Lowered Oxygen Tension Induces Expression of the Hypoxia Marker MN/Carbonic Anhydrase IX in the Absence of Hypoxia-inducible Factor 1α Stabilization: A Role for Phosphatidylinositol 3′-Kinase¹

The tumor-associated protein CAIX⁴ is an isoenzyme of the CA family (1) whose expression was identified in a large number of human tumors but not in the corresponding normal tissue (Refs. 2 and 3 and the references therein). Although its exact role in carcinogenesis is not known, recently it was suggested that tumor-associated transmembrane CA isoenzymes (CAIX and CAXII) may facilitate acidification of the extracellular milieu surrounding cancer cells and in this way promote tumor growth and spread (3, 4). In vitro experiments with acetazolamide, a potent CA inhibitor, proved the importance of CA activity for the invasive capacity of renal cancer cells in vitro (5).

In comparison with normal tissues, solid tumors are poorly oxygenated and contain hypoxic regions. Observations that expression of the transmembrane CA CAIX is induced under hypoxic conditions in cultured cells and of expression of CAIX in hypoxic regions of human tumors are suggestive of the potentially important role of CAIX in tumor adaptation to hypoxic conditions (3, 6). By catalyzing the reversible hydration of carbon dioxide, hypoxia-inducible CAIX activity may influence tumor pH_e (3). It is widely accepted that pH_e in tumors is acidic due to increasing lactate production by glycolysis (7), and CAs may contribute to maintaining more neutral intracellular pH at the expense of lowering pH_e (3). CAIX was postulated as an intrinsic marker of tumor hypoxia, and its clinical usefulness was indicated because the scoring of its expression in clinical samples is rapid, reproducible, and simple (8).

Preliminary characterization of the CA9 promoter identified a functional HRE (6). Under hypoxic conditions, HREs are transactivated by the key mediator of hypoxic response, HIF-1 (9). However, it is now clear that a number of genes whose regulatory regions contain a HRE and are activated via stabilization of HIF-1 may also be regulated in a HIF-1-independent fashion (10, 11). In the case of CAIX, recent studies showed significant overlap between CAIX expression and regions of hypoxia in solid tumors, but expression of CAIX extended beyond the hypoxic regions (6, 12). This partial discordance raised questions about the mechanism(s) governing CAIX expression and the role of HIF-1 in regulation of CAIX expression at intermediate O₂ levels.

Relatively little is known about the molecular mechanisms controlling CAIX expression. Functional analysis of the CA9 upstream sequence suggested that the (−173/+31) region (designated as the CA9 promoter) contains the critical regulatory elements (13). Within this region, five PRs, PR1–5, were identified by DNase I protection assay under normoxic conditions (13). Detailed mutational analysis identified SP1/SP3 and AP1 as the critical factors binding PR1 and PR2, respectively (14).

A functional TACGTGCA HRE was identified immediately upstream of the CA9 transcription start in the position (−3/−10) on the antisense strand (6). HIF-1, which binds HRE, is a heterodimeric transcription factor consisting of the regulated HIF-1α and constitutive HIF-1β subunits (15). Activity of the α subunit can be increased through stabilization by either low oxygen concentration (16) or inactivation of the VHL tumor suppressor gene (17). VHL forms part of the ubiquitin-ligase complex that targets the HIF-1α subunit for oxygen-dependent proteolysis (17). In renal carcinoma cells defective for VHL, constitutive CAIX up-regulation is associated with the loss of O₂-dependent regulation, consistent with the critical function of VHL in the regulation of HIF-1 (6). In addition to stabilization of the HIF-1α protein, the HIF-1α promoter can be transactivated under apparently normoxic conditions by a number of stimuli, including cytokines (18, 19), hormones (20, 21), and growth factors (22, 23).

To increase our knowledge of the regulation of CAIX expression, we have used the experimental model of sparse and dense cultures for studying O₂-related mechanisms of inducible CAIX expression. The dependence of CAIX expression on cell density in vitro is striking: the protein is absent in sparse, rapidly proliferating HeLa cells; whereas its synthesis is induced in dense, overcrowded (albeit still proliferating) cultures that express CAIX protein in significant amounts (2). As a part of our ongoing effort to understand the molecular mechanisms governing CAIX expression, we were interested in mechanism(s) that were differentially activated in sparse/dense cultures. In this study, we have identified the role of the PI3K pathway in the induction of CAIX expression in dense HeLa cultures. We have also mapped the cis elements in the CA9 promoter that are necessary for the density-dependent induction of CAIX expression.

Numbers in parentheses indicate position relative to the CA9 transcription start. Reagent kits, enzymes, Abs, and other reagents were used according to each manufacturer’s recommendations.

pBMN5, MN6, and MN7 constructs contain the (−173/+31), (−45/+31), and (−30/+31) CA9 fragments, respectively, in the pBLCAT6 vector (24). pBMN5PR1mut and pBMN5HREmut are pBMN5 derivatives with point mutations (underlined) in the PR1 [(−45/−24); GGCTTGCTCCTAACCCACCCAG] and HRE [(−19/+8); CGTTTCCAATGCTTTTACAGCCCGTAC] CA9 regions, respectively. The (−173/+31) fragment was also cloned in the firefly luciferase pGL2 basic vector (Promega). pRL-CMV (Promega) was used as an internal control for the luciferase construct. pBLCAT5, used as a control for cell density-dependent activation, contains the herpes simplex virus thymidine kinase promoter (24). Plasmids CD2p110 and CD2p110KD expressing a constitutively active mutant of PI3K p110 catalytic subunit and its kinase-dead mutant, respectively, were kindly provided by Dr. D. A. Cantrell (Cancer Research UK, London, United Kingdom; Ref. 25).

The cervical carcinoma cell line HeLa and the fibrosarcoma cell line HT 1080 were grown in DMEM (BioWhittaker) supplemented with 10% FCS (Life Technologies, Inc.) and 0.16 mg/ml gentamicin (Sigma). The Chinese hamster ovary cell lines C4.5 (control) and Kal3.5 (HIF-1α deficient; Ref. 26), kindly provided by Dr. P. J. Ratcliffe, were grown in Ham’s F-12 medium (Life Technologies, Inc.) supplemented with 10% FCS, l-glutamine (2 mm), penicillin (50 IU/ml), and streptomycin sulfate (50 μg/ml). The effects of cell density, stirring, pH, glucose, inhibitors, hypoxia-mimic CoCl₂, and dense culture-conditioned medium on levels of endogenous CAIX were tested on the cells that had been seeded at 10,000 cells/cm² and grown for 3 days. These cells were then plated at varying densities, incubated with or without gentle stirring at 37°C, and harvested after 24 h. The effect of medium height was investigated on cells plated at 160,000 cells/cm² for 12 h. The effect of PI3K inhibitors LY294002 (Calbiochem) and wortmannin (Sigma) and the nitric oxide synthase inhibitor S-isopropylisothiourea (Alexis Corp.) was tested on cells seeded at 160,000 cells/cm² for 24 h. Controls for LY294002 and wortmannin were treated with appropriate volumes of DMSO vehicle. HIF-1α levels were tested in cells seeded at varying densities in the presence or absence of 240 μm CoCl₂ with or without stirring for the indicated intervals. Dense culture-conditioned medium was prepared by plating cells at 160,000 cells/cm², changing the medium after 24 h, and incubating cells for an additional 24 h. Conditioned medium was gently added to cells plated at 20,000 cells/cm² and incubated for 24 h.

For transient cotransfections of HeLa, C4.5, and Kal3.5 cells with the (−173/+31) CA9 fragment in pGL2 vector and pRL-CMV plasmid, Effectene Transfection Reagent (Qiagen) was used. Transfectants were treated with the indicated concentrations of LY294002 for 48 h and harvested 52 h after transfection. Luciferase (firefly and Renilla) activities were assayed with the dual-luciferase reporter assay system (Promega) in a Monolight 2010 luminometer. For mapping of the minimal cell density-dependent CA9 region, a modification of the standard transient transfection procedure (GenePORTER 2; Gene Therapy Systems) was used. After 4 h at 37°C, the cells transfected with CA9 fragments in pBLCAT6 vector were rinsed with PBS and trypsinized, and the suspension was split into halves that were seeded at 20,000 and 160,000 cells/cm² and cultivated for an additional 48 h. CAT activity was assayed with the CAT ELISA kit (Roche). CAT activities from at least three independent experiments were normalized against protein concentration determined with the BCA Protein Assay Reagent (Pierce).

For CAIX protein, total cellular protein was prepared from control and appropriately treated cells 24 h after plating by lysis with radioimmunoprecipitation assay buffer [7.5 mm phosphate buffer (pH 7.2), 140 mm NaCl, 1% Triton X-100, 0.1% sodium deoxycholate, 1 mm phenylmethylsulfonyl fluoride, 10 μg/ml leupeptin, and 10 μg/ml aprotinin] at 4°C for 10 min. For HIF-1α, approximately 450,000 cells were seeded at varying densities. Control and treated cells were scraped into 100 μl of 1× sample buffer [62.5 mm Tris (pH 6.8), 2% SDS, 10% glycerol, 50 mm dithiotreitol, and 0.01% bromphenol blue] and passed through a needle several times to decrease the viscosity of the lysate. Samples containing 30 μg of protein (for CAIX protein) or 50 μl of lysate (for HIF-1α) were electrophoretically separated by 8% SDS-PAGE and transferred onto an Immobilon-P transfer membrane (Millipore). The membrane was blocked in ST buffer [20 mm Tris (pH 7.5), 150 mm NaCl, and 0.1% Tween 20] containing 5% nonfat milk for 1 h, incubated with 0.75 μg/ml anti-CAIX monoclonal Ab M-75 (27) or rabbit anti-HIF-1α Ab (Santa Cruz Biotechnology) in blocking buffer for 1 h at room temperature, and washed three times with ST buffer for 5 min. The membrane was then incubated with horseradish peroxidase-conjugated antimouse or antirabbit Ab (1:5000; Santa Cruz Biotechnology) for 1 h at room temperature and washed three times with ST buffer for 5 min, and immune complexes were detected with SuperSignal Chemiluminiscent Substrate (Pierce). To verify equal protein loading, the membrane was reprobed with a monoclonal anti-β-actin Ab (Sigma).

HeLa cells (3 × 10⁶) were seeded at 20,000 and 160,000 cells/cm² and grown for 24 h. Cells were rinsed with ice-cold PBS, scraped, and lysed for 5 min on ice with 0.5 ml of cell lysis buffer [20 mm Tris (pH 7.5), 150 mm NaCl, 1 mm EDTA, 1 mm EGTA, 1% Triton X-100, 2.5 mm Na PP_I, 1 mm β-glycerolphosphate, 1 mm Na₃VO₄, 10 μg/ml leupeptin, and 1 mm phenylmethylsulfonyl fluoride]. The lysate was sonicated four times (5 s each) on ice and centrifuged for 10 min. Tyrosine-phosphorylated proteins were immunoprecipitated by rocking 100 μg of total protein with 2 μg of monoclonal anti-phosphotyrosine Ab (Cell Signaling Technology) at 4°C overnight. The immune complexes were bound to protein A-Sepharose CL-4B (Amersham Pharmacia Biotech) for 3 h at 4°C. The immunoprecipitates were washed twice with PBS containing 1% NP40 and 200 μm Na₃VO₄; once with 0.5 m LiCl, 100 mm Tris (pH 7.4), and 200 μm Na₃VO₄; and twice with 10 mm Tris (pH 7.4), 100 mm NaCl, 1 mm EDTA, and 200 μm Na₃VO₄. The beads were mixed with 10 μg of phosphoinositide (Sigma), [γ-³²P]ATP (0.37 MBq, 925 TBq/mmol; ICN), and 30 μl of 40 mm HEPES (pH 7.5), 20 mm MgCl₂ and incubated at 37°C for 30 min. The reaction was stopped by the addition of 40 μl of 4 m HCl and 160 μl of CHCl₃/CH₃OH (1:1). The organic phase was recovered and applied to a silica gel thin layer chromatography plate precoated with 1% potassium oxalate (Analtech). The plate was developed in CHCl₃/CH₃OH/H₂O/NH₄OH (60:47:11.3:2), dried, and autoradiographed.

It has been well established that there is a positive correlation between CAIX protein expression and increasing cell density in vitro (1, 2). Cell density-dependent induction of CAIX expression in combination with the extraordinary stability of this protein required specific conditions for the current experimental study. To prevent accumulation of CAIX that would hamper subsequent analysis, we seeded HeLa cells at 10,000 cells/cm² and cultured them for at least 3 days. This prolonged cultivation in sparse conditions allowed for the turnover of the preexisting CAIX protein without replacement by de novo synthesis of new protein. These cells were then plated at various densities, and the effect of increasing crowding on CAIX levels was tested for 24 h. As can be seen in Fig. 1, CAIX was induced in cultures that were seeded at 80,000 cells/cm² or greater density, whereas in cultures of lower density, CAIX was not detectable. Similar results were obtained with the HT 1080 fibrosarcoma cell line (data not shown). We further characterized the process of density-dependent CAIX induction by testing its dependence on serum growth factors and anchorage. High cell density induced CAIX expression even in the absence of serum, albeit with diminished efficiency. In the absence of serum, cells seeded at 160,000 cells/cm² expressed CAIX at the same level as cells seeded at 80,000 cells/cm² in the presence of serum (Fig. 1, Lane 6). However, we observed a decreased yield of total protein from the cells grown in the absence of serum, which suggests a slower proliferation of these cells. This means that by the time of harvesting, the cells in the absence of serum had reached a lower density than the control cells, and the observed lower CAIX induction thus may reflect differences in the final culture density. The fact that cell density-mediated activation of CAIX expression functions in the absence of serum confirms that this process is independent of pathways initiated by growth factors contained in serum. Cell density-mediated CAIX expression is also anchorage independent: the cells that were not allowed to attach by cultivation on agarose-coated plates expressed the same amount of CAIX as the cells cultivated under normal adherent conditions (Fig. 1, Lane 7).

In certain cell types, cell density-dependent expression of VEGF (another HIF-1-regulated gene) is regulated by a soluble factor, and the conditioned medium from dense cells stimulated VEGF expression in sparse cells (28). In another report, nitric oxide was identified as the diffusible soluble factor produced by dense cultures and was capable of transactivating a HRE-driven reporter construct (29). Therefore, we also tested the possible involvement of a soluble factor in density-dependent CAIX induction in HeLa cells. Conditioned medium from dense cultures (seeded at 160,000 cells/cm²) was added to sparse cells (seeded at 20,000 cells/cm²) and incubated for 24 h. Western blotting revealed that conditioned medium did not activate CAIX expression in sparse cultures (data not shown). At all concentrations tested (20, 40, 80, and 100 μm), the inhibitor of nitric oxide synthase, S-isopropylisothiourea, did not have any effect on CAIX production in densely cultured HeLa cells (data not shown). Thus, the cell density-mediated induction of CAIX expression in HeLa cells is not mediated by nitric oxide or some other soluble factor. Similarly, we were unable to observe any CAIX up-regulation due to decrease in pH or glucose concentration (Fig. 2). Increasing medium acidification, in fact, had a toxic effect on HeLa cells, resulting in lowered cell density and lower CAIX expression (Fig. 2,A). Effect of decreased glucose concentration was tested on cells grown in the presence of dialyzed FCS. Over the whole range of glucose concentrations tested, lowered glucose concentration did not affect CAIX expression (Fig. 2 B). We conclude that cell density-mediated CAIX expression is not affected by lowered pH or decreased glucose concentrations.

Recently, it has been reported that increased HIF-1α levels (compared with sparse cultures) were found in dense cultures of some cell lines possessing constitutively active HIF-1 under normoxic conditions (29, 30). This led to the suggestion that confluence (dense cultures) under normoxic conditions induces HIF-1α levels due to mild hypoxia from cell-to-cell contact (30). The above-mentioned reports prompted us to investigate the possible involvement of HIF-1α in cell density-induced CAIX expression in HeLa cells. To this end, we analyzed HIF-1α levels in lysates prepared from cells plated at different densities by Western blotting. Under these conditions, HIF-1α Ab recognized M_r 115,000 and M_r 105,000 bands constitutively present at a relatively low level. The amount of this protein did not change appreciably after 4, 8 (data not shown), and 24 h over the whole range of densities tested (Fig. 3). Cultures of various densities were also treated with the hypoxia mimic CoCl₂ (16). In contrast to control cells, Western blotting of the lysates from CoCl₂-treated cells revealed markedly increased amounts of the M_r 115,000 protein and moderately increased amounts of the M_r 105,000 protein. As judged from the intensity of these bands, even in CoCl₂-treated cells the levels of corresponding HIF-1α forms are independent of cell density (Fig. 3). Moreover, these levels did not change after 4, 8, or 24 h (Fig. 3), and apparently, HIF-1α expression remains constant during the hypoxia mimic treatment. We conclude that HeLa cells constitutively express a small amount of labile HIF-1α that is not regulated by cell density and that the activation of cell density-mediated CAIX expression is independent of HIF-1α stabilization.

Participation of the PI3K pathway in regulation of HIF-1-inducible VEGF expression has been firmly established (31, 32, 33). Therefore, we wished to investigate whether the PI3K pathway contributes to the cell density-mediated induction of CAIX in HeLa cells. Initially, we studied the association between PI3K activity and CAIX expression by analyzing the steady-state levels of PI3K activity in sparse and dense HeLa cultures. Cells were plated at 20,000 cells/cm² and 160,000 cells/cm². After a 24-h incubation, the cells were harvested, cell lysates were immunoprecipitated with an anti-phosphotyrosine Ab, and PI3K activity was assayed in the immunoprecipitates in the presence of γ-ATP and phosphatidylinositol. PI3K activity in dense cultures is two to three times higher than that observed in sparse cultures (Fig. 4). This experiment was repeated three times with similar results (data not shown). Thus, increased steady-state PI3K activity correlates with induction of CAIX in dense cultures.

Increased steady-state PI3K activity in dense cultures revealed a positive correlation between PI3K activity and CAIX expression in dense cultures. To investigate the functional involvement of the PI3K pathway in regulation of CAIX expression in dense cultures, we tested the effects of specific pharmacological inhibitors of PI3K activity on the level of endogenous CAIX protein. The cells were seeded at 160,000 cells/cm² and treated with DMSO (control) or increasing LY294002 concentrations for 24 h, and CAIX levels were evaluated by Western blotting. LY294002 displayed powerful, dose-dependent inhibition, and, as can be seen in Fig. 5,A, the two highest concentrations completely abrogated CAIX expression. This inhibition was specific because there was no effect on β-actin levels within the range of inhibitor concentrations tested (Fig. 5 A). Similar inhibition was observed with wortmannin, another specific inhibitor of PI3K (data not shown).

To further investigate this observation, we treated dense cultures in the presence or absence of the hypoxia mimic, CoCl₂, and in the presence or absence of LY294002. LY294002 (15 μm) was added to the culture medium 30 min before the addition of the hypoxia mimic. The cells were then harvested 24 h later. As seen in Fig. 5 B, in the presence of hypoxia mimic and LY294002, there was a decrease in active HIF-1α protein compared with hypoxia mimic alone. There was also a decrease in the amount of CAIX protein compared with hypoxia mimic only. However, in the case of dense culture alone, there was substantially more CAIX protein, even though there were barely detectable levels of active HIF-1α protein. And, in the presence of LY294002, there was almost complete absence of CAIX protein. The decrease in HIF-1α protein seen in the condition of hypoxia mimic plus LY294002 is possibly a result of LY294002-mediated inhibition of transcriptional activation of the HIF-1α promoter (34). However, this point will require further investigation. Thus, these results further support our conclusion that high density culture conditions induce significant CAIX protein in the absence of significant levels of HIF-1α protein and that this induction is almost completely inhibited by the PI3K inhibitor LY294002.

The (−173/+31) CA9 promoter appears to contain cis elements essential for CA9 transcription at high cell density. To test the effects of PI3K inhibition on the CA9 promoter, HeLa cells were transiently transfected with a reporter construct driven by the (−173/+31) CA9 fragment and treated with LY294002. Reporter assays confirmed that there is a dose-dependent inhibition of expression from the CA9 promoter by LY294002, whereas the control cytomegalovirus promoter was unaffected by the tested inhibitor concentrations (Fig. 5 C). This confirms that the PI3K-dependent mechanism activated in dense cultures specifically targets the (−173/+31) CA9 promoter region.

To gain further insight into the role of PI3K in the regulation of CA9 transcription, we studied the effects of overexpression of the constitutively active p110 PI3K subunit on CA9 promoter activity. We transiently cotransfected HeLa cells with the CA9 promoter reporter construct and a construct expressing either the constitutively active p110 PI3K subunit or its kinase-dead mutant. Compared with the kinase-dead mutant, the constitutively active p110 transactivated the CA9 promoter 3-fold (Fig. 6).

We have shown above that CAIX expression in dense HeLa cultures is not mediated by a soluble factor, decrease in pH, or glucose concentration and is not associated with stabilization of HIF-1α. Other investigators have recently shown that, in comparison with pO₂ in equilibrated medium (16%; 116 mmHg), pO₂ close to the surface of plated cells 48 h after plating was 13% (96 mmHg) for sparse and 9% (70 mmHg) for dense LNCaP human prostate carcinoma cells (29). Although some changes in pO₂ occur near the cell surface of dense cultures, these changes are relatively small and are not as low as those seen in experimentally induced hypoxic responses (0.1–1%). Despite this, elimination of the pO₂ gradient by gentle shaking prevented activation of a HRE-driven reporter, leading to the conclusion that the formation of this small gradient is an important component of cell density-mediated transcriptional activation (29). We also studied the effects of oxygenation on CAIX expression. A pO₂ gradient was generated by incubating cells plated at 160,000 cells/cm² with varying volumes of medium. We found that covering cells with 1.5 mm of medium already induced CAIX expression after 12 h. Increasing the medium height up to 3 mm led to further induction of CAIX expression (Fig. 7,A). Above 3 mm, no further induction was observed (Fig. 7,A), indicating that 3 mm of medium sufficiently limits O₂ diffusion for activation of the CA9 promoter. This suggested that O₂ deprivation could play a role in cell density-mediated induction of CAIX. In agreement with our previous result, there was no effect on HIF-1α activity (data not shown). We also tested the effects of reoxygenation by gentle stirring on CAIX expression. Cells were plated at 160,000 cells/cm², covered with 1 cm of medium, allowed to attach for 5 h, and stirred for an additional 12 h. Reoxygenation by stirring abolished CAIX induction (Fig. 7 B), confirming the role of a pO₂ gradient in cell density-dependent CAIX expression.

The observation that stirring of dense cultures prevented CAIX induction indicated that the mechanism of induction is likely to involve changes in pO₂. Because hypoxic induction also involves O₂ deprivation, we wished to investigate whether both processes are mediated by the same mechanism. Initially, we studied induction of CAIX expression in HeLa cells plated at varying densities in the presence of the hypoxia-mimic CoCl₂. Results shown in Fig. 8,A demonstrate that the CoCl₂-induced CAIX expression is dependent on cell culture density, generating an additive effect of hypoxia- and cell density-dependent CAIX expression. This additive effect suggested that the corresponding mechanisms are independent. This proposition was confirmed by studying the effects of stirring on CoCl₂-induced CAIX expression in dense culture. Hypoxia mimic CoCl₂, by competing with Fe²⁺ for binding the O₂ sensor, produces a state equivalent to that of chronic hypoxia even in the presence of normoxic levels of O₂ (16). The unstirred CoCl₂-treated cell culture exhibited the additive effect of hypoxia and cell density (Fig. 8,B, Lane 2), whereas stirred cells expressed significantly decreased levels of CAIX (Fig. 8,B, Lane 4). Both stirred and unstirred cells treated with CoCl₂ displayed the same levels of HIF-1α (Fig. 8 B, Lanes 2 and 4). In our opinion, these important results lend themselves to the following conclusions: (a) hypoxia-induced CAIX expression and cell density-induced CAIX expression are mediated by two mechanisms that are additive; (b) CAIX expression in stirred CoCl₂-treated cells represents the net contribution of hypoxia-induced stabilization of HIF-1α; and (c) despite being dependent on a lowered O₂ level, the cell density-mediated mechanism does not depend on the O₂ sensor that is linked with stabilization of HIF-1α. Taken together, these results confirm that cell density-induced CAIX expression is brought about by pericellular mild hypoxia that is independent of HIF-1α stabilization.

We were also interested in defining the minimal cell density-dependent region of the CA9 promoter. To identify cis elements in the CA9 promoter that are required for cell density-mediated activation, we used deletion analysis and a simple modification of the standard transfection protocol. The cells were transfected at the previously optimized cell density and allowed to incorporate DNA for 4 h. Then the cells were trypsinized, seeded at 20,000 and 160,000 cells/cm², and cultivated for an additional 68 h. This modification allowed us to test the promoter fragment activity from one transient transfection mixture (hence, presumably having the same transfection efficiency) in conditions of varying densities. Cell density-dependent inducibility of the (−173/+31) CA9 promoter construct and its progressively deleted mutants was expressed as the ratio of reporter activity in dense and sparse cultures and is presented in Fig. 9 as the fold induction. The pBMN5 (PR1–5) construct displayed more than 100-fold induction. The MN6 (PR1 + HRE) construct was also activated by cell density (Fig. 9), although not as efficiently as pBMN5. Further deletion abolished density-dependent inducibility; activation of the (−30/+31) MN7 fragment, retaining only the HRE, is similar to activation of the herpes simplex virus thymidine kinase promoter (pBLCAT5; Fig. 9). In contrast to activation of the CA9 promoter by hypoxia, where HRE appeared to be sufficient for induction (6), the results of the deletion analysis clearly show that HRE on its own is not sufficient for cell density-dependent transactivation of CA9.

Next, we studied the effects of point mutations in PR1 and HRE within the context of the CA9 promoter on inducibility by cell density. The construct with a mutated SP1 binding site in the PR1 position (PR1mut) completely failed to respond to induction by cell density, displaying inducibility in the same range as MN7 or pBLCAT5 constructs (Fig. 9). Similarly, cell density-dependent inducibility of the construct with mutated HRE (HREmut) was significantly impaired, showing only a slight but reproducible increase above that of pBLCAT5 (Fig. 9). The results of mutation analysis are in good agreement with the results of deletion analysis. Both approaches highlight the significant contribution of PR1 in cell density-dependent activation of CAIX expression; in the absence of functional PR1, this activation is lost. At the same time, by revealing the importance of intact HRE for cell density-dependent CAIX expression, the mutational analysis expanded on the results of the deletion analysis. Although not sufficient on its own, binding to HRE appears to be necessary for density-dependent CAIX expression.

We therefore conclude that the cell density-induced mechanism requires cooperation between factors bound to PR1 and factors bound to the neighboring HRE. Thus, the (−45/+31) fragment, containing the HRE and PR1, appears to be the minimal CA9 region that can be induced by cell density.

Mutational analysis demonstrated that for cell density-mediated CA9 promoter activation even in the absence of HIF-1α stabilization, the functional HRE was required. This suggested that HIF-1 activity could be involved in cell density-dependent transcriptional activation. To test this hypothesis, we used HIF-1α-deficient Kal3.5 cells (26) and investigated the activity of the CA9 promoter under conditions of increased cell density. In transiently transfected control C4.5 cells, the CA9 promoter was activated by CoCl₂ and increased cell density (Fig. 10). In Kal3.5 cells, however, the promoter was not activated by CoCl₂ and was moderately activated by cell density (1.5×, compared to 8–9× in C4.5 cells; Fig. 10). Thus, we conclude that there is an essential role for some minimal level of HIF-1 activity for cell density-induced CAIX expression.

Under conditions of true hypoxia (0.1–1.0% O₂)-mediated induction of CAIX, HIF-1 binding to HRE in the CA9 promoter immediately upstream of the transcription start appears to be critical (6). However, there is now an increasing body of evidence that a number of hypoxia-activated genes may also be regulated in a HIF-1α-independent fashion (10, 11). In the case of CAIX, recent studies have shown substantial but incomplete colocalization of CAIX expression and regions of hypoxia in tumors, with CAIX-positive areas extending beyond regions of hypoxia (6, 12). This partial discordance suggested the existence of additional pathway(s) that controls CAIX expression under pathophysiological conditions in tumors independently of HIF-1α stabilization.

Compared with cells grown in sparse culture, cells in dense cultures have to adjust to greatly changed conditions by down- and up-regulating expression of appropriate genes. In doing so, they alter the expression pattern of cell surface receptors, transcription factors, enzymes, oncogenes, and growth factors (28). CAIX also ranks among the proteins regulated by cell density; it is not detectable in sparse HeLa cultures, but it is present in large amounts in dense HeLa cultures. Because of this strikingly tight correlation with cell density, we chose this cervical cell line as a model system for studying mechanisms participating in CAIX regulation. Investigation of differences between sparse and dense HeLa cultures should primarily identify pathways that are differentially activated in these cultures and are required for CAIX expression. Although not fully reflecting the situation inside tumors, this model of transformed cells is easy to study in vitro, and conclusions from this comparative study could be of certain relevance to carcinogenesis in vivo.

Study of the process of cell density-induced CAIX expression in HeLa cells allowed us to draw the following conclusions: (a) this process is anchorage independent because growing cells on agarose-coated plates had no effect on CAIX levels; (b) this process is not mediated by a soluble factor because conditioned medium from dense cultures had no effect on CAIX levels in sparse cells; and (c) this process is serum independent because serum-starved cells still expressed CAIX, albeit with a reduced efficiency. Although these results appeared to implicate the importance of cell-cell contacts, additional experiments confirmed that cell contacts per se do not initiate CAIX expression. Instead, we proved the importance of microenvironmental factors that prevail in dense cultures. The microenvironmental factors could be eliminated by gentle stirring, and of these factors, we considered acidosis, glucose deprivation, and lowered O₂ concentrations, all of which can have an effect on VEGF expression (35, 36). Lowered medium pH as well as decreased glucose concentration failed to elicit any effect on CAIX expression in HeLa cells. Apparently, the pO₂ gradient produced by depletion and limited diffusion of O₂ in dense cultures (37) is the major causative factor in the CAIX induction by cell density. However, the pO₂ gradient produced under these conditions is modest (9% in dense cultures compared with 13% in sparse cultures; Ref. 29) and well above the concentrations used for hypoxic inductions (0.1–1%). This argued that the effect of mild hypoxia in dense cultures is not mediated by HIF-1 but by a novel pathway that is turned on at intermediate O₂ levels that are higher concentrations than that required for HIF-1α stabilization.

Several lines of evidence suggest that transcription of genes coding for CAIX and the angiogenic factor VEGF may be regulated in similar ways. Both genes possess functional HREs; in the case of VEGF, it reads TACGTGGG [position (−975/−967); Refs. 6 and 9]. Expression of both proteins is positively correlated with increasing cell density in vitro (23, 28, 38). The factors that trigger VEGF induction in dense cultures may depend on the cells used: in human colon carcinomas, it appeared to be a soluble factor (28); whereas in U87 glioblastoma-astrocytoma cells, the presence of such a factor was ruled out (38). In the latter study, the mitogen-activated protein kinase pathway was implicated in the cell density-mediated induction of VEGF (38).

Although undoubtedly critical for hypoxic induction, it is noteworthy that HRE does not always play a critical role in VEGF regulation. VEGF promoter fragments with or without HRE were stimulated to the same level after treatment with epidermal growth factor (10). Differential expression of VEGF in human pancreatic adenocarcinoma cells was also independent of HRE but correlated strongly with differential levels of constitutive SP1 activity (11). Glioblastoma cells express high levels of VEGF in the absence of external stimuli, suggesting the existence of factors intrinsic to these cells and independent of the environment (10).

Analogies with VEGF prompted us to test the involvement of the PI3K pathway in cell density-dependent CAIX expression. Several lines of evidence confirm the crucial role of PI3K in the process of cell density-induced CAIX expression. We found that the dense HeLa culture displayed significantly higher PI3K activity than the sparse culture. Also, pharmacological inhibitors of PI3K activity, LY294002 and wortmannin, efficiently suppressed CAIX expression in dense cultures in a dose-dependent manner. In addition, we were able to demonstrate that the cell density-induced mechanism mediated by PI3K specifically targets the (−173/+31) CA9 promoter. Finally, the constitutively active PI3K p110 mutant transactivated the CA9 promoter.

The requirement of PI3K activity for cell density-dependent CAIX expression may provide a link between the tumor-associated nature of CAIX protein and the well-established role of the PI3K pathway in tumorigenesis (39). Mutations in several proteins in the PI3K/Akt and mammalian target of rapamycin/p70^S6K signaling pathways, which regulate cell survival (40), growth (41), and proliferation (42), are causally involved in a high percentage of common human malignancies (39). In addition, numerous human malignancies are associated with inactivating mutations in the tumor suppressor gene PTEN (43), a 3′-phosphoinositide phosphatase that dephosphorylates PI3K products. It is possible that tumors with deregulated PI3K activity may express CAIX constitutively. Additional studies will be required to investigate this possibility.

Hypoxia has a profound activating effect on CAIX expression, implicating HIF-1 as a strong regulator of the CA9 promoter (6). HIF-1 activity is dependent on the availability of its HIF-1α subunit (15, 16). The emerging theme from several recent reports is that there are two mechanistically distinct modes of increasing the total available amount of HIF-1α (44, 45). The first is mediated by the posttranslational stabilization of HIF-1α as a result of either hypoxic conditions (16) or defects in the VHL-dependent degradation pathway (17). The second appears to induce HIF-1 activity via transcriptional activation of HIF-1α (18, 19, 20, 21, 22). In the latter group, the role of the PI3K pathway has received much attention lately. Several earlier studies indicated that the PI3K pathway might be involved in the transcriptional activation of HIF-1α (31, 32, 33); however, evidence presented in more recent reports distinguishes induction of HIF-1 activity from activation of the PI3K pathway (44, 45). Chemical or genetic interference with PI3K completely prevented Akt activation without affecting the stabilization of HIF-1α by hypoxia (18, 46). Furthermore, the enhancement of VEGF induction by hypoxia found in cells with a constitutively active PI3K pathway is not mediated by HIF (10).

All these reports, together with the suggestion that cell-to-cell contact causes mild hypoxia (30) and the observation that HIF-1 activity depends on cell density in prostate cell lines (29), prompted us to investigate levels of HIF-1α in HeLa cultures of various densities. We were unable to detect any changes of HIF-1α levels as a consequence of increased density. At the same time, HIF-1α levels increased substantially in response to treatment with hypoxic mimics. The observation that the steady-state levels of HIF-1α are not subject to change at high cell density distinguishes the mechanisms of cell density-dependent induction of CAIX from that of induction due to hypoxia-mediated HIF-1α stabilization. However, this does not rule out participation of a minimal level of active HIF-1α.

The requirement for PI3K activation and a minimal level of active HIF-1α at high cell densities suggests that there are two separate but interdependent pathways for CAIX induction. PI3K activity and HIF-1α activity both appear to contribute to CAIX induction at high cell density and under conditions of true hypoxia. However, PI3K plays the major inducing role at high cell density, whereas HIF-1 takes on this major role under true hypoxic conditions.

We have identified the critical cis elements that are required for the cell density-dependent induction of the CA9 promoter. Deletion and mutational analysis revealed that PR1 and the HRE together are necessary and sufficient for density-dependent transactivation of the CA9 promoter. Characterization of PR1 was the subject of our recent study in which we defined PR1 as a functional SP1 site binding SP1/SP3 factors (14). For a long time, there was a widely held notion that SP1 sites represent constitutive promoter elements that support basal transcription from these promoters (47). However, there is now mounting evidence that SP1 sites can participate in modulation of gene activity by binding differentially expressed Krüppel-like factors with which they share binding sites. These factors play an important role in growth factor signal transduction pathways (48) and apoptosis (49) and can be the subject of modulation by oncogene proteins, tumor suppressors, and cell cycle control molecules (47). We are currently investigating the mechanism of transcriptional activation from the PR1-HRE module and the role of SP1/SP3 factors in this process.

Finally, by using HIF-1α-deficient Chinese hamster ovary cells, we proved that in the absence of HIF-1α, not only hypoxia-mediated but also cell density-mediated CAIX expression is lost. This shows that the small amount of constitutive HIF-1α present under these conditions plays an essential role in cell density-dependent activation of the CA9 promoter. Unlike the expression of VEGF, which can be independent of HRE/HIF-1 (10, 11), activation of CAIX expression requires functional HRE and HIF-1α activity.

In this report, we provide evidence for an alternative pathway that controls CAIX expression. Based on our results, we have developed the following model (Fig. 11): there are two pathways controlling CAIX expression, both of which involve HIF-1α. Both pathways are O₂ sensitive, and they differ in the O₂ threshold required for their initiation. The intermediate O₂ levels generated by high cell density do not increase the overall amount of HIF-1α but activate PI3K by an as yet undetermined mechanism, which, in turn, affects induction of CAIX expression. This induction itself is also dependent on a minimal level of HIF-1α activity. However, for efficient activation, cooperation between factors that bind both the PR1 and HRE domains is required. Under conditions of true hypoxia, stabilization of HIF-1α occurs that leads to a lower level of CAIX induction in the absence of PI3K activity, as seen in the LY294002 experiments.

This model is consistent with what is seen in human tumors, where CAIX, in addition to being expressed within regions of hypoxia, is also expressed in juxtaposed areas that are not hypoxic (as measured by pimonidazole binding; Ref. 12). These areas may well represent cells experiencing intermediate O₂ levels that have activated PI3K. If this phenomenon is generally true [we have observed it in many different human tumors (3)] and also holds true for other HIF-mediated “hypoxia-inducible” genes (e.g., VEGF), this has important implications for tumor cell survivability, angiogenesis, and invasion.

It may also explain why certain human tumors, other than those with defective VHL, constitutively express CAIX, i.e., due to constitutive PI3K activity. These possibilities are currently under investigation.

We thank Dr. D. A. Cantrell for plasmids expressing mutants of the PI3K p110 subunit. We also thank Dr. P. J. Ratcliffe for the generous gift of the Chinese hamster ovary cell lines C4.5 and Kal3.5.