p53 Inhibits Hypoxia-inducible Factor-stimulated Transcription*

p53 is required for hypoxia-induced apoptosisin vivo, although the mechanism by which this occurs is not known. Conversely, induction of the hypoxia-inducible factor-1 (HIF-1) transactivator stimulates transcription of a number of genes crucial to survival of the hypoxic state. Here we demonstrate that p53 represses HIF-1-stimulated transcription. Although higher levels of p53 are required to inhibit HIF than are necessary to transcriptionally activate p53 target genes, these levels of p53 are similar to those that stimulate cleavage of poly(ADP-ribose) polymerase, an early event in apoptosis. Transfection of full-length p300 stimulates both p53-dependent and HIF-dependent transcription but does not relieve p53-mediated inhibition of HIF. In contrast, a p300 fragment, which binds to p53 but not to HIF-1, prevents p53-dependent repression of HIF activity. Transcriptionally inactive p53, mutated in its DNA binding domain, retains the ability to block HIF transactivating activity, whereas a transcriptionally inactive double point mutant defective for p300 binding does not inhibit HIF. Finally, depletion of doxorubicin-induced endogenous p53 by E6 protein attenuates doxorubicin-stimulated inhibition of HIF, suggesting that a p53 level sufficient for HIF inhibition can be achieved in vivo. These data support a model in which stoichiometric binding of p53 to a HIF/p300 transcriptional complex mediates inhibition of HIF activity.

The p53 tumor suppressor protein mediates both growth arrest and apoptosis. Whereas p53-dependent growth arrest requires transcriptional activation of p21 WAF1/CIP1 (1-3), substantial data have accumulated that transactivating capability is not necessary for p53 to stimulate apoptosis (4 -6).
Hypoxia is perhaps the most physiologic inducer of p53 (7), and hypoxia-mediated apoptosis of tumors in vivo requires p53 (8). Indeed, p53 is the most frequently inactivated gene in solid tumors, and in animal tumor models, hypoxia selects against wild type p53 (8). However, the mechanism by which p53 mediates apoptosis in hypoxic tumor cells is not known. Hypoxic conditions in vitro as well as in vivo result in induction of hypoxia-inducible factor-1␣ (HIF-1␣), 1 the limiting, hypoxiainducible subunit of the HIF-1 transactivator (9). HIF-1, in turn, stimulates transcription of a number of genes important for tumor survival under hypoxic conditions in vivo, including vascular endothelial growth factor (VEGF), erythropoietin (Epo), and several glycolytic enzymes (10). Tumors in which hypoxia cannot induce HIF-1 transcriptional activity remain small and fail to metastasize (11).
We have recently shown that hypoxic induction of p53 requires concomitant induction of HIF-1␣, and that HIF-1␣ binds to and stabilizes p53 (12). We found previously that HIF-1␣ had no direct effect on p53 transcriptional activity. We now report that association of HIF-1␣ and p53 results in inhibition of HIF-stimulated transcription. This requires a higher p53 level than is necessary for transcriptional activation of several endogenous p53-responsive promoters but correlates well with the level of p53 necessary to cause apoptosis.

MATERIALS AND METHODS
Cell Lines-SKBr3 and MCF7 are human breast cancer cell lines obtained from American Type Culture Collection (Rockville, MD), and PC3M is a highly metastatic variant of the prostate cancer cell line, PC3. SKBr3 contains one mutated, transcriptionally inactive p53 allele, MCF7 contains transcriptionally active wild type p53 (wtp53), and PC3M cells are p53-null.
Reagents and Plasmids-Ad-LacZ, a ␤-galactosidase-expressing, replication-deficient adenovirus, and Ad-p53, a wtp53-expressing, replication-deficient adenovirus, were obtained from B. Vogelstein (Johns Hopkins Oncology Center). Viral titer was determined as described previously (13). Multiplicity of infection (MOI) is defined as the ratio of total number of viruses used in a particular infection per number of cancer cells to be infected (i.e. number of viruses per cell).
Transient Transfection Assay-8 ϫ 10 5 or 4 ϫ 10 6 cells were plated in T25 flasks or 6-well plates (Costar, Acton, MA), respectively. The next day, cells were transfected with plasmids in the presence of Lipo-fectAMINE (Life Technologies, Inc.). After 6 -12 h of incubation with the plasmid-lipid suspension, the medium was changed, and cells were grown for an additional 24 h, unless otherwise indicated. The cells were lysed and analyzed for luciferase activity as described previously (12).
Western Blot Analysis-Proteins were resolved with 8% SDS-poly-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. acrylamide gel electrophoresis for detection of p53, Mdm-2, and poly-(ADP-ribose) polymerase (PARP) as described previously (15).

RESULTS AND DISCUSSION
Wild Type p53 Abrogates HIF-1␣ Activity-We initially investigated the effects of wtp53 on HIF-1-responsive transcription in SKBr3 and PC3M cell lines. Transient expression of HIF-1 dramatically induced transcription of two HIF-dependent reporter constructs, Epo-Luc (containing an HIF-responsive element from the erythropoietin gene) and VEGF-Luc (containing HIF-responsive elements from the VEGF gene) in both cell lines, whereas co-transfection of wtp53 abrogated this induction ( Fig. 1, A, B, and D). Dose-response analysis showed near-maximal inhibition of HIF-1-stimulated transcription (using 0.5 g of HIF-1) by as little as 0.1 g of wtp53 in SKBr3 cells and 0.5 g of wtp53 in PC3M cells (Fig. 1, A and B).
The Transactivating Function of p53 Is Not Required for Suppression of HIF-1␣ Activity-In order to determine whether transactivating capability was essential for p53-mediated HIF inhibition, we next examined the HIF-1 inhibitory activity of the p53 "DNA contact" mutant p53-273H (16), which failed to transactivate three independent p53-responsive reporters: PG13-Luc, WWP-Luc, and Bax-Luc (data not shown). Although lacking transactivating capability, p53-273H repressed HIF-1driven transcription in both SKBr3 and PC3M cells (Fig. 1, A,  B, and D), demonstrating that neither DNA binding nor transactivation is essential for p53-mediated suppression of HIF-1 activity.
To confirm the specificity of this phenomenon, we tested the effects of both wtp53 and p53-273H on a control luciferase plasmid, pGL2-control (VEGF-Luc is a pGL2-based reporter and Epo-Luc is a pGL3-based reporter), in transiently transfected PC3M cells. At the concentrations used in these experiments, neither p53 construct significantly inhibited the SV40-driven luciferase activity generated from pGL2-controltransfected cells (Fig. 1C).
p53-mediated HIF Suppression Correlates with Stimulation of PARP Cleavage but Not with the Transactivating Function of p53-Although p53-mediated suppression of HIF activity can be clearly demonstrated, it requires higher p53 levels than are necessary to observe p53-mediated transactivation. Thus, transactivation of PG13-Luc in SKBr3 cells requires co-transfection of 10 -20-fold less p53 than is necessary to obtain significant HIF-1 inhibition ( Fig. 2A). Because transfection does not introduce p53 into 100% of the cells tested, it is difficult to compare suppression of HIF activity with the growth inhibitory and apoptotic activity of p53. Since we demonstrated previously that essentially 100% of cells infected with the p53containing adenovirus Ad-p53 express wtp53 protein following infection (minimum MOI ϭ 2; see Ref. 13), we used this approach to further explore the relationship of p53 expression to transactivation and HIF suppression, respectively.
We determined the MOI of Ad-p53 required in SKBr3 cells to yield maximal inhibition of Epo-Luc (Fig. 2B) and DNA synthesis (Fig. 2C), stimulation of PARP cleavage (Fig. 2D), induction of Mdm-2 protein (Fig. 2E), and stimulation of Bax-Luc (Fig. 2F). To control for any nonspecific effects of viral titer, cells were infected with 32 MOI of Ad-LacZ (second bar/lane in Fig. 2, B-F). PARP cleavage is an early event in apoptosis, and inhibition of DNA synthesis is a marker of growth arrest. The data demonstrate that near-maximal inhibition of DNA synthesis, Mdm-2 induction, and transactivation of Bax-Luc occur at an Ad-p53 MOI of 2-4. In contrast, both significant inhibition of HIF-responsive transcription and stimulation of PARP cleavage do not occur until an MOI of 16 -32 is reached. Infection with Ad-LacZ was consistently without effect. Thus, whereas the transactivating function of p53 correlates with induction of growth arrest, suppression of HIF activity correlates with the higher amount of p53 required to initiate apo- ptosis. Similar to what was found in the previous transfection experiment ( Fig. 2A), 8 -16-fold more p53 was needed to suppress HIF activity and stimulate PARP cleavage than was required to support transactivation of p53 target genes and to cause growth arrest.
These results are in agreement with our earlier observation that transactivation does not play a role in the inhibition of HIF activity by p53. Taken together with our recent observation that p53 co-precipitates with HIF-1␣ (12), the data fit a model in which direct association of HIF-1␣ with p53 results in HIF inhibition.
p53 Interaction with p300 Is Required for Inhibition of HIF Activity-Both HIF-1␣ and p53 bind to p300, and p300 is required for full activity of both transactivators (17)(18)(19)(20)(21). Thus, it was of interest to determine whether p300 plays any role in p53-mediated HIF inhibition. We first determined whether exogenous p300 could restore HIF activity in the presence of p53. We transfected PC3M cells with Epo-Luc and HIF-1␣, plus either CMV-␤-galactosidase (control plasmid) or wtp53, and full-length p300 (Fig. 3A). Although exogenous p300 augmented HIF-dependent transcriptional activity, as reported previously (17), it could not reverse p53-mediated inhibition of this activity.
Based on this model, a p300 fragment that bound p53 but not HIF-1␣ would be expected to block p53-mediated HIF repression. To test this, we repeated the above experiment but cotransfected with the p53-binding p300 fragment, p300(1514 -1922), instead of full-length p300. Since it lacks the HIF-1␣ binding site this fragment did not further stimulate HIF activity, but it markedly ameliorated p53-mediated inhibition of this activity (Fig. 3A). Co-transfection of p300(1514 -1922) did not affect the level of expression of p53 (not shown), but it did abrogate the ability of p53 to induce endogenous Mdm-2 (see Fig. 3A, inset), previously shown to be dependent on p53/p300 interaction (21). These data suggest that p300(1514 -1922) reversed p53-mediated inhibition of HIF-dependent transcription by sequestering p53 and preventing its association with the endogenous pool of p300.
In order to further explore the requirement for p300 in p53mediated inhibition of HIF-1, we tested the anti-HIF activity of a p53 double point mutant, p53(22/23), which is mutated at residues 22 and 23 in the amino-terminal transactivating domain (5,22) and does not bind to p300 (21). As can be seen in Fig. 3B, this mutant was unable to suppress HIF-1 activity in PC3M cells. Although 0.25 g of wtp53 markedly inhibited HIF-1, p53(22/23) was inactive even when transfected at 2 g. Both wtp53 and p53(22/23) were expressed to similar levels, as monitored by Western blotting (not shown). Taken together, these data demonstrate that p53 must bind to p300 in order to inhibit HIF-1.
Drug-induced Elevation of Endogenous wtp53 Suppresses HIF-1 Activity-Although exogenously supplied p53 inhibits HIF-1-driven transcription at levels that initiate apoptotic events, we wished to determine whether endogenous p53 could be induced to display similar activity. Thus, we transfected MCF7 cells, which contain wtp53, with Epo-Luc and HIF-1␣ and examined the effect on reporter activity of the DNA-damaging drug doxorubicin, a potent inducer of wtp53. Treatment with doxorubicin led to marked elevation of p53 and significantly blunted HIF transcriptional activity (Fig. 4). Co-transfection of a human papilloma virus E6-expressing plasmid together with Epo-Luc and HIF-1␣ reversed this inhibition (Fig.  4). In contrast, doxorubicin had no effect on HIF-1 activity in p53-null PC3M cells (data not shown).
In summary, we have demonstrated that p53 is able to repress HIF activity in a manner not requiring the transactivating function of p53. That a significantly greater amount of p53 is necessary to inhibit HIF than is required to transactivate p53 target genes or cause growth arrest can be explained by a stoichiometric model in which p53 must saturably bind to HIF/ p300-containing complexes in order to inhibit HIF. The physiologic relevance of this phenomenon is supported by two observations. First, the level of p53, which must be reached to mediate HIF repression, is equivalent to that at which PARP cleavage becomes detectable, and second, this level of p53 can be reached in vivo following exposure to a commonly used chemotherapeutic. FIG. 4. Endogenous wtp53 can be induced to levels that inhibit HIF activity. MCF7 cells were transfected with Epo-Luc and HIF-1␣ (0.5 g each) in the presence or absence of E6 (2.0 g) as indicated. Treatment with 200 ng/ml doxorubicin was begun 24 h after transfection and continued for an additional 20 h, at which time cells were lysed and Epo-Luc activity was determined (A). The activity obtained in the absence of E6 and doxorubicin was set at 100%. The relative increase in endogenous p53 steady-state protein level following doxorubicin is shown in B.