Chelation of cellular calcium modulates hypoxia-inducible gene expression through activation of hypoxia-inducible factor-1alpha.

Hypoxia-Inducible Factor-1 (HIF-1) is the key transcription factor in control of the expression of hypoxia-inducible genes needed by cells to adapt to decreased oxygen availability. Herein, we investigated the HIF-1alpha-mediated gene expression of carbonic anhydrase 9 (CA9) in response to hypoxia and changes of intracellular calcium levels in the neuroblastoma cell line SH-SY5Y. Decreasing the intracellular calcium level by BAPTA (1,2-bis(2-aminophenoxy)ethane-N,N,N',N'-tetraacetic acid) induced HIF-1alpha nuclear accumulation and enhanced HIF-1 DNA binding within 1 h of incubation. Like hypoxia, BAPTA stimulated HIF-1-dependent transcription by increasing the activity of the C-terminal transactivation domain of HIF-1alpha and greatly enhanced expression of the HIF-1 target gene CA9. Detailed analysis of HIF-1alpha accumulation revealed that BAPTA attenuated the interaction of HIF-1alpha with von-Hippel-Lindau protein thus decreasing proteasomal degradation of HIF-1alpha. Knock down of HIF-1alpha mRNA and protein by small interference RNA for HIF-1alpha revealed that both hypoxia and the BAPTA-induced gene expression of CA9 were strictly dependent on HIF-1alpha. In contrast, elevation of cytosolic calcium level by thapsigargin reduced the BAPTA-mediated effects. Measurements of intracellular calcium under hypoxia revealed a change in the cellular calcium distribution. BAPTA-dependent induction of HIF-1 activity was not caused by its in vitro capability to chelate iron. Instead, effective chelation of cellular calcium caused the accumulation of HIF-1alpha protein through inhibition of HIF-prolyl hydroxylases and activated HIF-1-dependent gene expression under normoxic conditions.

Hypoxia induces a variety of genes that act in concert to facilitate the supply of oxygen and nutrients and to promote cell growth and survival. Hypoxia-inducible factor-1 (HIF-1) 1 is of central importance for control of the expression of most of these genes (1). HIF-1 is a heterodimer consisting of one of three ␣-subunits (HIF-1␣ and isoforms HIF-2␣ or HIF-3␣) and the ␤-subunit (HIF-1␤, also called aryl hydrocarbon receptor nuclear translocator or ARNT) (2)(3)(4)(5). HIF-1␤ is constitutively found in the nucleus, whereas stabilization of HIF-1␣ and nuclear accumulation are induced by hypoxia. HIF-1␣ is constitutively synthesized but sent to destruction by the ubiquitinproteasome pathway in normoxia (6 -8). This process is mediated by binding of the von-Hippel-Lindau protein (pVHL) (9), which is the substrate-recognizing component of an E3 ubiquitin ligase complex (10 -12). Oxygen dependence of this process depends on the post-translational hydroxylation of HIF-1␣ at proline residues 564 and 402 by prolyl hydroxylases (13,14). Activity of these prolyl hydroxylases (PHDs) requires oxygen. Thus, under hypoxic conditions when prolyl hydroxylation ceases, HIF-1␣ is not recognized by pVHL and evades degradation. In addition, trans-activity of HIF-1␣ is regulated by oxygen-sensitive hydroxylation of the asparagine residue 803 by a related asparagine hydroxylase called FIH-1 (15). Under hypoxic conditions, HIF-1␣ accumulates, translocates into the nucleus, and forms an active HIF-1 complex that turns on expression of target genes such as vascular endothelial growth factor (VEGF), erythropoietin, glucose transporter-1, and carbonic anhydrase (CA9) (16,17).
In addition to this well established O 2 -dependent activation mechanisms of HIF-1␣, a role for changes in intracellular calcium levels in hypoxia-induced gene expression has been postulated. After prolonged hypoxic incubation, an increase in cytosolic calcium concentration was observed in some cell lines (18 -20) that was due to the release of calcium from intracellular stores (21,22). Increased calcium concentrations have been suggested to be involved in hypoxia-induced expression of tyrosine hydroxylase gene (23)(24)(25), VEGF, and NDRG-1/Cap43 (26,27). But unlike hypoxia or hypoxia-mimicking agents, the elevation of intracellular calcium concentration neither induced HIF-1␣ protein nor stimulated HIF-1-dependent transcription (27). Instead, increased intracellular calcium activated VEGF expression through an AP-1-dependent pathway during hypoxia (26,27). Recently Mottet et al. (20) reported that elevated calcium levels after prolonged hypoxia increased extracellular signal-regulated kinase 2/1 (ERK 2/1) activation and increased HIF-1 transcriptional activity but did not induce HIF-1␣ accumulation. However, HIF-1␣ accumulation and activation is instantaneous upon the onset of hypoxia (28), and it has not yet been clarified whether changes in intracellular calcium level play a role in this early phase of hypoxia. Herein, we studied the effect of decreased intracellular calcium level in SH-SY5Y neuroblastoma cells on HIF-1 activation and expression of a typical HIF-1 target gene, carbonic anhydrase 9, under normoxic and hypoxic conditions.
Cell Culture Experiments and Hypoxic Induction-SH-SY5Y neuroblastoma cells (American Type Culture Collection, Manassas, VA) were kept in RPMI 1640 medium (BioWhittaker, Cambrex, Verviers, Belgium) supplemented with 10% fetal calf serum, 100 units/ml penicillin, and 100 units/ml streptomycin in a normoxic atmosphere of 5% CO 2 (by vol.) in air. Renal clear cell carcinoma cells RCC4 were grown in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal calf serum, 100 units/ml penicillin, and 100 units/ml streptomycin. For cell culture experiments, cells were either exposed to normoxia or placed in a hypoxic incubator with 3% O 2 , 92% N 2 , 5% CO 2 (Heraeus incubator, Hanau, Germany) for variable periods of time. Severe hypoxia was established by using anaerobic-culture jars with hydrogen-and CO 2 -generating envelopes (BD Biosciences). For manipulation of intracellular calcium levels, cells were incubated in medium supplemented with either 5 M BAPTA-AM (stock solution: 50 mM in Me 2 SO) or 10 nM thapsigargin (stock solution: 1 mM in Me 2 SO) for the indicated periods of time.
Nuclear Extract Preparation and Electrophoretic Mobility Shift Assays-Nuclear protein extracts were prepared from 60-mm dishes of subconfluent cells by using the method of Schreiber (29). Doublestranded oligonucleotides containing the HIF-1 binding site from the hypoxia response element (HRE; 5Ј-GCC CTA CGT GCT GTC TCA-3Ј) of the erythropoietin 3Ј enhancer were end-labeled with [␥-32 P]dATP (ICN, Munich, Germany) using T4 polynucleotide kinase (Fermentas, St. Leon-Rot, Germany). DNA-protein binding reactions were carried out in a total volume of 20 l containing 5 g of nuclear extract, 30 fmol of 32 P-labeled oligonucleotides, and a nonspecific competitor (50 ng of calf-thymus DNA) in binding buffer (60 mM KCl, 1 mM EDTA, 1 mM dithiothreitol, 12 mM HEPES, 4 mM Tris, pH 7.9) for 30 min at room temperature. 1 g of monoclonal anti-HIF-1␣ antibody (BD Transduction Laboratories) was added for supershift detection of the HIF-1 complex and incubated overnight at 4°C. The products were analyzed by electrophoresis in 5% nondenaturing polyacrylamide gels. The dried gels were exposed to x-ray films.
Plasmid Constructs-The pH3SVL plasmid, containing six HIF-1binding sites from the transferrin 3Ј enhancer controlling the pGL3 vector (Promega, Heidelberg, Germany) was generously provided by Roland Wenger (30). The empty vector without HIF-1-binding sites (pGL3 promoter or promoter/enhancer vector; Promega) served as a control to exclude unspecific activation of the reporter gene. To measure the activation of HIF-1␣ by its C terminus the Gal4 chimeric activator/ reporter system was used in which the HIF-1␣ 775-826 C-terminal transactivation domain (C-TAD) is fused to the Gal4 DNA binding domain. The cytomegalovirus promoter ensures expression of the fusion protein. The reporter plasmid contains the luciferase gene under the control of Gal4 binding sites. Both plasmids of the chimeric activator/ reporter system were generously provided by Peter Ratcliffe (31).
Cell Transfection and Reporter Gene Assays-1 ϫ 10 7 SH-SY5Y cells were transfected with 10 g of pH3SVL plasmids or control plasmid (Promega) or with 5 g of Gal4-C-TAD and 15 g of GAL4-Luc plasmids together by electroporation method at 975 microfarads, 250 V, in 0.4-mm thick cuvettes using a Gene Pulser and Capacitance Extender apparatus (Bio-Rad). After recovering, the cells were split in 18 aliquots and grown in 6-well dishes overnight. After treatment with BAPTA-AM or thapsigargin under normoxic or hypoxic atmosphere for 4 or 6 h, cells were lysed in a buffer made of 2 mM dithiothreitol, 2 mM EDTA, 10% glycerin, 1% Triton X-100, 25 mM Tris, pH 7.8. Luciferase activity was determined with the luciferase assay system (Promega) and normalized to total cellular protein. Luciferase activity of the control vector without HIF-1 binding sites was not affected by hypoxia, BAPTA-AM, or thapsigargin treatment (data not shown). All transfections were at least done in three separate culture dishes; data are expressed as means Ϯ S.D.
Reverse Transcription and Quantitative Real Time PCR-Total RNA was extracted with the guanidinium isothiocyanate method as described previously (32). 1 g of total RNA was reverse transcribed with oligo(dT) and Moloney murine leukemia virus reverse transcriptase (Promega). Human CA9 cDNA and cDNA of the housekeeping gene 60 S acidic ribosomal protein were quantified by real time PCR using the qPCR TM Mastermix for SYBR Green I (Eurogentec, Verviers, Belgium) and the GeneAmp5700 sequence Detection System (PE Biosystems, Foster City, CA). The PCR reactions were set up in a final volume of 25 l with 0.5 l of cDNA, in 1ϫ reaction buffer with SYBR Green I, 10 pmol of forward and 10 pmol of reverse primer. The primers used were as follows: human CA9, forward 5Ј-CAC GTG GTT CAC CTC AGC AC-3Ј and reverse 5Ј-CAG CGA TTT CTT CCA AGC G-3Ј; human 60 S acidic ribosomal protein: forward: 5Ј-ACG AGG TGT GCA AGG AGG GC-3Ј and reverse: 5Ј-GCA AGT CGT CTC CCA TCT GC-3. The PCR amplification profile was as follows: 10 min at 95°C followed by 30 cycles 15 s at 95°C and 1 min at 60°C. Agarose gel electrophoresis, purification, and DNA sequencing confirmed the identity of the PCR products. 1:10 dilutions of purified PCR products starting at 1 pg to 0.1 fg were used as standards. The quantified cDNA of the CA9 was normalized to cDNA of the 60 S acidic ribosomal protein and expressed as normalized CA9 mRNA level. All RT-PCRs were done in triplicate from RNA from three separate culture dishes. Data are the means Ϯ S.D.
RNA Interference Treatment-For HIF-1␣ siRNA experiments 1 ϫ 10 7 SH-SY5Y cells were electroporated with 20 M double-stranded antisense oligonucleotides specific for HIF-1␣ mRNA (antisense: 5Ј-ACA CAC UGU GUC CAG UUA G-3Ј; sense: 5Ј-CUA ACU GGA CAC AGU GUG U-3Ј; GenBank TM accession number: HIF-1␣ NM_001530) or scrambled sequence siRNA control duplexes II (MWG-Biotech, Ebersberg, Germany). Control cells were electroporated without oligonucleotides under the same conditions. After electroporation the cells were split in 15 aliquots and grown in 6-well dishes (final concentration of siRNA oligonucleotides: 200 nM) for 24 h and then exposed to normoxic or hypoxic atmosphere for the indicated time. Cells were lysed, and total RNA or protein was extracted as described above.
In Vitro Protein Interaction Assay-The impact of BAPTA on recombinant HIF-1␣ prolyl hydroxylase 2 (PHD2) was studied by the in vitro protein interaction assays as described previously (33). The assay makes use of the interaction between the C-terminal oxygen degradation domain of HIF-1␣ (ODD, amino acids 549 -582) and pVHL, which requires the presence of hydroxyproline in the ODD. To test hydroxylasedependent effects of BAPTA on HIF-1␣-pVHL interaction in normoxia, we produced GalDBD-HIF-1␣549 -582, PHD2, and 35 S-labeled pVHL in T7-coupled rabbit reticulocyte lysate in vitro transcription/translation system (Promega). GalDBD-HIF-1␣549 -582 was purified with GalDBD antibodies conjugated to agarose beads (Santa Cruz Biotechnology) and suspended in a reaction buffer (5 mM KCl, 1.5 mM MgCl 2 , 20 mM Tris, pH 7.5) supplemented with cofactors (1 mM ascorbate, 1 mM ␣-ketoglutarate, 20 M FeCl 2 ). GalDBD-HIF-1␣549 -582 was then incubated for 30 min with BAPTA hexapotassium salt (final concentration, 10 -500 M) or either with 200 M CaCl 2 or FeCl 2 before recombinant PHD2 expressed in reticulocyte lysate was added. The reaction was terminated after 30 min by addition of desferoxamine mesylate (final concentration, 100 M). Streptavidine-agarose was added, and incubations went on for 16 h on an end-over-end rotator at 4°C. Unbound 35 S-pVHL was removed by washing, beads were boiled in SDS-PAGE loading buffer, and the supernatant was subjected to 15% SDS-PAGE. Bound 35 S-pVHL was detected by autoradiography. The VHL gene contains an Calcium Modulates HIF-1␣ Activity internal in-frame methionine at codon 54, thus two pVHL isoforms were regularly detected. In all experiments unprogrammed reticulocyte lysates served as negative controls. To test for hydroxylase-independent effects of BAPTA during protein interactions, we incubated a synthetic peptide resembling the biotinylated hydroxyproline ODD (Biosyntan, Berlin, Germany) with 35 S-labeled pVHL expressed in rabbit reticulocyte lysate.
Immunofluorescence Studies-SH-SY5Y cells were grown on poly-Dlysine-coated glass coverslips in 24-well dishes overnight. Subconfluent cells were fixed by ice-cold methanol/acetone (1:1) for 5 min on ice and blocked with 3% bovine serum albumin in PBS. As primary antibody the mouse monoclonal anti-HIF-1␣ (1:50 dilution, BD Transduction Laboratories) and as secondary antibody an Alexa-568-conjugated goat anti-mouse IgG (1:400, Molecular Probes) antibody was used. Coverslips were mounted on the slides with Mowiol (Calbiochem). Visualization was performed with a Nikon E1000 microscope (Nikon, Dü sseldorf, Germany), equipped with a charge-coupled digital camera (Optronics, Visitron Systems, Puchheim, Germany) and the image acquisition software EZ2000 (Coord, Utrecht, Netherlands). Fluorescence emission was encoded in gray levels shading from black (low intensity of emission) over gray to white (high intensity of emission).
Determination of the Effect of Intracellular Chelatable Iron on the Fluorescence of OG-BAPTA and PG SK by Laser Scanning Microscopy-SH-SY5Y cells were cultivated for 24 h on poly-D-lysine-coated glass coverslips. Cells grown on coverslips were transferred to a modified Pentz chamber, which was described earlier (34).  (35). Afterward, the cells were washed twice with HBSS and incubated for 5 min (experiments with OG-BAPTA-AM) or 15 min (experiments with PG SK) in the absence of the probes, respectively, and subsequently covered with HBSS (37°C). Fluorescence was measured with a laser-scanning microscope (LSM 510, Zeiss, Oberkochen, Germany) equipped with an argon and a helium/neon laser. The objective lens was a 63ϫ numerical aperture 1.40 Plan-Apochromat. The pinhole was set at 118 m, giving confocal optical slices of about 1.0 m in thickness. Green fluorescence of OG-BAPTA-AM and PG SK, respectively, excited at 488 nm using the argon laser power supply at 6.75 milliwatts was collected through a 505-nm long-pass filter at 10-min intervals. Image processing and evaluation were performed using the "Physiology Evaluation" software of the LSM 510 imaging system (36). The intracellular level of chelatable iron was manipulated 5-10 min after the beginning of the measurements by addition of the highly membrane-permeable FeCl 3 :8-hydroxyquinoline complex (8-HQ, 10.0 M) or membrane-permeable iron chelators (2.0 mM 1,10-phenanthroline, 5 mM 2,2Ј-dipyridyl, or 10 mM desferoxamine mesylate, Desferal®) to the supernatant (35,37).
Determination of the Effect of Chelatable Iron Ions on the Fluorescence of OG-BAPTA in Cell-free Systems-The effect of Fe 2ϩ and Fe 3ϩ on the fluorescence of OG-BAPTA (Oregon Green® 488 BAPTA-1, hexapotassium salt) was determined in two different media: 1) a simple buffered solution containing 2 mM ascorbate (freshly prepared) in imidazole buffer (10 mM), pH 7.2, and 2) a medium designed to simulate the composition of the cytosol (37). This "cytosolic" medium, which is rich in physiological iron-chelating ligands and reductants, contained 100 mM KCl, 5 mM Na 2 HPO 4 , 2 mM MgCl 2 , the full amino acid composition contained in minimum essential medium Eagle (from Sigma), 6.85 mM glucose, 1.5 mM lactate, 230 M citrate, 138 M pyruvate, 2.99 mM inorganic phosphate, 4 mM ATP, 4.5 mM glutathione, and 2 mM ascorbate (both freshly prepared), and 10 mM imidazole buffer, pH 7.2 (37). All components of the media were dissolved in double-distilled water. The final media were treated with Chelex 100 to minimize their heavy metal contamination. Aliquots (3 ml) of the "cytosolic" medium, the simple buffered solution or "interaction medium" (containing reaction buffer, cofactors, and reticulocyte lysate) from the in vitro HIF-1␣-pVHL interaction assay (diluted 1:100) were transferred into a modified Pentz chamber (37°C), and OG-BAPTA (5.0 M final concentration) was added and mixed with the media. The fluorescence of OG-BAPTA was determined as described for the cellular measurements (but at 30-s intervals) using the laser-scanning microscope. The quenching effect of Fe 2ϩ on the fluorescence of the dye was determined by adding 50 M ferrous ammonium sulfate from a freshly prepared stock solution (1 mM) in distilled water supplemented with 20 mM ascorbate (35). Fe 3ϩ was added as FeCl 3 or as ferrous ammonium sulfate from a normoxic stock solution (1.0 mM) without ascorbate. Due to the rapid autoxidation of Fe 2ϩ ions in aqueous solutions at neutral pH, this solution exclusively contained ferric iron ions as confirmed using the ferrous oxidation-xylenol orange assay (38). Experiments with Fe 3ϩ were performed in the absence of the reducing equivalents ascorbate and glutathione within the media.
Ca 2ϩ Measurements in SH-SY5Y Cells by OG-BAPTA-AM Fluorescence Microscopy-SH-SY5Y cells were grown to 80% confluence on poly-D-lysine-coated glass coverslips in 6-well dishes overnight and placed in a modified Pentz chamber. Cells were loaded with the calcium indicator Oregon Green® 488 BAPTA-1-AM (OG-BAPTA-AM, final concentration 5 M) in RPMI 1640 medium supplemented with 10% fetal calf serum, 100 units/ml penicillin, and 100 units/ml streptomycin for 20 min at 37°C. Cells were then washed twice and incubated for another 10 min. During the measurements performed by laser scanning microscopy (see above), the Pentz chamber was perfused with humidified, prewarmed (37°C) gas mixtures to achieve normoxic (21% O 2 , 74% N 2 , and 5% CO 2 ) or hypoxic (95% N 2 and 5% CO 2 ) conditions. Simultaneously to the intracellular OG-BAPTA fluorescence, the O 2 concentration in the medium was measured by using a Universal Oxygen Monitor (LICOX MCB, GMS, Mielkendorf, Germany).
Cell Viability-For the fluorescence microscopic studies, the uptake of the vital dye propidium iodide (5 g/ml) was routinely determined either during or at the end of the experiments to detect loss of cell viability. The red fluorescence of propidium iodide excited at 543 nm was collected through a 560-nm long-pass filter when laser scanning microscopy was used; using digital fluorescence microscopy, propidium iodide was detected at exc. ϭ 535 Ϯ 17.5 nm and em. Ն 590 nm. Toxicity of BAPTA-AM and thapsigargin was excluded as judged from the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay (39).
Statistics-All experiments with SH-SY5Y cells were repeated at least three times with cells from different subcultures; experiments in a cell-free system were repeated at least twice. Cellular microfluorographs and traces shown in the figures are representative for all the corresponding experiments carried out. The results are expressed as means Ϯ S.D.

RESULTS
Effects of BAPTA and Thapsigargin on the HIF-1␣ Accumulation-In normoxic SH-SY5Y neuroblastoma cells HIF-1␣ protein was absent. Hypoxia-induced HIF-1␣ accumulation was detectable after 30 min and was maximal after 4 h. Application of the calcium chelator BAPTA induced HIF-1␣ protein levels in normoxia and enhanced HIF-1␣ protein accumulation under hypoxic conditions. Maximum HIF-1␣ protein induction was observed after 1 h of incubation with BAPTA-AM (Fig. 1). In contrast, elevation of intracellular calcium using thapsigargin diminished hypoxia-induced HIF-1␣ protein levels after incubation from 0.5 to 4 h (Fig. 1).
Because BAPTA-induced HIF-1␣ accumulation was maximal after 1 h, we next investigated the effects of BAPTA-AM, thapsigargin, or both under normoxic and hypoxic conditions at this time point (Fig. 2). Western analysis of whole cell lysate showed that the HIF-1␣ protein expression was maximally stimulated under hypoxia plus BAPTA. Additional treatment with thapsigargin reduced the BAPTA-induced HIF-1␣ protein accumulation under normoxia and hypoxia ( Fig. 2A). Immunofluorescence was performed to study nuclear translocation of HIF-1␣ by BAPTA-AM treatment. As shown in Fig. 2B, BAPTA-induced HIF-1␣ nuclear accumulation under normoxic conditions and increased the hypoxic effect, which as expected was decreased by treatment with thapsigargin. These experiments clearly indicate that chelation of calcium induced HIF-1␣ protein accumulation and nuclear localization, whereas elevation of cytosolic calcium by inhibition of the Ca 2ϩ -ATPase attenuated this effect.
BAPTA Increases HIF-1␣ Levels by Inhibiting Its Degradation-Under normoxia BAPTA induced HIF-1␣ accumulation despite the presence of cycloheximide, an inhibitor of translation that prevents the de novo synthesis of HIF-1␣ (Fig. 3A) and stabilized HIF-1␣ during re-oxygenation (Fig. 3B). To test whether BAPTA affected proteasomal degradation of HIF-1␣, we compared the effect of BAPTA on HIF-1␣ with the one of the membrane-permeable proteasomal inhibitors, MG132. Inhibition of the proteasome by MG132 resulted in accumulation of high molecular bands that represent polyubiquitinated HIF-1␣ (33). Neither BAPTA nor hypoxia induced a similar accumulation of polyubiquitinated HIF-1␣ (Fig. 3C). In addition, MG132 inhibited proteasomal degradation of the short-lived p53, whereas BAPTA and hypoxia hardly affected p53 (Fig. 3C). This effect is most likely due to the stabilizing interaction of p53 with HIF-1␣ (40). Finally, BAPTA did not further increase HIF-1␣ in renal clear carcinoma cells lacking functionally active pVHL (Fig. 3D). Therefore, the accumulation of HIF-1␣  1. Effect of BAPTA and thapsigargin on HIF-1␣ protein accumulation. SH-SY5Y cells were exposed to normoxic (21% O 2 ) or hypoxic (3% O 2 ) conditions in the presence or absence of the Ca 2ϩ chelator BAPTA-AM (5 M) or the inhibitor of the endoplasmatic Ca 2ϩ ATPase thapsigargin (10 nM) for the indicated periods (0.5-4 h). Whole cell extracts (50 g) were used for Western blots for HIF-1␣ and ␣-tubulin as a loading control as described under "Experimental Procedures." Data are representative for comparable results from at least three separate experiments each.
induced by BAPTA under normoxic conditions suggests an inhibition of the HIF-1␣ degradation pathway mediated by PHDs/pVHL.
BAPTA Decreases HIF-1␣ Prolyl Hydoxylase Activity in Vitro-To study the hypothesis, that the very rapid accumulation of HIF-1␣ induced by BAPTA under normoxic conditions results from an inhibition of prolyl hydroxylation and subsequent pVHL-dependent proteasomal degradation, as it is known from hypoxic stabilization of HIF-1␣ protein (14,41), recombinant PHD2 was incubated with GalDBD-HIF-1␣549 -582 in the presence of BAPTA. Free Ca 2ϩ in the reaction mixture of the interaction assay was 5.64 Ϯ 0.25 M measured according to a previous study (42). Calculation of the free Ca 2ϩ in the presence of different BAPTA concentrations using the stochastic simulation method (43) revealed a similar free Ca 2ϩ in the reaction mixture with 200 M BAPTA as in whole cells treated with 5 M BAPTA-AM (data not shown). BAPTA significantly inhibited the activity of PHD2 under aerobic conditions, as indicated by the largely attenuated pVHL capture of GalDBD-HIF-1␣549 -582 (Fig. 4). The inhibitory effect on PHD2 was found to be dose-dependent (Fig. 4A) and was attenuated by the addition of Ca 2ϩ (Fig. 4B). In contrast, the addition of Fe 2ϩ did not attenuate the inhibition of PHD2 by BAPTA (Fig. 4C) indicating that the effect of BAPTA did not result from iron chelation (see also below). To exclude direct effects of BAPTA on the interaction of HIF-1␣ with pVHL, a synthetic 19-mer hydroxyproline564-ODD was incubated with pVHL. BAPTA, however, had no effect on this interaction (Fig. 4D).
BAPTA Increases the HIF-1 Binding to a Hypoxia Response Element-To test for activation of HIF-1␣, DNA-binding activity of the HIF-1 complex was studied using nuclear extracts from BAPTA-AM-treated cells (Fig. 5). BAPTA induced HIF-1 DNA-binding activity under normoxic conditions and increased DNA binding under hypoxia (complex i in Fig. 5). In contrast, thapsigargin failed to induce HIF-1-binding activity under normoxia and decreased BAPTA-induced DNA binding. To confirm the presence of HIF-1␣ in the DNA-bound complex, supershift experiments were performed with monoclonal HIF-1␣ antibodies that shifted the complete DNA-protein complex to a higher molecular mass (complex ss in Fig. 5).
BAPTA Enhances the Transcriptional Activity of HIF-1-Transcriptional activity of the BAPTA-AM-induced HIF-1 complex was determined in reporter gene assays using the luciferase gene under the control of six HIF-1-binding sites from the transferrin enhancer. After transient transfection, SH-SY5Y cells were incubated in normoxic or hypoxic atmosphere in the presence or absence of 5 M BAPTA-AM, 10 nM thapsigargin or both for 4 h (Fig. 6A). BAPTA induced a 2.6-fold increase in luciferase activity compared with normoxic controls. Hypoxia alone increased luciferase activity 3.1-fold, whereas BAPTA plus hypoxia showed a 27-fold increase over normoxic controls. Thapsigargin greatly attenuated the stimulatory effect of BAPTA-AM treatment and reduced hypoxic activation. To dissect whether the increase was solely dependent on HIF-1␣ accumulation (Fig. 1) or whether the trans-activity of HIF-1␣ was also influenced by BAPTA, the effect of BAPTA on the C-TAD of HIF-1␣ was tested. BAPTA stimulated HIF-1␣ trans- activity under hypoxia 6.9-fold compared with untreated control, whereas hypoxia stimulated C-TAD activity 2.7-fold. Again thapsigargin reduced the stimulatory effect of BAPTA treatment (Fig. 6B). The data indicate that in addition to accumulation of HIF-1␣, BAPTA also increases the activation of the transcription factor HIF-1.
Hypoxia-and BAPTA-induced Expression of CA9 Depends on HIF-1␣ Activation-To determine whether HIF-1 activation by BAPTA affects expression of HIF-1-regulated genes, changes in mRNA levels for carbonic anhydrase 9 (CA9) that are reported to be strictly regulated by HIF-1 (44) were studied. Using quantitative real time PCR analysis, we found hypoxia-inducible CA9 mRNA expression in SH-SY5Y that was greatly amplified by BAPTA treatment both under hypoxia and normoxia (Fig. 7).
To prove the critical role of HIF-1␣ in the up-regulation of CA9 gene expression by hypoxia or BAPTA in SH-SY5Y cells, HIF-1␣ mRNA and protein were reduced by siRNA treatment against HIF-1␣ (Fig. 8A). siRNA treatment greatly decreased the hypoxic and the BAPTA-dependent induction of CA9 mRNA levels to values close to those in normoxia (Fig. 8B). Scrambled oligonucleotides confirmed the specificity of the treatment, because they affected neither HIF-1␣ mRNA levels (Fig. 8A) nor CA9 expression.

Effects of Alterations of the Intracellular Chelatable Iron Concentration on the Fluorescence of OG-BAPTA and Phen-
Green SK-Several other chemicals, most prominently iron chelators, can mimic the BAPTA-mediated effects on hypoxic response. Although BAPTA has the highest specificity for Ca 2ϩ of all Ca 2ϩ chelators, one has to consider that BAPTA might bind Fe 2ϩ/3ϩ . To exclude that the effects of BAPTA on HIF-1␣ action are due to the chelation of intracellular iron, BAPTA was studied in SH-SY5Y cells as well as in a cell-free system for its iron-chelating properties. For these experiments, the fluorescent BAPTA-AM analogue OG-BAPTA-AM (Oregon Green 488-BAPTA-1-AM) was used and compared with the known iron indicator PhenGreen SK (35,36,45). The fluorescence of both indicators was homogenously and comparably distributed inside the cells. The chelation of iron by a fluorescent indicator always leads to a dynamic fluorescence quenching, because both the quantum yield of the indicator as well as the life span of fluorescence are decreased (for review see Ref. 35). Therefore, strong non-fluorescent iron chelators (which remove indicator-bound iron and thus "dequenche" the indicator's fluorescence (35)) were added to OG-BAPTA-AM-loaded SH-SY5Y cells to study whether the indicator is capable of chelating intracellular labile iron. The addition of the Fe 2ϩ chelator 2,2Ј-dipyridyl (5 mM, Fig. 9, A and B) and of other iron chelators (2.0 mM 1,10-phenanthroline or 10 mM desferoxamine mesylate; data not shown) to OG-BAPTA-AM-loaded SH-SY5Y cells had no effect on the fluorescence intensity of the indicator, indicating that no chelatable iron was bound to OG-BAPTA within these cells. Because the concentration of chelatable iron in SH-SY5Y cells may be too low to significantly quench OG-BAPTA fluorescence, we increased the concentration of the intracellular chelatable iron pool by the addition of the highly lipophilic and membrane-permeable Fe 3ϩ (1) ferrous state, did not decrease the intracellular fluorescence of OG-BAPTA (Fig. 9E). In contrast, and to our surprise FeCl 3 / 8-HQ led to a strong fluorescence increase (Fig. 9F), which most likely reflected an iron-mediated release of intracellular Ca 2ϩ (46,47). These data clearly demonstrate that OG-BAPTA is not able to complex labile iron ions but chelates Ca 2ϩ ions in SH-SY5Y cells.
In contrast to OG-BAPTA, intracellular PhenGreen SK fluorescence readily increased upon the addition of 2,2Ј-dipyridyl (5 mM) within 20 -30 min (Fig. 9, compare A and B with C and D) indicating that intracellular chelatable iron was bound to the indicator (35,36,45). Likewise, addition of the Fe 2ϩ chelator 1,10-phenanthroline (2 mM) and the Fe 3ϩ chelator desferoxamine mesylate (10 mM) increased PG SK fluorescence (data not shown). As expected and in contrast to OG-BAPTA, the intracellular fluorescence of PG SK was strongly quenched (Ϸ80% within 1 min) when the intracellular chelatable iron pool was increased by addition of FeCl 3 /8-HQ (10 M; Fig. 9, G and H). Neither loading of the cells with OG-BAPTA-AM or PG SK, the scanning procedures, nor the addition of the iron chelators and of FeCl 3 /8-HQ, respectively, was cytotoxic to SH-SY5Y cells.

Effect of Chelatable Iron Ions on the Fluorescence of OG-BAPTA in the Absence and Presence of Cellular Iron Chelating
Ligands in Cell-free Systems-Because BAPTA-AM has previously been reported to chelate both Fe 2ϩ and Fe 3ϩ (48) but herein did not affect chelatable iron within SH-SY5Y cells, we additionally studied the chelating properties of OG-BAPTA in a cell-free system. In line with Britigan et al. (48), the addition of Fe 2ϩ and Fe 3ϩ (50 M) rapidly and strongly decreased the fluorescence intensity of OG-BAPTA (5 M) in a simple buffered solution, indicating that OG-BAPTA can complex chelatable iron ions under certain conditions (Fig. 10A). Fe 3ϩ more efficiently quenched OG-BAPTA fluorescence than Fe 2ϩ , which well reflects the higher affinity of BAPTA for Fe 3ϩ as compared with Fe 2ϩ (K d ϭ 5 ϫ 10 Ϫ9 and 1 ϫ 10 Ϫ6 M (48)). Most impor-tantly, however, and in contrast to the quenching effect in a simple buffered solution, the addition of chelatable iron ions had no effect on the fluorescence of OG-BAPTA when the indicator was dissolved in a medium composed to simulate the cytosol (Fig. 10A). This indicates that the iron-binding site of OG-BAPTA cannot compete with the chelating ligands in the "cytosolic" medium (e.g. citrate, phosphate, ATP, and ascorbate), i.e. those that have been reported to form iron chelates within cells (35). In addition, we studied the chelating properties of OG-BAPTA under the conditions of the in vitro interaction assay (see Fig. 4). In contrast to the quenching effect in a simple buffered solution, the addition of chelatable iron ions had no effect on the fluorescence of OG-BAPTA in the "interaction medium" of the PHD assay containing reaction buffer supplemented with cofactors (1 mM ascorbate, 1 mM ␣-ketoglutarate, 20 M FeCl 2 ) and reticulocyte lysate. Moreover, addition of the Fe 2ϩ chelator 1,10-phenanthroline to the "interaction medium" did not increase OG-BAPTA fluorescence indicating that no intracellular chelatable iron was bound to the indicator under these conditions (Fig. 10B).
Effects of Acute Hypoxia on Intracellular Calcium Levels in SH-SY5Y Cells-Because Ca 2ϩ chelation by BAPTA mimicked hypoxic HIF-1 activation, we raised the question whether decreased O 2 concentrations affect the intracellular Ca 2ϩ level in SH-SY5Y cells. SH-SY5Y cells loaded with the fluorescent Ca 2ϩ indicator OG-BAPTA-AM were exposed to hypoxia while the intracellular indicator fluorescence and the pO 2 in the medium were simultaneously recorded. Acute exposure to hypoxia up to 120 min (3% to 0.2% O 2 ) did not change the basal overall Ca 2ϩ level of the cells under conditions (Fig. 11, A and  B) when HIF-1␣ protein accumulation was already induced. The addition of BAPTA-AM readily decreased the intracellular OG-BAPTA fluorescence, which was increased by thapsigargin, indicating that the indicator well responded to changes in intracellular Ca 2ϩ levels under the present experimental conditions (Fig. 11C). In further experiments, Ca 2ϩ measurements using the indicator OG-BAPTA or Fura-2 confirmed that total cellular Ca 2ϩ levels were not altered during 4 h of prolonged hypoxic superfusion (data not shown). However, microfluorographs with subcellular resolution revealed changes in the intracellular distribution of Ca 2ϩ ; during hypoxia Ca 2ϩ levels gradually increased within the nuclei, whereas they slightly decreased within the cytosol (Fig. 11D). DISCUSSION Calcium-dependent signaling is known to affect the expression of hypoxia-inducible genes (26,27,49). However, very little is known to what extent induction of hypoxia-inducible factor-1 (HIF-1), the master regulator of oxygen homeostasis, is affected by changes in intracellular Ca 2ϩ concentrations. Herein, for the first time we provide evidence that lowering intracellular Ca 2ϩ level can induce HIF-1␣ accumulation, activation and HIF-1 target gene expression.
Increased HIF-1␣ protein levels were obviously caused by inhibition of the PHD2 (Fig. 4) that hydroxylates HIF-1␣ and tags the protein for subsequent recognition by pVHL, polyubiquitination, and proteasomal degradation (50). Because the in vitro inhibition of PHD2 activity was also attenuated by the addition of excess Ca 2ϩ but not by iron, a role for Ca 2ϩ in the regulation of PHD2 activity has to be considered. Stochastic simulations of the Ca-BAPTA reaction under the condition of the interaction assay (43) revealed that inhibition of PHD2 occurred in the nanomolar range of free Ca (data not shown) implicating a calcium-dependent PHD activity within cells. It is unlikely that BAPTA itself simply obstructs recognition of HIF-1␣ by pVHL, because we found no effect of BAPTA on interaction of hydroxyproline-HIF-1␣ and pVHL (Fig. 4D). In- with or without 2 mM ascorbic acid) or in cytosolic medium (with or without ascorbate and glutathione; see "Experimental Procedures"), which contains physiological concentrations of intracellular iron-chelating ligands. Then the solutions (pH 7.3, 37°C) were transferred into modified Pentz chambers (37°C). The fluorescence of OG-BAPTA was recorded at 30-s intervals using laser-scanning microscopy ( exc ϭ 488 nm; em Ն 505 nm) and the same scanning parameters as for the cellular fluorescence measurements (cf. Fig. 9). After establishing the baseline fluorescence (5-10 min), either Fe 2ϩ or Fe 3ϩ (50 M) were added (arrows) from concentrated stock solutions of ferrous ammonium sulfate (1 mM; preincubated with or without 20 mM ascorbic acid) and rapidly mixed with the media. B, to study, wether BAPTA chelates iron ions under the in vitro conditions of the interaction assay, further experiments were performed using interaction medium containing reaction buffer supplemented with cofactors (1 mM ascorbate, 1 mM ␣-ketoglutarate, 20 M FeCl 2 ) and reticulocyte lysate. OG-BAPTA (5 M) was dissolved in interaction medium that has been diluted by a factor of ϫ100 with distilled water to obtain the same relations of the ingredients present within the in vitro interaction assay with 500 M BAPTA (see "Experimental Procedures" and Fig. 4). The fluorescence of OG-BAPTA was recorded as described above. After establishing the baseline fluorescence (5-10 min), either FeCl 2 or the iron chelator 1,10-phenanthroline was added (arrows) from concentrated stock solutions. The relative emission intensities are given in arbitrary units (a.u.). Traces shown are representative of at least three separate experiments.
terestingly, Oehme et al. (51) have described a putative PHD-4 that possesses an EF-hand motif indicating Ca 2ϩ -dependent regulatory functions. However, further experiments are necessary to determine whether active Ca 2ϩ -binding sites exist in PHDs and whether these binding sites influence the activity of prolyl hydroxylases. Because BAPTA also increased the activity of the C-TAD of HIF-1␣, it is very likely that FIH-1 was inhibited like the PHDs, although this was not explicitly tested in this study.
Nuclear accumulation of HIF-1␣ by BAPTA was accompanied by increased binding of the HIF-1 complex to DNA (Fig. 5) and activation of the HRE-driven reporter gene (Fig. 6). Again, this effect of BAPTA is most likely related to its Ca 2ϩ -chelating capacity, because in both assays co-treatment with thapsigar- FIG. 11. Effect of the pO 2 on the intracellular Ca 2؉ level of SH-SY5Y cells. SH-SY5Y cells were grown on coverslips and placed in a modified Pentz chamber. Cells were loaded with 5 M of the Ca 2ϩ indicator Oregon Green 488 BAPTA-1-AM (OG-BAPTA) in medium (37°C) as described above. The Pentz chamber was perfused with humidified, prewarmed (37°C) gas mixtures to achieve normoxic (21% O 2 , 74% N 2 and 5% CO 2 ) or hypoxic (0% O 2 , 95% N 2 and 5% CO 2 ) conditions after establishing the baseline of OG-BAPTA fluorescence (arrows). A, the O 2 concentration in the medium was continuously monitored with a polarographic probe. The insets in A show the HIF-1␣ nuclear accumulation observed by immunofluorescence after 120min normoxic or hypoxic gas-flow in a parallel experiment without loading the cells with OG-BAPTA. B, OG-BAPTA fluorescence ( exc ϭ 488 nm; em Ն 505 nm; in a.u., arbitrary units) was measured every 10 min by laser scanning microscopy. The insets in B show the OG-BAPTA staining of SH-SY5Y cells after 120 min of normoxic or hypoxic gas-flow. In C 10 nM thapsigargin or 50 nM BAPTA-AM was added (arrows) after establishing the OG-BAPTA baseline fluorescence of cells kept under normoxic conditions. D, the intracellular Ca 2ϩ distribution was imaged by laser scanning microscopy of OG-BAPTA loaded cells. Cells were incubated under normoxia (4 h) or under hypoxia (1% O 2 for 2 or 4 h) and then loaded with 5 M OG-BAPTA as described under "Experimental Prcoedures" under normoxic or hypoxic conditions, respectively. Scale bar in D represents 50 m. Data are the means Ϯ S.D. At some data points the S.D. is hidden by the signs. gin reduced the stimulation by BAPTA. Calcium chelation not only controlled a transfected luciferase reporter gene but also the endogenous HIF-1 target gene CA9 (Fig. 7), which is considered a typically HIF-1-regulated gene (44). The critical role of HIF-1␣ accumulation for the BAPTA effect is underlined by the loss of induction of HIF-1-dependent gene expression after HIF-1␣ expression was knocked down by siRNA. Likewise, hypoxic induction of CA9 mRNA expression was completely lost (Fig. 8), which confirms the work of Wykoff et al. (44) and justifies our choice of CA9 as a typical HIF-1-regulated gene. Moreover, the strict dependence of the BAPTA-induced CA9 expression on HIF-1 precludes a significant contribution of other transcription factors like AP-1 that have been found to be influenced by changes in intracellular Ca 2ϩ (27). Interestingly, BAPTA-induced CA9 expression had also functional consequences for the cells: during 24 h of incubation BAPTA accelerated the acidification of the culture medium indicating a higher CA9 activity as pointed out by Wykoff et al. (44). Thapsigargin attenuated the acidification of the cell culture medium as one would expect from the antagonistic on BAPTA-induced CA9 expression (data not shown).
Despite the consistency of our data, discrepancies exist with respect to earlier studies on the role of intracellular Ca 2ϩ in HIF-1 and HIF-1-dependent gene activation. Salnikow et al. (27) have reported that expression of VEGF, a well-known HIF-1-target gene, was induced by elevated Ca 2ϩ . However, this effect did not depend on HIF-1 but rather on AP-1 activity (27). Nevertheless, in another study increased VEGF secretion under hypoxia was linked to increased Ca 2ϩ levels activating HIF-1 transcriptional activity in HepG2 hepatoma cells (20). It is of note, however, that this effect was observed after 16 h of hypoxia and was attributed to elevated Ca 2ϩ , because it was reduced by BAPTA treatment at that time point. Our data demonstrate that the induction of HIF-1␣ by BAPTA occurs much earlier (Fig. 1). In fact, the early accumulation and activation of HIF-1 found here coincides with the instantaneous activation of HIF-1 after the onset on hypoxia (28) and leads to the induction and expression of the strictly HIF-1-regulated CA9 gene. This effect can be most likely attributed to the chelation of free Ca 2ϩ and was attenuated by increasing the Ca 2ϩ concentration using thapsigargin (Fig. 11).
In previous studies the Ca 2ϩ chelator BAPTA has been demonstrated to bind also Fe 2ϩ and Fe 3ϩ (48), and, consequently, protection of cells from iron-dependent injury has been attributed to the iron-chelating properties of BAPTA rather than to an effect of Ca 2ϩ chelation (48). Because BAPTA mimicked hypoxic HIF-1 induction under normoxic conditions like iron chelators, e.g. desferoxamine or 2,2Ј-dipyridyl (52), we have considered that iron chelation might have caused the effects we have observed. Indeed, we found that the BAPTA-AM derivative OG-BAPTA-AM is capable of chelating iron ions in a simple imidazole buffer as observed from the fluorescence quenching of the indicator upon addition of Fe 2ϩ and Fe 3ϩ (Fig. 10). However, in SH-SY5Y cells (Fig. 9), OG-BAPTA, unlike the well known iron indicator PG SK, did not respond to experimental changes in the concentration of the intracellular chelatable iron pool, indicating that within cells BAPTA does not complex iron ions. The inability of OG-BAPTA to complex chelatable iron within cells obviously reflects its apparently low affinity for Fe 2ϩ and Fe 3ϩ (K d ϭ 5 ϫ 10 Ϫ6 and 1 ϫ 10 Ϫ9 M (48)). Within cells, however the metabolically and catalytically reactive "chelatable iron" is bound to a variety of highly abundant organic ligands with a comparable or higher affinity for iron ions than that of BAPTA (53)(54)(55). This notion is strongly supported by experiments with the intracellular iron chelating ligand ATP and a "cytosol-like" medium in our experiments.
Under these in vivo-like conditions fluorescence of OG-BAPTA became insensitive to iron ions (Fig. 10A). Further support of a specific Ca 2ϩ -chelating effect is provided by the experiments using the "interaction medium" of the PHD assay (Fig. 10B). Collectively, HIF-1␣ induction by BAPTA most likely results from chelation of intracellular free Ca 2ϩ but not iron ions and is thus clearly different from the hypoxia mimicking effects of desferoxamine or 2,2Ј-dipyridyl.
Herein, we provide evidence that lowering intracellular Ca 2ϩ concentration activates HIF-1 through inhibition of hydroxylation of HIF-1␣. In that way, BAPTA acts like iron chelators (13) or NO donors (33) that cause accumulation of HIF-1␣ protein through inhibition of prolyl hydroxylases and thus mimic hypoxic signaling. Interestingly, our data indicate that total cellular Ca 2ϩ levels were not affected during 120 min of acute or up to 4 h of prolonged hypoxic exposure, although HIF-1␣ levels were already increased (Fig. 11). Instead we found an intracellular change in the distribution of Ca 2ϩ between the cytosolic and the nuclear compartment. Ca 2ϩ levels increased within the nuclei most prominently after 4 h of hypoxia resulting in decreased cytosolic levels, whereas the overall Ca 2ϩ did not change. Because PHD2 is mainly localized within the cytosol (56), a decrease in Ca 2ϩ availability may be a physiological factor, that modulates PHD activity. In this context, the effect of BAPTA could mimic this change in cellular Ca 2ϩ distribution by chelating Ca 2ϩ within the cytosolic compartment. On the other hand, however, PHD1 and -3 are localized within the nucleus where Ca 2ϩ may be essential for their activity during reoxygenation subsequent to hypoxia. Therefore, the Ca 2ϩ dependence of the regulation of PHDs and, consequently, of gene expression needs to be studied in future projects.