Lea Rahtu-Korpela,1 Sara Karsikas,1 Sohvi Hörkkö,2,3 Roberto Blanco Sequeiros,4

Eveliina Lammentausta,4 Kari A. Mäkelä,5 Karl-Heinz Herzig,5 Gail Walkinshaw,6 Kari I. Kivirikko,1

Johanna Myllyharju,1 Raisa Serpi,1 and Peppi Koivunen1

HIF Prolyl 4-Hydroxylase-2Inhibition Improves Glucose andLipid Metabolism and ProtectsAgainst Obesity and MetabolicDysfunctionDiabetes 2014;63:3324–3333 | DOI: 10.2337/db14-0472

Obesity is a major public health problem, predisposingsubjects to metabolic syndrome, type 2 diabetes, andcardiovascular diseases. Specific prolyl 4-hydroxylases(P4Hs) regulate the stability of the hypoxia-induciblefactor (HIF), a potent governor of metabolism, withisoenzyme 2 being the main regulator. We inves-tigated whether HIF-P4H-2 inhibition could be usedto treat obesity and its consequences. Hif-p4h-2–deficient mice, whether fed normal chow or a high-fat diet, had less adipose tissue, smaller adipocytes,and less adipose tissue inflammation than their litter-mates. They also had improved glucose tolerance andinsulin sensitivity. Furthermore, the mRNA levels ofthe HIF-1 targets glucose transporters, glycolyticenzymes, and pyruvate dehydrogenase kinase-1 wereincreased in their tissues, whereas acetyl-CoA con-centration was decreased. The hepatic mRNA levelof the HIF-2 target insulin receptor substrate-2 washigher, whereas that of two key enzymes of fatty acidsynthesis was lower. Serum cholesterol levels and denovo lipid synthesis were decreased, and the micewere protected against hepatic steatosis. Oral admin-istration of an HIF-P4H inhibitor, FG-4497, to wild-typemice with metabolic dysfunction phenocopied thesebeneficial effects. HIF-P4H-2 inhibition may be a noveltherapy that not only protects against the development

of obesity and its consequences but also reverses theseconditions.

Hypoxia-inducible factor (HIF) regulates the expression ofnumerous hypoxia-regulated genes (1–3). The HIF-a sub-unit isoforms HIF-1a and HIF-2a are synthesized consti-tutively, and hydroxylation of two critical prolinesgenerates 4-hydroxyproline residues that target HIF-afor degradation in normoxia. In hypoxia, this hydroxyl-ation is inhibited so that HIF-a evades degradation andforms a functional dimer with HIF-b (1–3).

The hydroxylation of HIF-a is catalyzed by HIF prolyl4-hydroxylase (P4H) isoenzymes 1–3 (also known asPHDs 1–3 and EglNs 2, 1, and 3) (1–6) and a transmem-brane, P4H-TM (3,7), with HIF-P4H-2 being the mainoxygen sensor in the HIF pathway (1–3). Hif-p4h-2 nullmice die during embryonic development, whereas Hif-p4h-1and Hif-p4h-3 null mice are viable (8). Broad-spectrumconditional Hif-p4h-2 inactivation leads to severe eryth-rocytosis, hyperactive angiogenesis, and dilated cardiomy-opathy (3,9,10). We have generated Hif-p4h-2 hypomorphicmice (Hif-p4h-2gt/gt) that express decreased amounts ofwild-type Hif-p4h-2 mRNA and show stabilization of Hif-as (11). These mice appear healthy and have a normal life

1Biocenter Oulu, Faculty of Biochemistry and Molecular Medicine, Oulu Center forCell-Matrix Research, University of Oulu, FIN-90014 Oulu, Finland2Nordlab Oulu, Oulu University Hospital, FIN-90220 Oulu, Finland3Department of Medical Microbiology and Immunology, Medical Research Center,University of Oulu, FIN-90014 Oulu, Finland4Department of Radiology, Oulu University Hospital and University of Oulu, FIN-90029 Oulu, Finland5Biocenter Oulu, Department of Physiology, University of Oulu, FIN-90014 Oulu,Finland6FibroGen, Inc., San Francisco, CA

Corresponding author: Peppi Koivunen, peppi.koivunen@oulu.fi.

Received 21 March 2014 and accepted 24 April 2014.

This article contains Supplementary Data online at http://diabetes.diabetesjournals.org/lookup/suppl/doi:10.2337/db14-0472/-/DC1.

© 2014 by the American Diabetes Association. Readers may use this article aslong as the work is properly cited, the use is educational and not for profit, andthe work is not altered.

3324 Diabetes Volume 63, October 2014

OBESITY

STUDIES

span. They have no increased erythrocytosis and show nosigns of hyperactive angiogenesis or dilated cardiomyopa-thy; instead, they are protected against myocardial infarc-tion and ischemia-reperfusion injury (11,12).

Many studies have demonstrated that hypoxia reducesbody weight (13–15). The only obvious abnormality foundin the Hif-p4h-2gt/gt mice was that their weights were 85–90% of those of the wild type (11). We focused initially onthis difference and found that these mice have less adi-pose tissue than their littermates. This finding promptedus to study their lipid and glucose metabolism, especiallybecause obesity is a major public health problem andincreases the risk of metabolic syndrome, type 2 diabetes,and cardiovascular diseases. The data show that the Hif-p4h-2gt/gt mice, whether fed normal chow or a high-fatdiet (HFD), have major alterations in their adipose tissuesand metabolism, including improved glucose tolerance andinsulin sensitivity, reduced serum cholesterol levels, andprotection against hepatic steatosis.

Small-molecule HIF-P4H inhibitors have been devel-oped for the treatment of anemias and ischemic diseases,for example (3). Our Hif-p4h-2gt/gt mouse data suggestthat pharmacological HIF-P4H-2 inhibition could also bebeneficial for the treatment of obesity and metabolic syn-drome. We therefore administered FG-4497 to wild-typemice with a metabolic dysfunction, which stabilizes HIF-ain cultured cells and in vivo and increases erythropoiesisin animals with no apparent toxicity (3,16,17). The datademonstrate that FG-4497 administration phenocopiedthe beneficial effects of genetic HIF-P4H-2 deficiency onadipose tissues and metabolism.

RESEARCH DESIGN AND METHODS

Animal ExperimentsGeneration of C57BL/6 Hif-p4h-2gt/gt mice has been de-scribed (11). All experiments were performed according toprotocols approved by the Provincial State Office ofSouthern Finland. All data obtained with the Hif-p4h-2gt/gt mice were compared with those of their wild-typelittermates. The mice were fed a standard rodent diet orHFD (18% and 42% kcal fat, respectively) (TekladT.2018C.12 and TD88137; Harlan Laboratories). FG-4497 was dissolved in 0.5% sodium carboxymethyl cellu-lose (Spectrum) and 0.1% polysorbate 80 (Fluka), and thesolvent was also used as a vehicle. Both were administeredorally.

Histological AnalysesFive-micrometer sections of formaldehyde-fixed paraffin-embedded tissue samples were stained with hematoxylin-eosin (H-E) and viewed and photographed with a LeicaDM LB2 microscope and Leica DFC320 camera or NikonEclipse 50i microscope and DS-5M-L2 camera. Represen-tative pictures (five to eight per mouse) were taken, andareas of 100 adipocytes were quantified with Nikon NIS-Elements BR 2.30 imaging software. Macrophage infiltra-tion was analyzed by an anti-CD68 antibody (ab955;

Abcam) and EnVision Detection System (Dako). Thenumber of macrophage aggregates was calculated fromfive to eight fields per sample. Hepatic steatosis wasscored (0 to ++++) from H-E–stained sections.

Western BlottingNE-PER extraction reagents (Thermo Fisher Scientific)were used to prepare nuclear fractions. Samples of 30–100 mg were resolved by SDS-PAGE, blotted, and probedwith the following primary antibodies: Hif-1a (NB100-479; Novus Biologicals), Hif-2a (ab199; Abcam), andb-actin (NB600-501; Novus Biologicals).

Quantitative PCR AnalysesTotal RNA from tissues was isolated with E.Z.N.A. TotalRNA Kit II (Omega Bio-Tek) or TriPure Isolation Reagent(Roche Applied Science) and reverse transcripted with aniScript cDNA Synthesis Kit (Bio-Rad). Quantitative PCRwas performed with iTaq SYBR Green Supermix with ROX(Bio-Rad) in a Stratagene Mx3005 thermocycler or C1000Touch Thermal Cycler and CFX96 Touch Real-Time PCRDetection System (Bio-Rad) with the primers shown inSupplementary Table 1.

MRI AnalysisTissue fat content was determined using the 3-pointDixon method and a 3-T clinical scanner (Skyra; Siemens,Erlangen, the Netherlands). The sequence used was T1turbo spin echo Dixon (field of view 3203 320; repetitiontime 720; echo time 20, 24 slices). The torso of the animalwas scanned, and the region of interest was drawn toseparate the regional volume of subcutaneous fat for pos-sible correlation with other variables.

Glucose and Insulin Tolerance Tests andDeoxyglucose Uptake TestGlucose tolerance test (GTT) and insulin tolerance test(ITT) were performed on mice fasted for 6–12 h andanesthetized with fentanyl/fluanisone and midazolam.For GTT, mice were injected intraperitoneally with1 mg/g glucose, and blood glucose concentrations weremonitored with a glucometer. Serum insulin values weredetermined with Rat/Mouse Insulin ELISA kit (EZRMI-13K; Millipore), and HOMA-IR scores were calculatedfrom the glucose and insulin values. For ITT, mice wereinjected intraperitoneally with 1 IU/kg insulin (HumulinRegular; Lilly), and blood glucose concentrations weredetermined as for GTT. For deoxyglucose uptake test,mice were injected intraperitoneally with 0.6 mCi/g14C-deoxyglucose (PerkinElmer) combined with 1 mg/gglucose and killed 60 min later. Fifty-microgram piecesof tissues and 50-mL blood samples were homogenizedin 1:1 (or in the case of white adipose tissue [WAT],2:1) chloroform:methanol and centrifuged. The pelletwas re-extracted, and the supernatants were combinedand scintillated for 14C activity.

Determination of Serum LipidsSerum total cholesterol, HDL cholesterol, and triglyceridelevels were determined by an enzymatic method (Roche

diabetes.diabetesjournals.org Rahtu-Korpela and Associates 3325

Diagnostics). The Friedewald equation (18) was used tocalculate LDL + VLDL cholesterol concentrations.

Determination of Tissue Acetyl-CoA ConcentrationSnap-frozen tissues were homogenized in precooled 8%perchloric acid in 40% (volume for volume) ethanol andcentrifuged. The supernatant was neutralized, and theamount of acetyl-CoA was measured with a PicoProbeAcetyl CoA Assay Kit (ab87546; Abcam).

Lipogenesis AnalysisMice were terminally anesthetized with fentanyl/fluanisoneand midazolam and;30-mg pieces of WAT and liver wereexcised and incubated for 90 min at 37°C with 40 mCi[14C]acetate (PerkinElmer) in DMEM pregassed with 95%O2 and 5% CO2. Lipids were extracted after saponificationand scintillated for 14C activity (19).

Statistical AnalysesStudent two-tailed t test was used for comparisons be-tween groups, and Student t test for paired samples wasused for pairwise testing. Fisher exact test was used tocalculate significance for the difference between geno-types in liver steatosis analyses. Areas under the curvewere calculated by the summary measures method. Alldata are mean 6 SEM. P , 0.05 was considered statisti-cally significant.

RESULTS

Hif-p4h-2gt/gt Mice Are Lighter Than Their LittermatesDespite No Alterations in Food Intake and PhysicalActivityHif-p4h-2gt/gt mice have their Hif-p4h-2 gene disrupted bya gene trap insertion cassette. Small amounts of wild-typeHif-p4h-2 mRNA were nevertheless generated from thegene-trapped alleles (11), its smallest relative amountsin the Hif-p4h-2gt/gt tissues studied being found in theheart and skeletal muscle, whereas the highest proportionwas found in the liver (11). We analyzed these values inthe skeletal muscle and liver of the mice in the presentstudy and found that the relative proportions of wild-typeHif-p4h-2 mRNA in these Hif-p4h-2gt/gt tissues were;25% and ;60%, respectively (Supplementary Fig. 1A).We found previously that the amount of Hif-p4h-2 pro-tein in Hif-p4h-2gt/gt skeletal muscle, heart, and kidneycorresponds to that of the wild-type mRNA (11). Hif-1awas well stabilized in the heart and weakly stabilized inthe skeletal muscle and kidney of the Hif-p4h-2gt/gt mice,whereas Hif-2a was weakly stabilized in the heart andskeletal muscle (11). We also found a weak stabilizationof Hif-2a, but not of Hif-1a, in the Hif-p4h-2gt/gt liverwith a more sensitive antibody (Fig. 1A).

The weight of 5-week-old Hif-p4h-2gt/gt mice fed nor-mal chow (18% kcal fat) was lower than that of their wild-type littermates, and the Hif-p4h-2gt/gt mice gained lessweight during the subsequent 10 weeks (Fig. 1B). Theweight difference persisted later, with the weight of 1-year-old Hif-p4h-2gt/gt mice being 80–85% of that ofwild type (Fig. 1B). No difference was found between

the genotypes in tibia length or liver weight relative totibia length (Supplementary Fig. 1B).

No difference was found in the amount of food intakebetween the Hif-p4h-2gt/gt and wild-type mice in metabolichome cage analyses whether expressed per mouse or bodyweight, and no difference was found in the physical ac-tivity between these mice (Supplementary Fig. 2). Therewas a trend for increased O2 consumption and CO2 pro-duction when expressed per body weight in the Hif-p4h-2gt/gt mice, but this reached statistical significance only inthe case of O2 consumption in the dark, with no differ-ence in respiratory exchange ratio between the genotypes(Supplementary Fig. 2).

Hif-p4h-2gt/gt Mice Have Less Adipose Tissue andSmaller AdipocytesThe amount of gonadal WAT in the Hif-p4h-2gt/gt mice,expressed relative to tibia length to correct for bodyweight differences, was one-half of that in the wild-typelittermates (Fig. 1C). Furthermore, the adipocytes in theHif-p4h-2gt/gt WAT were smaller (Fig. 1D) and the amountof wild-type Hif-p4h-2 mRNA in the Hif-p4h-2gt/gt WATwas ;50% of that in wild-type tissue (SupplementaryFig. 1A). The amount of Hif-p4h-2 protein in the Hif-p4h-2gt/gt WAT was correspondingly reduced (Supplemen-tary Fig. 1C), and its Hif-1a but not Hif-2a was stabilizedto a low extent (Fig. 1A). The weight of the brown adiposetissue (BAT) was likewise reduced (Fig. 1E), and the wild-type Hif-p4h-2 mRNA level in the Hif-p4h-2gt/gt BAT wasonly ;10% of that in the wild-type tissue (SupplementaryFig. 1A). No difference in the mRNA level of uncouplingprotein 1 (Ucp1) was found between the Hif-p4h-2gt/gt andwild-type BAT (Supplementary Fig. 1D). MRI analyses oflive anesthetized mice indicated that the amount of sub-cutaneous fat was also reduced in the Hif-p4h-2gt/gt mice(Fig. 1F).

Hif-p4h-2gt/gt Mice Have a Reduced Number ofAdipose Tissue Macrophage AggregatesObesity is associated with chronic adipose tissue in-flammation and formation of macrophage aggregatesaround adipocytes (20,21). Chronic 21-day hypoxiadecreases the number of such aggregates in 1-year-oldmice (15). We found in the present study that their num-ber is significantly lower in the WAT of 1-year-old Hif-p4h-2gt/gt mice than in wild-type mice (Fig. 1G). As in themice subjected to chronic hypoxia (15), increased adipo-nectin mRNA level and decreased leptin and chemochine(C-C motif) ligand 2 mRNA levels were found in the Hif-p4h-2gt/gt WAT (Supplementary Fig. 3A), whereas no in-crease was seen in serum adiponectin level, but serumleptin level was decreased (Supplementary Fig. 3B).

Hif-p4h-2gt/gt Mice Have Reduced Serum CholesterolLevels and Are Protected Against Hepatic SteatosisThe levels of serum total cholesterol and HDL and LDL +VLDL cholesterol were lower in the Hif-p4h-2gt/gt micethan in the wild-type mice, with the HDL/LDL + VLDLratio being increased, whereas the triglyceride level was

3326 HIF-P4H-2 in Obesity and Metabolism Diabetes Volume 63, October 2014

not altered significantly (Fig. 2A). Livers of 1-year-oldwild-type mice had steatosis, even though these micewere not fed an HFD, whereas the Hif-p4h-2gt/gt micewere protected against its development (Fig. 2B).

Hif-p4h-2gt/gt Mice Have Improved Glucose Toleranceand Insulin Sensitivity

The glucose tolerance of 1-year-old Hif-p4h-2gt/gt mice wasmarkedly better than that of their littermates, and eventhe fasting blood glucose levels (at 0 min) were signifi-cantly lower than in the wild type (Fig. 3A and Supple-mentary Fig. 4A). The fasting serum insulin levels andHOMA-IR scores were likewise lower, indicating that thelower blood glucose levels were due to increased insulinsensitivity (Fig. 3B and Supplementary Fig. 4B). A similar,though less marked, difference in glucose tolerance wasseen between the genotypes in 4–5-month-old mice (Fig.3A). Correspondingly, 3–4-month-old Hif-p4h-2gt/gt micehad lower blood glucose levels than wild-type mice in anITT (Fig. 3C). To learn which Hif-p4h-2gt/gt tissues wereresponsible for the increased glucose intake, we studiedthe uptake of 14C-deoxyglucose in fasting mice and foundan increased uptake relative to the wild type in the skel-etal muscle (Fig. 3D).

Wild-Type But Not Hif-p4h-2gt/gt Mice DevelopMetabolic Dysfunction With AgeA comparison of data between 1-year-old and 4–5-month-old wild-type mice indicated that the former had largeradipocytes, a larger number of adipose tissue macrophageaggregates, increased fasting blood glucose and serum in-sulin levels, and higher HOMA-IR scores, whereas none ofthese were significantly different between 1-year-old and4–5-month-old Hif-p4h-2gt/gt mice (Fig. 4A–E). The 1-year-old wild-type mice thus had relative insulin resistance andmetabolic dysfunction compared with the younger mice,whereas the Hif-p4h-2gt/gt mice were protected againsttheir development.

Hif-p4h-2gt/gt Mice Have Altered Expression ofGlucose and Lipid Metabolism GenesThe mRNA levels of the HIF-1a targets Glut1 and severalenzymes of glycolysis (22) were increased in the Hif-p4h-2gt/gt skeletal muscle and WAT (Fig. 5A and B) but not inHif-p4h-2gt/gt liver (Supplementary Fig. 5A). The mRNAlevel of the glucose-regulated Glut2 (23) was lower inthe liver of the Hif-p4h-2gt/gt mice than in wild-typemice, presumably because of their improved glucose tol-erance, whereas the mRNA level of the insulin-regulated

Figure 1—Hif-p4h-2gt/gt mice are lighter than their wild-type (wt) littermates and have less adipose tissue, smaller adipocytes, anda reduced number of macrophage aggregates in WAT. A: Western blot analyses of Hif-1a and Hif-2a protein in nuclear fractions of liverand WAT. b-Actin was used as a loading control. B: Weight of 5-week-old female wt and Hif-p4h-2gt/gt (gt/gt) mice fed normal chow andtheir weight gain during the subsequent 10 weeks and weight of 1-year-old female and male mice. C: Weight of gonadal WAT in 1-year-old male mice. D: Cross-sectional area of adipocytes in the WAT of 1-year-old male mice. Scale bar = 100 mm. E: Weight of sub-cutaneous BAT relative to tibia length in 1-year-old female mice. F: MRI analyses of the amount of subcutaneous adipose tissue in4-month-old female mice. G: Number of macrophage aggregates in gonadal WAT of 1-year-old male wt and gt/gt mice. Adipocytessurrounded by macrophage aggregates (*). Scale bar = 100 mm. Data are mean 6 SEM (n = 4–10 per group). *P < 0.05, **P< 0.01, ***P =0.0001. S.c., subcutaneous.

diabetes.diabetesjournals.org Rahtu-Korpela and Associates 3327

Glut4 (24) was higher in the skeletal muscle and WAT ofthe Hif-p4h-2gt/gt mice, probably because of their in-creased insulin sensitivity (Fig. 5A and B). The mRNAlevel of the HIF-1a target Pdk1, which inhibits pyruvatedehydrogenase activity (22), was increased in the Hif-p4h-2gt/gt skeletal muscle and WAT (Fig. 5A). The PpargmRNA level was likewise increased in the Hif-p4h-2gt/gt

skeletal muscle and WAT (Fig. 5A), this change beingsimilar to that seen in the WAT of mice with adipocyte-specific Hif-p4h-2 deletion (25), whereas the PparamRNA level was slightly decreased in the Hif-p4h-2gt/gt

liver (Fig. 5B). The mRNA levels of the lipolysis markersLipe and Pnpla2 were increased in the Hif-p4h-2gt/gt WAT(Fig. 5A), suggesting that an increased lipolysis may havecontributed to the decreased amount of WAT in thesemice. The mRNA levels of Srebp1c, which regulates lipo-genesis and fatty acid synthesis, and its targets Acca andFas, enzymes of fatty acid synthesis, were lower in theHif-p4h-2gt/gt liver, whereas the mRNA level of the Ldlreceptor was similar in the Hif-p4h-2gt/gt and wild-typelivers (Fig. 5B). The mRNA level of the HIF-2a targetIrs2 (26), which regulates Srepb1c and hepatic lipid accu-mulation (27), was increased in the Hif-p4h-2gt/gt liver,whereas the mRNA level of Irs1 was not altered (Fig.5B). To study whether the increased mRNA levels led toincreased protein levels, we analyzed Glut4, Gadph, andPdk1 in WAT by Western blotting and found increasedlevels of all three proteins in the Hif-p4h-2gt/gt WAT (Sup-plementary Fig. 5B).

Hif-p4h-2gt/gt Mice Have Reduced Acetyl-CoA Levelsand De Novo LipogenesisTo study whether the presumed decreased conversion ofpyruvate to acetyl-CoA as a result of an increased Pdk1expression actually decreased the amount of acetyl-CoA,we measured its amount in skeletal muscle, WAT, andliver and found a decreased concentration in all three Hif-p4h-2gt/gt tissues (Fig. 5C). We also studied whether thedecreased Acca and Fas mRNA and acetyl-CoA levels de-creased de novo lipogenesis by incubating fresh tissueslices with [14C]acetate and measuring the incorporationof radioactivity into extractable lipids and found de-creased lipogenesis in the Hif-p4h-2gt/gt WAT (P = 0.03)and liver (P = 0.06) (Fig. 5D).

Hif-p4h-2gt/gt Mice Are Protected Against HFD-Induced Metabolic Changes and SteatosisTo study whether Hif-p4h-2gt/gt mice are protected againstobesity-induced changes in glucose metabolism, 6-month-old mice were fed an HFD (42% kcal fat) for 6 weeks. Theweight gain of the Hif-p4h-2gt/gt and wild-type mice duringthe HFD treatment was similar, the Hif-p4h-2gt/gt micethus remaining lighter than their littermates (Fig. 6A).The adipocytes were smaller and the number of macro-phage aggregates lower in the Hif-p4h-2gt/gt WAT than inthe wild-type WAT (Fig. 6B and C and Supplementary Fig.6A). The glucose tolerance of the HFD-fed Hif-p4h-2gt/gt

mice was better than that of the HFD-fed wild-type mice,and their fasting blood glucose levels (at 0 min) werelikewise lower (Fig. 6D), whereas the lower serum insulinvalues (by 15%) and HOMA-IR scores (by 36%) in theHFD-fed Hif-p4h-2gt/gt mice were not statistically signifi-cant (Supplementary Fig. 6B). Livers of all 6-month-oldHFD-treated wild-type mice had steatosis (Fig. 6E), withfour of the nine having a steatosis score of ++++, whereasonly four of the eight Hif-p4h-2gt/gt mice had steatosis (P =0.03), with only one of these having a score of +++ andnone having ++++. Thus, the Hif-p4h-2gt/gt mice were pro-tected against the development of HFD-induced hepaticsteatosis.

Pharmacological Hif-p4h Inhibition ReversesMetabolic Dysfunction Both in Aged Wild-Type Miceand in Mice Fed an HFDFG-4497 inhibits all three HIF-P4Hs competitively withrespect to 2-oxoglutarate, with similar IC50 values (7).We studied whether its oral administration can be usedto reverse metabolic dysfunction in two models: 1) 1-year-old wild-type mice fed normal chow that wereshown to have metabolic dysfunction (Fig. 4) and 2)3.5-month-old wild-type mice fed HFD for 6 weeks be-fore the administration of 60 mg/kg FG-4497 on days 1,3, and 5 of each week was begun (7). This FG-4497 dosestabilizes Hif-1a and Hif-2a in mouse kidney and liverand increases serum erythropoietin concentration aboutsixfold (7).

After a 1-week adjustment period, FG-4497 adminis-tration to 1-year-old mice fed normal chow reduced the

Figure 2—Hif-p4h-2gt/gt mice have decreased serum cholesterollevels and are protected against hepatic steatosis. A: Serum totalcholesterol, HDL cholesterol, and LDL + VLDL cholesterol levels;HDL/LDL + VLDL cholesterol ratios; and triglyceride (TG) levels of7–11-month-old male wild-type (wt) and Hif-p4h-2gt/gt (gt/gt) miceafter fasting for 2 h (n = 19 for wt and n = 11 for gt/gt for totalcholesterol, HDL cholesterol, and TG values and n = 14 for wtand n = 5 for gt/gt for LDL + VLDL and HDL/LDL + VLDL cholesterolvalues). Data are mean 6 SEM. B: H-E–stained liver sections of 1-year-old male mice (n = 7 for both groups). Scoring of steatosis isshown. Scale bar = 200 mm. *P < 0.05, **P = 0.005, ***P = 0.001.s, serum.

3328 HIF-P4H-2 in Obesity and Metabolism Diabetes Volume 63, October 2014

weight of these mice during the subsequent 5 weeks by;1.3 g, whereas the vehicle-treated mice gained ;0.6 g(Fig. 7A). The adipocytes were smaller and the number ofmacrophage aggregates lower in the WAT of the FG-

4497–treated than those of the vehicle-treated mice(Fig. 7B and C). The serum total cholesterol level andthe HDL and LDL + VLDL cholesterol levels of the FG-4497–treated mice were significantly decreased, whereas

Figure 3—Hif-p4h-2gt/gt (gt/gt) mice have improved glucose tolerance, insulin sensitivity, and increased deoxyglucose uptake into skeletalmuscle. A: GTTs of 1-year-old and 4–5-month-old female wild-type (wt) and gt/gt mice. The 0-min value was determined after fasting for12 h (n = 9–17 per group). B: Serum insulin levels and HOMA-IR scores determined from the GTT of the 1-year-old mice in A. C: ITT of 3–4-month-old female mice. The 0-min value was determined after fasting for 6 h (n = 5–6 per group). Data are relative to glucose values at 0 min. D:Deoxyglucose uptake test. Eight sibling pairs of wt and gt/gt mice were fasted for 12 h. 14C-deoxyglucose was then injected intraperitoneally, themice were killed 60 min later, and their tissues were homogenized and analyzed for radioactivity. Disintegrations per min per milligram for eachtissue were compared between the wt and gt/gt members of each sibling pair. Data are mean6 SEM. *P< 0.05, **P< 0.01. b, blood; s, serum.

Figure 4—One-year-old wild-type (wt) but not Hif-p4h-2gt/gt (gt/gt) mice showed metabolic dysfunction relative to corresponding youngermice. A: Cross-sectional area of adipocytes. B: Number of macrophage aggregates. C: Fasting (12-h) blood glucose values. D: Fasting(12-h) serum insulin values. E: HOMA-IR scores in 4–5-month-old (young [y]) and 1-year-old (old [o]) wt and gt/gt mice (n = 5–14 per group).Data are mean 6 SEM. *P < 0.05, **P < 0.01, ***P < 0.001. b, blood; s, serum.

diabetes.diabetesjournals.org Rahtu-Korpela and Associates 3329

their HDL/LDL + VLDL ratio was increased (Fig. 7D). Thefasting blood glucose levels of the FG-4497–treated micewere likewise decreased, whereas the decreases in theserum insulin level by ;25% and HOMA-IR score by

;75% (Fig. 7E) were not statistically significant (P =0.11 in both cases).

In the other model, 2-month-old mice were fed normalchow or HFD for 6 weeks, after which the mice fed

Figure 5—Hif-p4h-2gt/gt (gt/gt) mice have altered expression of genes of glucose and lipid metabolism and reduced acetyl-CoA levels andshow reduced lipogenesis. A and B: Quantitative PCR analyses of the mRNA levels of HIF target and lipid metabolism genes in the skeletalmuscle and WAT (A) and liver (B) of gt/gt mice relative to the wild type (wt). Acca, acetyl-CoA carboxylase a; ApoB, apolipoprotein B; Fas,fatty acid synthase; Hk1, hexokinase-1; Irs1 and Irs2, insulin receptor substrates 1 and 2; Ldha, lactate dehydrogenase a; Ldlr, LDLreceptor; Lipe, hormone-sensitive lipase; Pdk1 and Pdk4, pyruvate dehydrogenase kinases 1 and 4; Pfkl, phosphofructokinase; Pgk1,phosphoglycerate kinase-1; Pnpla2, patatin-like phospholipase domain-containing protein 2; Pparg and Ppara, peroxisome proliferator–activated receptors g and a; Srebp1c, sterol regulatory element–binding protein 1c. The expression of each gene was studied relative tob-actin (n = 6–9 per group). C: Level of acetyl-CoA in the skeletal muscle, WAT, and liver. D: Incorporation of radioactivity from [14C]acetateinto extractable lipids as an indication of lipogenesis in the WAT and liver of 4–5-month-old male wt and gt/gt mice, expressed per milligramtissue wet weight (n = 4–7 per group in C and n = 9–10 per group in D). Data are mean 6 SEM. *P < 0.05, **P < 0.01, ***P < 0.001.

Figure 6—Hif-p4h-2gt/gt (gt/gt) mice are protected against HFD-induced metabolic changes and steatosis. A: Weight of 6-month-oldfemale wild-type (wt) and gt/gt mice before and after administration of HFD for 6 weeks (n = 7–10 per group). B: Cross-sectional areaof WAT adipocytes. C: Number of macrophage aggregates (aggregates/field) in gonadal WAT. D: GTT after the 6-week HFD administration.The 0-min value was determined after fasting for 12 h. E: H-E–stained liver sections of these HFD-fed mice. Scoring of steatosis is shown.Scale bar = 200 mm. Data are mean 6 SEM. *P < 0.05, **P < 0.005. b, blood.

3330 HIF-P4H-2 in Obesity and Metabolism Diabetes Volume 63, October 2014

normal chow were given vehicle, and those fed HFD weregiven either HFD and vehicle or HFD and FG-4497 for4 weeks. During the initial 6-week period, the HFD-fedmice gained more weight than those fed normal chow(Fig. 7F). The FG-4497 treatment decreased the weightof the HFD-fed mice by ;0.6 g, whereas their vehicle-treated controls gained ;3.0 g (FG-4497 vs. vehicle mice

P = 0.02) (Fig. 7F). The WAT weight of the FG-4497–treated HFD mice was lower than that of their controls(Fig. 7G). The glucose tolerance of the FG-4497–treatedHFD mice was better than that of their vehicle-treatedcontrols (Fig. 7H), and their fasting serum insulin levelsand HOMA-IR scores were significantly decreased(Fig. 7I).

Figure 7—Pharmacological Hif-p4h inhibition reverses metabolic dysfunction in both aged mice and mice fed HFD. A–E: One-year-oldmale wild-type mice fed normal chow (NC) were given vehicle or 60 mg/kg of FG-4497 on days 1, 3, and 5 of each week for 6 weeks (n = 6–7 per group). A: Weight gain of the mice at 6 weeks relative to their weights at 1 week (adjustment period). B: Cross-sectional area of WATadipocytes. Scale bar = 100 mm. C: Number of WAT macrophage aggregates (*) (aggregates/field). Scale bar = 100 mm. D: Serum totalcholesterol, HDL cholesterol, and LDL + VLDL cholesterol levels and HDL/LDL + VLDL cholesterol ratios measured after fasting for 2 h. E:Blood glucose and serum insulin levels and HOMA-IR scores determined from the samples in D. F–I: Two-month-old male wild-type mice(n = 8–10 per group) were fed NC or HFD (42% kcal fat) for 6 weeks, after which the mice fed NC were given vehicle and those fed HFDeither HFD and vehicle or HFD and FG-4497 for 4 weeks as in A–E. F: Weight of the mice fed NC or HFD before and after the administrationof vehicle or FG-4497. G: Weight of gonadal WAT of HFD-fed mice after the 4-week vehicle or FG-4497 treatment. H: GTT of the HFD-fedmice after the 4-week vehicle or FG-4497 administration. The 0-min value was determined after fasting for 12 h. I: Serum insulin levels andHOMA-IR scores determined from the GTT in H. Data are mean 6 SEM. *P < 0.05, **P < 0.01, ***P < 0.001. b, blood; s, serum; Veh,vehicle.

diabetes.diabetesjournals.org Rahtu-Korpela and Associates 3331

DISCUSSION

The data indicate that Hif-p4h-2gt/gt mice fed either a nor-mal chow or an HFD have less adipose tissue, smalleradipocytes, a decreased number of adipose tissue macro-phage aggregates, and lower serum cholesterol levels thantheir littermates. They are also protected against hepaticsteatosis and show increased glucose tolerance and insulinsensitivity.

The Hif-p4h-2gt/gt mice had higher levels of glucosetransporters and glycolysis enzymes in their skeletal mus-cle and WAT, with similar higher levels previously foundin their hearts (11). Uptake of deoxyglucose in skeletalmuscle was also increased. Furthermore, Pdk1 expressionwas increased in the Hif-p4h-2gt/gt skeletal muscle andWAT, as has also been found in the Hif-p4h-2gt/gt heart(11). A higher Pdk1 level may further increase glycolysisby inhibiting the entry of pyruvate into the citric acidcycle (22). Thus, it appears that glycolysis is increasedin several Hif-p4h-2gt/gt tissues, contributing to the overallimproved glucose tolerance (Supplementary Fig. 7). Thesechanges agree with the established consequences of thestabilization of HIF-1a (22). HIF-P4H-1 and -3 have beenreported to also have enzyme-specific substrates otherthan HIF-1a and HIF-2a (1–3), and thus, changes inthe levels of those two enzymes may also influence HIF-independent pathways. Such substrates have so far notbeen identified for HIF-P4H-2; therefore, it seems likelythat most, if not all, of the metabolic changes found in theHif-p4h-2gt/gt mice were mediated by Hif-a.

Obesity is associated with a chronic low-grade in-flammation that predisposes to insulin resistance. Adi-pose tissue macrophages are believed to play a key role inobesity-induced insulin resistance (20,21). They infiltrateobese adipose tissue and along with the hypertrophiedadipocytes, release cytokines and adipokines that contrib-ute to the proinflammatory response (20). Macrophage-derived proinflammatory factors block insulin action inadipocytes by downregulating the expression of the insulin-regulated GLUT4 and impairing insulin-stimulatedGLUT4 transport to the plasma membrane (20). Becausewe found decreased size of adipocytes, a reduced numberof adipose tissue macrophages, and increased Glut4 ex-pression in the Hif-p4h-2gt/gt mice, it seems likely thatthese changes contribute to increased insulin sensitivityin these mice (Supplementary Fig. 7).

Increased expression of the Hif-2a target Irs2 in theliver of mice with acute hepatic Hif-p4h-3 deletion wasaccompanied by a decreased Srebp1c expression (26).The present results indicating weak stabilization of Hif-2a, increased expression of Irs2, and decreased expression ofSrebp1c and its targets Acca and Fas in the Hif-p4h-2gt/gt

liver agree with those data. These changes are probablyresponsible for the decreased fatty acid synthesis and denovo lipogenesis found in the Hif-p4h-2gt/gt liver and WAT.The lack of acetyl-CoA in Hif-p4h-2gt/gt tissues, which ispresumably due to pyruvate dehydrogenase inhibition, is

likely to contribute to the decreased lipogenesis (Supple-mentary Fig. 7).

Extensive liver-specific stabilization of Hif-2a leads tohepatic steatosis (26,28). However, the Hif-p4h-2gt/gt micein the present study showed no steatosis but were insteadprotected against it. Liver-specific stabilization of Hif-2aby acute Hif-p4h-3 deletion likewise did not lead to he-patic steatosis, suggesting that low-level hepatic Hif-2astabilization, as found in the present Hif-p4h-2gt/gt mice,has beneficial effects, whereas extensive hepatic Hif-2astabilization leads to steatosis (26,28).

Liver-specific stabilization of Hif-1a and Hif-2aappears to have no effect on hepatic cholesterol synthesisor intestinal cholesterol absorption, but extensive liver-specific Hif-2a stabilization increases hepatic and serumcholesterol levels (28,29) as a result of decreased choles-terol oxidation to bile acids (29). However, the Hif-p4h-2gt/gt mice with low-level hepatic stabilization of Hif-2ahad decreased serum cholesterol levels. The decreasedamount of acetyl-CoA is likely to contribute to the lowserum cholesterol level in the Hif-p4h-2gt/gt mice (Supple-mentary Fig. 7), but other mechanisms may also beinvolved.

Mice with adipocyte-specific Hif-p4h-2 deletion alsohave less WAT, smaller adipocytes, a lower number ofadipose tissue macrophages, and improved glucose toler-ance (25); however, such changes were seen only in thosefed an HFD, and no changes were reported in serumcholesterol levels, suggesting that Hif-p4h-2 deficienciesin several tissues play an important role in metabolicchanges in Hif-p4h-2gt/gt mice. Acute hepatic Hif-p4h-3 de-letion has also been reported to improve glucose toleranceand insulin sensitivity, but no data were available on itseffects on weight gain or serum cholesterol levels (26).

Inhibition of Hif-1a by disruption of its gene in adi-pocytes (30) or administration of its inhibitor (31) orantisense oligonucleotides (32) attenuates the consequencesof an HFD in mice. Currently, no explanation is availablefor the discrepancy between those data and the beneficialeffects of Hif-a stabilization by Hif-p4h-2 deficiency, butthe additional stabilization of Hif-2a has been suggestedto possibly play an important role (25).

Administration of FG-4497 to mice in two models ofmetabolic dysfunction led to changes very similar to thoseseen in the Hif-p4h-2gt/gt mice, indicating that HIF-P4H-2inhibition may not only protect against the developmentof obesity and metabolic dysfunction but also reversethem. FG-4497 inhibits all three HIF-P4Hs, but in viewof the changes found in the Hif-p4h-2gt/gt mice, it wouldseem possible to obtain a similar effect with a compoundthat specifically inhibits HIF-P4H-2. Of note, administra-tion of another pan-HIF-P4H-inhibitor, FG-4592, cur-rently in clinical trials for treatment of anemia ofchronic kidney disease, also lowers serum cholesterol lev-els and increases the HDL/LDL ratio (33,34), thus sup-porting the view that HIF-P4H-2 inhibition may indeed

3332 HIF-P4H-2 in Obesity and Metabolism Diabetes Volume 63, October 2014

be a useful strategy for the treatment of obesity and itsconsequences.

Acknowledgments. The authors thank T. Aatsinki, R. Juntunen,E. Lehtimäki, S. Rannikko, and M. Siurua for excellent technical assistance.Funding. This study was supported by Academy of Finland grants 200471and 202469 (to J.M.); Center of Excellence 2012–2017 grant 251314 (to J.M.); theS. Jusélius Foundation (to J.M. and P.K.); Academy of Finland grants 120156,140765, 218129, and 266719 (to P.K.); and the Emil Aaltonen Foundation (to P.K.).Duality of Interest. G.W. is a senior cell biology director at FibroGen, Inc.K.I.K. is a scientific founder and consultant of FibroGen, Inc., which develops HIF-P4H inhibitors as potential therapeutics. K.I.K. and J.M. own equity in thiscompany, and the company has sponsored research in the laboratory headedby K.I.K. and currently supports research headed by J.M. No other potentialconflicts of interest relevant to this article were reported.Author Contributions. L.R.-K., S.K., and R.S. contributed to the researchand data analysis. S.H. contributed expertise in serum lipid analyses. R.B.S. andE.L. contributed to the MRI analyses. K.A.M. and K.-H.H. contributed to themetabolic home cage experiments and analysis. G.W. provided the FG-4497and made useful suggestions. K.I.K. contributed to generating the Hif-p4h-2gt/gt

mouse line and to the study design, data analysis, and writing of the manu-script. J.M. contributed to generating the Hif-p4h-2 gt /gt mouse line and to thediscussions. P.K. contributed to generating the Hif-p4h-2 gt /gt mouse line andstudy supervision and to the study design, data analysis, and writing of themanuscript. P.K. is the guarantor of this work and, as such, had full access toall the data in the study and takes responsibility for the integrity of the data andthe accuracy of the data analysis.

References1. Kaelin WG Jr, Ratcliffe PJ. Oxygen sensing by metazoans: the central role ofthe HIF hydroxylase pathway. Mol Cell 2008;30:393–4022. Semenza GL. Hypoxia-inducible factors in physiology and medicine. Cell2012;148:399–4083. Myllyharju J, Koivunen P. Hypoxia-inducible factor prolyl 4-hydroxylases:common and specific roles. Biol Chem 2013;394:435–4484. Epstein AC, Gleadle JM, McNeill LA, et al. C. elegans EGL-9 and mammalianhomologs define a family of dioxygenases that regulate HIF by prolyl hydroxyl-ation. Cell 2001;107:43–545. Bruick RK, McKnight SL. A conserved family of prolyl-4-hydroxylases thatmodify HIF. Science 2001;294:1337–13406. Ivan M, Haberberger T, Gervasi DC, et al. Biochemical purification andpharmacological inhibition of a mammalian prolyl hydroxylase acting on hypoxia-inducible factor. Proc Natl Acad Sci U S A 2002;99:13459–134647. Laitala A, Aro E, Walkinshaw G, et al. Transmembrane prolyl 4-hydroxylaseis a fourth prolyl 4-hydroxylase regulating EPO production and erythropoiesis.Blood 2012;120:3336–33448. Takeda K, Ho V, Takeda H, Duan LJ, Nagy A, Fong GH. Placental but not heartdefects are associated with elevated hypoxia-inducible factor alpha levels in micelacking prolyl hydroxylase domain protein 2. Mol Cell Biol 2006;22:8336–83469. Takeda K, CowanA, FongGH. Essential role for prolyl hydroxylase domainprotein2 in oxygen homeostasis of the adult vascular system. Circulation 2007;116:774–78110. Minamishima YA, Moslehi J, Bardeesy N, Cullen D, Bronson RT, Kaelin WGJr. Somatic inactivation of the PHD2 prolyl hydroxylase causes polycythemia andcongestive heart failure. Blood 2008;111:3236–324411. Hyvärinen J, Hassinen IE, Sormunen R, et al. Hearts of hypoxia-induciblefactor prolyl 4-hydroxylase-2 hypomorphic mice show protection against acuteischemia-reperfusion injury. J Biol Chem 2010;285:13646–1365712. Kerkelä R, Karsikas S, Szabo Z, et al. Activation of hypoxia response inendothelial cells contributes to ischemic cardioprotection. Mol Cell Biol 2013;33:3321–3329

13. Yun Z, Maecker HL, Johnson RS, Giaccia AJ. Inhibition of PPAR gamma 2gene expression by the HIF-1-regulated gene DEC1/Stra13: a mechanism forregulation of adipogenesis by hypoxia. Dev Cell 2002;2:331–34114. Quintero P, Milagro FI, Campión J, Martínez JA. Impact of oxygen availabilityon body weight management. Med Hypotheses 2010;74:901–90715. van den Borst B, Schols AM, de Theije C, et al. Characterization of theinflammatory and metabolic profile of adipose tissue in a mouse model of chronichypoxia. J Appl Physiol (1985) 2013;114:1619–162816. Hsieh MM, Linde NS, Wynter A, et al. HIF prolyl hydroxylase inhibition re-sults in endogenous erythropoietin induction, erythrocytosis, and modest fetalhemoglobin expression in rhesus macaques. Blood 2007;110:2140–214717. Bernhardt WM, Gottmann U, Doyon F, et al. Donor treatment witha PHD-inhibitor activating HIFs prevents graft injury and prolongs survival in anallogenic kidney transplant model. Proc Natl Acad Sci U S A 2009;106:21276–2128118. Friedewald WT, Levy RI, Fredrickson DS. Estimation of the concentration oflow-density lipoprotein cholesterol in plasma, without use of the preparativeultracentrifuge. Clin Chem 1972;18:499–50219. Soh J, Iqbal J, Queiroz J, Fernandez-Hernando C, Hussain MM. MicroRNA-30c reduces hyperlipidemia and atherosclerosis in mice by decreasing lipidsynthesis and lipoprotein secretion. Nat Med 2013;19:892–90020. Harford KA, Reynolds CM, McGillicuddy FC, Roche HM. Fats, inflammationand insulin resistance: insights to the role of macrophage and T-cell accumu-lation in adipose tissue. Proc Nutr Soc 2011;70:408–41721. Fuentes E, Fuentes F, Vilahur G, Badimon L, Palomo I. Mechanisms ofchronic state of inflammation as mediators that link obese adipose tissue andmetabolic syndrome. Mediators Inflamm 2013;2013:13658422. Semenza GL. Regulation of oxygen homeostasis by hypoxia-inducible factor 1.Physiology (Bethesda) 2009;24:97–10623. Im SS, Kang SY, Kim SY, et al. Glucose-stimulated upregulation of GLUT2gene is mediated by sterol response element-binding protein-1c in the hep-atocytes. Diabetes 2005;54:1684–169124. Abel ED, Peroni O, Kim JK, et al. Adipose-selective targeting of the GLUT4gene impairs insulin action in muscle and liver. Nature 2001;409:729–73325. Matsuura H, Ichiki T, Inoue E, et al. Prolyl hydroxylase domain protein 2plays a critical role in diet-induced obesity and glucose intolerance. Circulation2013;127:2078–208726. Taniguchi CM, Finger EC, Krieg AJ, et al. Cross-talk between hypoxia andinsulin signaling through Phd3 regulates hepatic glucose and lipid metabolismand ameliorates diabetes. Nat Med 2013;19:1325–133027. Taniguchi CM, Ueki K, Kahn R. Complementary roles of IRS-1 and IRS-2 inthe hepatic regulation of metabolism. J Clin Invest 2005;115:718–72728. Rankin EB, Rha J, Selak MA, et al. Hypoxia-inducible factor 2 regulateshepatic lipid metabolism. Mol Cell Biol 2009;29:4527–453829. Ramakrishnan SK, Taylor M, Qu A, et al. Loss of von Hippel-Lindau protein(VHL) increases systemic cholesterol level through targeting hypoxia-inducible factor2a and regulation of bile acid homeostasis. Mol Cell Biol 2014;34:1208–122030. Jiang C, Qu A, Matsubara T, et al. Disruption of hypoxia-inducible factor 1 inadipocytes improves insulin sensitivity and decreases adiposity in high-fat diet-fed mice. Diabetes 2011;60:2484–249531. Sun K, Halberg N, Khan M, Magalang UJ, Scherer PE. Selective inhibition ofhypoxia-inducible factor 1a ameliorates adipose tissue dysfunction. Mol Cell Biol2013;33:904–91732. Shin MK, Drager LF, Yao Q, et al. Metabolic consequences of high-fat dietare attenuated by suppression of HIF-1a. PLoS One 2012;7:e4656233. Bakris GL, Yu K-P, Leong R, Shi W, Lee T, Saikali K, Henry E, Neff TB.Effects of a novel anemia treatment, FG-4592 - an oral hypoxia-inducible prolylhydroxylase inhibitor (HIF-PHI) on blood pressure and cholesterol in patients withchronic kidney disease. J Clin Hypertens 2012;14:48934. Myllyharju J. Prolyl 4-hydroxylases, master regulators of the hypoxia re-sponse. Acta Physiol (Oxf) 2013;208:148–165

diabetes.diabetesjournals.org Rahtu-Korpela and Associates 3333