A Conserved Family of Prolyl-4-Hydroxylases That Modify HIF

See allHide authors and affiliations

Science  09 Nov 2001:
Vol. 294, Issue 5545, pp. 1337-1340
DOI: 10.1126/science.1066373


Mammalian cells respond to changes in oxygen availability through a conserved pathway that is regulated by the hypoxia-inducible factor (HIF). The alpha subunit of HIF is targeted for degradation under normoxic conditions by a ubiquitin-ligase complex that recognizes a hydroxylated proline residue in HIF. We identified a conserved family of HIF prolyl hydoxylase (HPH) enzymes that appear to be responsible for this posttranslational modification. In cultured mammalian cells, inappropriate accumulation of HIF caused by forced expression of the HIF-1α subunit under normoxic conditions was attenuated by coexpression of HPH. Suppression of HPH in cultured Drosophila melanogaster cells by RNA interference resulted in elevated expression of a hypoxia-inducible gene (LDH, encoding lactate dehydrogenase) under normoxic conditions. These findings indicate that HPH is an essential component of the pathway through which cells sense oxygen.

The ability of cells to recognize and respond to a low-oxygen environment (hypoxia) is critical in many physiological and pathophysiological conditions (1). Almost all mammalian cells express components of a hypoxia response pathway that is conserved in both flies (2, 3) and worms (4). The hypoxia-inducible transcription factor HIF lies at the heart of this pathway. HIF is a heterodimer composed of two members of the basic-Helix-Loop-Helix (bHLH)–containing PER-ARNT-SIM (PAS) domain family, HIF-1α and HIF-1β, also known as the aryl hydrocarbon receptor nuclear translocator (ARNT) (5). Under normoxic conditions, HIF-1α is constitutively expressed. However, this subunit is rapidly targeted for proteasome-mediated degradation (6–8) through a protein-ubiquitin ligase complex containing the product of the von Hippel Lindau tumor suppressor protein (pVHL) (9–12). pVHL recognizes the oxygen degradation domain (ODD) of HIF-1α only under normoxic conditions (13–15). When cells are exposed to a hypoxic environment, this degradation pathway is blocked, thereby allowing HIF-1α to accumulate and migrate to the nucleus, where it activates hypoxia-responsive genes [reviewed in (16)].

pVHL recognizes the ODD in HIF-1α through a conserved proline residue that is hydroxylated exclusively under normoxic conditions (13–15). Cellular extracts prepared under normoxic conditions contain a prolyl-4-hydroxylase activity capable of modifying a proline-containing peptide derived from the ODD, whereas extracts prepared under hypoxic conditions or in the presence of “hypoxia mimics” such as CoCl2 or the iron chelator deferoxamine mesylate are deficient in this activity (13–15). As is the case for known prolyl-4-hydroxylases, this activity is enhanced by supplementation with Fe2+, ascorbate, and 2-oxoglutarate (14, 15).

The best characterized prolyl-4-hydroxylases modify collagen as it matures along the exocytotic pathway (17). These enzymes are typically composed of two α and two β subunits, with the α subunit responsible for the prolyl hydroxylase activity. Three lines of evidence rule out these enzymes as the HIF prolyl hydroxylase. First, the substrate context of the modified proline residues in collagen is different from that surrounding the relevant proline residue in HIF (13–15). Second, the collagen-modifying enzymes reside within the endoplasmic reticulum rather than the cytoplasm, as expected for the HIF prolyl hydroxylase (17). Finally, two recombinant isoforms of the collagen-modifying enzymes expressed from baculovirus were reported to show no activity against the HIF substrate (14). To identify prolyl-4-hydroxylases, we queried the GenBank database for sequences related to the catalytic α subunit of the collagen-modifying prolyl-4-hydroxylases. Of the several families of putative prolyl hydroxylases in the database, five contained human homologs. Furthermore, each of these families contained conserved amino acid residues believed to bind Fe2+ and 2-oxoglutarate (17).

To investigate whether any of these candidate enzymes specify HIF-1α–directed prolyl hydroxylase activity, we cloned a representative member of each family into the pcDNA3.1/V5-His expression vector (18). These five polypeptides, designated candidates A, B, C, and D, and one candidate subsequently designated HIF prolyl hydroxylase-1 (HPH-1), were in vitro transcribed/translated in a rabbit reticulocyte lysate system. To measure prolyl hydroxylase activity (19), we incubated the translation products in the presence of ascorbate, 2-oxoglutarate, and FeSO4 with a biotinylated peptide derived from the human HIF-1α ODD that contains the target proline residue (biotin-Acp-DLDLEALAP*YIPADDDFQL) (20). When this proline residue (P*) is hydroxylated, this peptide binds to human pVHL (13–15). Streptavidin-coated agarose beads were used to precipitate [35S]-labeled human pVHL associated with the biotinylated peptide and measured by scintillation counting (Fig. 1A). Minimal prolyl hydroxylase activity was observed in rabbit reticulocyte lysate programmed to express a LacZ control, demonstrating that the majority of endogenous prolyl hydroxylase activity reported for reticulocyte lysate is inactivated during the 1-hour incubation. Of the five candidate enzymes assayed, only one, HPH-1, enhanced pVHL association with the biotinylated peptide. In addition to HPH-1, the human genome encodes two highly related HPH-1 paralogs, designated HPH-2 and HPH-3 (Fig. 2). Like HPH-1, both of these enzymes catalyzed the hydroxylation of the key proline residue in the HIF-1α peptide substrate as measured by the [35S]-pVHL pull-down assay (Fig. 1B).

Figure 1

HPH enzymes hydroxylate a human HIF-1α–derived peptide. (A) [35S]-VHL pull-down assays of a bi- otin-Acp-DLDLEALAPYIPADDDFQL (20) peptide corresponding to residues 556 to 574 of human HIF-1α. Poly-peptides encoded by LacZ and each candidate gene were synthesized by in vitro transcription/translation in rabbit reticulocyte lysate before incubation with the substrate. (B) [35S]-VHL pull-down assays of the peptide substrate incubated with each in vitro transcribed/translated HPH enzyme or HPH-1 containing the indicated point mutations. (C) [35S]-VHL pull-down assays of the peptide substrate incubated with 0.5 μg of recombinant HPH-2 in the presence or absence of 2 mM ascorbate (ASC), 2 mM 2-oxoglutarate (2-OG), 100 μM FeSO4 (Fe+2), and 1 mM CoCl2. Bars represent the average values from three experiments ± SD. (D) Mass spectrometry analysis of the peptide substrate bearing the target proline residue (left) or the Pro to Ala mutation (right) after incubation in the presence of active recombinant HPH-2.

Figure 2

Alignment of conserved regions of human HPH-1, -2, and -3 with the Drosophila dmHPH sequence (20). Identical residues are shaded in black, and similar residues are shaded in gray. Asterisks indicate critical amino acids for Fe2+ binding in the corresponding prolyl hydroxylase enzymes that modify collagen.

Several amino acid residues have been shown through site-directed mutagenesis studies to be involved in coordinating Fe2+within the active site of the collagen-modifying prolyl-4-hydroxylases (17). These same residues are also present in the HPH enzymes (indicated by asterisks in Fig. 2). To test the role of these residues in HPH-1–mediated prolyl hydroxylation, we individually changed His135, Asp137, and His196to Ala (H135A, D137A, and H196A, respectively). Each of these mutations eliminated the prolyl hydroxylase activity of the native HPH-1 enzyme (Fig. 1B).

We next expressed the human HPH in bacterial cells and purified soluble protein by affinity chromatography (21). As shown in Fig. 1C, purified HPH hydroxylated the HIF-1α peptide substrate, and its activity was greatly stimulated by the addition of ascorbate, 2-oxoglutarate, and FeSO4. Mass spectrometry analysis of the resultant peptide product indicated a mass increase of 16 daltons, consistent with hydroxylation of the proline residue (Fig. 1D). A control peptide in which the target proline residue was replaced by alanine was not modified (22). Furthermore, HPH activity was inhibited by an excess of CoCl2, which is known to induce the hypoxic response pathway by stabilizing HIF under normoxic conditions, possibly by competing with Fe2+ for occupancy within the active site of HPH (Fig. 1C). Together, these data suggest that HPH-1, -2, and -3 are prolyl hydroxylases that modify HIF-1α.

Previous studies have indicated that amino acids in close proximity to the proline residue targeted for hydroxylation (Pro564 in human HIF-1α) influence HIF-1α modification and pVHL binding. Specifically, mutation of Leu562 to Ala (L562A) (13), Ala563 to Gly (A563G) (15), and Tyr565 to Ala (Y565A) (14) prevents prolyl hydroxylation by the endogenous HIF prolyl hydroxylase activity in cellular extracts. Conversely, mutation of Pro567 to Gly exerts a slight stimulatory effect in pVHL pull-down assays (14). To examine the substrate specificity of HPH-1, we synthesized individual biotinylated peptide substrates that contained each of these four point mutations. To distinguish the effects of these mutations on pVHL binding from their effects on proline hydroxylation, we synthesized each peptide with the target proline residue (Pro564) already hydroxylated. Fully hydroxylated peptides corresponding to the wild-type and variant ODD sequences were all recognized by [35S]-pVHL, although the A563G and Y565A mutations did partially compromise pVHL association (Fig. 3A). As reported for the endogenous HIF prolyl hydroxylase, HPH-1 generated by in vitro transcription/translation (Fig. 3B), as well as recombinant HPH-2 purified from Escherichia coli (Fig. 3C), did not modify peptides containing the L562A, A563G, or Y565A mutations. However, a peptide containing the Pro567 to Gly mutation was an equal, if not better, substrate for the human HPH enzymes.

Figure 3

Substrate specificity of the HPH enzymes. (A) Human VHL binds to all hydroxylated peptide substrates. Shown is a [35S]-VHL pull-down assay of the hydroxylated biotin-Acp-DLDLEALAP*YIPADDDFQL peptide (LAPYIP) (20) and hydroxylated peptides containing the indicated point mutations (underlined). (B) [35S]-VHL pull-down assays of the unmodified peptide substrates incubated with the in vitro transcribed/translated HPH-1 enzyme. (C) [35S]-VHL pull-down assay of the unmodified peptide substrates incubated with the recombinant HPH-2 enzyme. Bars represent the average values from three experiments ± SD.

To test whether HPH enzymes function in the hypoxic response pathway in intact cells, we cotransfected (23) human embryonic kidney 293 cells with a hypoxia-responsive luciferase reporter (24) and increasing amounts of a vector expressing human HIF-1α under the control of the constitutive human cytomegalovirus (CMV) immediate early promoter. Forced overexpression of HIF-1α can overcome the degradation pathway, resulting in accumulation of HIF under normoxic conditions and subsequent induction of the HRE-containing HIF reporter gene (24). Induction of the HIF-dependent luciferase reporter by either 30 ng (Fig. 4A) or 100 ng (Fig. 4B) of HIF-1α expression vector was not affected by cotransfection of 3 ng of CMV-driven vectors expressing LacZ or candidates A, B, C, and D (Fig. 4, A and B). By contrast, HPH-1 attenuated HIF-1α activation of the HRE-driven reporter, as might be expected if the prolyl hydroxylase activity was rate-limiting for HIF-1α degradation under these conditions (Fig. 4, A and B, left panels). Mutation of His135, Asp137, or His196 to Ala eliminated the interfering activity of HPH-1 in these transfection assays (Fig. 4, A and B, left panels). Neither HPH-1 nor the other candidate enzymes affected expression of a constitutive, CMV-driven luciferase reporter (Fig. 4C), demonstrating that HPH-1 specifically attenuates HIF-1α–mediated induction of the hypoxia-responsive reporter. Finally, the ability of HPH-1 to diminish induction of the HIF-responsive luciferase reporter was lost when cells were incubated in an atmosphere containing 0.5% O2. These data suggest that the transiently overexpressed HPH-1 enzyme is inhibited by a low-oxygen environment (Fig. 4, A and B, right panels).

Figure 4

Cotransfection of HPH-1 specifically attenuates overexpressed HIF-1α activity. Human 293 cells were transfected with 30 ng of a HIF-dependent luciferase reporter and 30 ng (A) or 100 ng (B) of a constitutive HIF-1α expression vector in the presence of 3 ng of HPH-1, other candidate prolyl hydroxylases, or inactive HPH-1 mutants. Cells cotransfected with the reporter, HIF-1α and Lac Z, or HPH-1 were also incubated in an atmosphere containing 0.5% O2 for 15 hours starting 5 hours after transfection [(A) and (B), right panels]. (C) Three nanograms of each candidate expression vector was cotransfected with 0.5 ng of a constitutive luciferase reporter driven by the CMV promoter. Bars represent the average values from three experiments ± SD.

We also examined the role of HPH in the fruit flyDrosophila melanogaster. We identified a singleHPH gene in the Drosophila genome (gene CG1114), hereby designated dmHPH (Fig. 2). Drosophilahomologs of HIF-1α (Sima), ARNT (dARNT), and pVHL (dVHL) have been previously identified (2, 3, 25,26), and the activity and stability of Sima have been shown to be regulated by oxygen (2). Together, these observations suggest that Drosophila has a hypoxic response pathway analogous to that in mammalian cells. We incubated double-stranded RNAs corresponding to Sima or dmHPH with the KC167 cell line derived from Drosophila embryos to eliminate expression of these genes by RNA interference (RNAi) (27). After incubation under normoxic conditions, we prepared total RNA from the cells and examined it by Northern blotting. These results confirmed that RNAi substantially reduced the levels of mRNAs encoding Sima or dmHPH (Fig. 5A).

Figure 5

Reduction of dmHPH expression by RNAi induces expression of the hypoxia-responsive dmLDHgene. KC167 cells were treated with double-stranded RNAs corresponding to Sima or dmHPH. (A) Northern blots showing (left) Sima mRNA levels in untreated cells (−) or in cells treated with double-stranded Sima RNA (+) and (right) dmHPH mRNA levels in untreated cells (−) or in cells treated with double-strandeddmHPH RNA (+). (B) Northern blot showing total RNA isolated from untreated cells (20% O2), cells treated with double-stranded Sima or dmHPHRNAs under normoxic conditions, or untreated cells incubated under hypoxic conditions (1% O2). Blots were hybridized with [32P]-labeled probes specific for dmLDH or the ribosomal protein 49 gene, RP49 (loading control). Relative dmLDH levels normalized to normoxic levels are indicated below.

To determine the effects of the partial loss of function of these gene products, we examined the expression of the gene encoding lactate dehydrogenase (dmLDH), which is a HIF-dependent hypoxia-inducible gene in mammalian cells (28,29). As shown in Fig. 5B, untreated KC167 cells incubated in 1% O2 for 15 hours accumulated 7.5 times more dmLDH mRNA than cells maintained under normoxic conditions (20% O2). KC167 cells treated with double-strandedSima RNA showed a twofold reduction in dmLDH mRNA levels. By contrast, RNAi-mediated reduction of dmHPH mRNA resulted in a 2.5-fold increase in dmLDH mRNA levels under normoxic conditions.

These data suggest that the HPH enzymes are bona fide HIF prolyl hydroxylases and act as integral regulators of the HIF-dependent hypoxia response pathway. In previous studies, a rat protein likely to be homologous to HPH-1, designated SM-20, has been reported to induce apoptosis (30). It is noteworthy that the ability of HPH-1 to attenuate HIF-1α induction of the hypoxia-responsive luciferase reporter is not attributable to its effect on cell viability. Cotransfection of HPH-1 with a CMV-driven luciferase reporter did not attenuate luciferase expression (Fig. 4C). Moreover, HPH-1 did not attenuate the HRE-driven luciferase reporter under hypoxic culture conditions (right panels of Fig. 4, A and B). Finally, cell cultures transfected with the HPH-1 expression vector showed no overt evidence of apoptosis.

The three human HPH enzymes identified here may be attractive targets for drug discovery. For example, a recent study with mice has revealed that transgenic expression of a modified form of HIF-1α lacking the ODD in basal keratinocytes results in increased dermal vascularization (31). In contrast to the leaky vasculature that is induced by overexpression of single vasculogenic growth factors such as VEGF (32–34), the HIF-induced vascular bed is stable (31). The more substantive neovascularization resulting from constitutive HIF-1α expression may reflect the fact that this transcription factor activates not only VEGF gene expression but also other genes important for the formation of new blood vessels [reviewed in (35)]. Selective inhibitors of the HPH enzymes may therefore merit investigation as new drugs for therapeutic angiogenesis.

  • * To whom correspondence should be addressed. E-mail: smckni{at}


Stay Connected to Science

Navigate This Article