Transforming Growth Factor β1 Induces Hypoxia-inducible Factor-1 Stabilization through Selective Inhibition of PHD2 Expression*

The hypoxia-inducible transcription factor-1 (HIF-1) is central to a number of pathological processes through the induction of specific genes such as vascular endothelial growth factor (VEGF). Even though HIF-1 is highly regulated by cellular oxygen levels, other elements of the inflammatory and tumor microenvironment were shown to influence its activity under normal oxygen concentration. Among others, recent studies indicated that transforming growth factor (TGF) β increases the expression of the regulatory HIF-1α subunit, and induces HIF-1 DNA binding activity. Here, we demonstrate that TGFβ acts on HIF-1α accumulation and activity by increasing HIF-1α protein stability. In particular, we demonstrate that TGFβ markedly and specifically decreases both mRNA and protein levels of a HIF-1α-associated prolyl hydroxylase (PHD), PHD2, through the Smad signaling pathway. As a consequence, the degradation of HIF-1α was inhibited as determined by impaired degradation of a reporter protein containing the HIF-1α oxygen-dependent degradation domain encompassing the PHD-targeted prolines. Moreover, inhibition of the TGFβ1 converting enzyme, furin, resulted in increased PHD2 expression, and decreased basal HIF-1α and VEGF levels, suggesting that endogenous production of bioactive TGFβ1 efficiently regulates HIF-1-targeted genes. This was reinforced by results from HIF-1α knock-out or HIF-1α-inhibited cells that show impairment in VEGF production in response to TGFβ. This study reveals a novel mechanism by which a growth factor controls HIF-1 stability, and thereby drives the expression of specific genes, through the regulation of PHD2 levels.

Overexpression of hypoxia-inducible transcription factor-1 (HIF-1) 5 is a hallmark of most cancers and is generally associ-ated with increased tumor growth, angiogenesis, and patient mortality (1)(2)(3). In xenograft models, tumor growth and angiogenesis are stimulated by HIF-1 overexpression and inhibited by the loss of HIF-1 (4 -6). The essential role of HIF-1 in tumor progression is attributed to its dominant role in regulating various genes associated with the angiogenic process, cell growth/ survival, as well as invasion/metastasis. HIF-1 is a heterodimeric transcription factor composed of HIF-1␣ and HIF-1␤ subunits, however, the biological activity of HIF-1 is determined by the expression and activity of HIF-1␣ (7). Whereas the HIF-1␤ protein is readily found in all cell types, HIF-1␣ is rapidly degraded under normoxia and is therefore virtually undetectable. Indeed, oxygen-dependent prolyl hydroxylases modify HIF-1␣ and permit its recognition by the von Hippel-Lindau tumor suppressor (VHL), which functions together with elongin B, elongin C, and Cullin 2 as an E3 ubiquitin ligase complex that mediates HIF-1␣ proteasomedependent degradation. Under hypoxic conditions, prolyl hydroxylation is inhibited, leading to HIF-1␣ stabilization (8 -10). In mammalian cells, three HIF-1 prolyl hydroxylase isoforms, PHD1, PHD2, and PHD3, have been identified and shown to hydroxylate two proline residues (Pro 402 /Pro 564 ) within the oxygen-dependent degradation domain (ODDD) of HIF-1␣ (11). The concomitant expression of all three PHDs by several cell lines and tissues raised the possibility that they might have distinctive roles (12,13). Using small interfering RNAs, it was indeed demonstrated that PHDs display distinct functions in vivo and that PHD2 is the critical oxygen sensor controlling the low steady-state levels of HIF-1␣ in normoxia (14). Although the mechanisms controlling PHDs expression levels are not well characterized, recent data indicating that PHD2 and PHD3 mRNA levels, but not PHD1, are increased under hypoxic conditions, suggest divergences in their regulation (15)(16)(17).
Based on the strong influence of low oxygen concentration on HIF-1␣ stabilization, several studies were conducted to demonstrate a correlation between HIF-1␣ expression and hypoxic regions within tumors. However, the reported patterns of HIF-1␣ staining were often not consistent with hypoxic regions and expression profiles surprisingly showed homogenous distribution of HIF-1␣ within the tumor mass (18,19). More recently, it was proposed that the control points regulating HIF-1␣ levels are multiple and depend on the tumor microenvironment. In fact, in addition to hypoxia, other stimuli such as growth factors, hormones, and cytokines were demonstrated to induce HIF-1␣ accumulation and activity under normal oxygen conditions (20 -22). Although little is known of the mechanisms involved, data accumulated so far suggests that these agents mainly act at the level of HIF-1␣ gene transcription and/or protein synthesis (23)(24)(25)(26).
Numerous changes that promote tumor progression are observed within the tumor microenvironment. When tumors grow, they produce a wide array of pro-tumorigenic molecules. Among others, many common tumors are known to be associated with enhanced levels of transforming growth factor-␤1 (TGF␤1) (27). This growth factor is overexpressed by tumor cells and is also directly released by infiltrating leukocytes, such as tumor-associated macrophages, a major component of the inflammatory circuit that participates in tumor growth, angiogenesis, and metastasis (28). TGF␤1 is well known to play an important role in the tumorigenic process by enhancing the expression of metalloproteases and angiogenic factors that favor tumor invasion and vascularization (29 -32). The proangiogenic action of TGF␤1 is attributed, in part, to its capacity to induce the expression of the angiogenic factor VEGF, a key gene target of the transcription factor HIF-1 (33,34). Moreover, TGF␤1 was shown to regulate the expression of its own converting enzyme, furin, which is a recently identified HIF-1-regulated gene; this results in augmented processing of the TGF␤1 precursor into its bioactive form (35,36). This regulatory cycle is of potential importance in the induction, as well as, in the activation of numerous factors implicated in the pathogenesis of cancer.
Among recent advances underlining the contribution of TGF␤1 to tumor progression is the finding that this growth factor can influence the accumulation of HIF-1␣ in normoxic conditions. Increased HIF-1␣ expression levels and DNA binding activity were observed in fibrosarcoma HT1080 cells and in vascular smooth muscle cells stimulated with TGF␤1 (37,38). The exact molecular mechanism involved in the HIF-1␣ increase remains unknown. Herein, we demonstrate that TGF␤1 induces HIF-1␣ accumulation and activity in tumor cell lines, under normoxic conditions, through a mechanism involving stabilization of the HIF-1␣ subunit. More importantly, we show that TGF␤1 attenuates HIF-1␣ ODDD prolyl hydroxylation, an event associated with a marked and specific decrease of PHD2 prolyl hydroxylase gene and protein expression levels. Therefore, our findings reveal a novel mechanism by which an inflammatory and tumor growth factor controls the stability of HIF-1␣ through the specific regulation of prolyl hydroxylase expression levels.
Cell Culture and Transfection-The hepatoma cell line HepG2 and the fibrosarcoma HT1080 cells were obtained from the American Type Culture Collection (ATCC, Rockville, MD). The wild type and HIF-1␣ knock-out mouse embryonic fibroblasts (MEF) were a generous gift from Dr. Randall Johnson (University of California-San Diego, La Jolla, CA). All cell lines were cultured in complete media consisting in minimal essential medium (Invitrogen) supplemented with 10% fetal bovine serum (Sigma) in a humidified 95% air, 5% CO 2 incubator at 37°C. HT1080 were stably transfected with pcDNA3-␣1-PDX vector encoding the furin inhibitor ␣1-antitrypsin PDX, pcDNA3-HIF-1␣DN vector encoding a dominant-negative form of HIF-1␣, or with the control empty vector pcDNA3. Three clones expressing different amounts of ␣1-PDX (PDX-A, PDX-B, and PDX-C), one clone expressing a dominant negative form of HIF-1␣ (HIF-1␣DN), and one control clone (CTL) were obtained. HepG2 cells were stably transfected with pcDNA3-PHD1, pcDNA3-PHD2, or pcDNA3-PHD3 vectors encoding the different HIF-1␣ prolyl hydroxylases or with the empty pcDNA3 vector as a control. Three cell pools overexpressing each of the PHDs (PHD1, PHD2, and PHD3), and one control clone (CTL) were obtained. All stably transfected cells were maintained in complete media containing 600 g/ml G-418 as a selecting agent. Cells were serumstarved at least 2 to 3 h prior stimulation with TGF␤1.
Western Blot Analysis-HIF-1␣, PHD1, PHD2, and PHD3 were detected by Western blot analysis. Cells were serum starved and stimulated as indicated. Nuclear cell extracts were prepared as previously described (36). Total cell lysates and immunoblotting were performed as previously described (40). The membranes were probed overnight with primary antibodies. The secondary antibody was a peroxidase-conjugated anti-rabbit (Amersham Biosciences). Blots were developed using ECL Western blotting detection reagent (Amersham Biosciences).
Northern Blot Analysis-Cells were serum-starved and stimulated as indicated. Total cellular RNA was extracted from cells according to the previously described TRI Reagent protocol (41). Northern blot analysis was performed as previously described using a human ␣1-antitrypsin 32 P-labeled riboprobe (35).
Luciferase Assays-HepG2 cells were transiently transfected by the CaPO 4 precipitation technique using a Mammalian Cell Transfection Kit (Specialty Media, Lavallette, NJ) as previously described (42). Briefly, cells were transfected with 2 g/well of different luciferase reporter constructs, including a plasmid encoding CMV-Luc-HIF-1␣ ODDD (generously provided by Dr. Richard K. Bruick, University of Texas Southwestern Med-ical Center, TX), the previously described PRE-tk-LUC and 3TP-Lux vectors (43)(44)(45), as well as the human PHD2 proximal promoter construct pGL3b (1454/3172) P2PWT (generously provided by Dr. Eric Metzen, University of Luebeck, Luebeck, Germany), in the presence or absence of a Smad7 encoding vector. The pRL-SV Renilla reniformis luciferase expression vector was used as a control for transfection efficiency (Promega, Madison, WI). Twenty hours following transfection, cells were serum starved 2-3 h and stimulated with 0 to 5 ng/ml TGF␤1 for 4 to 16 h in the presence or absence of 10 M SB431542, which was added 15 min prior to the addition of TGF␤1. Cell lysates were assayed for luciferase activity using the Dual-Glo luciferase system (Promega). Data are presented as the -fold induction Ϯ S.E. of normalized relative luciferase units consisting of the ratio of luciferase activity produced by the studied luciferase vectors over R. reniformis luciferase activity.
Real-time PCR-HepG2 cells were seeded and the following day they were serum-starved and stimulated with 0 or 5 ng/ml TGF␤1 for 4 to 16 h in the presence or the absence of 10 M of the Smad chemical inhibitor SB431542, which was added 15 min prior to the addition of TGF␤1. Total RNA was isolated using the TRI Reagent protocol as described above. Reverse transcription was achieved using random decamer primers (Ambion). The following primer pairs were selected for PHD2, (forward) 5Ј-GCACGACACCGGGAAGTT-3Ј, (reverse) 5Ј-CCAGCTTCCCGTTACAGT-3Ј; and for 18 S, (forward) 5Ј-AGGAATTGACGGAAGGGCAC-3Ј, (reverse) 5Ј-GTGCAGCCCCGGACATCTAAG-3Ј. The quantitative real-time PCR was performed by Rotor-Gene 3000 (Corbett Research, Kirkland, Québec, Canada). The reaction mixtures containing SYBR Green were generated following the manufacturer's protocol. The cycling program was: initial denaturation at 95°C for 5 min, 40 cycles of amplification with an annealing temperature of 60°C for 45 s, and an extension at 72°C for 30 s. Results are expressed as copy numbers of PHD2 transcripts over 18 S (ϫ10 Ϫ3 ).
Measurement of VEGF-Cells were seeded at 60,000 cells/ well in 24-well plates and serum starved 3-4 h prior to overnight stimulation with 0 or 5 ng/ml of TGF␤1. VEGF concentrations in conditioned media were measured using a commercial Quantikine mouse VEGF immunoassay (R&D Systems, Minneapolis, MN) or human VEGF colorimetric enzyme-linked immunosorbent assay kit (Pierce).

TGF␤1 Induces HIF-1␣ Accumulation and Activity in Normoxic
Conditions-There is considerable evidence indicating that during tumor growth, large amounts of TGF␤1 are found within the tumor microenvironment and contribute to tumor progression (27). TGF␤ is described as a pro-angiogenic molecule, owing in part to its capacity to induce the expression of VEGF, a key target of the transcription factor HIF-1 (33,34). Following these observations, we first tested the ability of TGF␤1 to regulate HIF-1␣ expression, the HIF-1 regulatory subunit. Results expressed in Fig. 1A indicate that HIF-1␣ nuclear accumulation is increased up to 12-fold following a 6-h stimulation with 0.5 and 5 ng/ml TGF␤1 when compared with untreated cells, whereas stimulation with 10 ng/ml resulted in a lesser effect. In further experiments, we therefore used 5 ng/ml, a concentration that corresponds to physiological levels found within tumor and inflammatory microenvironments (46,47). Kinetics of HIF-1␣ protein accumulation in response to TGF␤1 is shown in Fig. 1B. Results indicate that an increase in HIF-1␣ nuclear expression was detected at 2 to 4 h, with maximum levels obtained at 8 and 16 h following stimulation.
To determine whether the HIF-1 protein induced by TGF␤1 is transcriptionally active, we used a luciferase reporter gene driven by 3 repeats of a specific hypoxia-responsive element retrieved from the erythropoietin gene (PRE-tk-LUC) (43,44). As presented in Fig. 1C, TGF␤1 treatment resulted in a significant increase in HIF-1-inducible luciferase reporter activity, FIGURE 1. Induction of HIF-1␣ protein in response to TGF␤1. HepG2 and/or HT1080 cells were serum starved and stimulated or not with TGF␤1 (A) at concentrations ranging between 0 and 10 ng/ml for 6 h or with 5 ng/ml TGF␤1 (B) for time periods ranging from 1 to 16 h. Nuclear extracts (100 g/lane) were resolved on 7.5% SDS-PAGE gels and immunoblotted using a rabbit antiserum specific to human HIF-1␣ or an anti-actin antibody as an internal control. C, HepG2 cells were transiently transfected with 2 g/well of PRE-tk-LUC promoter construct. Cells were incubated overnight in serumfree media in the presence or absence of 5 ng/ml TGF␤1 before luciferase activity measurement. Data are expressed as the mean Ϯ S.E., n ϭ 3.
TGF␤1 Induces HIF-1␣ Stability-The regulation of protein stability is a key event in controlling HIF-1␣ accumulation because the protein is subjected to rapid degradation under normoxic conditions (48,49). Therefore, we next examined whether TGF␤1 affects HIF-1␣ half-life. For this, cells were incubated for 6 h with 5 ng/ml TGF␤1 or 200 M CoCl 2 , a well described HIF-1␣ stabilizing agent (blocks HIF-1␣ degradation), prior to the addition of cycloheximide (CHX) to block ongoing protein synthesis (50). In the presence of CHX, the half-life of HIF-1␣ was ϳ30 min in CoCl 2 -treated cells, whereas it was ϳ15 min in TGF␤1-treated cells ( Fig. 2A). As reported, the addition of CHX to unstimulated cells resulted in a rapid decay of HIF-1 expression levels ( Fig. 2A) (51). Considering that in normoxic conditions HIF-1␣ protein half-life is less than 5 min (52), these results indicate that HIF-1␣ stabilization contributes to the accumulation of HIF-1␣ protein by TGF␤1.
We further investigated the impact of TGF␤1 on the stability of HIF-1␣ by transfecting HepG2 cells with a reporter plasmid expressing a protein that contains the HIF-1␣ ODDD fused with luciferase (CMV-Luc-ODDD). It is well established that HIF-1␣ degradation in normoxic conditions depends on posttranslational modifications that consist in the hydroxylation of the prolines 402 and 564 located within the ODDD (8,9,53). Thus, as previously demonstrated, the levels of the fusion protein, monitored by measuring luciferase activity, are negatively regulated by the hydroxylation of Pro 402 /Pro 564 , which permits HIF-1␣ binding to the von Hippel-Lindau and subsequent proteasomal degradation (50). Following the transfection with CMV-Luc-ODDD, HepG2 cells were stimulated for different time periods with 0, 2, or 5 ng/ml TGF␤1 or with 200 M CoCl 2 , used as a control. Results presented in Fig. 2B indicate that TGF␤1 increased luciferase-ODDD stabilization with maximal levels attained at 8 h of stimulation. Together, these results indicate that TGF␤1 treatment increases HIF-1␣ protein stabilization, possibly through impaired prolyl hydroxylation.
TGF␤1 Inhibits PHD2 Expression Levels-It has been established that prolyl hydroxylation by iron-dependant prolyl hydroxylases is a prerequisite for HIF-1␣ degradation in normoxia. Three mammalian prolyl hydroxylases (PHD1, PHD2, and PHD3) have been shown, in vitro, to hydroxylate key pro-FIGURE 2. TGF␤1 increases HIF-1␣ protein stability. A, HepG2 cells were serum-starved and cultured in the presence or absence of 200 M CoCl 2 or 5 ng/ml TGF␤1 for 6 h. Following stimulations, 5 M CHX was added to cell cultures for time periods ranging from 15 to 45 min. Nuclear extracts were analyzed by Western blotting using a rabbit antiserum specific to human HIF-1␣ or an anti-actin antibody as an internal control. The densitometry ratio of HIF-1␣ nuclear accumulation/actin is represented were 100% equal to HIF-1␣ levels in CHX non-stimulated cells. B, HepG2 cells were transiently transfected with 2 g/well of CMV-Luc-HIF-1␣ ODDD construct. Cells were serum-starved and incubated for time periods ranging from 4 to 16 h with 0 to 5 ng/ml TGF␤1 or 200 mM CoCl 2 before luciferase activity measurement. Data are expressed as the mean Ϯ S.E., n ϭ 2 to 5; *, p Ͻ 0.05, compared with non-stimulated cells.
line residues within the HIF-1␣ molecule (11). Because HIF-1␣ hydroxylation is impaired in TGF␤1-stimulated cells (Fig. 2B), we therefore evaluated the impact of TGF␤1 on the expression of HIF-1␣-associated prolyl hydroxylases. For this, HepG2 and HT1080 cells were incubated with TGF␤1 and the expression of PHD1, PHD2, and PHD3 was analyzed by immunoblotting. As presented in Fig. 3A, TGF␤1 dramatically decreased PHD2 protein levels in both cell lines with a sustained effect from 4 to 16 h. The TGF␤ effect is selective because no significant inhibition in PHD1 or PHD3 protein expression could be observed.
To determine whether the observed reduction of PHD2 protein levels is responsible for the accumulation of HIF-1␣ by TGF␤1, we performed a reconstitution experiment by overexpressing in HepG2 cells each of the PHD-cDNA under the control of a CMV promoter. Cell transfectants were stimulated with 0 or 5 ng/ml TGF␤1 and HIF-1␣ expression levels were measured. The respective overexpression of PHD1, PHD2, and PHD3 by each cell pool was confirmed by Western blotting (Fig. 3B). As presented in Fig. 3C, HIF-1␣ accumulation in response to TGF␤1 was specifically impaired in PHD2 overexpressing cells, indicating that the re-expression of PHD2 is sufficient to counter the effect of TGF␤1. These results, which are in line with previous data demonstrating that PHD2 is critical for controlling the low steady-state levels of HIF-1␣ in normoxia (14), designate the inhibition of PHD2 expression levels as the mechanism underlying TGF␤1-induced HIF-1␣ stabilization.
TGF␤1 Inhibits PHD2 Gene Expression through the Smad Signaling Pathway-Little is known about the cellular aspects controlling PHD2 expression. Only few stimuli, including hypoxia, insulin, and insulin-like growth factor-1, were shown (in contrast to TGF␤1) to increase PHD2 expression through transcriptional regulation (15)(16)(17)54). To gain further insight into the mechanism by which TGF␤1 negatively regulates PHD2 expression, we measured PHD2 mRNA levels in response to TGF␤1. As illustrated in Fig. 4A, real-time PCR analysis indicated that TGF␤1 treatment inhibited PHD2 mRNA expression by ϳ60% at the 4-and 8-h time points, with a milder effect observed at 16 h.
To determine whether TGF␤1-suppressed PHD2 mRNA expression is due to an effect on PHD2 transcription, we studied the regulation of the human PHD2 promoter. Previous characterization of the human PHD2 promoters revealed that in several cell lines, the PHD2 gene is exclusively transcribed from a promoter region located in a Cpg island near the coding sequence (17). Based on this information, a luciferase gene reporter construct driven by this promoter sequence was transfected in HepG2 cells to study the impact of TGF␤1 on PHD2 transcription. As presented in Fig. 4B, a decrease in luciferase activity to ϳ55% of control cells was observed following stimulation with TGF␤1, indicating that the PHD2 promoter is negatively regulated by TGF␤1.
Activation of TGF␤ receptors following ligand binding elicits downstream signaling events such as the phosphorylation of Smad proteins. In particular, Smad2 and Smad3 are defined has the cognate Smads activated in response to TGF␤1. These proteins form a complex with Smad4 that translocates to the nucleus to regulate the expression of target genes (55). To determine whether Smad proteins are involved in the suppression of PHD2 mRNA expression, HepG2 cells were pretreated 15 min prior stimulation with TGF␤1, with a chemical Smad inhibitor (SB431542), which specifically abrogates TGF␤1 receptor-mediated phosphorylation/activation of Smad2 and Smad3 (56). The capacity of this inhibitor to block TGF␤1induced transcription in our assays was confirmed using a luciferase reporter gene driven by 3 repeats of a specific Smad binding element retrieved from the plasminogen activator inhibitor-1 gene (3TP-Lux) (Fig. 4C) (45). The addition of SB431542 to TGF␤1-stimulated cells efficiently restored PHD2 mRNA expression (Fig. 4D), indicating the involvement of the Smad2/3 signaling pathway. Interestingly, cell pretreatment with SB431542 recovered PHD2 mRNA expression to levels even higher than the control. This effect might be attributed to FIGURE 3. TGF␤1 decreases PHD2 expression. HepG2 and HT1080 cells were serum-starved and stimulated or not with 5 ng/ml TGF␤1 for time periods ranging from 4 to 16 h. A, total cell lysates (75 g/lane) were resolved on 7.5 (PHD1 and PHD2) or 10% (PHD3) SDS-PAGE gels and immunoblotted using antibodies specific for PHD1, PHD2, or PHD3 and an anti-actin antibody as an internal control. B, HepG2 cells were stably transfected with pcDNA3-PHD1, pcDNA3-PHD2, or pcDNA3-PHD3 vectors or with pcDNA3 (empty vector). The overexpression of PHDs by each cell pool was analyzed by Western blotting as described above. C, cell transfectants were stimulated 16 h with 0 or 5 ng/ml TGF␤1 and nuclear extracts were analyzed by Western blotting using a rabbit antiserum specific to human HIF-1␣ or an anti-actin antibody as an internal control. AUGUST 25, 2006 • VOLUME 281 • NUMBER 34 the inhibition of signaling events mediated by endogenously produced TGF␤1, as demonstrated later in this study.

TGF␤1 Inhibits PHD2 Expression
We next examined whether Smad proteins are implicated in the suppression of PHD2 transcription by TGF␤1. For this, we analyzed PHD2 promoter activity after pretreatment with SB431542 or transfection of the inhibitory Smad, Smad7, a natural inhibitor of the phosphorylation of Smad2 and Smad3 and their association with Smad4 (55). The presence of Smad inhib-itors blocked the capacity of TGF␤1 to inhibit PHD2 promoter activity (Fig. 4E), indicating that the human PHD2 promoter is under negative regulation by TGF␤1 through activation of the Smad signaling pathway. In addition, we studied whether this signaling pathway is related to the reduced expression of PHD2 protein as well as the accumulation of HIF-1␣ in response to TGF␤1. As depicted in Fig. 5, A and B, SB431542 impaired both TGF␤1-mediated inhibition of PHD2 protein expression as well as HIF-1␣ accumulation in HepG2 and HT1080 cell lines. Together, these results demonstrate that TGF␤1 mediates its action through the activation of Smads, and indicate a correlation between the signaling pathway involved in HIF-1␣ protein accumulation as well as in decreased PHD2 expression levels.
Endogenous TGF␤1 Regulates HIF-1␣ and PHD2 Expression Levels-TGF␤1 is produced as an inactive precursor (pro-TGF␤1) that requires limited proteolytic cleavage to produce the mature and bioactive growth factor (57). Previous data obtained in our laboratory indicate that furin, the best characterized member of the mammalian proprotein convertases family, is the most important enzyme responsible for pro-TGF␤1 proteolytic processing (58). The autocrine production of bioactive TGF␤1 has been shown to mediate, in several cancer cells, the high constitutive expression of cancer-related genes, such as MMP-2 and MMP-9 (59 -61). To assess the possibility that endogenous TGF␤1 is involved in basal HIF-1␣ expression, we used a cell model previously established in our laboratory consisting of HT1080 cells stably transfected with ␣1-PDX, a serpin modified to acquire potent furin inhibition capacity (clones PDX-A, PDX-B, and PDX-C), or with an empty pCDNA3 vector (control clone) (62). Expression of ␣1-PDX resulted in the inhibition of the release of bioactive TGF␤1 in cellular supernatants (58,63). As illustrated in Fig.  6A, basal nuclear accumulation of HIF-1␣ observed in the HT1080 CTL clone was inhibited in cells expressing ␣1-PDX (PDX-A). We also used dec-RVKR-cmk, a synthetic furin inhibitory peptide that mimics the proprotein recognition site FIGURE 4. TGF␤1 impacts on PHD2 expression through the Smad signaling pathway. A, HepG2 cells stimulated with 0 or 5 ng/ml TGF␤1 for 4 -16 h. Total RNA was isolated and reverse transcribed using random decamer primers. Quantification of PHD2 mRNA levels was performed by real-time PCR. Results are expressed as the copy number Ϯ S.E. (n ϭ 3) relative to 18 S; *, p Ͻ 0.05, compared with non-stimulated cells. B, HepG2 cells were transiently transfected with the PHD2 promoter-Luc construct. Cells were serum-starved and stimulated 8 h with 0 or 5 ng/ml TGF␤1 before luciferase activity measurement. Data are expressed as the mean Ϯ S.E., n ϭ 3; *, p Ͻ 0.05, compared with non-stimulated cells. C, HepG2 cells were transiently transfected with the 3TP-Lux vector. Cells were serum-starved and stimulated overnight with 0 or 5 ng/ml TGF␤1 before luciferase activity measurement. Data are expressed as the mean Ϯ S.E., n ϭ 3, *, p Ͻ 0.05; compared with non-stimulated cells. D, HepG2 cells stimulated with 0 or 5 ng/ml TGF␤1 for 6 h in the presence or absence of 10 M SB431542. Following total RNA isolation and reverse transcription quantification of PHD2, mRNA levels were performed by real-time PCR. Results are expressed as copy number Ϯ S.E. (n ϭ 3) relative to 18 S, *, p Ͻ 0.05; compared with non-stimulated cells. E, HepG2 cells were transiently transfected with the PHD2 promoter-Luc construct and either pCMV5 (empty vector) or Smad7-encoding plasmid. Cells were serum-starved and stimulated 8 h with 0 or 5 ng/ml TGF␤1 in the presence or absence of 10 M SB431542 before luciferase activity measurement. Data are expressed as the mean Ϯ S.E., n ϭ 3, *, p Ͻ 0.05; compared with non-stimulated cells. (64). Similarly, treatment of the CTL clone with dec-RVKRcmk resulted in a dose-dependent decrease in nuclear HIF-1␣ protein levels. Because furin has the capability to process/activate other growth factors such as platelet-derived growth factor, which has been shown to also induce normoxic accumulation of HIF-1␣ in a particular cell system (65), we performed a growth factor reconstitution experiment to verify whether the impact of furin inhibition on basal HIF-1␣ expression in HT1080 cells was mediated by TGF␤1 (22,37,66). Stimulation of the PDX-A clone with TGF␤1 reconstituted HIF-1␣ expression to basal levels, suggesting that impaired bioactivation of TGF␤1 was responsible for the decrease in HIF-1␣ expression observed following expression of the furin inhibitor ␣1-PDX. Moreover, stimulation of HepG2 cells with platelet-derived growth factor-AA and platelet-derived growth factor-BB failed to induce HIF-1␣ expression (data not shown). To define whether the effect of endogenous TGF␤1 on HIF-1␣ accumulation is related to variations in PHD2 levels, we analyzed PHD2 protein expression in the same experimental conditions as used above. Results demonstrate that the inhibition of TGF␤1 processing by overexpression of ␣1-PDX or treatment with dec-RVKR-cmk increased PHD2 basal expression, whereas the addition of exogenous TGF␤1 to PDX-A clone reduced PHD2 expression levels (Fig. 6A). Thus, endogenous production of bioactive TGF␤1 increases the constitutive expression of HIF-1␣ in HT1080 cancer cells and is associated with decreased amounts of PHD2 protein.
TGF␤1-induced Expression of VEGF Is Partly Mediated by HIF-1-Several studies have demonstrated that TGF␤1 enhances VEGF production, a key target of HIF-1␣, in a wide variety of tumors in vivo and cells lines in vitro (67)(68)(69). We therefore examined whether the impaired production of bioactive TGF␤1 in ␣1-PDX overexpressing HT1080 cells could impact VEGF production. As presented in Fig. 6B, decreased levels of VEGF in cells supernatants were measured in cells expressing ␣1-PDX. The reduction of basal VEGF production was proportional to the expression of the furin inhibitor by the different PDX clones. Stimulation of the highest ␣1-PDX expressing clone (PDX-A) with TGF␤1 reconstituted VEGF production to levels similar to those produced by parental and CTL (empty vector) HT1080 cultures (Fig. 6C). Therefore, endogenous expression of TGF␤1 regulates both HIF-1 and VEGF expression in HT1080 cells.
The possible involvement of HIF-1 in VEGF expression induced by TGF␤1 was studied using two different HIF-1-inhibited cell systems. The first one consisted of HT1080 cells stably transfected with a dominant-negative form of HIF-1␣ (HIF-1␣DN). This truncated version of HIF-1␣, lacking both the N-terminal basic domain FIGURE 5. TGF␤1 effects are mediated by the Smad signaling pathway. HepG2 and HT1080 cells were serum-starved and treated or not with 10 M SB431542 15 min prior stimulation with 5 ng/ml TGF␤1 for 6 h. A, nuclear cell extracts (100 g/lane) were resolved on 7.5% SDS-PAGE gels and immunoblotted using a rabbit antiserum specific to human HIF-1␣ or an anti-actin antibody as an internal control. B, total cell lysates (75 g/lane) were resolved on 7.5% SDS-PAGE gels and immunoblotted using an antibody specific to human PHD2 or an anti-actin antibody as an internal control. FIGURE 6. Inhibition of TGF␤1 maturation decreases basal HIF-1␣ protein expression and increases PHD2 levels. HT1080 cells were stably transfected with the furin inhibitor ␣1-PDX (clones PDX-A, PDX-B, and PDX-C) or the empty vector (control clone CTL). A, cells were treated with 0 to 100 M of the convertase inhibitor decanoyl-Arg-Val-Lys-Arg-chloromethylketone (DEC-RVKR) or with 5 ng/ml of TGF␤1 for 6 h. Nuclear extracts (100 g/lane) were analyzed by Western blotting using a rabbit antiserum specific to human HIF-1␣ or an anti-actin antibody as an internal control. Total cell lysates (75 g/lane) were immunoblotted using an antibody specific to human PHD2 or an anti-actin antibody as an internal control. B, for quantitative determination of VEGF, cells were seeded at 6 ϫ 10 5 cells/well of 24-well plates, serum-starved, and cultured overnight in serum-free media. Measurement of VEGF within cell conditioned media was performed using an enzyme-linked immunosorbent assay specific to human VEGF. The expression of ␣1-PDX was analyzed by Northern blotting using a human ␣1 antitrypsin 32 P-labeled riboprobe. Data are expressed as the mean Ϯ S.E., n ϭ 2. C, HT1080 WT, control, and PDX-A cells were treated with 0 or 5 ng/ml TGF␤1 for 16 h prior VEGF measurement. Results are expressed as a representative experiment of three. AUGUST 25, 2006 • VOLUME 281 • NUMBER 34 required for DNA binding and the C-terminal transactivation domain, heterodimerizes with endogenous HIF-1␤, generating biologically inactive complexes that are unable to bind DNA (34). Northern blot analysis confirmed the expression of HIF-1␣DN (Fig. 7A, inset). As presented in Fig. 7A, a marked inhibition of VEGF levels was detected in HT1080 cells expressing HIF-1␣DN. Stimulation of the HIF-1␣DN cells with TGF␤1 partly reconstituted VEGF expression up to control levels, indicating that HIF-1 signaling is one of the mechanisms by which TGF␤1 induces VEGF. In the second system, we studied TGF␤1-induced VEGF expression in WT or HIF-1␣-deficient MEF. When stimulated with TGF␤1, a 16.6-fold increase in VEGF production was observed in WT MEF cells compared with a 7.1-fold increase in HIF-1␣ Ϫ/Ϫ cells (Fig. 7B). Immunoblot analysis demonstrated an efficient increase of HIF-1␣ in HIF-1␣-proficient MEF cells treated with TGF␤1 (Fig. 7C).

TGF␤1 Inhibits PHD2 Expression
Moreover, treatment of WT MEF cells with YC-1, a chemical inhibitor known to inhibit HIF-1␣ expression, resulted in impaired VEGF production in response to TGF␤1 (data not shown) (70,71). Therefore, these results clearly indicate that the HIF-1 complex induced by TGF␤1 contributes to the mechanism by which TGF␤1 stimulates VEGF production.

DISCUSSION
An increasing body of evidence indicates that in addition to hypoxia, several non-hypoxic stimuli enhance HIF-1␣ expression levels. Recently, the growth factor TGF␤1 emerged as an agonist that induces HIF-1␣ accumulation under normoxic conditions in different cell lines such as HT1080 and vascular smooth muscle cells (37,38). However, the mechanism underlying this regulation remained unexplored. In this study, we investigated the molecular mechanisms used by the proinflammatory and tumor growth factor, TGF␤1, to increase HIF-1␣ protein expression in normoxia. Here, we demonstrate that TGF␤1 utilizes a novel mechanism to induce HIF-1␣ protein accumulation in cancer cells, involving increased HIF-1␣ protein stabilization through decreased PHD2 expression. Our data represent the first evidence that links negative regulation of the oxygen-sensing enzyme PHD2 by a (patho-) physiological effector (TGF␤1) to changes in HIF-1␣ hydroxylation and degradation.
The regulatory subunit of HIF-1, HIF-1␣, is continuously degraded under normoxic conditions, but stabilized under hypoxia, which blocks its ubiquitin/proteasome-dependent degradation through the inhibition of prolyl hydroxylase activity (8 -10). However, several growth factors, hormones, and cytokines were shown to up-regulate HIF-1␣ protein levels in normoxia through mechanisms that generally differ from those used under low oxygen concentration. These stimuli generally act by increasing HIF-1␣ gene transcription and/or mRNA translation without affecting protein stability (25,26,72), and thus, can cooperate with hypoxia to induce HIF-1␣ accumulation in an additive manner (22). Data obtained from our laboratory indicate no additive effect on HIF-1␣ accumulation or activity when HepG2 cells were stimulated with TGF␤1 under hypoxic conditions (data not shown). This suggested that, as opposed to other non-hypoxic stimuli, TGF␤1 does not affect HIF-1␣ production but rather shares mechanistic similarities with hypoxia, involving increased protein stability, to induce HIF-1␣ accumulation.
We report here that TGF␤1 specifically targets PHD2 mRNA and protein expression. Decreased PHD2 expression in response to TGF␤1 is associated with inhibition of HIF-1␣ ODDD prolyl hydroxylation, which resulted in increased protein stability. Within the mammalian prolyl hydroxylase family, PHD2 is now regarded as the main cellular oxygen sensor that regulates HIF-1␣ degradation in normoxia. An elegant set of experiments using PHD2 small interfering RNA demonstrated that slight reduction of the PHD2 protein is sufficient to impact on HIF-1␣ stability in normoxia in a battery of human cell lines, whereas PHD1 and PHD3 silencing failed to increase HIF-1␣ expression levels (14). Consistent with this, the increase in HIF-1␣ accumulation due to TGF␤1 treatment was abrogated by overexpressing the PHD2 enzyme, whereas no significant The expression of HIF-1␣DN was analyzed by Northern blotting using a human HIF-1␣DN 32 P-labeled riboprobe. HT1080 control and HIF-1␣DN-expressing cell-conditioned media were obtained following overnight stimulation with 0 or 5 ng/ml TGF␤1 and assayed for VEGF quantification. Results are expressed as a representative experiment of two different ones. B, wild type (WT ) and HIF-1␣ knock-out MEF were seeded at 6 ϫ 10 5 cells/well of 24-well plates. Cells were cultured overnight in serum-free media in the presence or the absence of 5 ng/ml TGF␤1. Quantitative determination of VEGF in cell supernatants was assessed by using an enzyme-linked immunosorbent assay specific to murine VEGF. Data are expressed as the mean Ϯ S.E., n ϭ 2. C, WT and HIF-1␣ knock-out MEF cells were serum starved and stimulated with 0 or 5 ng/ml TGF␤1 for 6 h. Nuclear extracts (75 g/lane) were resolved on 7.5% SDS-PAGE gels and immunoblotted using a rabbit antiserum specific to human HIF-1␣ or an anti-actin antibody as an internal control.
impact was observed when similar experiments were performed with PHD1 or PHD3. Together with the decrease in HIF-1␣ ODDD prolyl hydroxylation, these results clearly identify the inhibition of PHD2 expression as the underlying mechanism involved in TGF␤1-induced HIF-1␣ protein stability.
Very little information is available on the molecular control of PHD genes. Our study revealed that TGF␤ decreased PHD2 gene transcription levels through the activation of Smad proteins. This signaling pathway seems to be central to the TGF␤ effect, because its inhibition restored PHD2 mRNA and protein levels and impaired HIF-1␣ protein accumulation in response to the growth factor. An increasing number of genes, such as the telomerase reverse transcriptase, c-myc, E-cadherin, and several genes involved in inflammation, where reported to be transcriptionally inhibited by TGF␤ (73)(74)(75). However, the mechanisms involved are likely complex and have therefore not been fully characterized. An analysis of the proximal regulatory region of the PHD2 gene indicated that it harbors three candidate Smad-binding elements (5Ј-agac-3Ј) that could mediate direct inhibitory effects (73). However, electrophoretic mobility shift assays using oligonucleotide encompassing each of these sites failed to reveal the formation of Smad-specific complexes (data not shown), suggesting an indirect action of Smad proteins on the PHD2 gene. Consistent with this possibility, the Smad signaling pathway was shown to down-regulate E-cadherin expression through the induction of repressors such as the Snail and Smad-interacting protein SIP1 (76,77). A preliminary analysis of the PHD2 gene indicates the presence of CAC-CT(G) clusters that could, in principle, act as a SIP1 cis-acting sequence (76). Efforts are presently deployed to address the possible role of Smad-induced repressors in PHD2 regulation.
Recent studies established that PHD2 is a direct HIF-1 target gene. A consensus HIF-1 binding site was found within the PHD2 promoter sequence and this cis-acting sequence was shown to be responsible, at least in part, for the induction of human PHD2 gene by hypoxia (16,17). It was therefore somewhat surprising to observe that TGF␤, which rapidly increased HIF-1 levels (starting at 2 h), inhibited PHD2 protein for up to 16 h. This could be explained by recent results indicating that the induction of PHD2 by hypoxia has both HIF-1-dependent and -independent components. Indeed, short-term hypoxia (4 h) resulted in HIF-1-independent induction of PHD2, whereas PHD2 accumulation by prolonged hypoxia (16 h) was only partly inhibited by small interfering RNA-mediated degradation of the HIF-1␣ subunit (15). Therefore, it is conceivable that TGF␤, in contrast to hypoxia, does not recruit the cofactors involved in short-term induction of PHD2 levels. Whether prolonged treatment with TGF␤ modulates PHD2 levels through HIF-1 induction is unknown, but our observation that PHD2 transcription was inhibited to a greater extent at 4 -6 h compared with 16 h suggests a balance between the inhibitory Smad and the stimulatory HIF-1 pathways for PHD2 transcription.
NO donors and reactive oxygen species-inducing agents were previously shown to increase HIF-1␣ stability through the inhibition of PHD enzymatic activity (78). Because TGF␤1 was previously demonstrated to induce the production of reactive oxygen species in multiple cell types (79,80), we examined the possibility that in addition to the regulation of PHD2 expression, TGF␤1 also regulates the enzymatic activity of prolyl hydroxylases. Results obtained from intracellular reactive oxygen species measurement assays using a cellpermeant molecule that fluoresces when oxidized by peroxides (carboxy-H 2 DCFDA), indicated that, in our experimental conditions, HepG2 cells do not produce significant levels of reactive oxygen species in response to TGF␤1 (data not shown). Moreover, the complete inhibition of TGF␤1-induced HIF-1␣ accumulation observed in PHD2 overexpressing cells indicates that TGF␤1 does not impair the ability of PHD2 to inhibit HIF-1␣ stability. These observations, together with the fact that stimuli that down-regulate prolyl hydroxylases activity increase HIF-1␣ expression more rapidly (less than 30 min) as compared with TGF␤1 (Ϸ2-4h) (78) indicate that regulation of PHDs activity does not account for the action of TGF␤1 on HIF-1␣ accumulation.
Several studies have reported normoxic HIF-1␣ expression in several tumors in vivo and cell lines in vitro (81)(82)(83). Interestingly, numerous tumor cells express high levels of both TGF␤1 and its convertase furin, an enzyme recently identified in our laboratory as a HIF-1␣-regulated target (84 -87). In this context, we set out to investigate whether the endogenous production of bioactive TGF␤1 by tumor cells could be involved in regulating HIF-1␣ normoxic expression. As expected, impairment in the release of bioactive TGF␤1 through inhibition of furin resulted in a marked decrease in HIF-1␣ expression levels. This effect was recovered following the addition of TGF␤1, FIGURE 8. Schematic diagram of the mechanisms by which TGF␤1 regulates HIF-1␣. Tumor and inflammatory milieu express large amounts of TGF␤1. TGF␤1 will use selected signal transduction pathways to increase HIF-1␣ accumulation by increasing HIF-1␣ protein half-life through inhibition of PHD2 expression levels. The HIF-1 complex will then impact on inflammation and/or tumor progression by enhancing genes implicated in angiogenesis, cell invasion, and chemotaxis as well as cell growth and survival.
consistent with the idea that endogenous levels of TGF␤1 efficiently regulates basal HIF-1␣ expression. Such a process may permit up-modulation of furin expression and create an activation/regulation cycle favoring the production of bioactive TGF␤1 and the accumulation of HIF-1␣.
HIF-1-mediated transcription of VEGF, a potent angiogenic and survival factor, is a primary mechanism whereby hypoxia induces angiogenesis. The potent angiogenic factor, TGF␤, was also shown to induce VEGF production in several cell lines through transcriptional activation of its encoding gene. The possible involvement of HIF-1 in this regulation has been proposed, however, the exact contribution of HIF-1 in TGF␤1induced VEGF production remained unexplored (38). Using two different HIF-1-inhibited cell systems, we demonstrated that TGF␤1-induced HIF-1 also plays a significant role (40 -50%) in VEGF production in response to TGF␤1. This result provides an additional mechanistic link between TGF␤1 and VEGF expression, and suggests that the same transcription factor HIF-1 lies at the heart of VEGF production triggered by both the hypoxic and the inflammatory zones of the tumor and/or inflamed tissue.
Finally, our data suggest that HIF-1 may act as a protumorigenic and proinflammatory factor under normoxia in part through TGF␤-mediated inhibition of PHD2 levels, a mechanism distinct from the ones known to be used by other growth factors and proinflammatory molecules. Because a combination of cytokines and growth factors such as tumor necrosis factor-␣ and interleukin-1 are frequently detected in the tumor and/or inflammatory milieu, it can be suggested that multiple HIF-1 regulation checkpoints coexist to raise HIF-1 to levels needed to meet the increase in biologic and metabolic demand of the tissue. This would result in an increased expression of specific genes such as VEGF (Fig. 8). Because TGF␤ is abundantly and widely expressed by both inflammatory and tumor cells, and in view of recent advances indicating that this growth factor is key in the promotion of both tumor growth and inflammation, it appears that agents that counteract or overcome prolyl hydroxylase inhibition may provide a therapeutic advantage.