Hypoxia-inducible Factor 2α Regulates Expression of the Mitochondrial Aconitase Chaperone Protein Frataxin*

Mice lacking Epas1, encoding the transcription factor Hypoxia-inducible Factor 2α (HIF-2α), exhibit an apparent mitochondrial disease state. Similarities between knock-outs of Epas1 and of Sod2, encoding the mitochondrial antioxidant enzyme manganese superoxide dismutase, led to the identification of Sod2 as a HIF-2α target gene. However, Sod2 levels in Epas1–/– liver are intermediate between that of Sod+/– and Sod2–/– mice, which have subtle or severe phenotypes, respectively. This suggests that additional HIF-2α target genes besides Sod2 contribute to the Epas1–/– mitochondrial disease state. To define the nature of the mitochondrial defect in Epas1–/– liver, we performed biophysical, biochemical, and molecular studies. In the setting of decreased Sod2 levels and increased oxidative stress, we found reduced respiration, sensitized mitochondrial permeability transition pore opening, intact electron transport chain activities, and impaired mitochondrial aconitase activity. Mitochondrial aconitase protein levels were preserved, whereas mRNA and protein levels for frataxin, the oxidative stress-regulated mitochondrial aconitase chaperone protein, were markedly reduced in Epas1–/– livers. The mouse Fxn promoter was preferentially activated by HIF-2α through a consensus HIF-responsive enhancer element. In summary, the studies reveal that Fxn, like Sod2, is a nuclear-encoded, mitochondrial-localized HIF-2α target gene required for optimal mitochondrial homeostasis. These findings expand upon the previously defined role of HIF-2α in the cellular response to oxidative stress and identify a novel link of HIF-2α with mitochondrial homeostasis.

Members of the PAS domain protein family sense and respond to diverse environmental stimuli (1). Hypoxia-inducible factor 2␣ (HIF-2␣), 2 encoded by the EPAS1 gene, is a member of the bHLH-PAS domain protein transcription factor family that shows high sequence similarity with HIF-1␣, a transcription factor with a prominent role in the transcriptional response to hypoxia. Like HIF-1␣, HIF-2␣ is activated by hypoxia via a common signaling pathway. In normoxia, HIF␣ members are regulated by a family of Fe(II) and 2-oxoglutaratedependent prolyl hydroxylases, ultimately resulting in proteasomal degradation of prolyl-hydroxylated HIFs (2,3). Under hypoxic conditions, prolyl hydroxylases are inhibited, allowing HIF␣ members to escape degradation and become transcriptionally active.
Besides hypoxia, other environmental stimuli activate HIF members, including mitochondrial reactive oxygen species and/or altered cellular redox state (4 -11). However, the susceptibility of HIF-1␣ and HIF-2␣ to select environmental stresses differs in a HIF␣ isoform-specific manner. This is due in part to specific activation determinants for, biochemical properties of, and cellular proteins associated with HIF-1␣ and HIF-2␣. These factors contribute to unique HIF target gene selectivity and biological roles as evident from transgenic mouse studies.
The Epas1 Ϫ/Ϫ phenotype closely resembles that of Sod2 Ϫ/Ϫ mice, a mouse model of mitochondrial disease (19,20). Sod2 encodes manganese superoxide dismutase, the mitochondriallocalized antioxidant enzyme responsible for mitochondrial reactive oxygen species homeostasis. The similarities between the Epas1 Ϫ/Ϫ and Sod2 Ϫ/Ϫ strains suggested a shared biochemical etiology, increased oxidative stress, as well as a possibly common molecular etiology, decreased Sod2 gene expression, for the mitochondrial disease-like state in Epas1 Ϫ/Ϫ mice. Consistent with this hypothesis, biochemical and histological experiments revealed increased superoxide production and decreased Sod2 mRNA expression in Epas1 Ϫ/Ϫ liver (17). HIF-2␣ over-expression stimulated transcriptional activity of the Sod2 promoter and treatment of Epas1 Ϫ/Ϫ juvenile mice or embryos with an antioxidant enzyme mimetic rescued several aspects of the Epas1 Ϫ/Ϫ phenotype, similar to results obtained from antioxidant enzyme mimetic treatment of Sod2 Ϫ/Ϫ mice (21).
Even though the Epas1 Ϫ/Ϫ and Sod2 Ϫ/Ϫ phenotypes overlap considerably, liver Sod2 expression in Epas1 Ϫ/Ϫ mice is intermediate between that of Sod2 ϩ/Ϫ mice, which have subtle mitochondrial defects (22,23), and that of Sod2 Ϫ/Ϫ mice, which have absent Sod2 levels and more severe mitochondrial, as well as gross pathological, defects (19,20). Although a Sod2 threshold effect may explain the severity of the Epas1 Ϫ/Ϫ phenotype, an alternative etiology for the overt disease in Epas1 Ϫ/Ϫ mice is reduced expression of additional HIF-2␣ target genes involved in mitochondrial homeostasis.
In this study, we investigated the biophysical, biochemical, and molecular bases for the apparent mitochondrial defect in Epas1 Ϫ/Ϫ liver, a prominent site for pathology in Epas1 Ϫ/Ϫ mice. Our studies revealed increased oxidative stress, impaired mitochondrial respiration, intact electron transport activity, and reduced activities of two Krebs tricarboxylic acid cycle components, mitochondrial aconitase (mAco) and ␣-ketoglutarate dehydrogenase (KGDH). Whereas the KGDH defect in Epas1 Ϫ/Ϫ liver mitochondria is in all likelihood due to direct effects of oxidative stress upon KGDH, the mechanism for mAco inhibition involves reduced expression of a HIF-2␣ target gene encoding frataxin, a mAco chaperone protein, and a resultant increased susceptibility of mAco to oxidative stressinduced inactivation.

Reagents and Chemicals
All chemicals were from Sigma.

Mouse Studies
All animal procedures were in accordance with UTSWMC Institutional Animal Care and Use Committee guidelines. Epas1 Ϫ/Ϫ mice obtained from crosses of Epas1 ϩ/Ϫ 129S6/ SvEvTac and C57BL/6J strains (17) were housed in a standard 12-h light:12-h dark cycle and fed ad libitum (4% fat chow for mating pairs; 11% fat chow for pregnant dams and newborn to 1-month mice).

Liver Mitochondria Isolation Procedures
Oxygen Consumption Studies, Mitochondrial NADH Oxidase, Aconitase, KGDH, and Mitochondrial ATPase Assays-Liver mitochondria were isolated from 1-month-old mice by differential centrifugation (24). Minced liver was homogenized in 15 vol:wt ice-cold Buffer H1 (225 mM mannitol, 75 mM sucrose, 10 mM MOPS, 1 mM EGTA, 0.5% fatty acid-free bovine serum albumin, pH 7.2), pelleted at 1,000 ϫ g, and the supernatant centrifuged at 10,000 ϫ g to pellet mitochondria. The pellet was washed with Buffer H2 (Buffer H1 without EGTA), pelleted, resuspended in Buffer H2, and gently homogenized with a Tef-lon-glass homogenizer. Protein concentration was determined by Lowry method (Pierce. The suspension was used immediately for polarographic oxygen consumption studies or frozen at Ϫ80°C for enzymatic assays. Liver Oxidative Phosphorylation Enzyme Assays-Liver was washed, blotted, weighed, and minced in 10 vol:wt Homogenization Buffer (120 mM KCl, 5 mM MgCl 2 , 1 mM EGTA, 20 mM HEPES, 0.5% fatty acid-free bovine serum albumin, 50 units/ml heparin, pH 7.2). Minced liver was homogenized and centrifuged at 1,000 ϫ g, and aliquots of the supernatant were snapfrozen and stored at Ϫ80°C for Complexes I or III activity assays. The remainder of the supernatant was freeze-thawed three times for Complexes II or IV or citrate synthase activity assays. Complexes I through IV mean activities were normalized to mean citrate synthase (CS) activity within each genotype.

Polarographic Measurement of Oxygen Consumption
State III and IV respiration rates of isolated liver mitochondria were measured using a Clark-type oxygen electrode in a magnetically stirred chamber at 30°C with succinate (3.1 mM) plus rotenone (5 M) or glutamate plus malate (3.1 mM each) as substrates (24,26). Respiratory Control Ratio (RCR) was calculated as the ratio of state III to state IV respiratory rates.

NADH Oxidase Assay
Electron transport chain (ETC) activity was measured as NADH oxidation with sonicated submitochondrial particles (SMPs) (30) (⑀ 340 nm ϭ 6.22 mM Ϫ1 cm Ϫ1 ). NADH oxidase activ-ity was normalized to CS activity for each particular sample with a mean calculated from the normalized values according to genotype.

Mitochondrial Aconitase Assay
Mitochondria were used for mitochondrial aconitase activity determinations at 30°C by formation of cis-aconitate (31-33) (⑀ 240 nm ϭ 3.6 mM Ϫ1 cm Ϫ1 ). Mitochondrial aconitase activity was normalized to CS activity for each particular sample with a mean calculated from the normalized values according to genotype.

KGDH Assay
Freeze-thawed mitochondria prepared for the mitochondrial aconitase assay were used for KGDH activity measurements (34) as the rate of NADϩ reduction (⑀ 340 nm ϭ 6.22 mM Ϫ1 cm Ϫ1 ). KGDH activity was normalized to CS activity for each particular sample with a mean calculated from the normalized values according to genotype.

Preparation of SMPs for Superoxide Generation Assays
SMPs (submitochondrial particles) were prepared as described (35) with modifications. Frozen mitochondrial suspension was thawed on ice and adjusted to 20 mg/ml with 0.25 M sucrose.
EDTA was added to 2 mM and the pH adjusted to 8.0 -8.5. The mitochondrial suspension was sonicated for 10 s, cooled for 30 s, and sonicated for 10 s at 30% output level (Fisher Scientific Sonic Dismembrator 550 Model, Pittsburgh, PA). One ml of ddH 2 o was added and the unbroken mitochondria pelleted at 10,600 ϫ g for 10 min. The supernatant was transferred and centrifuged at 105,000 ϫ g for 1 h. The pellet was washed with 500 l of 0.25 M sucrose and pelleted again at 105,000 ϫ g for 40 min. The final pellet was resuspended in 50 -125 l of 0.25 M sucrose, frozen at Ϫ80°C, and assayed within a week of preparation.

Mitochondrial F 1 F 0 -ATPase Assay
Mitochondrial F 1 F 0 -ATPase activity was measured by modification of the method described previously (36). Freeze-thawed mitochondrial extract (30 g) or KH 2 PO 4 standards were incubated in Assay Buffer (50 mM KCl, 5 mM MgCl 2 , 20 mM Tris-Cl, pH 7.4, and 1.25 M rotenone) at 37°C for 5 min. F 1 F 0 -ATPase activity was stimulated by addition of Mg-ATP to 1 mM. Oligomycin (2 M) and ouabain (1 mM) were used to determine nonmitochondrial ATPase activity. Absorbance was measured at 400 nm after 5 min, and F 1 F 0 -ATPase activity was calculated as the difference in phosphate amounts between the 2-and 4-min time points normalized to protein.

Superoxide Generation Assay
Superoxide (O 2 . ) generation rates were calculated by ferricytochrome c reduction using acetylated cytochrome c as the substrate and liver SMPs as the O 2 . source (37,38). Superoxide dismutase (SOD) added to parallel samples confirmed SODsensitive cytochrome c reduction.

RNA and Protein Extract Preparation
Total RNA prepared from individual liver samples matched for age, sex, and genotype was used to generate cDNA. Where indicated, cDNA was pooled according to genotype. For liver protein extracts, frozen tissue was minced in 50 ml/gm ice-cold harvesting buffer (25 mM MOPS, 10 mM EDTA, pH 7.4) supplemented with 1 g/ml protease inhibitor mixture (Sigma).

Real-time Reverse Transcription PCR Assays
Real-time reverse transcription PCR identified changes in Sod2, Fxn, and Actb gene expression using an ABI Prism 7700 sequence detection system. Expression was compared using the threshold cycle method normalized to Actb.

Western Blot Analyses
Whole liver or mitochondrial protein preparations were separated by SDS-PAGE gels and transferred to nitrocellulose membranes or polyvinylidene difluoride membranes for mitochondrial aconitase and frataxin proteins, respectively (39). The blots were blocked in Blotto-Tween solution (5% nonfat dry milk, 1% Tween in phosphate-buffered saline) for 1 h and incubated overnight in Blotto-Tween with a rabbit anti-manganese superoxide dismutase polyclonal antibody (SOD-110; Stressgen, Victoria BC, Canada), rabbit anti-rat mitochondrial aconitase (1:5,000 dilution), or rabbit anti-human frataxin (1:1,000 dilution) primary antibody (40 -42). A horseradish peroxidase-coupled goat anti-rabbit secondary antibody (Pierce) and a chemiluminescent developer (SuperSignal West Pico Chemiluminescent Substrate; Pierce) were used to detect primary antibody binding. For cytochrome c analyses, a mouse anti-cytochrome c monoclonal antibody (1:1000 dilution; BD Biosciences) was used for the primary antibody and detected with a goat antimouse peroxidase-conjugated IgG. Western images were acquired by film and stored digitally for subsequent analysis.

Statistics
We report data as the mean with standard deviation (S.D.) for means generated from individual samples or with standard error of the mean (S.E.) for means generated from group means. Comparisons were made between groups of equal size by Student's t tests for paired data using Excel (Microsoft, Redmond, WA).

RESULTS
As we have previously found (17), the absence of HIF-2␣ results in reduced Sod2 mRNA (Fig. 1A) and protein (Fig. 1B) levels in liver. The reduced Sod2 gene expression had functional consequences as superoxide generation for Epas1 Ϫ/Ϫ liver SMPs were elevated compared with Epas1 ϩ/ϩ liver SMPs (Fig. 1C). The MPTP opening, a biophysical property affected by oxidative stress, was evaluated under basal conditions or after calcium (Ca 2ϩ ) stimulation. In the absence of Ca 2ϩ , there was no difference between MPTP opening of Epas1 Ϫ/Ϫ and Epas1 ϩ/ϩ liver mitochondria. However, in the presence of 250 M Ca 2ϩ , Epas1 Ϫ/Ϫ liver mitochondria underwent more rapid MPTP opening compared with Epas1 ϩ/ϩ liver mitochondria (Fig. 1D).
To determine whether increased oxidative stress was associated with impaired respiratory function, we analyzed respiration of isolated Epas1 ϩ/ϩ liver mitochondria using oxygen polarography (26). Complex I-linked state III rates were lower, whereas complex II-linked state III respiratory rates were not significantly different for Epas1 Ϫ/Ϫ liver mitochondria (Table  1). State IV respiratory rates were similar or slightly higher for Epas1 Ϫ/Ϫ liver mitochondria for Complex I as well as Complex II substrates. Lower state III and normal/higher state IV rates resulted in lower RCR of Epas1 Ϫ/Ϫ mitochondria. For Complex I substrates, there was a 25% reduction in RCR, and for Complex II substrates there was an 18% reduction in RCR.
We next examined ETC activities. In contrast to oxygen consumption, individual ETC complex activities were not impaired in Epas1 Ϫ/Ϫ liver mitochondria (Table 2). Complex V activity, measured as mitochondrial F 1 F 0 -ATPase activity, was not impaired (Table 3), eliminating this as an etiology for lower state III respiration rates. To assess the integrated ETC func-

HIF-2␣ knock-out liver mitochondria have impaired respiratory coupling
The ADP-limited (State III) or ADP-replete (State IV) respiration rates (nano atoms oxygen/min/mg protein) using liver mitochondria isolated from Epas1 ϩ/ϩ (WT) or Epas1 Ϫ/Ϫ (KO) liver were measured using an oxygen electrode set-up with Complex I-linked substrates (glutamate plus malate, Glut ϩ Mal) or Complex II-linked substrates (succinate in the presence of rotenone, Succ). The RCR for Complex I-or Complex II-linked substrates, calculated as the ratio of State III to State IV respiratory rates, is shown as is the ADP:O ratio (26). Each value represents the mean with SD of n ϭ 5/genotype. p values were determined by Student's paired t test (one-tailed) with p Ͻ 0.05 designated with a single asterisk (*) and p Ͻ 0.10 by a double asterisk (**).  tion, we performed an NADH-linked assay that depends on cooperation among ETC complexes, ubiquinone, and cytochrome c, but does not require Krebs tricarboxylic acid cycle activity. We found no impairment of NADH oxidase activity for Epas1 Ϫ/Ϫ compared with Epas1 ϩ/ϩ liver SMPs (Table 3). The aforementioned data suggested that a defect in Krebs tricarboxylic acid cycle rather than ETC function might be responsible for the impaired respiration of Epas1 Ϫ/Ϫ liver mitochondria. We measured activities of two Krebs tricarbox-ylic acid cycle enzymes, ␣-ketoglutarate dehydrogenase (KGDH) and mitochondrial aconitase (mAco), affected in oxidative stress states. KGDH activity was decreased by 33% in Epas1 Ϫ/Ϫ compared with Epas1 ϩ/ϩ liver mitochondria (Table  4). There was an even more dramatic reduction in mAco activity, which was reduced by 45% in Epas1 Ϫ/Ϫ compared with Epas1 ϩ/ϩ liver mitochondria.

Glut
To evaluate whether reduced mAco activity in Epas1 Ϫ/Ϫ liver mitochondria was due to lower mAco protein levels, Western blot studies were performed. There was no evidence of decreased mAco protein levels in Epas1 Ϫ/Ϫ liver samples. Instead, there was a slight increase in mAco protein levels ( Fig.  2A) as well as an increase in mAco degradation products, the latter observed with extreme oxidative stress states (44 -46). Of note, we could not restore mAco activity by reactivation with dithiothreitol and iron (data not shown), in contrast to the restoration of mAco activity in Sod2 ϩ/Ϫ liver mitochondria (47).
Protein levels for mAco ( Fig. 2A) and KGDH (data not shown) were preserved in Epas1 Ϫ/Ϫ liver. Whereas the susceptibility of KGDH to oxidative stress results from direct modification of KGDH residues or moieties (48,49), the effect of oxidative stress upon mAco function includes alterations of frataxin, a chaperone protein that retards irreversible iron-sulfur cluster disassembly, and subsequent inactivation of mAco (50). We postulated that mAco inhibition in Epas1 Ϫ/Ϫ liver involved reductions in frataxin levels. Indeed, Western blot analyses revealed a substantial reduction in frataxin protein levels in Epas1 Ϫ/Ϫ liver to levels Ͻ50% of Epas1 ϩ/ϩ liver (Fig. 2B).
The reduction in frataxin protein levels in Epas1 Ϫ/Ϫ livers indicated that frataxin might be a direct HIF-2␣ target gene.
Steady-state frataxin mRNA levels were reduced in Epas1 Ϫ/Ϫ versus Epas1 ϩ/ϩ livers (Fig. 2C). To identify cis-acting elements through which HIF members might modulate frataxin gene expression, we examined the upstream regulatory region of the mouse Fxn gene for presence of HIF-responsive enhancer elements (HREs). We found a consensus HRE located ϳ2 kb upstream from the start of transcription. To assess the ability of HIF members to regulate frataxin gene expression, we constructed mouse Fxn promoter reporters that retain or lack the upstream HRE (Fig. 3A). The mouse Fxn promoter reporters were transfected into HEK293 or HepG2 cells along with expression plasmids encoding constitutively active HIF-1␣ or HIF-2␣. Although HIF-1␣ and HIF-2␣ activated a synthetic HIF-responsive reporter to a comparable extent (data not shown), HIF-2␣ preferentially activated the HRE-containing Fxn promoter reporter in HEK293 (Fig. 3B) FIGURE 2. Frataxin expression is reduced in HIF-2␣ knock-out liver mitochondria. A, aconitase protein levels determined by Western blots of equivalent protein aliquots from Epas1 ϩ/ϩ (WT) or Epas1 Ϫ/Ϫ (KO) liver samples. The bar graph represents mean of densitometry measurements with S.D. of n ϭ 6/genotype. p value determined by Student's paired t test (two-tailed) was p Յ 0.01. Shown is a representative example of one of six independent WT/KO pairs examined in this study. B, frataxin protein levels determined by Western blots of equivalent protein aliquots from Epas1 ϩ/ϩ (WT) or Epas1 Ϫ/Ϫ (KO) liver samples used in panel A. The bar graph represents mean of densitometry measurements with S.D. of n ϭ 6/genotype. p value determined by Student's paired t test (two-tailed) was p Յ 0.0001. Shown is a representative example of one of six independent WT/KO pairs examined in this study. C, bar graph of frataxin mRNA levels from Epas1 ϩ/ϩ (WT) or Epas1 Ϫ/Ϫ (KO) liver samples as determined by real-time reverse transcription PCR. Each value represents the mean with S.D. of n ϭ 7/genotype. p value determined by Student's paired t test (two-tailed) was p ϭ 0.02. F 0 -ATPase and NADH oxidase activities are maintained in HIF-2␣ knock-out liver mitochondria F 1 F 0 -ATPase activity (mol/min/mg protein) and NADH oxidase activity (nmol/ min/mg protein) were measured using mitochondria isolated from Epas1 ϩ/ϩ (WT) or Epas1 Ϫ/Ϫ (KO) livers. Each value represents the mean with SD of n ϭ 4/genotype for F 1 F 0 -ATPase assays or n ϭ 8/genotype for NADH oxidase assays. p values were determined by Student's paired t test (two-tailed) with p Ͻ 0.05 designated with an asterisk (*).   and HepG2 cells (Fig. 3C). Deletion of the HRE resulted in complete loss of HIF-2␣ inducibility in HEK293 cells and in substantial reductions as well as loss of the preferential stimulatory action of HIF-2␣ in HepG2 cells.

DISCUSSION
Previous studies of Epas1 Ϫ/Ϫ mice revealed an in vivo role for HIF-2␣ in the regulation of several major antioxidant enzyme genes, including Sod2 (17). Comparisons of Epas1 Ϫ/Ϫ to Sod2deficient liver mitochondrial studies indicate Sod2 deficiency likely plays a role in several mitochondrial abnormalities observed in Epas1 Ϫ/Ϫ liver mitochondria. Epas1 Ϫ/Ϫ liver mito-chondria have increased oxidative stress as evidenced by increased O 2 . generation and sensitized MPTP opening.
Increased O 2 . generation was observed with Sod2 ϩ/Ϫ liver mitochondria (51), whereas MPTP opening sensitization was seen with both Sod2 ϩ/Ϫ (22) as well as Sod2 Ϫ/Ϫ (24) liver mitochondria. Mitochondria oxygen consumption, an integrative assessment of mitochondrial function (52,53) and metabolic capacity (54), is lower for complex I-linked substrates, and RCRs are lower for both complex I-as well as complex II-linked substrates in Epas1 Ϫ/Ϫ liver mitochondria. Similar patterns of mitochondrial respiration were observed in Sod2 ϩ/Ϫ (22) and Sod2 Ϫ/Ϫ (24) liver mitochondria. Oxidative stress can impair mitochondrial function by disrupting the function of iron-sulfur cluster-containing factors. Oxidative stress-induced mAco inhibition is the most sensitive biomarker of mitochondrial oxidative stress, occurs faster than inactivation of the iron-sulfur center containing complexes I and II (55), and involves both reversible and irreversible inactivation processes (46). The pattern of respiration in Epas1 Ϫ/Ϫ liver mitochondria, greater reduction of state III rates for complex I as opposed to complex II substrates, is consistent with reduced Krebs tricarboxylic acid cycle function. Indeed, mAco and KGDH activities are impaired in Epas1 Ϫ/Ϫ liver mitochondria whereas ETC enzyme activities remain intact. In comparison, mAco activity was reduced in Sod2 ϩ/Ϫ livers without a reduction in mAco protein levels (22) or ETC function (24). In contrast to Sod2 ϩ/Ϫ liver mitochondria (22), mAco activity in Epas1 Ϫ/Ϫ liver mitochondria cannot be reactivated. Furthermore, mAco protein is degraded in Epas1 Ϫ/Ϫ liver, suggesting a greater extent of oxidative damage and/or an enhanced susceptibility to oxidative stress.
For mouse knock-outs of transcription factors, including HIF-1␣ and HIF-2␣ (12)(13)(14)(15)(16), the phenotype is often a consequence of impaired regulation of multiple target genes linked to common physiological processes. Epas1 Ϫ/Ϫ mice have gross pathological features most similar to Sod2 Ϫ/Ϫ mice (24), yet Sod2 levels in Epas1 Ϫ/Ϫ liver are closer to Sod2 ϩ/Ϫ mice, which have clinically silent mitochondrial impairment and no overt histological abnormalities (22). This suggests that although reduced Sod2 levels are abnormal, particularly during oxidative stress states when Sod2 expression normally is increased, other factors besides Sod2 deficiency likely contribute to the apparent mitochondrial defect in Epas1 Ϫ/Ϫ liver. Frataxin gene expression is significantly depressed in Epas1 Ϫ/Ϫ liver. Frataxin protects mAco iron-sulfur clusters from disassembly, irreversible inactivation, and, possibly, degradation (46,50). This raises the possibility that frataxin is a HIF-2␣ target gene required for optimal mAco activity. Consistent with this hypothesis, the regulatory region for the mouse frataxin gene is preferentially activated by HIF-2␣ rather than HIF-1␣.
From these and earlier studies (17), we conclude frataxin joins Sod2 as HIF-2␣ target genes that maintain mitochondrial homeostasis. The biological role of HIF-2␣ complements that of HIF-1␣, a key transcriptional regulator of enzymes controlling aerobic and anaerobic glucose metabolism. The consequences arising from decreased expression of Sod2, frataxin, and other HIF-2␣ target genes in select tissues of Epas1 Ϫ/Ϫ mice are induction and/or augmentation of mitochondrial dysfunction, especially in a setting of A, schematic representation of a wild type mouse frataxin promoter construct (WT mFxn Prom/Luc, Ϫ1947 to ϩ5) or a 5Ј-deletion derivative that lacks the HRE (⌬HRE mFxn Prom/Luc, Ϫ1892 to ϩ5) used in the transfection studies. Numbers refer to nucleotide positions relative to the transcriptional start as annotated in the mouse genome (NIH/NCBI). B, reporter activity from HEK293 cells transfected with WT or ⌬HRE mFxn Prom/Luc plus or minus constitutively active (P1P2N) HIF-1␣ or HIF-2␣ expression plasmids. The bar graph represents the mean with S.E. of n ϭ 3 sets of transfections performed in triplicate for each set. p value determined by Student's paired t test (two-tailed) was p ϭ 0.02 for activation of the WT mFxn Prom/Luc reporter by HIF-2␣ compared with HIF-1␣. There was no significant difference (nsd) between activation of the ⌬HRE mFxn Prom/Luc reporter by HIF-2␣ compared with HIF-1␣. C, reporter activity from HepG2 cells transfected with WT or ⌬HRE mFxn Prom/Luc reporter plasmids plus or minus constitutively active (P1P2N) HIF-1␣ or HIF-2␣ expression plasmids. The bar graph represents the mean with S.E. of n ϭ 3 sets of transfections performed in triplicate for each set. p value determined by Student's paired t test (two-tailed) was p ϭ 0.016 for activation of the WT mFxn Prom/Luc reporter by HIF-2␣ compared with HIF-1␣. There was no significant difference (nsd) between activation of the ⌬HRE mFxn Prom/Luc reporter by HIF-2␣ compared with HIF-1␣.
increased oxidative stress. The proposed HIF-2␣-mediated regulatory response to oxidative and possibly other environmental stresses includes increases in Sod2 expression, to maintain optimal reactive oxygen species and/or mitochondrial homeostasis, and in frataxin expression, to preserve the function of the proposed cellular redox sensor/effector mAco (56). Because the upstream regulatory regions of frataxin are poorly conserved among mammals (57), it is not known whether HIF-2␣ regulation of mouse frataxin gene expression can be extrapolated to humans. Future investigations will be needed to address this question and to identify other regulatory processes that link HIF-2␣ biology to the intermediary metabolic processes of mice and men.