Activation of the HIF Prolyl Hydroxylase by the Iron Chaperones PCBP1 and PCBP2

doi:10.1016/j.cmet.2011.08.015

Cell Metab. Author manuscript; available in PMC 2012 May 29.

Published in final edited form as:

Cell Metab. 2011 November 2; 14(5): 647–657.

doi: 10.1016/j.cmet.2011.08.015

PMCID: PMC3361910

NIHMSID: NIHMS379548

Activation of the HIF Prolyl Hydroxylase by the Iron Chaperones PCBP1 and PCBP2

Anjali Nandal,¹ Julio C. Ruiz,² Poorna Subramanian,³ Sudipa Ghimire-Rijal,³ Ruth Ann Sinnamon,³ Timothy L. Stemmler,³ Richard K. Bruick,² and Caroline C. Philpott¹^*

¹Liver Diseases Branch, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD 20892, USA

²Department of Biochemistry, University of Texas Southwestern Medical Center, Dallas, TX 75390, USA

³Department of Biochemistry and Molecular Biology, Wayne State University School of Medicine, Detroit, MI 48201, USA

^*Correspondence: Email: carolinep/at/intra.niddk.nih.gov

The publisher's final edited version of this article is available at Cell Metab

See other articles in PMC that cite the published article.

SUMMARY

Mammalian cells express dozens of iron-containing proteins, yet little is known about the mechanism of metal ligand incorporation. Human poly (rC) binding protein 1 (PCBP1) is an iron chaperone that binds iron and delivers it to ferritin, a cytosolic iron storage protein. We have identified the iron-dependent prolyl hydroxylases (PHDs) and asparaginyl hydroxylase (FIH1) that modify hypoxia-inducible factor α (HIFα) as targets of PCBP1. Depletion of PCBP1 or PCBP2 in cells led to loss of PHD activity, manifested by reduced prolyl hydroxylation of HIF1α, impaired degradation of HIF1α through the VHL/proteasome pathway, and accumulation of active HIF1 transcription factor. PHD activity was restored in vitro by addition of excess Fe(II), or purified Fe-PCBP1, and PCBP1 bound to PHD2 and FIH1 in vivo. These data indicated that PCBP1 was required for iron incorporation into PHD and suggest a broad role for PCBP1 and 2 in delivering iron to cytosolic nonheme iron enzymes.

INTRODUCTION

Mammalian cells express hundreds of metalloproteins. Most contain the abundant metals iron and zinc, while others contain various trace metals such as copper, manganese, molybdenum, and cobalt (Waldron et al., 2009). These metals are essential nutrients because metal cofactors activate enzymes and proteins that perform critical functions in virtually every major cellular process (Dupont et al., 2010). Several factors complicate incorporation of the correct metal ion into a metalloprotein. First, the binding sites for different metals within metalloproteins can be structurally very similar, and incorporation of the noncognate metal ion is easily achieved in vitro for many of these proteins. Second, pools of “free” metal ions in cells may be vanishingly small and the metals largely unavailable, as most zinc and copper ions are tightly bound to cytosolic proteins (Outten and O'Halloran, 2001). Third, redox-active metals, such as iron and copper ions, can catalyze the production of damaging reactive oxygen species, and cells must maintain tight control over these metals in order to use them while simultaneously avoiding their toxic effects. Fortunately, the majority of metalloproteins receive the correct metal ion in vivo, as incorporation of the wrong metal ion typically inactivates the protein.

Although the incorporation of the appropriate metal ion(s) into cellular metalloproteins is a critical, essential process, the mechanism by which most metalloproteins receive their specific cofactor is unknown. Some proteins rely on metallochaperones: proteins that specifically bind metal ions and deliver them to target enzymes and transporters through direct protein-protein interactions (Rosenzweig, 2002). Metallochaperones delivering nickel and copper have been described in prokaryotes and eukaryotes, but much less is known about the delivery of iron and zinc. Frataxin, the protein lacking in the neurodegenerative disease Friedreich's Ataxia, is a mitochondrial protein that is thought to function as an iron chaperone for the assembly of iron-sulfur clusters (Stemmler et al., 2010).

More recently, we identified poly (rC) binding protein 1 (PCBP1) as a cytosolic iron chaperone that delivers iron to ferritin (Shi et al., 2008). In mammals, ferritin is a heteropolymer consisting of 24 subunits of heavy (H) and light (L) peptides that assemble into a hollow sphere into which iron is deposited (Crichton, 2009; Hintze and Theil, 2006). PCBP1 binds Fe(II) with micromolar affinity in a 3 Fe:1 PCBP1 molar ratio. PCBP1 binds ferritin in vivo and can enhance iron incorporation into ferritin in vitro and in vivo. Mammalian cells lacking PCBP1 exhibit defects in the incorporation of iron into ferritin as well as an increase in the labile pool of cytosolic iron and an increase in the iron-mediated degradation of iron-regulatory protein 2.

PCBP1 (also called α-CP1 or hnRNP E1) has previously been found to function as an RNA- and DNA- binding protein (Chaudhury et al., 2010; Makeyev and Liebhaber, 2002; Ostareck-Lederer and Ostareck, 2004). PCBP1 is one member of a family of four homologous proteins containing three heterogeneous nuclear ribonucleoprotein K-homology (KH) domains, an ancient and conserved RNA binding module. PCBP1, an intronless gene, likely arose from the retrotranspositon of a splice variant of PCBP2 mRNA, and became fixed in the genome because it encoded a unique function not shared by the other PCBPs. PCBP1 and 2 bind to cytosolic and viral RNAs, thereby affecting their translation or stability. PCBPs also have a role in transcriptional regulation and participate in several protein-protein interactions.

Numerous cellular proteins require iron for activity. Iron in the form of heme and iron-sulfur clusters are cofactors for proteins involved in a host of metabolic and regulatory functions. Enzymes of the “nonheme” iron families directly coordinate iron ions as cofactors. These families include the diiron monooxygenases, such as the δ-9-fatty acid desaturase and the small subunit of ribonucleotide reductase (Shanklin et al., 2009). A second family is the Fe(II)- and 2-oxoglutarate (2-OG)-dependent dioxygenases (Kaelin and Ratcliffe, 2008; Loenarz and Schofield, 2008; Ozer and Bruick, 2007). This family is a large, evolutionarily conserved class of enzymes that can oxidatively modify a variety of substrates. In mammals, four members of this class regulate the activity of the transcription factors that control the mammalian response to hypoxia.

Hypoxia-inducible factor (HIF) is a heterodimeric transcription factor that binds DNA at specific sites, termed hypoxia response elements (HREs), and activates the expression of more than 100 genes involved in the adaptation to reduced oxygen levels (Kaelin and Ratcliffe, 2008; Loenarz and Schofield, 2008; Ozer and Bruick, 2007). Under hypoxic conditions, the alpha subunit (HIF1α or HIF2α) accumulates and binds to the beta subunit (HIF1β, also called ARNT) to form the active transcription factor. Under conditions of normoxia or hyperoxia, HIF1α is hydroxylated on proline residues 402 and 564, which allows the protein to be recognized by the von Hippel-Lindau tumor suppressor protein (pVHL), thus targeting HIF1α for ubiquitin-mediated degradation in the proteasome (Ivan et al., 2001; Jaakkola et al., 2001). Three HIF prolyl hydroxylases, PHD1, 2, and 3 (also called HPH-3, -2, and -1 or EGLN2, 1, and 3, respectively) mediate the hydroxylation of proline residues on HIF1α (Bruick and McKnight, 2001; Epstein et al., 2001), although PHD2 is responsible for nearly all (>95%) of the activity in cultured cells (Berra et al., 2003). HIF1α is also hydroxylated on Asn803 by an asparaginyl hydroxylase, factor inhibiting HIF (FIH1) (Hewitson et al., 2002; Lando et al., 2002a). Hydroxylation of Asn803 inhibits the binding of transcriptional coactivators with HIF1α and represents a second mechanism for inhibition of HIF activity (Lando et al., 2002b). The activities of the HIF hydroxylases are regulated by the availability of the cosubstrates, 2-OG and oxygen. Because the hydroxylases exhibit changes in oxygen binding and activity over the range of oxygen concentrations present in tissues, these enzymes are hypothesized to function directly as oxygen sensors.

The activities of the HIF hydroxylases may also be regulated by the availability of iron. HIF hydroxylase activity is stimulated by the addition of Fe(II) in vitro, and, in cultured cells, activity is inhibited by iron chelators (Kaelin and Ratcliffe, 2008). In mice, HIF2α accumulates in duodenal enterocytes in response to iron deprivation, which may reflect a localized decrease in HIF hydroxylase activity (Shah et al., 2009). Cellular factors that control the incorporation of iron into the HIF hydroxylases are unknown.

Here we have addressed the question of whether PCBP1, or its paralog PCBP2, is involved in the delivery of iron to the Fe(II)-dependent prolyl and asparaginyl hydroxylases regulating HIF. We found that iron-deprived cells lacking PCBP1 or PCBP2 exhibited increased levels of HIF1α, which was due to a decrease in prolyl hydroxylation and VHL-mediated degradation. The loss of prolyl hydroxylase activity was traced to a decrease in iron loading of the enzyme, which could be restored with recombinant PCBP1. PCBP1 physically interacted with PHD2, indicating that PCBP1 likely acts as an iron chaperone for PHD2. Our studies also suggest a direct role for PCBPs in the activation of the asparaginyl hydroxylase FIH1.

RESULTS

Accumulation of Transcriptionally Active HIF1α in Cells Lacking PCBPs

The activity of the iron-dependent PHDs is reflected in the abundance of HIF1α protein. In Huh7 cells grown under normoxia, PHDs are fully active, and HIF1α from nuclear extracts was barely detectable (Figure 1A, left panel). Overnight treatment (16 hr) of cells with the iron chelator desferrioxamine B (DFO) inhibited the PHDs and led to a large accumulation of HIF1α. We examined the effects of PCBP depletion on HIF1α levels in the absence of DFO treatment by transfecting Huh7 cells with small interfering RNAs (siRNAs) directed against PCBP1 and 2 and with control siRNA. Depletion of PCBP1 or PCBP1 and 2 together resulted in a 2-fold to 3-fold increase in HIF1α protein when compared to the cells treated with control siRNA, without a significant change in HIF1α levels in cells treated with siRNA against PCBP2 (Figures 1A, right panel, and and1B1B).

Figure 1

Accumulation of HIF1α and Increased HIF Activity in Cells Depleted of PCBPs

We tested whether the increase in HIF1α protein was associated with an increase in HIF1α transcriptional activity. A HeLa cell line that contains a stably integrated copy of the firefly luciferase coding sequence under the control of the HRE exhibits luciferase activity in proportion to the activity of the HIF transcription factors (Tian et al., 1997). Treatment of this cell line with iron or untreated cells exhibited very low luciferase activity, while treatment with DFO for 16 hr led to a 27-fold increase in luciferase activity (Figure 1C). This cell line was then depleted of PCBP1, 2, or both PCBP1 and 2 without DFO treatment. Cells lacking PCBP1 exhibited a 5-fold increase in luciferase activity, while cells treated with siRNAs for both PCBP1 and 2 exhibited a 1.6-fold increase in activity when compared to control cells. Consistent with Figure 1A, these data indicated that, in the absence of iron chelation, cells lacking PCBP1 exhibited both an increase in HIF1α levels and an increase in HIF-dependent transcription.

Increased Effects of PCBP Depletion in Mildly Iron-Deficient Cells

While treatment of Huh7 cells with DFO for 8–14 hr markedly inhibited cellular PHDs, brief treatments of 2–4 hr produced a small accumulation of HIF1α, suggestive of mild iron deficiency and partial inhibition of PHD (Figure S1C). We hypothesized that cells depleted of PCBP might show increased sensitivity to mild iron deficiency and briefly treated cells depleted of PCBPs with DFO, then examined the levels of HIF1α (Figures 1D and and1E).1E). When cells subjected to a single transfection of PCBP siRNA were compared to cells treated with control siRNA, cells depleted of PCBP1 exhibited a 7-fold increase in the level of HIF1α. Cells depleted of both PCBP1 and 2 exhibited similarly high levels of HIF1α, while cells lacking only PCBP2 exhibited a variable increase in HIF1α levels. Further depletion of PCBP2 and PCBP1 and 2 together was achieved by two sequential treatments with siRNA (Figures 1D and 1E, lanes marked “x2”), which resulted in 9-fold increases in HIF1α levels. We measured the effects of siRNA treatment on PCBPs by western blotting and by quantitative real-time PCR, and confirmed that our siRNA transfections produced depletion of PCBP1 and 2 in Huh7 cells, although simultaneous depletion of PCBP1 and 2 was less efficient than depleting them individually (Figures 1F, S1A, and S1B). To rule out off-target effects of the siRNAs, we transfected Huh7 cells with alternative siRNAs directed against other regions of the PCBP mRNAs and found that these siRNAs also increased HIFα (Figures S1E and S1F). Although plasmid transfection efficiency of cells previously treated with siRNA was low, the effects of PCBP1 depletion could be partially rescued by expression of a mutant PCBP1 not targeted by the siRNA (Figure S1D).

Increased Half-Life of HIF1α in PCBP-Depleted Cells

The elevated levels of HIF1α in cells lacking PCBPs could be due to either an increase in HIF1α gene expression or a decreased rate of HIF1α protein degradation. We examined the levels of HIF1α mRNA in cells depleted of PCBPs using real-time PCR (Figure 2A). Changes in HIF1α mRNA levels in PCBP-depleted cells were small and did not account for the observed changes in protein levels. We next examined the half-life of HIF1α protein in Huh7 cells depleted of PCBP1 and 2 versus control cells. Cells were treated with DFO for 2 hr, cycloheximide was added to block new protein synthesis, cells were collected at intervals, and nuclear extracts were examined by western blotting (Figure 2B). HIF1α protein levels were quantitated and the half-life was calculated (Figure 2C). Consistent with previous reports, HIF1α was rapidly degraded in control cells (t_1/2 = 8 min), while HIF1α exhibited much greater stability in cells depleted of PCBP1 and 2 (t_1/2 = 50 min), although the presence of DFO during the cycloheximide incubation may also have contributed to the stability of HIF1α. These data indicated that the accumulation of HIF1α in cells depleted of PCBP1 was due to impaired degradation of the protein.

Figure 2

Increase in HIF1α Half-Life in Cells Lacking PCBPs

Reduced Hydroxylation of HIF1α in Cells Lacking PCBPs

VHL-mediated degradation of HIF1α is dependent on the hydroxylation of proline residues 402 and 564 on the oxygen-dependent degradation domain of HIF1α (Ivan et al., 2001; Jaakkola et al., 2001). Cells lacking PCBPs may exhibit impaired degradation of HIF1α because of impaired hydroxylation of these proline residues; therefore, we measured the hydroxylation of Pro564 in A549 cells lacking PCBPs. A549 cells were selected because they expressed higher levels of PHD2 mRNA than Huh7 cells (see Figure 4A) and HIF1α-OH was readily detectable. A549 cells were similar to Huh7 cells in that transfections of siRNA for PCBP1, 2, and PCBP1 and 2 together also produced markedly elevated levels of HIF1α in the presence of DFO (Figure S2A). Depletion of PCBPs in A549 cells was also efficient, although again simultaneous depletion of PCBP1 and 2 was less efficient (Figures 3D, S2C, and S2D). We measured hydroxylation of Pro564 by blocking proteasome-mediated degradation with the inhibitor MG132 and detecting hydroxylated HIF1α using an antibody that specifically recognizes the hydroxylated form (HIF1α-OH). Brief treatment of A549 cells with 25 μM DFO did not change HIF1α-OH levels (Figure S2B), and DFO (25 μM) was therefore included in experiments measuring HIF1α-OH levels. Nuclear extracts were probed for HIF1α-OH, total HIF1α, and CREB (Figures 3A and and3B).3B). HIF1α-OH was readily detected in cells treated with no or control siRNA, while cells depleted of PCBP1 exhibited dramatically reduced levels of HIF1α-OH and cells lacking PCBP2 or PCBP1 and 2 exhibited moderately reduced HIF1α-OH. The specificity of the antibody for HIF1α-OH was demonstrated by the absence of signal in untreated cells (−MG132), which also had no detectable HIF1α, and the absence of signal in cells treated for 16 hr with DFO alone (+DFO), which had abundant HIF1α. Although treatment with MG132 inhibited the degradation of HIF1α in all conditions, cells depleted of PCBP1 exhibited a higher level of total HIF1α. Despite the higher level of total HIF1α, the decrease in HIF1α-OH was still detected, confirming that depletion of PCBPs was associated with a loss of HIF1α prolyl hydroxylation.

Figure 4

Decrease in PHD2 Metallation after Depletion of PCBPs

Figure 3

Impaired HIF1α Hydroxylation and VHL Binding after Depletion of PCBPs

The activity of the HIF PHDs can be measured in vitro using an assay that relies on the specific interaction of pVHL with a HIF1α-derived peptide containing hydroxyproline 564, but not with a peptide containing an unmodified proline (Aprelikova et al., 2009; Tuckerman et al., 2004). We used HEK293T cell lysates containing PHD to hydroxylate a HIF1α synthetic peptide corresponding to residues 556–574, which had been bound to magnetic beads. VHL protein was incubated with the peptide, and the amount of pVHL captured was measured by western blotting. HEK293T cells were selected because they expressed more PHD2 than Huh7 cells, and depletion of PCBPs was very efficient with two sequential transfections of siRNAs (Figures 3D and S2F). Cells were treated with siRNAs for PCBP1 and 2 and control siRNA, briefly treated with DFO, then lysed and assayed for PHD activity by the pVHL capture method (Figures 3C and S2E). Cells transfected with no or control siRNA exhibited readily detectable PHD activity, while cells lacking PCBP1 and 2, individually or in combination, exhibited much lower levels of activity (38%–13% of Control). The specificity of the hydroxylated peptide for pVHL capture was confirmed using a hydroxylated synthetic peptide as a positive control (Peptide-OH) and lysate from cells treated overnight with DFO as a negative control (+DFO). HEK293T cells lacking either PCBP1, 2, or PCBP1 and 2 exhibited similar losses of PHD activity. Consistent with this observation, the increase in the total amount of HIF1α in HEK293T cells lacking PCBP1 and/or 2 was also similar (Figure 3C).

Reduced Metallation of PHD2 in Cells Lacking PCBPs

The reduced hydroxylation of HIF1α in cells lacking PCBPs could be explained by reduced levels of PHD mRNA or protein or by reduced specific activity of the enzyme. We measured mRNA levels of PHD2 by real-time PCR and found only small differences (less than 2-fold) between control cells and cells lacking PCBPs in Huh7, A549, or HEK293T cell lines (Figure 4A). PHD2 protein levels did not significantly change in Huh7, HEK293, and A549 cells after depletion of PCBPs (Figures 4B, ,4C,4C, and and5A).5A). These data suggested that depletion of PCBPs reduced the specific activity of PHD2 rather than affecting protein levels. The activity of PHD2 is dependent on the incorporation of iron into the enzyme. Therefore, we measured the amount of iron bound to PHD2 from cells labeled in vivo with [⁵⁵Fe]. HEK293T cells were depleted of PCBPs, then transiently transfected to overexpress a FLAG-epitope-tagged version of PHD2, and labeled with [⁵⁵Fe] in the absence of DFO. PHD2 with its bound iron was recovered by immunoprecipitation and measured by scintillation counting (Figure 4C). The amount of iron bound to PHD2 was reduced by 78%, 62%, and 73%, respectively, in cells lacking PCBP1, PCBP2, or PCBP1 and 2. There was no difference in the amount of PHD2 protein immunoprecipitated (Figure 4C, top panel). These data suggested that the loss of PHD activity in cells lacking PCBPs was due to a failure to incorporate the Fe(II) cofactor.

Figure 5

Requirement of PCBPs for Iron-Dependent PHD Activity

Restoration of PHD Activity with Iron or Purified PCBP1

We employed a more quantitative assay to further characterize the loss of PHD activity in cells lacking PCBPs without DFO treatment. Peptides corresponding to residues 556–574 of HIF1α were immobilized, then lysates from A549 cells depleted of PCBPs were incubated with the peptides in the presence of 2-OG and ascorbate. The amount of hydroxylation of Pro564 was measured using an antibody that specifically recognized Hydroxypro564 in the HIF1α-derived peptide. When the hydroxylase assay included a low concentration of Fe(II) that partially activated PHD, robust activity was detected in control lysates, and essentially no activity was detected in lysates from cells treated overnight with DFO (Figure 5A). Lysates from cells depleted of PCBP1, 2, and PCBP1 and 2, but not treated with DFO, exhibited a 67%–71% reduction in PHD activity, confirming that PCBPs were required for full PHD activity in cells in the absence of DFO treatment.

PHD2 that lacks the iron cofactor can be activated in vitro with high concentrations of Fe(II) (Figure S4). We tested whether PHD activity could be restored by treating the PCBP lysates with Fe(II) in vitro. Lysates from control cells exhibited full PHD activity when assayed with or without exogenous iron (Figure 5B), suggesting that the PHD from these cells was fully metallated. In contrast, lysates from cells depleted of PCBP1 and PCBP1 and 2 exhibited virtually no PHD activity in the absence of exogenous iron, but were restored to nearly full activity upon addition of exogenous iron. These data suggested that the loss of PHD activity could be fully explained by the absence of the Fe(II) cofactor from the enzyme in the PCBP-depleted lysates. We next tested whether addition of purified, iron-loaded PCBP1 could restore PHD activity. Purified recombinant PCBP1 (P1) was incubated anaerobically with Fe(II) and added to lysates from cells depleted of PCBPs. The lysates were then assayed for PHD activity (Figure 5C). Iron-loaded bovine serum albumin (BSA) and unliganded Fe(II) were used as controls, and the final concentration of iron (10 μM) was the same for all samples. Purified, iron-loaded PCBP1 was able to restore PHD activity to near control levels in lysates from cells lacking PCBP1 and to partially restore activity in lysates from cells lacking PCBP1 and 2. In contrast, addition of iron-loaded PCBP1 to lysates from cells lacking PCBP2 was no more effective than iron-loaded albumin or iron alone in restoring PHD activity. These data indicated that iron-loaded PCBP1 could restore PHD activity in vitro to cells lacking PCBP1, but not to cells lacking only PCBP2. This finding suggested that PCBP1 and 2 had nonredundant functions and may cooperate in the delivery of iron to PHD.

In Vivo Interactions between PCBP1 and PHD2

PCBP1 directly binds to ferritin in the process of delivering iron for mineralization, and, in yeast cells, this interaction is dependent on the presence of iron (Shi et al., 2008). We examined the in vivo binding of PCBP1 to PHD2 in HEK293T cells by testing for the presence of PHD2 in immune complexes precipitated using antibodies against PCBP1 (Figure 6A). PHD2 was readily detected in immune complexes obtained after immunoprecipitation of PCBP1, and no PCBP1 or PHD2 was detected in immunoprecipitates using nonspecific IgY as a control antibody. The effects of iron manipulation on PHD2 binding to PCBP1 were small and not consistently detected. Iron manipulation had no effect on the total levels of PHD2 or PCBP1 in the whole-cell lysates. These data indicated a physical interaction between PCBP1 and PHD2.

Figure 6

Physical Interaction of PCBP1 with PHD2

Our data suggested that both PCBP1 and 2 were involved in the delivery of iron to PHDs, and that their functions were not redundant. To determine whether PCBP1 and 2 physically interacted, we again immunoprecipitated PCBP1 from HEK293T cells that had been treated with iron or DFO and tested for the presence of PCBP2 in the immune complexes (Figure 6B). PCBP2 was detected in precipitates of PCBP1, but not in control precipitates using nonspecific IgY, demonstrating that at least a portion of PCBP1 and PCBP2 are bound together as a complex in vivo.

Genetic and Physical Interactions of PCBP1 with FIH1

FIH1 is an asparaginyl hydroxylase of the same Fe- and 2-OG-dependent dioxygenase family as the PHDs, and we questioned whether FIH1, similarly to PHD, also required PCBPs for Fe-dependent activation. FIH1 hydroxylates a conserved asparagine residue in the carboxyl-terminal transactivation domains (CADs) of HIF1α and HIF2α (Lando et al., 2002a; Lando et al., 2002b). This modification does not lead to degradation of the CADs, but instead prevents the association of transcriptional coactivators with HIF, thereby blocking the activity of the HIFα CADs. Fusion of the CADs to the Gal4 DNA-binding domain (GalDBD) permits the measurement of FIH1 hydroxylase activity through the capacity of FIH1 to inhibit the GalDBD/HIF2α CAD-dependent transcription of a Gal-responsive luciferase reporter. HEK293 cells were depleted of PCBPs, then cotransfected with plasmids expressing GalDBD/HIF2α CAD, wild-type FIH1, and a luciferase reporter under the control of a Gal4-responsive promoter (5XGRE) (Figure 7A). Mutation of Asn851 to Ala in HIF2α blocks hydroxylation of the Asn residue by FIH1, resulting in constitutive activation of the HIF2α CAD. Therefore, a plasmid containing the Asn851Ala substitution was also transfected. Cells were not treated with DFO. In cells treated with control siRNA, FIH1 was fully active, and luciferase activity was very low (Figure 7B). In contrast, control cells expressing the Asn851Ala CAD exhibited 10-fold higher luciferase activity. Cells depleted of PCBP1 exhibited a 5-fold increase in luciferase activity when compared to the control, and cells depleted of PCBP1 and 2 exhibited a 3-fold increase in activity. As with PHD2, depletion of PCBPs had no effect on FIH1 protein levels (Figure 7C) and depletion of PCBPs by siRNA was confirmed (Figure S5A). Similar to Figures 1A and and1C,1C, these data suggested that cells lacking PCBP1 had reduced FIH1 activity even without treatment with DFO.

Figure 7

Genetic and Physical Interactions of PCBP1 with FIH1

We tested whether FIH1 was bound to PCBP1 in HEK293 cells overexpressing FIH1 by immunoprecipitating endogenous PCBP1 and examining immune complexes for coprecipitation of FIH1 (Figures 7D and S5D). Similar to PHD2, FIH1 was detectable in immune complexes containing PCBP1, but not in control IgY immunoprecipitations. Unlike PHD2, only immune complexes from cells treated with iron contained significant coprecipitated FIH1. The migration of FIH1 as a doublet was frequently observed in lysates from transfected and untransfected cells (Figure S5C). Coimmunoprecipitation of endogenous FIH1 with PCBP1 was also detected in cells treated with iron (Figures 7D and S5D, right panels, “– pFIH1”). These data suggested that PCBP1 also acted as an iron chaperone for FIH1.

DISCUSSION

Although most nonheme iron enzymes receive their metal cofactor through an unknown mechanism, our studies indicated that PHDs and FIH1, the prolyl and asparagyl hydroxylases that modify HIF1α, depend on members of the PCBP family of iron chaperones/RNA-binding proteins to incorporate Fe(II) into their active sites, especially in cells briefly exposed to iron limitation. Cells lacking PCBP1 or 2 exhibited reduced PHD activity, which resulted in less prolyl hydroxylation of HIF1α, less degradation of HIF1α through the pVHL-proteasome pathway, and accumulation of HIF1α protein. Because the loss of PHD activity was associated with a loss of the iron cofactor, PCBP1 must be involved in the metallation of PHD. Previously, we have shown that depletion of PCBP1 in Huh7 cells does not result in a loss of iron uptake activity and is associated with an increase in the labile iron pool (Shi et al., 2008); therefore, cellular iron for metallation of PHD must be present. We found that iron-loaded PCBP1 specifically restored activity to inactive PHD in vitro and that PCBP1 and PHD2 physically interacted in vivo by coimmunoprecipitation. Thus, we propose that PCBP1 is an iron chaperone for PHD as well as for ferritin.

We have previously shown that expression of PCBP2 in yeast cells containing human ferritin activates the iron deficiency response of yeast, indicating that PCBP2 can disrupt iron homeostasis in yeast (Shi et al., 2008). Preliminary data indicate that PCBP2 can bind both iron and ferritin, suggesting that PCBP2 can also function as an iron chaperone (A.N., T.L.S., and C.C.P., unpublished data). Here, we show that cells lacking PCBP2 also exhibit loss of PHD activity and accumulation of HIF1α, indicating that PCBP2 is also an iron chaperone for PHDs. The loss of FIH1 activity in cells lacking PCBP1 and the binding of PCBP1 to FIH1 in vivo suggest that PCBP1 acts as an iron chaperone for this second type of Fe- and 2-OG-dependent oxygenase. The role of PCBP2 in the activation of FIH1 is not yet clear.

In these studies, PCBP1 and 2 were required in cultured cells for hydroxylation and degradation of HIF1α, but the effects were much greater in cells transiently exposed to iron limitation. In the absence of DFO, the effect of PCBP1 depletion on HIF1α degradation was reproducible but small, and no effect of PCBP2 depletion was observed. In contrast, in vitro measurement of PHD activity in lysates from A549 cells not treated with DFO indicated that depletion of PCBP1 or 2 resulted in a dramatic decrease in activity. There are multiple possible explanations for these results: (1) When iron is present in abundance, as it is in cell culture media, some metallation of PHDs could proceed in a PCBP-independent manner; (2) PHD activity may be present in excess in vivo, and large decreases in PHD activity may be required before HIF1α hydroxylation decreases and HIF1α accumulation occurs; (3) PCBPs may be required for the remetallation of PHDs when the active site iron is lost, and iron limitation with DFO could accelerate the turnover of iron in PHD active sites. The sensitivity of PHDs to DFO treatment in vivo would suggest that the active site iron is readily exchangeable with available cytosolic iron pools. The loss of PHD activity in lysates lacking PCBP1 or 2 may reflect a role for PCBPs in maintaining iron in the active site in vitro, as well.

Our data do not suggest, however, that PCBP1 and 2 can functionally substitute for each other. Cells lacking PCBP1 or 2 contain wild-type levels of the other paralog. Furthermore, while addition of purified, iron-loaded PCBP1 to lysates lacking PCBP1 fully restored PHD activity, addition of PCBP1 did not restore activity to lysates lacking only PCBP2. PCBP1 and 2 may function as a hetero-oligomeric complex to deliver iron to targets such as ferritin and PHD. PCBP1 and 2 bind to each other when in complex with mRNA (Makeyev and Liebhaber, 2002), and we have confirmed this interaction. Our data suggest a model in which PCBP1, in complex with PCBP2 or another PCBP family member, binds iron and interacts with target nonheme iron enzymes to donate metal to the active site. The differences in HIF1α accumulation that occurred with depletion of PCBP1 versus 2 may reflect the activity of PCBP1 in complex with residual PCBP2 or with PCBP3 or PCBP4, which are expressed at low levels in cells.

Our studies indicate that ferritin, PHD, and FIH1 receive iron from PCBPs. Ferritin and the 2-OG dependent hydroxylases are structurally dissimilar, but have certain characteristics in common that may point to mechanisms of iron donation. Both ferritin and PHD2 can bind iron after folding into their native conformations. Thus, PCBPs could interact with these targets posttranslationally to donate iron. The H-chain of ferritin catalyzes the oxidation of Fe(II) to Fe(III) for ferritin core mineralization, and the ferroxidase center structurally and mechanistically resembles the catalytic centers of the oxo-bridged diiron family of monooxygenases (Crichton, 2009; Hintze and Theil, 2006). However, PCBP1 likely does not provide iron directly to the ferroxidase sites of ferritin, as these sites are located on the interior surface of the ferritin sphere. Iron ions gain access to the ferroxidase sites through pores lined with hydrophilic residues. PCBP1 could donate Fe(II) to the His, Asp, and Glu residues that line these funnel-shaped channels. The active site of PHD2 and other enzymes of this class is solvent-exposed and coordinates a single Fe(II) through a His-Xaa-Asp/Glu-Xaa(n)-His triad located deep in the active site pocket (Loenarz and Schofield, 2008; Ozer and Bruick, 2007). Thus, the iron ligands and the coordination environment in these two enzymes are similar.

Given that PCBP1 serves as an iron chaperone for these two diverse target enzymes, we suggest that it is highly likely that other enzymes of the Fe(II)- and 2-OG oxygenase class will require PCBPs for metallation. Sequence analysis of the human genome indicates the presence of more than 60 enzymes of this class, many of which have not been functionally characterized, but are predicted to oxidatively modify a broad range of substrates (Loenarz and Schofield, 2008; Ozer and Bruick, 2007). In addition to its role in HIF regulation, FIH functions in mice as a regulator of metabolism, likely by hydroxylating asparagine residues on proteins other than HIF (Zhang et al., 2010). This class also includes the collagen prolyl and lysyl hydroxylases, mutations in which cause connective tissue diseases in humans. A subclass of the 2-OG oxygenase family is defined by the presence of the jumonji C domain and members of this subclass catalyze the oxidative demethylation of mono-, di-, and trimethylated lysine residues located in histone proteins. Another subclass of this family resembles the AlkB demethylase of E. coli. Eight members of this subclass have been identified in humans, with activities that include the dealkylation and oxidative demethylation of modified bases in both DNA and RNA.

PCBPs may act as iron chaperones for the diiron monooxygenases as well as the 2-OG-dependent dioxygenases and represent the major distributors of iron for the metallation of cytosolic nonheme iron enzymes. The mechanism by which PCBPs acquire cytosolic iron is unknown. In yeast, cytosolic monothiol glutaredoxins are required to make iron available to iron-requiring enzymes (Mühlenhoff et al., 2010). Whether PCBPs interact with these glutaredoxins awaits further study.

EXPERIMENTAL PROCEDURES

Cell Culture

Huh7, A549, HEK293, and HEK293T cells were grown in high glucose Dulbecco's modified Eagle's medium with 10% fetal bovine serum, penicillin G, 50 U/ml, and streptomycin, 50 μg/ml (Gibco-BRL).

Protein Depletion by siRNA

PCBP1 and 2 were depleted using Stealth Select RNAi (Invitrogen, sequences in Table S1). A nontargeting, scrambled sequence siRNA pool was used as a control. Cells were transfected with 50 pmol of siRNA and harvested at 3 days after transfection. For some experiments, two sequential transfections spaced 24 hr apart were performed.

RNA Extraction and RT PCR

RNA was isolated from cells using the RNeasy RNA Isolation kit (QIAGEN). For reverse transcription, 1 μg of total RNA was used in a reaction mixture containing dNTPs and superscript II (Invitrogen). Reverse transcription was performed for 10 min at 25°C, 60 min at 42°C, and 5 min at 70°C. Quantitative real-time PCR was performed using SYBR Green on an ABI 7500 system according to the manufacturer's protocols (primers in Table S2). The HIF1α, FIH1, and PHD2 values were normalized to β-actin according to Pfaffl's mathematical model for relative quantification in real-time PCR (Pfaffl, 2001). Actin quantitation did not vary with siRNA or DFO treatment.

Immunoblot Analysis

Nuclear and cytoplasmic extracts were prepared using a nuclear extract kit (Active Motif). Nuclear extracts were used for HIF and CREB western blots. Membranes were probed using mouse anti-HIF1α (Transduction, 1:1000) or mouse anti-CREB (Cell Signaling, 1:1000) antibodies. For detection of hydroxylated HIF1α, A549 cells were treated with 100 μM MG132 for 6 hr followed by 25 μM DFO for 2 hr. Nuclear extracts (40 μg) were analyzed by western blotting with a rabbit antibody that specifically recognizes HIF1μ hydroxylation at proline 564 (gift of O. Aprelikova). Whole-cell lysates or cytoplasmic extracts were used for all other western blotting experiments. Whole-cell extracts were prepared by lysis with buffer A (25 mM Tris [pH 8]; 40 mM KCl; 1% Non-idet P40; 0.1% Triton; 1 mM DTT; 1× protease inhibitor cocktail, Roche). Membranes were incubated with the indicated primary antibodies, followed by detection with horseradish peroxidase-linked secondary antibodies (Amer-sham) and enhanced chemiluminescence substrates (Pierce). Antibodies used for western blotting were rabbit anti-PHD2 and anti-FIH1 (Novus Biologicals), mouse anti-PCBP2 (Novus Biologicals), mouse anti-HA (Covance), and horseradish peroxidase-conjugated, mouse anti-FLAG (Sigma). Antibody against recombinant human PCBP1 (Shi et al., 2008) was raised in chickens and purified from yolks by the manufacturer (Covance) and used at 1:2000–1:5000. Western blots were quantitated using ImageJ.

Luciferase Activity Assays

HeLa 3XHRE Luc cells (Tian et al., 1997) were cultured and PCBP1 and/or PCBP2 were depleted as described above. Firefly luciferase activity was assayed using Dual-Luciferase reporter system (Promega). Samples were read in Lumat LB 9507 luminometer (Berthold Technologies). For FIH activity, PCBPs were depleted in HEK293 cells, then plasmids encoding the GalDBD/HIF2α CADs, G5E1b-Luc reporter, pcDNA3.1-FIH1-myc-HisA, and pRL-TK (Lando et al., 2002b) were transfected for 18 hr prior to measurement of luciferase activity. Activity was reported as the ratio to renilla luciferase control.

HIF Half-Life Determination

Huh7 cells depleted of PCBP1 and 2 were treated with 25 μM DFO for 2 hr. Cycloheximide was added at 60 μg/ml, and the cells were collected at intervals. HIF1α was analyzed in nuclear extracts.

VHL Capture Assay

HEK293T cells were depleted of PCBP1 and 2, treated with 25 μM DFO for 2 hr, then lysed by sonication. HA-pVHL was synthesized by in vitro transcription/translation reactions using TNT T7 Quick Coupled Rabbit Reticulocyte Lysate kit (Promega). The VHL capture assay was performed as described (Aprelikova et al., 2009; Tuckerman et al., 2004) in buffer containing 2-OG, ascorbate, and FeCl₂.

PHD2 Assays

Biotinylated peptides derived from the HIF1α oxygen-dependent degradation domain (Biotin-Acp-DLDLEALAPYIPADDDFQL or a hydroxylated control Biotin-Acp-DLDLEALAP[OH]YIPADDDFQL) were immobilized on NeutrAvidin-coated 96-well plates. A549 cells were depleted of PCBP1 and/or 2, then harvested and resuspended in 1 ml hypotonic buffer (20 mM HEPES, 5 mM NaF, 10 mM Na₂MoO₄ or Na₃VO₄, 0.1 mM EDTA, protease inhibitor cocktail, and 2 mM DTT) for 15–20 min, then 0.5% NP-40 was added, and the samples were vortexed for 10 s. Clarified lysates (50 μg/well) were incubated in reaction buffer containing 20 mM Tris-Cl (pH 7.5), 5 mM KCl, 1.5 mM MgCl₂, 2 mM DTT, 0–100 μM ferrous sulfate, 0.5 mM 2-OG and 1 mM ascorbate for 45 min at room temperature. Purified recombinant PCBP1 or albumin (30 μM) was loaded with 100 μM ferrous sulfate for 2 hr at 4°C in an anoxic chamber (Coy Laboratory Products), then added to lysates at a final concentration of 3 μM protein/10 μM Fe(II). Peptide hydroxylation was detected using a polyclonal rabbit antibody raised against a hydroxylated HIF peptide, followed by addition of a goat anti-rabbit HRP-conjugated secondary antibody (Santa Cruz). Luminescence was measured in an EnVision plate reader (PerkinElmer).

Protein Purification

Human PCBP1 was cloned into a pCDF-sumo fusion construct vector (courtesy of Zhe Yang) and transformed into BL21* Escherichia coli competent cells containing streptomycin resistance. Cells were grown in LB media at 37°C to an OD₆₀₀ of 0.6, induced with a IPTG concentration of 0.1 mM and then grown for 22 hr at 15°C. Cells were collected, resuspended in 30 ml of buffer A (20 mM Tris [pH 7.9], 250 mM NaCl, 30 mM imidazole, 20% glycerol, and 2.5 mM TCEP) with an EDTA-free protease inhibitor cocktail tablet (Roche), 30 mgs of lysozyme, and were broken using a French Press. The extract was spun at 21,000 rpm for 30 min and loaded on a nickel column. Separation was carried out using a gradient between buffer A and buffer B (20mM Tris [pH 7.9], 250 mM NaCl, 500 mM imidazole, 20% glycerol, 2.5 mM TCEP), with the protein eluting at 110–230 mM imidazole. The sumo tag was cleaved by adding 1/1000 concentration sumo protease and incubating overnight at 4°C. Protein was buffer exchanged with buffer A before passing on the nickel column again to separate the tag from the untagged protein. The protein was run on a S-200 gel filtration column (GE Healthcare) eluting at a molecular weight of ~75 kDa, consistent with a protein as a dimer.

Immunoprecipitation

HEK293T cells were depleted of PCBPs, transiently transfected with p3xFLAG-CMV-PHD2 (1 μg), then labeled overnight with 2 μM of ⁵⁵Fe(II):NTA (1:4 molar ratio). The cells were lysed in 50 mM Tris-HCl, (pH 7.5); 150 mM NaCl; 0.5% NP-40; and protease (Sigma) and phosphatase inhibitors (Pierce). PHD2 was immunoprecipitated with anti-FLAG antibody and protein A dynabeads (Invitrogen). IPs from mock-transfected cells were used to measure nonspecific ⁵⁵Fe background. Beads were washed, and retained ⁵⁵Fe was measured by scintillation counting (LS 6500, Beckman Coulter). For coimmunoprecipitations, anti-PCBP1 antibody or bulk chicken IgY (Gallus Immunotech) was coupled to magnetic M280 tosylactivated dynabeads (Invitrogen) using 100 μg of antibody. Beads were then blocked in 0.5% BSA and washed with PBS (pH 7.4) containing 0.1% BSA prior to use. Cells were treated 2 hr or overnight with either 20 μM ferric chloride or 100 μM DFO. The cells were lysed in buffer A and lysates (3 mg) were incubated with beads, washed, and the immune complexes were analyzed by western blotting.

Supplementary Material

Supp

Click here to view.^{(677K, pdf)}

ACKNOWLEDGMENTS

We thank Olga Aprelikova, Hye-sik Kong, and Len Neckers for generously providing technical suggestions, cell lines, plasmids, and antibodies. These studies were supported by the Intramural Research Program of NIDDK, NIH (A.N. and C.C.P), and NIH DK068139 (P.S., S.G.-R., and T.L.S.). R.K.B. is the Michael L. Rosenberg Scholar in Medical Research and was supported by the Burroughs Wellcome Fund, NIH grant CA115962, and a National Center for Research Resources Grant (C06-RR15437-01). J.C.R. was supported by the Sara and Frank McKnight Fund for Biomedical Research.

Footnotes

SUPPLEMENTAL INFORMATION Supplemental Information includes five figures and two tables and can be found with this article online at doi:10.1016/j.cmet.2011.08.015.

REFERENCES

Aprelikova O, Pandolfi S, Tackett S, Ferreira M, Salnikow K, Ward Y, Risinger JI, Barrett JC, Niederhuber J. Melanoma antigen-11 inhibits the hypoxia-inducible factor prolyl hydroxylase 2 and activates hypoxic response. Cancer Res. 2009;69:616–624. [PMC free article] [PubMed]
Berra E, Benizri E, Ginouvès A, Volmat V, Roux D, Pouysségur J. HIF prolyl-hydroxylase 2 is the key oxygen sensor setting low steady-state levels of HIF-1alpha in normoxia. EMBO J. 2003;22:4082–4090. [PubMed]
Bruick RK, McKnight SL. A conserved family of prolyl-4-hydroxylases that modify HIF. Science. 2001;294:1337–1340. [PubMed]
Chaudhury A, Chander P, Howe PH. Heterogeneous nuclear ribonucleoproteins (hnRNPs) in cellular processes: Focus on hnRNP E1's multifunctional regulatory roles. RNA. 2010;16:1449–1462. [PubMed]
Crichton RR. In: Intracellular Iron Storage and Biomineralization. In Iron Metabolism: From Molecular Mechanisms to Clinical Consequences. Crichton RR, editor. Wiley; West Sussex: 2009. pp. 183–215.
Dupont CL, Butcher A, Valas RE, Bourne PE, Caetano-Anollés G. History of biological metal utilization inferred through phylogenomic analysis of protein structures. Proc. Natl. Acad. Sci. USA. 2010;107:10567–10572. [PubMed]
Epstein AC, Gleadle JM, McNeill LA, Hewitson KS, O'Rourke J, Mole DR, Mukherji M, Metzen E, Wilson MI, Dhanda A, et al. C. elegans EGL-9 and mammalian homologs define a family of dioxygenases that regulate HIF by prolyl hydroxylation. Cell. 2001;107:43–54. [PubMed]
Hewitson KS, McNeill LA, Riordan MV, Tian YM, Bullock AN, Welford RW, Elkins JM, Oldham NJ, Bhattacharya S, Gleadle JM, et al. Hypoxia-inducible factor (HIF) asparagine hydroxylase is identical to factor inhibiting HIF (FIH) and is related to the cupin structural family. J. Biol. Chem. 2002;277:26351–26355. [PubMed]
Hintze KJ, Theil EC. Cellular regulation and molecular interactions of the ferritins. Cell. Mol. Life Sci. 2006;63:591–600. [PubMed]
Ivan M, Kondo K, Yang H, Kim W, Valiando J, Ohh M, Salic A, Asara JM, Lane WS, Kaelin WG., Jr. HIFalpha targeted for VHL-mediated destruction by proline hydroxylation: implications for O2 sensing. Science. 2001;292:464–468. [PubMed]
Jaakkola P, Mole DR, Tian YM, Wilson MI, Gielbert J, Gaskell SJ, Kriegsheim Av, Hebestreit HF, Mukherji M, Schofield CJ, et al. Targeting of HIF-alpha to the von Hippel-Lindau ubiquitylation complex by O2-regulated prolyl hydroxylation. Science. 2001;292:468–472. [PubMed]
Kaelin WG, Jr., Ratcliffe PJ. Oxygen sensing by metazoans: the central role of the HIF hydroxylase pathway. Mol. Cell. 2008;30:393–402. [PubMed]
Lando D, Peet DJ, Gorman JJ, Whelan DA, Whitelaw ML, Bruick RK. FIH-1 is an asparaginyl hydroxylase enzyme that regulates the transcriptional activity of hypoxia-inducible factor. Genes Dev. 2002a;16:1466–1471. [PubMed]
Lando D, Peet DJ, Whelan DA, Gorman JJ, Whitelaw ML. Asparagine hydroxylation of the HIF transactivation domain a hypoxic switch. Science. 2002b;295:858–861. [PubMed]
Loenarz C, Schofield CJ. Expanding chemical biology of 2-oxoglutarate oxygenases. Nat. Chem. Biol. 2008;4:152–156. [PubMed]
Makeyev AV, Liebhaber SA. The poly(C)-binding proteins: a multiplicity of functions and a search for mechanisms. RNA. 2002;8:265–278. [PubMed]
Mühlenhoff U, Molik S, Godoy JR, Uzarska MA, Richter N, Seubert A, Zhang Y, Stubbe J, Pierrel F, Herrero E, et al. Cytosolic monothiol glutaredoxins function in intracellular iron sensing and trafficking via their bound iron-sulfur cluster. Cell Metab. 2010;12:373–385. [PubMed]
Ostareck-Lederer A, Ostareck DH. Control of mRNA translation and stability in haematopoietic cells: the function of hnRNPs K and E1/E2. Biol. Cell. 2004;96:407–411. [PubMed]
Outten CE, O'Halloran TV. Femtomolar sensitivity of metalloregulatory proteins controlling zinc homeostasis. Science. 2001;292:2488–2492. [PubMed]
Ozer A, Bruick RK. Non-heme dioxygenases: cellular sensors and regulators jelly rolled into one? Nat. Chem. Biol. 2007;3:144–153. [PubMed]
Pfaffl MW. A new mathematical model for relative quantification in real-time RT-PCR. Nucleic Acids Res. 2001;29:e45. [PMC free article] [PubMed]
Rosenzweig AC. Metallochaperones: bind and deliver. Chem. Biol. 2002;9:673–677. [PubMed]
Shah YM, Matsubara T, Ito S, Yim SH, Gonzalez FJ. Intestinal hypoxia-inducible transcription factors are essential for iron absorption following iron deficiency. Cell Metab. 2009;9:152–164. [PMC free article] [PubMed]
Shanklin J, Guy JE, Mishra G, Lindqvist Y. Desaturases: emerging models for understanding functional diversification of diiron-containing enzymes. J. Biol. Chem. 2009;284:18559–18563. [PubMed]
Shi H, Bencze KZ, Stemmler TL, Philpott CC. A cytosolic iron chaperone that delivers iron to ferritin. Science. 2008;320:1207–1210. [PMC free article] [PubMed]
Stemmler TL, Lesuisse E, Pain D, Dancis A. Frataxin and mitochondrial FeS cluster biogenesis. J. Biol. Chem. 2010;285:26737–26743. [PubMed]
Tian H, McKnight SL, Russell DW. Endothelial PAS domain protein 1 (EPAS1), a transcription factor selectively expressed in endothelial cells. Genes Dev. 1997;11:72–82. [PubMed]
Tuckerman JR, Zhao Y, Hewitson KS, Tian YM, Pugh CW, Ratcliffe PJ, Mole DR. Determination and comparison of specific activity of the HIF-prolyl hydroxylases. FEBS Lett. 2004;576:145–150. [PubMed]
Waldron KJ, Rutherford JC, Ford D, Robinson NJ. Metalloproteins and metal sensing. Nature. 2009;460:823–830. [PubMed]
Zhang N, Fu Z, Linke S, Chicher J, Gorman JJ, Visk D, Haddad GG, Poellinger L, Peet DJ, Powell F, Johnson RS. The asparaginyl hydroxylase factor inhibiting HIF-1alpha is an essential regulator of metabolism. Cell Metab. 2010;11:364–378. [PMC free article] [PubMed]