Role of Iron (II)-2-Oxoglutarate-dependent Dioxygenases in the Generation of Hypoxia-induced Phosphatidic Acid through HIF-1/2 and von Hippel-Lindau-independent Mechanisms*

Hypoxia-inducible factors (HIF-1/HIF-2) govern the expression of critical genes for cellular adaptation to low oxygen tensions. We have previously reported that the intracellular level of phosphatidic acid (PA) rises in response to hypoxia (1% O2). In this report, we have explored whether components of the canonical HIF/von Hippel-Lindau (VHL) pathway are involved in the induction of PA. We found that hypoxia induces PA in a cell line constitutively expressing a stable version of HIF-1α. PA induction was also found in HIF-1α- and 2α-negative CHO Ka13 cells, as well as in HIF-β-negative HepaC4 cells. These data indicate that HIF activity is neither sufficient nor necessary for oxygen-dependent PA accumulation. PA generation was also detected in cells deficient for the tumor suppressor VHL, indicating that the presence of VHL was not required for the induction of PA. Here we show that PA accumulation also occurs at moderate hypoxia (5% O2), although to a lesser extent to that seen at 1% O2, revealing that PA is induced at the same hypoxia range required to activate HIF-1. Prolyl hydroxylases (PHD) and asparaginyl hydroxylase (FIH) belong to the iron (II) and 2-oxoglutarate-dependent dioxygenase family and have been proposed as oxygen sensors involved in the regulation of HIFs. Chemical inhibition of these activities by treatment with iron chelators or 2-oxoglutarate analogs also results in a marked PA accumulation similar to that observed in hypoxia. Together these data show that PA accumulation in response to hypoxia is both HIF-1/2- and VHL-independent and indicate a role of iron (II)-2-oxoglutarate-dependent dioxygenases in the oxygen-sensing mechanisms involved in hypoxia-driven phospholipid regulation.

Conditions ranging from severe to moderate oxygen supply deficiency (hypoxia) to tissues occurs under different pathophysiological conditions (1,2). Cells trigger a compensatory response that is critical for cellular survival under these lowered oxygen conditions (3,4). The best characterized response to hypoxia is the induction of HIF 1 transcription factors (5) that control oxygen-dependent induction of a series of genes including those encoding the vascular endothelial growth factor (VEGF), erytropoietin (EPO), and glycolytic enzymes (6, 7) among others. HIFs are heterodimers composed of ␣ and ␤ subunits, which belong to the basic helix-loop-helix PAS (Per, Arnt, Sim) family of transcription factors (8). The ␤ subunit, also known as the aryl hydrocarbon receptor nuclear translocator, ARNT, is constitutively expressed, while the ␣ subunit is tightly oxygen-regulated. There are three types of ␣ subunits, HIF-1␣, HIF-2␣ (endothelial PAS domain; EPAS), and HIF-3␣ (9), encoded by different genes. In normoxia, the ␣ subunits of HIF are very unstable proteins since they interact with the tumor suppressor von Hippel-Lindau (VHL) (10), which is a component of the E3 ubiquitin ligase complex (11) that leads to ubiquitinization and subsequent proteasome-dependent degradation of HIF ␣ subunits (12,13). VHL recognizes HIF ␣ subunits through two independent sites that contain the common motif LXXLAP present in both HIF-1␣ and HIF-2␣ (12). It has been recently demonstrated that hydroxylation of specific proline residues (underlined) into this conserved box is required for the interaction of VHL with HIFs (14,15). A novel family of mammalian proline hydroxylases: PHD1, PHD2, and PHD3, which belong to the iron (II)-2-oxoglutarate-dependent dioxygenase family, have been involved in the hydroxylation of HIF proline residues (14,16,17). In the absence of oxygen, these post-translational modifications do not occur; pVHL cannot recognize HIF ␣ subunits and as a consequence, HIF is acutely induced. Thus, under hypoxia, HIF␣ proteins are accumulated, migrate to the nucleus, associate with ␤ subunits, and lead to the subsequent formation of the HIF-1␣/␤ and HIF-2 ␣/␤ heterodimers that bind to DNA at the specific HRE (hypoxia response elements) of different genes (18). More recently it has been reported that hypoxia-dependent transactivation activity of HIFs is controlled by hydroxylation of an asparaginyl residue in the C-terminal transactivation domain. FIH has been identified as the asparaginyl hydroxylase involved in this hydroxylation (19,20). Therefore PHDs and FIH have been * This work was supported by grants from the Fondo de Investigaciones Sanitarias (CO3/01), and from the Ministerio de Ciencia y Tecnología (SAF 2001-0215). 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.
‡ named as HIF hydroxylases and have been proposed as the oxygen sensors that control HIF activity (21)(22)(23).
In order to respond to multiple extracellular stimuli, cells often generate lipid second messengers (24). Phosphatidic acid (PA) has been proposed as one of these intracellular second messengers (25) and has been suggested to participate in different cellular events such as cell proliferation and actin polymerization (26 -29) as well as in inflammatory responses and cytokine secretion (30). PA has also been implicated to modulate the activity of numerous enzymes such as Raf kinase (31,32), protein phosphatase I (33), as well as the mammalian target of rapamycin, mTOR (34).
We have previously observed that hypoxia (1% O 2 ) leads to an increase in the intracellular PA level, most probably through the action of a DGK activity (35). In this work we have characterized the oxygen-dependent mechanisms involved in the generation of PA induced after oxygen deprivation. Here we show that hypoxia-dependent PA accumulation does not require the activity of HIF-1 nor HIF-2 transcription factors or that of VHL. Moreover, we also provide evidence for the role of iron (II)-2-oxoglutarate-dependent dioxygenase in oxygen mechanisms involved in PA generation by low oxygen tension.
Cell Culture, Cell Treatments, and Hypoxic Conditions-Human embryonic kidney cells 293-T, as well as UMRC cells (36) and 786-O (37), derived from human renal carcinomas were grown in RPMI 1640 medium with GLUTAMAX-I (Invitrogen, Life Technologies Ltd.). The 786-O-PP13 clone was derived from 786-O-WT8 cells upon stable expression of P402A, P564G mutant HIF-1, and grown in the presence of hygromycin B (Roche Applied Science). HeLa cells were grown in Dulbecco's minimal essential medium (Biochrom KG, Berlin, Germany). CHO 4.5 and Ka13 cells (38) were grown in Ham's Nutrient Mixture F-12 (Euroclone); Hepa C1 and C4 cells were maintained in ␣-modified Eagle's medium ␣MEM (Biochrom). All the cells were cultured in the presence of 10% (v/v) fetal calf serum (Labtech International Ltd., Woodside, UK). Cells were routinely cultured in 95% air, 5% CO 2 (normoxic conditions) at 37°C. To expose cells to hypoxia, they were placed into an airtight chamber with inflow and outflow valves that were infused with a mixture of 1% O 2 , 5% CO 2 , 94% N 2 (S. E. Carburos Metalicos S. A., Madrid, Spain) or placed in an in vivo 400 hypoxia workstation (Ruskinn Technology, West Yorkshire, UK). In all experiments, cells were plated at 70 -90% confluence, and when completely attached they were exposed to normoxia or hypoxia. In those experiments in which inhibitors were used, these compound were added 30 min before other treatments.
Measurement of [ 32 P] Phosphatidic Acid-Cells were cultured in phosphate-free medium supplemented or not with 10% (v/v) fetal calf serum (extensively dialyzed against 0.9% (w/v) NaCl) for 90 min before the addition of [ 32 P]orthophosphate (100 Ci/ml) for an additional 90 min. Thereafter, the cells were exposed to normoxia or hypoxia. Phospholipids were then extracted by the method of Bligh and Dyer (39) and analyzed as previously described (35). Quantification of the band corresponding to [ 32 P]PA was performed using the Image Reader v1.8 software (Science Lab software, Fuji Photo Film).
Transient Transfection and Luciferase Assays-Ka13 cells were transfected with 3.3 g of expression plasmid for HIF-1␣ or pcDNA3 without the insert in the presence of 0.4 g of the reporter plasmid p9HIF1-Luc (35) and 0.3 g of pRLTK (Promega). Transfections were performed using Lipofectin (Invitrogen). 7-8 h after transfection cells were split into the appropriate number of wells (35-mm plates). After 12 h of expression, the cells were incubated under either normoxic or hypoxic 1% O 2 conditions for 5-6 h. Finally, cells were harvested and Western blot assays as well as luciferase assays were performed in each case. Firefly and Renilla luciferase activities were determined using a dual luciferase system (Promega), and firefly luciferase activity was normalized based on the Renilla luciferase activity.
Western Blotting-Proteins from total cell lysates were resolved using 8 -10% polyacrylamide-SDS gel. Proteins were then transferred to a nitrocellulose membrane (Bio-Rad), blocked with 5% nonfat dry milk in TBS-T (50 mM Tris, pH 7.6, 150 mM NaCl, 0.1% Tween-20), and incubated overnight at 4°C with the indicated antibodies. Immunolabeling was detected by enhanced chemiluminescence (ECL, Amersham Biosciences) and visualized with a digital luminescent image analyzer (FUJIFILM LAS-1000 CH).

PA Generation Induced by Hypoxia Is HIF-1/2-independent-
We have previously reported that hypoxia (1% O 2 ) induces PA accumulation in different cell types (35). The fact that low oxygen tension induces both PA accumulation and HIF-1␣ protein stabilization raised the possibility that the intracellular PA increment depends on HIF activation after exposure to hypoxia. To assess this possibility, we used the 786-O-PP13 clone, which constitutively expresses HIF-1␣ protein. These cells derive from the wild-type 786-O, that express HIF-2 but not HIF-1␣ (Fig. 1, top panel), by stable transfection of a mutated version of HIF-1 at the 402 and 564 proline residues. This construct is constitutively stabilized, because it cannot be hydroxylated by HIF hydroxylases, and therefore VHL cannot recognize this protein. This clone expresses HIF-dependent genes despite the oxygen concentrations (40). To rule out whether this cell line has constitutively elevated PA levels, we analyzed PA content in these cells. We found that 786-O as well as the mutated version 786-O-PP13 had similar PA basal levels (data not shown), and both were able to induce PA in response to hypoxia (Fig. 1, bottom panel). From these results we can conclude that HIF-1␣ stabilization per se is not sufficient to trigger PA accumulation.
To further study whether PA induction depends on HIF activity, we decided to use the previously described CHO 4.5 cell line, that does not express HIF-2␣ and the CHO Ka13, obtained from CHO 4.5 by mutagenesis and which also lacks HIF-1␣ expression (Fig. 2B, top panel). Therefore CHO Ka13 cells do not exhibit inducible HRE-dependent transcriptional activity (38). However, the oxygen mechanisms involved in the activation of HIF are functional in these cells, since an exog-enously transfected HIF-1␣ subunit is regulated by hypoxia ( Fig. 2A, top panel) and is able to drive the transcription of an HRE-linked reporter plasmid ( Fig. 2A, bottom panel). Importantly, we found that exposure of the Ka13, as well as the parental cell line 4.5, to hypoxia (1% O 2 ) results in an increase in the level of PA (Fig. 2B) despite the lack of HIF-1/2. In addition, we also utilized the widely used HepaC4 cells, that do not express a functional HIF␤ subunit, in contrast to the parental HepaC1 cells, and as a consequence, HIF proteins are not able to transactivate target genes (41,42). Here we found that PA was also induced in response to hypoxia in both cell lines (Fig. 2C). Together all the data indicated that neither HIF-1 nor HIF-2 factors are necessary for PA accumulation by hypoxia.
VHL Function Is Not Required for PA Induction in Response to Hypoxia-Since VHL plays a critical role in the canonical oxygen-sensing pathway, we investigated its role in hypoxia- FIG. 2. Functional HIF activity is not required for PA accumulation by hypoxia. A, CHO Ka13 cells were transiently transfected with human HIF-1␣ expression plasmid (pHIF-1␣) or with the empty vector (pCDNA3) in the presence of the reporter plasmids p9HIF-1 Luc and pRLTK, and subsequently exposed to hypoxia (Hx, 1% O 2 ) or left at normoxia (N) for 5-6 additional hours. HIF-1␣ exogenous protein level was determined by Western blot assay (top panel), and luciferase activity was analyzed and normalized by Renilla activity in each case. We assigned the value of 1 to the control normoxic cells and the rest of samples were represented as fold over the control. One experiment representative of three independent experiments is shown (bottom panel). B, [ 32 P]orthophosphate metabolically labeled CHO 4.5 and CHO Ka13 cells were exposed to hypoxia, (Hx, 1% O 2 ) for 5-6 h or left in normoxia (N) for the same period of time. The endogenous level of HIF-1␣ protein was determined by Western blotting as a control of hypoxia stimulation (top panel). We analyzed the PA content (bottom panel) as described in Fig. 1. Each bar represents the PA fold induction (mean Ϯ S.D.) over normoxia from three independent experiments performed in triplicate. C, Hepa C1 and Hepa C4 cells were [ 32 P]orthophosphate metabolically labeled and exposed to hypoxia (Hx, 1% O 2 ) for 5-6 h or maintained at normoxia (N). Cellular phospholipid extraction and PA analysis were performed as described above. Each bar represents the PA fold induction (mean Ϯ S.D.) from three independent experiments performed in triplicate. Hypoxia stimulation was controlled by Western blotting against HIF-1␣ (data not shown).  A and B, top panels). PA content was analyzed as described in Fig. 1. Each bar represents the average PA fold induction over normoxia from two independent experiments performed in duplicate (A and B, bottom panels). mediated PA elevation. For this purpose, we used the previously described UMRC (36) and 786-O (37) cell lines, which lack the vhl tumor suppressor gene. The major HIF isoform expressed by UMRC cells is 1␣, whereas the 786-O cell line expresses mainly HIF-2␣ (data not shown). The absence of VHL results in a constitutive accumulation of HIF factors in normoxia (13,36), which was not further increased upon hypoxic conditions (Fig. 3, A and C). Interestingly, we found a significant intracellular PA induction by hypoxia in both cell lines regardless of the presence or absence of VHL (Fig. 3, B  and D). These results indicate that the presence of VHL is not necessary for hypoxia-dependent PA accumulation. It is important to note that in the case of UMRC6 VHL (Ϫ) cells, which exhibit HIF-1␣ protein that is completely stabilized (Fig. 3C), the PA basal level was significantly lower than that in UMRC3.4 VHL (ϩ) cells (Fig. 3D). These data support our observation above, suggesting that HIF activation is neither sufficient nor necessary for hypoxia-induced PA accumulation. Taken together the experiments described so far indicate that neither HIF-1/2 nor VHL are involved in the induction of PA triggered by hypoxia.
PA Accumulation Is Induced by Moderate Hypoxia-Since we have observed PA accumulation at hypoxia (1% O 2 ) and we and other investigators (35,43) could observe HIF-1␣ stabilization at conditions ranging from 1 to 5% O 2, (Fig. 4, top panels), we decided to analyze whether hypoxia-induced PA accumulation also occurred at this range of oxygen tension. For this purpose, we exposed [ 32 P]orthophosphate metabolically labeled 293-T and HeLa cells to 1 and 5% oxygen concentrations. As shown in Fig. 4 (bottom panels), the accumulation of PA was detected under both hypoxic conditions in both cell lines. Importantly, we found that there was a good correlation between the levels of HIF and PA induced by different degrees of hypoxia, 1% O 2 giving a greater accumulation of PA (Fig. 4). These data suggest that the oxygen-sensing mechanisms controlling HIF could be involved in hypoxia-induced PA generation.
Inhibition of Iron (II)-2-Oxoglutarate-dependent Dioxygenases Induces PA Accumulation-HIF proline hydroxylases (PHDs) and HIF asparaginyl hydroxylase (FIH) are iron (II)-2-oxoglutarate-dependent dioxygenases previously proposed to act as oxygen sensors that directly govern HIF activity (14,16,17). The similar oxygen dependence of PA induction and HIF-1␣ protein stabilization, led us to ask whether cellular iron (II)-2-oxoglutarate-dependent dioxygenase activity could function to regulate hypoxia-induced PA accumulation. First, we measured the intracellular PA level in the presence of the iron chelator agent deferoxamine (Df) and the transition metal cobalt chloride (CoCl 2 ), which are widely used as HIF activators. Both compounds are potent inhibitors of PHD and FIH because of the iron requirement of these enzymatic activities (21)(22)(23). As it is shown in Fig. 5, the treatment of 293-T cells with these agents, at the same concentration required to accumulate HIF-1␣ (top panel), was able to elevate the intracellular level of PA (bottom panel). These results indicate that these compounds mimic hypoxia not only by their positive effect upon HIF-1␣ expression, but also by increasing the level of intracellular PA.
In addition to iron and oxygen, the above mentioned family of dioxygenases also requires 2-oxoglutarate to catalyze the hydroxylation reaction. Therefore, to further investigate the possible role of these enzymes in PA accumulation, we used a series of structurally distinct 2-oxoglutarate analogs previously described. As shown in Fig. 6A, treatment of 293-T cells with dimethyl oxalylglycine (DMOG) (16), increased HIF-1␣ protein levels (top panel), and the same doses induced PA accumulation (bottom panel). To support these data we also used the known dioxygenases inhibitors L-mim (44) and 3,4-DHB (45). It has been recently described that exposure of cells to these compounds induces HIF-1␣ protein since they are 2-oxoglutarate analogs (46). 293-T cells treated with 3,4 DHB or L-mim stabilized the HIF-1␣ subunit in a dose-dependent manner (Fig. 6,  B and C, top panels). Importantly, a parallel increase in the PA level was also observed (Fig. 6, B and C, bottom panels). Therefore we conclude that chemical inhibition of iron (II)-2-oxoglutarate-dependent dioxygenase activity mimic hypoxia-driven PA generation. These findings and the fact that the generation of PA occurs at the same oxygen tensions that activate HIF-1 (see Fig. 4), led us to propose that HIF hydroxylases or other iron (II)-2-oxoglutarate-dependent dioxygenases with similar oxygen sensitivity are involved in PA accumulation after oxygen deprivation. DISCUSSION We have previously reported that hypoxia-induced PA accumulation is involved in HIF-1 expression (35). This fact led us to hypothesize that oxygen-sensing mechanisms involved in PA accumulation are upstream and independent of HIF-1 activation. Here we have probed our speculation using different approaches. First we found that the expression of a stably version of HIF-1␣ (786-O-PP13) did not trigger PA induction. Second, the PA increase occurred in the absence of HIF-1 and HIF-2 activity in Ka13 cells as well as in Hepa C4. Third, in VHL- negative cells, which have constitutive HIF activity, we did not observe a constitutively elevated PA level. Moreover, unlike HIF, PA was induced by hypoxia in VHL-deficient cells. These data reveal the existence of HIF-1/2-and VHL-independent mechanisms involved in PA generation upon cellular exposure to hypoxia.
Our findings are in agreement with previous reports indicating the existence of several hypoxia-induced processes that are HIF-independent. Some of these events are the hypoxia-driven induction of heme oxygenase-1 expression (38,47), phosphorylation of the translation initiation factor eIF2␣ (48), ion channel activity regulation (49), down-regulation of mammalian target of rapamycin mTOR and its targets (50), as well as the induction of EGR gene expression (51), which are not impaired in HIF-deficient cells, such as Ka13 or Hepa C4, which are widely used as models to demonstrate HIF-independent processes (47,52,53).
Because of the existence of these HIF-independent mechanisms in response to hypoxia and the role of PA as intracellular second messenger, our data raise the question of whether hypoxia-induced PA accumulation may have a role, not only on HIF-1␣ activation (35) but also on other HIF-independent processes triggered by hypoxia. Particularly interesting cases are mTOR and Raf-1, previously identified as intracellular targets of PA. In this regard it has been reported that PA generation upon mitogen stimulation plays a role in the regulation of these two intracellular kinases (31,32,34). There are several reports indicating the regulation of mTOR activity in response to hypoxia (50,54,55). The possibility that hypoxia-induced PA generation could be involved in the regulation of mTOR should be considered. Moreover Raf-1 has been described to be activated by hypoxia in endothelial cells (56), indicating the possibility that PA could also be mediating this induction. Therefore future experiments will be designed to answer the question of whether PA generated in hypoxia constitutes a key intracellular regulator of certain hypoxia-induced processes. Because the most therapeutic approaches against pathological hypoxia are focused in HIF, the study and characterization of parallel and independent pathways will be of great interest.
HIF hydroxylases (PHDs and FIH) have been proposed as oxygen sensors involved in the regulation of HIF transcription factors (21)(22)(23). The use of previously recognized inhibitors of these dioxygenase activities result in a strong induction of HIFs (16,46). Here we found that cellular exposure to variable doses of these inhibitors are able to induce HIFs and also resulted in marked PA generation. These data suggest that hypoxia and subsequent inhibition of hydroxylase activity could lead to PA accumulation. In addition, the fact that PA is tightly modulated by oxygen tension (see Fig. 4), suggests that one possible candidate to account for this PA accumulation could be HIF hydroxylases (PHD and/or FIH), since they are much more sensitive to small changes in O 2 than other iron (II)-2-oxoglutarate-dependent dioxygenases (57). However, our results do not exclude the possibility that other members of the iron (II)-2-oxoglutarate-dependent dioxygenases family, different from PHD or FIH, could be involved in PA induction by hypoxia. Further analysis is required to fully determine the involvement of these enzymatic activities on PA generation by hypoxia.
Our data also suggest that the cellular mechanisms involved in hypoxia-induced PA generation are putative new targets for iron (II)-2-oxoglutarate-dependent dioxygenases. We could therefore suggest that enzymatic activities involved in PA metabolism could be regulated by these novel post-translational modifications. This possibility could extend the range of effects mediated by iron (II)-2-oxoglutarate-dependent dioxygenases after exposure to hypoxia in an HIF-independent manner. Our results indicate that cellular hypoxia is not totally orchestrated by HIF transcription factors and lead us to hypothesize, for the first time, the existence of new targets for this novel family of dioxygenases that control PA generation in response to low oxygen tension.