives) or cell pellets (created by washing and centrifuging cell lines, which had been scraped from tissue culture flasks) was sectioned onto 3-aminopropyltriethoxy-silane (Sigma Chemical Co.) -coated slides. Sections were dewaxed and rehydrated before being blocked in 10% horse serum. A rabbit polyclonal antibody to BNIP3, which has been described previously (15) , or a mouse monoclonal antibody to human CD68 (Dako) was applied to the sections at 1:500 and 1:10. Biotinylated horse antimouse IgG or goat antimouse IgG (1:200) and avidin-biotin complex AP conjugate were applied consecutively for 30 min each at room temperature and visualized using AP substrate (Vectastain).

Probe Production.

Regions of BNIP3 and NIX selected to avoid areas of homology were amplified by reverse transcription-PCR from cDNA synthesized from MCF-7 cells subjected to hypoxia. BNIP3 cDNA was amplified between bp 277 and 431 using 5′-ACCAACAGGGCTTCTGAAAC-3′ as the upstream primer and 5′-GAGGGTGGCCGTGCGC-3′ as the downstream reverse complement primer. NIX cDNA was amplified between bp 716 and 798 using 5′-AGTAGCTTATTTGAACTTGAGACCATTG-3′ as the upstream primer and 5′-TGAGGGTTACTGGAATTGGATATGTA-3′ as the downstream reverse complement primer. The purified PCR products were labeled for Northern blotting with [³²P]dCTP (T7 Quickprime kit; Pharmacia), and unincorporated label was separated from the probe by running the mixture through a NICK column (Pharmacia) followed by precipitation in 5 m ammonium acetate and ethanol using yeast tRNA as a carrier. For in situ hybridization the purified PCR products were cloned into pCR-script SK (Stratagene, Cambridge, United Kingdom) and sequenced to confirm their identity and orientation. Riboprobes were transcribed (MAXIscript in vitro transcription kit, Ambion AMS Ltd., Witney, Oxon, United Kingdom) from linearized plasmids with [³³P]UTP (Amersham) before phenol extraction and ethanol precipitation.

RNA Preparation and Northern Blotting.

Total RNA was prepared according to Chomczynski and Sacchi (16) and assessed by absorbance at 260/280 nm. Aliquots (20 μg) were electrophoresed in 1% agarose gels containing formaldehyde and transferred to Hybond N membranes by capillary blotting in 10 × SSC [1 × SSC consists of 150 mm sodium chloride and 15 mm tri-sodium citrate (pH 7.0)]. After fixation, blots were incubated overnight at 68°C with ³²P-labeled cDNA probes and washed in several changes of 1 × SSC/0.1% SDS before exposing to X-ray film for ≤7 days. The consistency of RNA loading and transfer was assessed by staining of the 28S rRNA with ethidium bromide.

In Situ Hybridization.

The in situ hybridization protocol used in this study has been described previously (17) . Briefly, the riboprobes were diluted to 30,000 cpm/μl in hybridization buffer [50% deionized formamide, 0.3 m sodium chloride, 10 mm Tris (pH 6.8), 10 mm sodium phosphate (pH 6.8), 5 mm EDTA (pH 8.0), 1× Denhardt’s solution, 10% dextran sulfate, 50 mm DTT, and 1 mg/ml yeast tRNA] and incubated on the sections for 18 h at 55°C. The slides were then washed and treated with RNase A before being coated with autoradiographic emulsion and exposed to film for 21 days at 4°C.

Results

Expression of BNIP3 Protein in Human Cell Lines.

The 18 human cell lines described above, representing epithelial tumors, sarcomas, endothelium, lymphocytes, and macrophages, were subjected to 0.1% hypoxia or normoxia for 16 h before making protein extractions. Western blot analysis for BNIP3 revealed that an increase of both the M_r 30,000 and 60,000 forms of BNIP3 protein occurred after hypoxia in 15 of the 18 cell lines (results from 6 cell lines are shown in Fig. 1A ⇓ ). The induction was concordant for both protein forms. Of these cell lines, 12 were derived from carcinomas (SKBR, T47D, MDA468, MDA231, MCF7, SKOV3, EJ, MKN45, C32, LS174T, RCC4, and DU145), 2 from macrophages (U937 and THP-1) and 1 from endothelial cells (HUVEC). The RZM (lymphocyte) and HT1080 (fibrosarcoma) cell lines expressed very low levels of BNIP3 protein under normoxic conditions, but protein levels did not increase in response to hypoxia (Fig. 1A) ⇓ . Additional analysis of the MCF7 cell line revealed that BNIP3 protein production was up-regulated within 8 h of exposure to hypoxia and persisted for ≥24 h in the continued presence of hypoxia (data not shown).

Hif-1α-dependent Induction of BNIP3 Protein.

The RCC4 cell line, which is derived from renal carcinoma cells defective for VHL, showed constitutive up-regulation of BNIP3 under normoxic conditions (Fig. 1B) ⇓ . When VHL was reintroduced into this cell line, levels of BNIP3 in normoxia were suppressed and became inducible by hypoxia (Fig. 1B) ⇓ . To additionally check the role of Hif-1α in BNIP3 induction, CHO KA13 and C4.5 cells were subjected to normoxia or hypoxia for 16 h before being analyzed immunohistochemically for BNIP3 protein. Neither cell line expressed BNIP3 protein under normoxic conditions, although after treatment with hypoxia, BNIP3 expression was markedly increased in the C4.5 cells. The KA13 cells, which are defective for Hif-1α, showed little, if any, increase in BNIP3 protein after hypoxia (Fig. 2) ⇓ .

Expression of NIX and BNIP3 mRNA in Human Cell Lines.

The HT1080 cell line and 4 of the cell lines showing increased protein levels under hypoxia (EJ, T47D, MDA231, and MDA468) were subjected to normoxia or hypoxia for 16 h before extracting total RNA. Northern blot analysis identified transcripts of 4.5 kb and 1.4 kb (NIX) and 5 kb and 1.7 kb (BNIP3), which, in concordance with the BNIP3 protein expression, increased under hypoxia in all of the cell lines except HT1080 (Fig. 3) ⇓ . The increase in expression was concordant for both transcripts of NIX and BNIP3.

Expression of NIX and BNIP3 mRNA in Human Normal Breast and Breast Tumor Tissue.

RNA was extracted from tumor and distant normal breast obtained from mastectomy samples and analyzed for NIX and BNIP3 expression using Northern blotting. Levels of NIX and BNIP3 mRNA were higher in the tumor samples compared with normal tissue in 5/5 and 3/5 cases respectively (Fig. 3) ⇓ . The expression of BNIP3 mRNA in the remaining 2 cases was unchanged. Both transcripts of NIX mRNA were present in human tissue, but the 5-kb transcript of BNIP3 was only detected in the tumor sample of 1 of the cases.

Localization of BNIP3 mRNA in Human Tumors.

To localize BNIP3-expressing cells in human tissue, formalin-fixed blocks of normal breast and tumor from 2 of the patients described above, as well as blocks from various other types of human epithelial tumors, were subjected to in situ hybridization analysis. BNIP3 mRNA was detectable in 5/9 tumors, consisting of 1/2 SSC head and neck carcinomas, 1/2 ovarian carcinomas, 1 pancreas carcinoma, and 2 breast carcinomas. BNIP3 expression was not detected in 1 lung carcinoma, 1 lymphoma, and 2 case of normal breast tissue. BNIP3 mRNA was expressed on epithelial carcinoma cells in perinecrotic areas of the tumor (Fig. 4) ⇓ in all of the samples except 1 breast carcinoma, where hybridization was seen in epithelial cells from a well-vascularized area of the tumor (data not shown). No specific hybridization was detected when the sections were hybridized with the sense control probe for BNIP3 (Fig. 4) ⇓ .

Discussion

In this study we have demonstrated that the cell death factors BNIP3 and NIX are hypoxically inducible in a wide range of human epithelial, endothelial, and macrophage cell lines but not in lymphocyte or fibrosarcoma cell lines. This result has been confirmed at the protein level for BNIP3 and at the mRNA level for both BNIP3 and NIX. Time course studies in a breast carcinoma cell line indicated that BNIP3 protein markedly increased after 8 h of hypoxia; this result demonstrates that induction is relatively rapid in tumor cells and contrasts with a recent study where BNIP3 protein was only detectable in a CHO-K1 cell line after 4 days of hypoxic culture (13) .

It is probable that the hypoxic induction of BNIP3 in human cells is mediated via Hif-1α, because RCC4 cells lacking wild-type pVHL have high levels of BNIP3 protein under both normoxic and hypoxic conditions. Reintroduction of pVHL to this cell line restores degradation of Hif-1α under normoxic conditions (6) and reduces BNIP3 expression. In addition, BNIP3 protein is not markedly induced under hypoxic conditions in CHO cells defective for Hif-1α. These results confirm recent data that suggested a HIF-1-dependent response based on mutational analysis of the BNIP3 promoter (13) .

Importantly, our study has also demonstrated that mRNAs encoding NIX and in most cases BNIP3 are expressed at higher levels in clinical material from human breast tumors when compared with normal breast tissue. This result is consistent with up-regulation of the HIF-1 pathway in human tumors but somewhat surprising in light of other studies that have shown down-regulation of BNIP3 in keloid cells compared with normal tissue (18) and human T-cell leukemia virus type I injected cells (19) . These differences may relate to different patterns of microenvironmental hypoxia. In our material, in situ hybridization analysis of RNA expression in human tumors revealed that it is expressed by perinecrotic areas of tumor, which result from hypoxic stress.

Areas of necrosis are commonly found in solid tumors and correlate with poor prognosis. Also, cell death by necrosis is seen more commonly than apoptosis in hypoxic tumors. BNIP3 activates caspase-independent necrosis-like cell death as a consequence of opening the mitochondrial permeability transition pore (20) and may be the pathway mediating hypoxia-induced necrotic cell death in cancer. HIF-1 activation regulates many pathways advantageous to tumor growth such as angiogenesis, glycolysis, and glucose uptake (5) , although our results suggest that activation of HIF-1 during the evolution of cancer also coselects pathways such as BNIP3/NIX that have the potential for antitumor effects.

Most tumor cells retain the ability to undergo apoptosis in response to hypoxic stress (1) , although paradoxically, this loss of apoptotic-sensitive cells leads to the selection of viable cells that are more resistant to treatment and contribute to tumor relapse (3) . Hockel et al. (21) have determined that a subset of hypoxic cervical carcinomas have a low apoptotic index and that these tumors are highly aggressive. The mechanism by which hypoxia selects for cells resistant to apoptosis is unclear. Striking up-regulation of the BNIP3/NIX gene products by hypoxia and enhanced expression in clinical tumors suggests that additional analysis of this pathway in normal and tumor tissue may be helpful in understanding this important process.

Note Added in Proof

Since this work was submitted, Guo et al. (22) have demonstrated the induction of BNIP3 by hypoxia in rat cardiomyocytes. This shows that this pathway can also be regulated by hypoxia in non-carcinoma tissue.