Contribution of HIF-P4H isoenzyme inhibition to metabolism indicates major beneficial effects being conveyed by HIF-P4H-2 antagonism

Hypoxia-inducible factor (HIF) prolyl 4-hydroxylases (HIF-P4Hs 1–3) are druggable targets in renal anemia, where pan-HIF-P4H inhibitors induce an erythropoietic response. Preclinical data suggest that HIF-P4Hs could also be therapeutic targets for treating metabolic dysfunction, although the contributions of HIF-P4H isoenzymes in various tissues to the metabolic phenotype are inadequately understood. Here, we used mouse lines that were gene-deficient for HIF-P4Hs 1 to 3 and two preclinical pan-HIF-P4H inhibitors to study the contributions of these isoenzymes to the anthropometric and metabolic outcome and HIF response. We show both inhibitors induced a HIF response in wildtype white adipose tissue (WAT), liver, and skeletal muscle and alleviated metabolic dysfunction during a 6-week treatment period, but they did not alter healthy metabolism. Our data indicate that HIF-P4H-1 contributed especially to skeletal muscle and WAT metabolism and that its loss lowered body weight and serum cholesterol levels upon aging. In addition, we found HIF-P4H-3 had effects on the liver and WAT and its loss increased body weight, adiposity, liver weight and triglyceride levels, WAT inflammation, and cholesterol levels and resulted in hyperglycemia and insulin resistance, especially during aging. Finally, we demonstrate HIF-P4H-2 affected all tissues studied; its inhibition lowered body and liver weight and serum cholesterol levels and improved glucose tolerance. We found very few HIF target metabolic mRNAs were regulated by the inhibition of three isoenzymes, thus suggesting a potential for selective therapeutic tractability. Altogether, these data provide specifications for the future development of HIF-P4H inhibitors for the treatment of metabolic diseases.

Hypoxia-inducible factor (HIF) prolyl 4-hydroxylases (HIF-P4Hs 1-3) are druggable targets in renal anemia, where pan-HIF-P4H inhibitors induce an erythropoietic response. Preclinical data suggest that HIF-P4Hs could also be therapeutic targets for treating metabolic dysfunction, although the contributions of HIF-P4H isoenzymes in various tissues to the metabolic phenotype are inadequately understood. Here, we used mouse lines that were gene-deficient for HIF-P4Hs 1 to 3 and two preclinical pan-HIF-P4H inhibitors to study the contributions of these isoenzymes to the anthropometric and metabolic outcome and HIF response. We show both inhibitors induced a HIF response in wildtype white adipose tissue (WAT), liver, and skeletal muscle and alleviated metabolic dysfunction during a 6-week treatment period, but they did not alter healthy metabolism. Our data indicate that HIF-P4H-1 contributed especially to skeletal muscle and WAT metabolism and that its loss lowered body weight and serum cholesterol levels upon aging. In addition, we found HIF-P4H-3 had effects on the liver and WAT and its loss increased body weight, adiposity, liver weight and triglyceride levels, WAT inflammation, and cholesterol levels and resulted in hyperglycemia and insulin resistance, especially during aging. Finally, we demonstrate HIF-P4H-2 affected all tissues studied; its inhibition lowered body and liver weight and serum cholesterol levels and improved glucose tolerance. We found very few HIF target metabolic mRNAs were regulated by the inhibition of three isoenzymes, thus suggesting a potential for selective therapeutic tractability. Altogether, these data provide specifications for the future development of HIF-P4H inhibitors for the treatment of metabolic diseases.
The transcriptional hypoxia response is chiefly regulated by the hypoxia-inducible factor (HIF) system, in which three HIF prolyl 4-hydroxylase isoenzymes, HIF-P4Hs 1 to 3, (also known as PHDs 1-3 or EglN2, 1 and 3, respectively) provide the oxygen (O 2 )-sensing component for the system (1,2). HIF-P4Hs are iron and 2-oxoglutarate-dependent dioxygenases that hydroxylate one or two proline residues in the HIFα subunit (3). The resulting 4-hydroxyproline residue acts as an earmark for pVHL binding and proteasomal degradation of HIFα (3). The hydroxylation is largely dependent on the cellular oxygenation status, as the HIF-P4Hs have a very low affinity (high K m value) for O 2 , making them excellent sensors for hypoxia (3). The stabilized HIFα (one of the three isoforms, HIF1α-3α) forms a transcriptionally active dimer with HIFβ, which binds to the regulatory region of the HIF target gene and upregulates its transcription (1,2). Altogether several hundred HIF target genes have been identified, with those for erythropoietin (EPO) and vascular endothelial growth factor being among the most studied, indicating that the HIF response aims to restore cellular oxygenation and O 2 delivery by inducing erythropoiesis and angiogenesis (1,2,4). However, an even more central process to target in order to survive under hypoxia is energy metabolism, since mitochondrial oxidative phosphorylation (OXPHOS) is the most O 2consuming process in the cell (5). Thus the HIF target genes also upregulate genes that increase glucose intake and the nonoxygen-demanding glycolytic metabolism and downregulate OXPHOS, for instance (5,6).
The HIF-P4Hs have been shown to be druggable targets, since several small-molecule inhibitors that typically compete with the binding of 2-oxoglutarate have been accepted since 2018 for the treatment of anemia in chronic kidney disease (CKD), first in Asia and very recently also in Europe (7). These antagonists that inhibit all three HIF-P4Hs upregulate endogenous renal EPO production (also hepatic EPO production in kidney-deficient patients) and support efficient iron metabolism, as many HIF target genes support iron intake and transfer (3,8). Via EPO the HIF-P4H inhibitors also indirectly downregulate hepcidin (HAMP) levels, which are typically high in CKD and other inflammatory conditions, and inhibit cellular iron recycling (9). Interestingly, data from clinical trials with patients having CKD have shown that the HIF-P4H inhibitors also alter serum lipid values (10). Specifically, they reduce total cholesterol, non-high-density lipoprotein (HDL) cholesterol, and triglyceride levels (10). As these levels are elevated in dyslipidemia, a condition found in metabolic syndrome, these results could be desirable in a patient cohort suffering from metabolic dysfunction (11). The clinical data are supported by preclinical studies in which beneficial effects on lipid and glucose metabolism and inflammation have been reported with preclinical HIF-P4H inhibitors in several metabolic disease models (12)(13)(14)(15)(16), although the individual contributions of the HIF-P4H isoenzymes to the metabolic phenotype and the various tissues are not well understood. Beneficial effects of genetic inhibition of HIF-P4H-2 and HIF-P4H-1 on metabolism have been reported, but the data on HIF-P4H-3 have been contradictory (12,(16)(17)(18)(19)(20). We therefore used mouse lines gene modified for HIF-P4Hs 1 to 3 and two preclinical pan-HIF-P4H inhibitors, FG-4497 and FG-4539, to study the contributions of inhibition of the isoenzymes to the metabolic outcome in the key metabolic tissues. The resulting data will be of assistance in the future development of therapeutics to treat obesity, metabolic syndrome, and fatty liver disease by exploiting the HIF pathway as a novel treatment strategy.

Results
HIF-P4H-1 loss has beneficial effects on metabolism upon aging, while HIF-P4H-3 deficiency has aggravating effects We have shown earlier that adiposity, WAT inflammation, fasting blood glucose levels, fasting serum insulin levels, and homeostatic model assessment for insulin resistance (HOMA-IR) scores all increase significantly in WT mice fed normal chow ad libitum and housed in a standard cage at 1 year of age as compared with young animals (4-5 months of age), whereas mice that are hypomorphic for HIF-P4H-2 (Hif-p4h-2 gt/gt ) are protected from these increases (16). We therefore used aging up to 1 year as a challenge for studying potential metabolic dysfunction in Hif-p4h-1 and Hif-p4h-3 knockout (KO) male mice (Fig. S1).
There were significant differences in body weight between the aged Hif-p4h-1 KO and Hif-p4h-3 KO mice at sacrifice and their WT littermates, the former having lower and the latter higher body weight (Fig. 1A). The Hif-p4h-3 KO mice also had significantly more gonadal WAT and heavier livers than their WT littermates, whereas the tendency for lower values in the Hif-p4h-1 KO mice relative to WT did not reach significance (Fig. 1, B and C). No difference in liver glycogen levels were detected between the Hif-p4h-1 KO or Hif-p4h-3 KO mice and their WT littermates (Fig. 1D), but the Hif-p4h-3 KO mice had higher liver triglyceride levels than their WT counterparts (Fig. 1E). There were no significant differences in blood hemoglobin (Hb) levels between the aged Hif-p4h-1 KO or Hif-p4h-3 KO mice and their WT littermates (Fig. 1F), but the aged Hif-p4h-1 KO mice had lower serum total cholesterol levels than their WT littermates and the Hif-p4h-3 KO mice had higher levels (Fig. 1G). No differences in serum triglyceride levels were detected between the Hif-p4h-1 or 3 KO mice and their littermates (Fig. 1H). The histologically analyzed adipocyte size did not differ between the aged Hif-p4h-1 KO or Hif-p4h-3 KO mice and WT, respectively, but there was an almost significant reduction in the number of macrophage aggregates in the Hif-p4h-1 KO WAT and a significant increase in the Hif-p4h-3 KO WAT (Figs. 1,I-L and S2).
We next analyzed HIF1α/2α protein levels and the expression levels of the key metabolic HIF target and some other selected mRNAs in the WAT, liver, and skeletal muscle of the aged Hif-p4h-1 KO and Hif-p4h-3 KO mice by comparison with their WT littermates. Stabilization of HIF1α in WAT and skeletal muscle was detected by Western blotting in Hif-p4h-1 KO mice (Fig. S3). The Hif-p4h-1 KO tissues showed upregulation of glucose transporter 1 (Glut1, Slc2a1) mRNA in the liver, phosphofructokinase l (Pfkl)-the central regulator of glycolysis-in all the tissues studied, lactate dehydrogenase (Ldha) in WAT and liver, pyruvate dehydrogenase kinase 1 (Pdk1) in skeletal muscle, peroxisome proliferator-activated receptor α (Ppara) and adiponectin (Adipoq) in WAT, and insulin receptor substrate 2 (Irs2) in the liver (Fig. 2, A-C). Interestingly, there was also upregulation of the HIF target Hif-p4h-2 and Hif-p4h-3 mRNAs in WAT and skeletal muscle of the Hif-p4h-1 KO mice as compared with WT (Fig. 2, A and C), which could suggest a compensatory attempt to overcome the Hif-p4h-1 loss and indicate an important role for isoenzyme 1 in WAT and skeletal muscle.
No stabilization of HIF1α or HIF2α was detected (data not shown) and none of the glycolytic HIF target mRNAs were significantly upregulated in any of the Hif-p4h-3 KO tissues studied (Fig. 2, D-F), but Adipoq mRNA was significantly downregulated in WAT and mRNAs for Glut1 (Slc2a1), Glut2 (Slc2a2), stearoyl-CoA desaturase-1 (Scd1), fatty acid synthase (Fasn), lipin-1 (Lpin1), and Hif-ph4-1 were downregulated in the livers of the Hif-p4h-3 KO mice relative to their WT littermates (Fig. 2, D and E). Downregulation of the central fatty acid synthesis mRNAs would suggest that the higher liver weight of the aged Hif-p4h-3 KO mice did not stem from higher lipogenic activity.
In order to evaluate the contribution of the mRNA levels to the observed differences in the metabolic outcome, association analyses were carried out. These indicated a negative association between the WAT glycolytic HIF target mRNAs, Ppara, Adipoq, Hif-p4h-2 and Hif-p4h-3 mRNA, and body weight, WAT and liver weight, liver triglyceride and serum total cholesterol levels, and the WAT macrophage count in the Hif-p4h-1 strain (Figs.S4A-S9A), suggesting that the increased levels observed in these mRNAs in the Hif-p4h-1 KO WAT contributed to the favorable changes in several of these metabolic parameters (Fig. 1). Similarly, a negative association between WAT Adipoq mRNA levels and body weight, liver weight, and the WAT macrophage count was seen in the Hif-p4h-3 strain, suggesting that the observed downregulation of its mRNA levels contributed to the detrimental outcome reflected in these parameters in the Hif-p4h-3 KO mice (Figs. S4B, S6B and S9B). The negative associations between hepatic Glut1, Glut2, and Hif-p4h-1 mRNA levels and body weight and WAT weight, between hepatic Glut1 mRNA and total cholesterol levels, and between hepatic Glut2 mRNA levels and the WAT macrophage count in the Hif-p4h-3 strain (Figs. S4B, S5B, S7B and S9B) suggest a link between lower levels of these mRNAs in the Hif-p4h-3 KO mice and their increased adiposity and WAT inflammation (Fig. 1).

HIF-P4Hs in metabolism
We next administered FG-4539 3 times a week for 6 weeks to 6-to 7-month-old WT C57BL/6N males comparing 30 mg/ kg and 60 mg/kg doses with the vehicle alone (Fig. S1). The mice were subjected to a glucose tolerance test (GTT) after 4 weeks of treatment and were sacrificed after 6 weeks of treatment. The mice receiving the inhibitor had a concentration-dependent increase in blood Hb levels at sacrifice, the higher dosage resulting in a 25% increase relative to the vehicle (Fig. 3A). There was an inhibitor dose-dependent decline in body weight, the weight change with the higher dosage being −7% (−2 g) in 6 weeks (Fig. 3, B and C). There were also reductions of about 25% and 15% in the WAT and liver weights of the FG-4539-treated mice, respectively, although the 60-mg/kg dosage did not reach statistical significance (Fig. 3, D and E). The >15% decline in serum total cholesterol levels reached significance even with the lower FG-4539 dosage, however (Fig. 3F). There was a decline in fasting blood glucose levels in the FG-4539-treated mice, but unlike the reductions in fasting serum insulin levels or the HOMA-IR score, neither this nor the decline in the area under the curve (AUC) for GTT reached statistical significance (Fig. 3, G-K).
Interestingly, when FG-4539 was administered at a dose of 60 mg/kg to a WT male cohort with a different background (C57BL/6N/Sv129) and having before treatment a 35% lower body weight, 70% less WAT, and 35% lighter livers than the C57BL6/N cohort, no significant differences in the anthropometric or metabolic parameters were detected relative to the vehicle despite the increase in Hb levels (Figs. 3 and S11). These data suggest that the pan-HIF-P4H inhibitors only reverse adverse metabolic outcomes, which is important regarding the safety of their use, e.g., for treating hyperglycemia.

Analysis of adiposity and serum lipid levels of the HIF-P4H-1-3 isoenzyme-deficient mouse lines treated with pan-HIF-P4H inhibitors indicated some differences between the isoenzymes
We then administered 60 mg/kg FG-4539 or vehicle thrice a week for 6 weeks to 6-to 7-month-old Hif-p4h-1 KO, Hif-p4h-2 gt/gt , and Hif-p4h-3 KO male mice and their WT male HIF-P4Hs in metabolism littermates (Fig. S1). A second cohort of the Hif-p4h-3 KO male mice was treated similarly with FG-4497 for comparison (Fig. S1). In order to distinguish the role of each isoenzyme in the regulation of metabolism, four comparisons were made: (1) between vehicle-treated WT and vehicle-treated gene-deficient mice, to determine the contribution of genetic inhibition (total loss in the KO mice and tissue-specific downregulation in the Hif-p4h-2 gt/gt mice); (2) between inhibitor-treated WT and inhibitor-treated gene-deficient mice, to determine the role of genetic inhibition in addition to pharmacological inhibition of all isoenzymes; (3) between vehicle-treated WT and inhibitor-treated WT mice, to show the effect of pharmacological inhibition of all isoenzymes; and (4) between vehicletreated gene-deficient mice and inhibitor-treated gene-deficient mice, which in the case of the KO mice would indicate the contribution of pharmacologic inhibition of the other isoenzymes but in the case of the Hif-p4h-2 gt/gt mice may also reflect the effect of further inhibition of HIF-P4H-2 by its pharmacological antagonist.
At sacrifice the Hif-p4h-2 gt/gt mice, which were in a different background from the others (C57BL6/N/Sv129) and were about 10 g lighter than the C57BL6/N mice, showed a genotype-mediated lower body weight that did not change with FG-4539 treatment (Fig. 4B). The inhibitor treatmentassociated weight loss seen in the C57BL6/N WT mice relative to those receiving the vehicle was not observed in the FG-4539-treated Hif-p4h-1 KO or FG-4497-treated Hif-p4h-3 KO mice (Fig. 4, A and D). There was also a trend for an increased WAT weight in the FG-4497-treated Hif-p4h-3 KO mice relative to WT (Fig. 4D), further supporting data from the aged Hif-p4h-3 KO WAT (Fig. 1B). The only significant difference in liver weight was a decrease in the FG-4539-treated Hif-p4h-2 gt/gt mice relative to WT, which probably stemmed from pharmacological inhibition of HIF-P4H-2 in addition to the hypomorphic genetic inhibition (40% in the liver (16)), since the genetic inhibition of either HIF-P4H-1 or HIF-P4H-3 did not lower the liver weight (Fig. 4). The genotype-mediated lower serum total cholesterol and HDL cholesterol levels in the Hif-p4h-2 gt/gt mice were lost with FG-4539 treatment, and indeed, the former was slightly but significantly higher in the inhibitor-treated Hif-p4h-2 gt/gt mice than in those receiving the vehicle (Fig. 4B). These data suggest that simultaneous inhibition of all three isoenzymes does not result in the highest reduction in serum cholesterol levels. Serum triglyceride levels were increased with FG-4539 treatment independent of genotype in the C57BL6/N/Sv129 background but not in C57BL6/N (Fig. 4). As expected, Hb levels in the WT mice were increased by both pan-HIF-P4H inhibitors (Fig. 4). Inhibition of HIF-P4H-2 appeared to mediate the largest increase in Hb levels, and inhibition of HIF-P4H-3 also contributed to this (Fig. 4).

Analysis of glucose tolerance and insulin sensitivity in the HIF-P4H-1-3 isoenzyme-deficient mouse lines treated with pan-HIF-P4H inhibitors indicated beneficial effects of HIF-P4H-2 inhibition but adverse effects of HIF-P4H-3 inhibition
When the mice were subjected to a GTT after 4 weeks of inhibitor treatment (Fig. S1) there were no differences in fasting glucose levels between the vehicle and FG-4539-treated Hif-p4h-1 KO and WT littermates (Fig. 5A), while the Hif-p4h-2 gt/gt mice had genotype-mediated lower fasting glucose levels A black asterisk denotes a statistical difference between genotypes in vehicle-treated mice, a green asterisk a statistical difference between genotypes in FG-4539 or FG4497-treated mice, a red hash a statistical difference between vehicle and FG-4539 or FG4497-treated WT mice and a blue hash a statistical difference between vehicle and FG-4539 or FG4497-treated Hif-p4h-1/3 KO or Hif-p4h-2 gt/gt mice. * or # p ≤ 0.05, ** or ## p <0.01, *** or ### p < 0.001, #### p < 0.0001. HDL, high-density lipoprotein; VEH, vehicle-treated; WAT, white adipose tissue.
in both the vehicle and FG-4539-treated groups (Fig. 5B). The Hif-p4h-3 KO mice had a genotype-mediated significant increase in fasting glucose levels, a trend for higher fasting insulin levels and significantly higher HOMA-IR levels in both the vehicle and FG-4497 treatments than did the WT mice (Fig. 5B). In agreement with the above data, the AUC of the GTT was genotype mediated and lower in the Hif-p4h-2 gt/gt mice and higher in the Hif-p4h-3 KO mice than in the WT littermates, while no difference was seen in the Hif-p4h-1 KO mice (Fig. 5). Altogether these data suggest that inhibition of HIF-P4H-1 does not contribute significantly to glucose intake and insulin sensitivity, whereas HIF-P4H-2 inhibition ameliorates these factors and HIF-P4H-3 inhibition exacerbates them.
The metabolic HIF target and other genes in HIF-P4H-1-3 isoenzyme-deficient adipose tissue, liver, and skeletal muscle are regulated by pan-HIF-P4H inhibitors Finally, we carried out large-scale analyses of the expression of mRNAs of metabolic HIF target genes and certain others in WAT, liver, and skeletal muscle of inhibitor-or vehicle-treated Hif-p4h-1/3 KO and Hif-p4h-2 gt/gt mice by comparison with WT (Fig. S1). It should be noted that the experimental setting varied, in that the final dose of FG-4539 was administered to the Hif-p4h-1 KO and Hif-p4h-3 KO mice 24 h before sacrifice, whereas the Hif-p4h-2 gt/gt mice received their final dose of FG-4539 and the Hif-p4h-3 KO mice that of FG-4497 6 h before sacrifice (Fig. S1). Depending on the half-life of the mRNAs this may have influenced the data.
The vehicle-treated mice showed genotype-mediated changes in very few mRNA levels in WAT, some of which differed from those affected in the WAT of the aged Hif-p4h-1 KO and Hif-p4h-3 KO (Figs. 6 and 2, A and D). Inhibitor treatment in addition to the genetic deficiency upregulated Ldha and Ppara mRNA in the Hif-p4h-2 gt/gt WAT compared with WT, and Glut1 (Slc2a1), hexokinase 1 (Hk1), Ldha, Pdk1, Ppara, Adipoq, Hif-p4h-1, and Hif-p4h-2 mRNAs in the FG-4497-treated Hif-p4h-3 KO WAT (Fig. 6). Where FG-4539 treatment upregulated many glucose metabolism HIF target genes in the WAT of the WT littermates of the Hif-p4h-2 gt/gt and Hif-p4h-3 KO mice, independent of the time when they received the final dose of the inhibitor, this treatment led to the opposite results in the WT littermates of the Hif-p4h-1 KO mice, in which Glut1 (Slc2a1), Pfkl, Ldha, Pdk1, and Hif-p4h-2 mRNA levels were significantly downregulated compared with the vehicle 24 h after the final dosage (Fig. 6A). This difference may suggest a rebound effect in these animals following the inhibitor treatment. Interpretation of the contribution of inhibition of each HIF-P4H isoenzyme to the mRNA levels was further complicated by the fact that the knockdown level of HIF-P4H-2 in the WAT of the Hif-p4h-2 gt/gt mice was 60%, agreeing with earlier data (16). Altogether, the data regarding WAT do suggest that all the HIF-P4H isoenzymes contributed A black asterisk denotes a statistical difference between genotypes in vehicle-treated mice, a green asterisk a statistical difference between genotypes in FG-4539 or FG4497-treated mice, a red hash a statistical difference between vehicle and FG-4539 or FG4497-treated WT mice, and a blue hash a statistical difference between vehicle and FG-4539 or FG4497-treated Hif-p4h-1/3 KO or Hif-p4h-2 gt/gt mice. * or #p ≤ 0.05, **p <0.01, ***p < 0.001, ****p < 0.0001. AUC, area under the curve; GTT, glucose tolerance test; HOMA-IR, homeostatic model assessment for insulin resistance; VEH, vehicle-treated.
to the expression of the metabolic HIF target genes. Surprisingly, and in contrast to the data on the aged Hif-p4h-3 KO WAT, several mRNAs such as Glut1 (Slc2a1), Hk1, Pdk1, Lep, and Adipoq were largely regulated by inhibition of HIF-P4H-3 alone, while HIF-P4H-2 inhibition also contributed to the upregulation of Ldha and Ppara mRNA and inhibition of all the isoenzymes to Pfkl mRNA upregulation (Fig. 6). As for the inflammatory chemokine ligand 2 (Ccl2) mRNA, this was lowered by inhibition of HIF-P4H-2 but increased by inhibition of HIF-P4H-3 or HIF-P4H-1, although none of the effects reached statistical significance (Fig. 6).

HIF1α stabilization associates with decreased oxygen consumption and ATP production and increased glycolysis
In order to gain further functional understanding on the contribution of inhibition of the individual HIF-P4Hs to metabolism we studied oxygen consumption and extracellular acidification rates of mouse embryonic fibroblasts (MEFs) generated from the Hif-p4h-1 and Hif-p4h-3 strains and WT MEFs treated with FG-4539. We have earlier reported that, in similar analyses, Hif-p4h-2 gt/gt MEFs, which have 80% knockdown of Hif-p4h-2 mRNA and normoxic stabilization of HIF1α, have about 40% reduced oxygen consumption and ATP production and about 30% increased glycolysis compared with WT (13). Our analyses showed significant differences in the basal and maximal oxygen consumption rate, ATP production, and proton leak between the strains; the FG-4539-treated WT MEFs had reduced rates by 35 to 50% compared with controls, the ATP production of the Hif-p4h-1 KO MEFs was down by about one-third, whereas no differences indicative for decreased OXPHOS between the Hif-p4h-3 KO and WT MEFs were detected (Fig. 9, A-C). In line, the spare respiratory capacity was slightly higher (131%) in the Hif-p4h-3 KO MEFs compared with WT, whereas it was slightly lower (78%) in the FG-4539-treated WT MEFs (Fig. 9, B and C). The rate of glycolysis and glycolytic capacity were significantly increased by 133 to 262% in the Hif-p4h-1 KO and FG-4539treated MEFs compared with controls, whereas no difference was detected in the Hif-p4h-3 KO MEFs (Fig. 9, D-F). The glycolytic reserve of the FG-4539-treated cells was 25% less than the controls (Fig. 9F). Western blot analysis of the HIF1α levels indicated normoxic stabilization of it in the HIF-P4H-1deficient and the FG-4539-treated MEFs but not in HIF-P4H-3-deficient MEFs (Fig. 9G). Altogether these data with the data on the Hif-p4h-2 gt/gt MEFs (13) suggest that normoxic HIF1α stabilization associates with the metabolic switch where oxidative phosphorylation is downregulated and glycolysis is upregulated, and which occurs with genetic HIF-P4H-1 and HIF-P4H-2-deficiency in MEFs and is also obtained with a pharmacologic pan-HIF-P4H inhibitor. . * or # p ≤ 0.05, ** or ## p <0.01, *** or ### p < 0.001, ****p < 0.0001. A black asterisk denotes a statistical difference between genotypes in vehicle-treated mice, a green asterisk a statistical difference between genotypes in FG-4539 or FG4497-treated mice, a red hash a statistical difference between vehicle and FG-4539 or FG4497-treated WT mice, and a blue hash a statistical difference between vehicle and FG-4539 or FG4497-treated Hif-p4h-1/3 KO or Hif-p4h-2 gt/gt mice. Gbe1, 1,4-α-glucan branching enzyme 1; Glut1/4, glucose transporter ¼; Hif-p4h-1-3, hypoxia-inducible factor prolyl-4 hydroxylase 1 to 3; Ppara, peroxisome proliferator-activated receptor α; Pdk1, pyruvate dehydrogenase kinase; Pdk4, pyruvate dehydrogenase kinase 4; Pfkl, phosphofructokinase l; Pparg, peroxisome proliferator-activated receptor γ; VEH, vehicle-treated.

Discussion
The current pharmacological treatment for metabolic dysfunction consists of a combination of several therapeutics, since multiple drugs are required to target specific conditions, i.e., obesity, dyslipidemia, insulin resistance, and fatty liver disease (22). Preclinical data derived mostly from genetic mouse models have suggested that deficiencies in HIF-P4H isoenzyme 1 (18), and especially isoenzyme 2 (12,16,20), mediate several effects that can be beneficial in diseases associated with metabolic dysfunction, whereas loss of HIF-P4H-3 can have both beneficial (17) and adverse (19) effects. Pan-HIF-P4H inhibitors, which antagonize all the isoenzymes equally, have now been approved for the treatment of anemia (https://www.ema.europa.eu/en/medicines/human/EPAR/evrenzo) (7), and the use of such inhibitors developed especially to target the erythropoietic response and erythropoiesis-relevant tissues has, in addition to reversing anemia in patients with CKD, been reported to lower their serum total cholesterol, non-HDL cholesterol, and triglyceride levels (10). These data suggest that beneficial metabolic effects may also be achieved with pan-HIF-P4H inhibitors. Furthermore, individuals experiencing environmental exposure to hypoxia by living at higher altitudes have lower fasting glucose levels and better glucose tolerance than those living close to sea level. Thus, demographic studies have associated living at high altitudes with a lower incidence of obesity and diabetes (23)(24)(25).
HIF-P4Hs are the master regulators of the HIF pathway (3). Many HIF target genes regulate glucose and lipid metabolism, and the HIF response mediates a metabolic reprogramming in which glucose intake (independent of insulin) and non-oxygendemanding glycolytic metabolism are upregulated but OXPHOS is downregulated (6). This comes at the expense of markedly less ATP being generated per glucose molecule, which in general is considered inefficient and a waste of resources but in view of the need for treating the global obesity epidemic stemming from overeating and general inactivity may be desirable (6). We therefore studied here the effects of two preclinical pan-HIF-P4H inhibitors, FG-4497 and FG-4539, on certain anthropometric and metabolic parameters and the expression of metabolic HIF target genes in WT mice. Our data show that both inhibitors are widely applicable for activating the HIF response in WAT, liver, and skeletal muscle, the key tissues for energy metabolism, and conveying the metabolic interplay. These inhibitors alleviated weight gain, lowered the weight of the WAT and liver, improved glucose tolerance, and lowered serum total cholesterol levels. Of importance, the inhibitors were safe to use; they did not cause hypoglycemia, and they only lowered body weight, adiposity, and HOMA-IR in obese and insulin-resistant mice but not in healthy ones. Moreover, similar effects were obtained with both inhibitors, indicating that neither was inferior to the other.
We also aimed to separate out the contribution of each HIF-P4H isoenzyme to the metabolic parameters studied here, in order to provide specifications on which isoenzyme and which tissue should be targeted to obtain the optimal outcome in the treatment of metabolic dysfunction by means of HIF-P4H inhibition. Although categorical conclusions are difficult to formulate, our data show that HIF-P4H-1 inhibition has quite neutral effects on overall metabolism but can provide protection from aging-associated obesity and hypercholesterolemia. Earlier, HIF-P4H-1 loss has been associated with skeletal muscle metabolism, in that it can lower oxygen consumption by reprogramming glucose metabolism from OXPHOS to a more anaerobic form through activation of a PPARα pathway (26). Although HIF-P4H-1 loss has been reported to impair oxidative muscle performance under healthy conditions, it does provide acute protection for myofibers against lethal ischemia via a reduction in oxidative stress (26). The hypoxia tolerance of Hif-p4h-1 KO skeletal muscles was mainly mediated by HIF2α and was not seen in Hif-p4h-3 KO or Hif-p4h-2 heterozygous mice (26). Our data here, and those published earlier on the Hif-p4h-2 gt/gt mice (12,16,27), in which the knockdown of Hif-p4h-2 varies tissue specifically from >90% in the heart to about 80% in skeletal muscle, 50% in WAT and 40% in liver (16,21), indicate that, of the three isoenzymes, inhibition of HIF-P4H-2 has the greatest effects on metabolism, which also agrees with the fact that it is the most abundant one (3). HIF-P4H-2 deficiency has been associated with lower body weight but did not alter the weight gain of nonobese 6-to 7-month-old mice. In agreement with earlier data showing HIF-P4H-2 deficiency to be protective against fatty liver disease of all etiologies (12,28), HIF-P4H-2 inhibition proved here to be associated with lower liver weight, and also with lower serum total and HDL cholesterol levels, the latter being the major lipoprotein in mice, thus differing from the situation in humans (29). Interestingly, our data suggest that pan-inhibition of HIF-P4Hs results in a lesser decline in serum total cholesterol levels than in HIF-P4H-2 deficiency alone, an effect that may well be linked to HIF-P4H-3 inhibition-mediated effects, since, at least upon aging, HIF-P4H-1 loss lowered cholesterol levels while HIF-P4H-3 loss increased them, and we identified downregulation of hepatic Glut1 mRNA in the Hif-p4h-3 KO liver in association with higher cholesterol levels. HIF-P4H-2 was the only isoenzyme whose inhibition improved glucose metabolism by lowering fasting glucose levels and improving glucose tolerance in GTT. These data agree with earlier published observations (16,20). In contrast to HIF-P4H-2, inhibition of HIF-P4H-3 mediated higher body weight and weight gain, higher WAT and liver weights, and higher liver triglyceride levels, especially upon aging. Also, HIF-P4H-3 inhibition resulted in hyperglycemia, higher insulin resistance, and glucose intolerance in GTT. In earlier studies, acute hepatocyte-specific HIF-P4H-3 loss has been shown to improve insulin sensitivity and ameliorate diabetes by specifically stabilizing HIF2α and the upregulation of Irs2 transcription and insulin-stimulated protein kinase B activation (17). HIF-P4H-3 overexpression has been associated with acceleration of the progression of atherosclerosis (30), while its developmental loss has been associated with protection against abnormal sympathoadrenal development and systemic hypotension (31).
In the present analyses of HIF target mRNA levels in WAT, pan-inhibition of HIF-P4Hs upregulated glucose intake and glycolytic metabolism mRNAs, indicating that WAT metabolism can be targeted via pharmacological inhibitors and that inhibition of all the HIF-P4H isoenzymes can contribute to this. Although not reaching significance, there was a trend for HIF-P4H-2 deficiency to be associated with lower adipose Ccl2 mRNA levels, while HIF-P4H-3 deficiency upregulated these levels. This is in line with the detection of more macrophage aggregates in the WAT of aged Hif-p4h-3 KO mice and their higher HOMA-IR scores and the fact that adipose tissue inflammation is closely associated with obesityinduced insulin resistance. Despite these negative effects of HIF-P4H-3 loss in WAT, its inhibition, like that of HIF-P4H-1 and HIF-P4H-2, upregulated adipose Glut1 (Slc2a1), and glycolytic mRNAs, suggesting that the counteractive effects may not have been mediated by adipose tissue HIF-P4H-3 inhibition per se but may have stemmed from systemic effects. The expression of Adipoq mRNA was conversely downregulated in the aged Hif-p4h-3 KO WAT and upregulated in the vehicle-treated Hif-p4h-3 KO WAT. One potential explanation for the difference could be the solvent meglumine used in treating these mice, as this has been associated with improved glucose tolerance and limiting long-term weight gain (32).
In the liver both inhibitors upregulated metabolic HIF target mRNAs, especially in the samples collected from animals sacrificed 6 h after the last dose, suggesting fast hepatic drug metabolism. Upregulation of Glut1 (Slc2a1) mRNA, glycolytic mRNAs, and the insulin sensitivity-increasing Irs2 mRNA was seen in association with HIF-P4H-1 and HIF-P4H-2 inhibition, whereas HIF-P4H-3 inhibition was associated with their downregulation. Despite these similarities between isoenzymes 1 and 2, and the differences between isoenzymes 1/2 and 3, only the inhibition of HIF-P4H-2 led to better glucose tolerance in GTT, suggesting that other mRNAs, and potentially other tissues, contributed to this. Interestingly, the major lipogenic mRNAs in the liver, Srebp1c, and its targets Acca, Fasn, and Scd1, which we have reported earlier to be downregulated in 1-year-old Hif-p4h-2 gt/gt mouse livers (16), and partially here, too, were also downregulated in Hif-p4h-3 KO liver upon aging and when challenged with FG-4497. Considering the higher liver weight and hepatic triglyceride content observed in the aged Hif-p4h-3 KO mice, it is likely that this downregulation served as feedback to limit further de novo lipogenesis.
Also, both inhibitors upregulated metabolic HIF target mRNAs in skeletal muscle, the effect being more widespread in the C57BL6/N/Sv129 background than in C57BL6/N. Glycolytic mRNAs were upregulated following HIF-P4H-1 and HIF-P4H-2 inhibition, and in the case of Pdk4 mRNA also in Hif-p4h-3 KO skeletal muscle when treated with FG-4497. Only HIF-P4H-2 inhibition upregulated skeletal muscle Glut4 (Slc2a4) mRNA levels. Even though the regulation of skeletal muscle insulin-dependent glucose intake by GLUT4 mainly occurs at the level of transporter location in the plasma membrane (33), our data show that upregulation of its mRNA levels in the Hif-p4h-2 gt/gt mice is associated with lower fasting glucose levels and better glucose tolerance. Moreover, the mRNAs for Ppara, a key regulator of fatty acid oxidation in skeletal muscle, and Pparg, the activation of which in skeletal muscle can have a significant protective effect on whole-body glucose homeostasis and insulin resistance (34), were only upregulated in skeletal muscle by HIF-P4H-2 inhibition. Interestingly, upregulation of Hif-p4h-3 mRNA was observed in Hif-p4h-1 KO and Hif-p4h-2 gt/gt skeletal muscle. In view of the overall better metabolic outcome achieved in these mouse lines, the above effect could suggest that HIF-P4H-3 loss or inhibition in skeletal muscle is especially detrimental.
Altogether, our data speak for beneficial systemic effects on metabolism achieved by HIF-P4H-2 inhibition, while HIF-P4H-1 inhibition conveys some beneficial effects but is mostly neutral and HIF-P4H-3 inhibition clearly has detrimental effects on glucose and lipid metabolism. Analyses of the data are also complicated by the endogenous feedback loop in the HIF system, in that Hif-p4h-2 and Hif-p4h-3 are endogenous HIF target genes (35)(36)(37)(38), i.e., the outcome may have been influenced by these isoenzymes potentially providing compensation for the loss of the catalytic activity of others. Moreover, direct comparison between the data for different mouse lines is not possible. It is very important, as also evidenced here, that comparisons between genotypes should only be made within the same strain and cohort. To overcome the systemic complexity, we studied the rates of OXPHOS and glycolysis in MEFs extracted from the HIF-P4H isoenzyme-deficient mice or WT cells treated with FG-4539. These data, together with our earlier published data on HIF-P4H-2-deficient MEFs (13), supported the in vivo findings confirming that the metabolic reprogramming is especially mediated by HIF-P4H-2 inhibition while HIF-P4H-1 inhibition can contribute to it but HIF-P4H-3 inhibition does not. Treatment with the pan-HIF-P4H inhibitor also conveyed the metabolic switch suggesting that inhibition of HIF-P4H-3 on top of HIF-P4H-1 and HIF-P4H-2 cannot revert the phenotype. In all cells, the metabolic reprogramming associated with normoxic HIF1α stabilization. Substrates other than HIFα have been reported to HIF-P4H-3 (38-40), but their hydroxylation has not been confirmed in vitro (41). We cannot therefore conclude here whether the deficiency of HIF-P4H-3 inhibition to stabilize HIF1α and phenocopy metabolic reprogramming was due to it acting on another substrate or something else, for example, the fact that it prefers HIF2α over HIF1α (3). Despite our intention to provide a comprehensive analysis of the contribution of each HIF-P4H isoenzyme to metabolism, some caveats remain. Our data nevertheless clearly support the further development of HIF-P4H inhibitors -preferably ones that are selective for isoenzyme 2 or enzymes ½-that would target the WAT, skeletal muscle, and liver for the treatment of metabolic dysfunction with the aim of limiting the erythropoietic response, especially that mediated by HIF-P4H-2 inhibition.

Mouse lines
The animal experiments were performed in accordance with protocols approved by the National Animal Experiment Board HIF-P4Hs in metabolism of Finland (License number ESAVI/8179/04.10.07/2017, 53/ 2017, OH10). The mice in all the experiments were housed in a standard environment with a temperature of 21 to 22 C and a 12-h day/night cycle, and their well-being was monitored daily. The mice had access to a standard rodent diet (Teklad 18% protein rodent diet, ENVIGO) and water ad libitum. For study-specific setups, see Fig. S1. The generation of the genetically modified mice has been described earlier, the Hif-p4h-1 KO strain in (42), the Hif-p4h-2 gt/gt strain in (27), and the Hif-p4h-3 KO strain in (43).

Collection and analysis of mouse tissues
The male mice in the baseline studies were weighed at baseline (8 m/o) and at sacrifice (1 year/o), while those in the pharmacological experiments were weighed every week. Weight change was determined with the formula (final weight baseline weight). All baseline blood (b) and serum (s) samples were obtained from the vena saphena. Fasting (f) samples were measured after a 12-h fast, b-Hb was measured using a Hb meter (HemoCue Hb 201+) and glucose with a glucose meter (Contour, Bayer). Serum was separated out by centrifugation at 3000g for 20 min at 4 C. Body weight, gonadal WAT, and liver weight were measured at the moment of sacrifice. Tissues were either snap-frozen in liquid nitrogen or fixed in formalin (4% formaldehyde, VWR) overnight for histological analysis. Blood samples were taken at sacrifice from the vena cava.

Analysis of blood and serum parameters
Insulin values were determined with a Rat/Mouse Insulin ELISA kit (EZRMI-13K; Millipore) and s-total cholesterol, s-HDL cholesterol, and s-triglyceride levels by enzymatic methods (Roche Diagnostics). s-EPO was measured using an R&D Systems Quantikine ELISA kit (MEP00B).

Analysis of liver glycogen and triglyceride content
Liver glycogen content was analyzed with the Glycogen Assay Kit (Cayman Chemical, Item No. 700480) using 100 mg of liver and liver triglyceride content by an enzymatic method (Roche Diagnostics), the absorbances of the colorimetric products being determined with the Infinite M1000 Pro Multimode Plate Reader (Tecan).

Histological and immunohistological analyses
Formalin-fixed tissues were embedded in paraffin, cut into 5 μm thin sections and stained with H&E. The slides were studied using a Hamamatsu NanoZoomer S60 slide scanner. WAT adipocyte size was measured using a VisioPharm custom APP. The total area of the region of interest was used as a control between samples. Macrophages were counted from H&E-stained WAT slides by selecting five hot spots from the total tissue area and counting the number of adipose cells surrounded by macrophage aggregates at 20× magnification. The positive identity of the macrophage aggregates was confirmed by immunohistological staining for CD68 (ab955, Abcam, 1:100).

mRNA extraction and PCR analyses
Liver and muscle mRNAs were isolated using TriPure Isolation Reagent (Roche Applied Science) and purified using an E.Z.N.A. Total RNA Kit I (Omega Bio-Tek). For WAT mRNA, the E.Z.N.A. Total RNA Kit II (Omega Bio-Tek) was used. mRNA was transcribed to cDNA using the iScript cDNA Synthesis Kit (Bio-Rad) with 500 ng of mRNA as a template. Quantitative PCR was performed with iTaq SYBR Green Supermix with ROX (Bio-Rad) in a C1000 Touch Thermal Cycler and a CFX96 Touch Real-Time PCR Detection System (Bio-Rad) with the primers shown in Table S1.

HIF-P4H inhibitors and treatment protocols
For the metabolic studies with pharmacological HIF-P4H inhibitors the mice were fed with a standard rodent diet (Teklad 19% protein extruded rodent diet, ENVIGO) and water ad libitum. One of the two pan-HIF-P4H inhibitors FG-4497 or FG-4539 (FibroGen, Inc) or the vehicle was administered orally to the mice 3 times a week (Fig. S1). The FG-4497 used in the pharmacological studies was dissolved in 0.5% sodium carboxymethyl cellulose and 0.1% polysorbate80, which was also used as a vehicle in the comparisons with FG-4497. FG-4539 was dissolved in water and a 1 M meglumine solution, which was likewise used as the vehicle in comparisons with FG-4539. Individual doses were determined each week on the basis of body weight.

Glucose tolerance test
GTT was performed after treating the mice for 4 weeks with the pan-HIF-P4H inhibitors or vehicle. The mice fasted for 12 h prior to GTT. Animals were anesthetized (s.c.) with fentanyl-midazolam (0.1 ml/10 g), and glucose (1 mg/g) was administered via an i.p. injection. B-glucose was measured (Contour, Bayer) from the vena saphena at baseline and 15, 30, 60, and 120 min time points. The treatment was continued for 2 weeks after the GTT, and then the mice were sacrificed. The AUC for GTT was calculated by the summary measures method, and the HOMA-IR scores from the fb-glucose and fsinsulin values were calculated by a formula = (insulin (pmol/ l))/(glucose (mmol/l))*156.65.

Seahorse XFp analysis
Real-time monitoring of oxygen consumption and extracellular acidification rates of Hif-p4h-1, Hif-p4h-3, and their corresponding WT MEFs was performed using a Seahorse XFp Analyzer, Seahorse XFp Cell Mito Stress Test Kit, and Seahorse XFp Glycolysis Stress Test Kit (Agilent) as described (13). For the experiments with pan HIF-P4H inhibitor, WT MEFs were treated with either dimethyl sulfoxide as a vehicle or with 50 μM FG-4539 overnight. All the data were normalized to either cell number or total protein concentration. All the assays were analyzed using the Seahorse XF Report Generator software (Wave, Agilent) and the metabolic parameters were derived from calculations based on the manufacturer's instructions.

Statistical analyses
Student's t test was used to compare statistical significances between two groups. All data are presented as means ± standard error of the mean (SEM). p ≤ 0.05 was considered statistically significant. In the figures statistical significance is indicated by asterisks: * or # = p ≤ 0.05, ** or ## = p <0.01, *** or ### = p < 0.001, **** or #### = p < 0.0001. Grubbs' test was used to determine outliers in univariate data sets. Pearson correlation coefficients with 95% confidence intervals were used to evaluate associations between metabolic parameters and mRNA expression levels.

Data availability
All the data used are provided in this article.
Supporting information-This article contains supporting information.