PGC-1α Coordinates Mitochondrial Respiratory Capacity and Muscular Fatty Acid Uptake via Regulation of VEGF-B

Vascular endothelial growth factor B (VEGF-B) has been shown to control endothelial fatty acid (FA) transcytosis and tissue accumulation of lipids through regulation of the FA transport proteins (FATPs) FATP3 and FATP4 (1). Reduction of VEGF-B levels improves the development of type 2 diabetes in diabetic rodent models by decreasing muscular lipid accumulation (2). However, the regulation of Vegfb expression is not well understood. Vegfb expression is high in mitochondria-dense tissues (3), and bioinformatic analysis showed that Vegfb is coexpressed with a set of nuclear-encoded mitochondrial genes, the so-called OXPHOS genes (1). Peroxisome proliferator–activated receptor γ coactivator 1α (Ppargc1a/PGC-1α) is a major regulator of mitochondrial energy metabolism, and in response to extrinsic factors, PGC-1α binds to and coactivates several transcription factors, including estrogen-related receptor α (Esrra/ERR-α), peroxisome proliferator–activated receptor γ (PPARγ), and nuclear respiratory factor 1 (NRF1) (4).

A common pathology in type 2 diabetes is dysfunctional and insufficient muscular mitochondria content (5), and therefore, elevated PGC-1α expression was expected to be beneficial by inducing mitochondrial biogenesis. Of note, mice overexpressing PGC-1α in skeletal muscle (muscle creatine kinase PGC-1α transgenic [MPGC-1αTG] mice) displayed reduced insulin sensitivity upon high-fat diet (HFD) feeding despite increased mitochondrial density and higher respiratory capacity (6,7). Given the coexpression of Vegfb with OXPHOS genes and the role of VEGF-B in ectopic lipid accumulation and insulin resistance, we stipulated that Vegfb expression may be regulated by PGC-1α as well. In the current study, we show that PGC-1α and ERR-α control the expression of Vegfb. This previously undescribed regulatory pathway allows PGC-1α to coordinate FA uptake with mitochondrial FA oxidation through regulation of VEGF-B levels.

MPGC-1αTG mice (6) were crossed with Vegfb-deficient mice (3) to ultimately create MPGC-1αTG//Vegfb^−/− mice, and only male age-matched mice from the F3 generation were used. For HFD studies, wild-type (WT), MPGC-1αTG, MPGC-1αTG//Vegfb^−/−, and Vegfb^−/− mice where fed 60% HFD (Research Diets) for 12 weeks starting from 5 weeks of age. The study was repeated twice. All animals had ad libitum access to food and water and were housed in standard cages in an environment with 12-h light/dark cycles. All mouse work was conducted in accordance to the Swedish Animal Welfare Board at Karolinska Institutet, Stockholm, Sweden.

Cell lines were maintained in DMEM supplemented with 10% FBS at 37°C in 5% CO₂. C2C12 was maintained as described (8). For nutrient deprivation studies, cells were starved in DMEM.

Putative consensus binding sites for ERR-α (ERR1_Q2) within the first kilobase (kb) promoter and first intron of Vegfb were analyzed using Evolutionary Conserved Region (ECR) Browser, Mulan, and multiTF databases (9) (http://ecrbrowser.dcode.org). The sequences within the ECRs were fed to Mulan and then continued with multiTF analysis with a predefined factor of 0.80.

The plasmids pCMX-ERR-α, pcDNA3.1/myc-HisA-PGC-1α-FLAG, pTK-LUC, and pCMX-β-Gal and corresponding empty vectors were used in the experiments. PCR fragments of the first kb Vegfb promoter and first Vegfb intron were generated with overhangs of Hind III and Sbf I restriction cleavage sites and 12 more random base pairs (bps). The fragments were digested with Hind III and Sbf I and then subcloned into pTK-LUC. Putative ERR-α binding sites were mutated into Nco I restriction cleavage site (CCATGG) by site-directed mutagenesis.

COS-1 cells were transfected in 24-well plates using Lipofectamine LTX according to the manufacturer’s instructions. pCMX-β-Gal was cotransfected to allow normalization of transfection efficiency, and 24 or 48 h posttransfection, the cells were harvested and lysed. OD₄₀₅ for β-galactosidase activity and chemiluminescence signal for luciferase activity were read by using a POLARstar Omega plate reader.

PGC-1α overexpression in the MPGC-1αTG mice is highest in glycolytic type II fibers (6); therefore, quadriceps femoris was used for all downstream analyses. Total RNA was extracted by using the RNeasy Mini Kit (QIAGEN); thereafter, cDNA was synthesized using iScript cDNA Synthesis Kit according to the manufacturer’s instructions. Real-time quantitative PCR (qPCR) was performed as previously described (2). Primer sequences are listed in Table 1.

Muscles were dissected, sectioned, and stained with Oil Red O (ORO), and the staining was quantified as previously reported (10).

Body weights and postprandial blood glucose levels were measured by intraperitoneal glucose tolerance tests (IPGTTs) and intraperitoneal insulin tolerance tests (IPITTs) performed in mice as previously described (2). Pooled data from at least two independent experiments, totaling 10–14 animals per genotype, were included in the analysis. Metabolic analyses of mouse plasma were performed as previously described (2). HOMA of insulin resistance (HOMA-IR) was calculated as (fasting glucose [mmol/L] × fasting insulin [mU/L])/22.5.

Muscles were flash frozen in liquid nitrogen, cryosectioned into 12-μm sections, and stained using standard procedures. The primary antibodies were goat anti-FATP3 (Santa Cruz Biotechnology, Santa Cruz, TX), rat anti-CD31 (BD Biosciences, San Jose, CA), rabbit anti-FATP4 (Santa Cruz Biotechnology), and goat anti-CD31 (R&D Systems, Minneapolis, MN). At least 10 frames per animal within each section were photographed with an AxioVision microscope (Carl Zeiss, Oberkochen, Germany) at ×40 magnification. For quantification of the stainings, the number of pixels per square micrometer was calculated using the AxioVision quantification program.

β-Oxidation activity was measured in isolated skeletal muscle preparations as previously described (1).

Ten milligrams of muscle tissue were analyzed by ultra-fast liquid chromatography–mass spectrometry (LC-MS)–based lipidomic analysis by the Swedish Metabolomics Centre at the Swedish University of Agricultural Sciences. The analyses were carried out as previously described (11).

In all figures, data are presented as mean ± SEM, with n referring to the number of independent data points. P values were calculated with two-tailed Student t test, and P < 0.05 was considered significant.

To investigate whether Vegfb can be induced by starvation, we monitored the expression levels of Ppargc1a and Vegfb in serum-deprived C2C12 myotubes (Supplementary Fig. 1). Eight hours of nutrient deprivation induced the expression of both Ppargc1a and Vegfb by 2.5- and 1.5-fold, respectively, and the induction was increased to 3.5- and 1.8-fold after 16 h of starvation (Supplementary Fig. 1A). The induced expression of Ppargc1a and Vegfb upon nutrient deprivation was confirmed in fibroblast (COS-1) and pericyte precursor (10T1/2) cell lines (Supplementary Fig. 1B and C).

We characterized the Vegfb promoter by cloning the first kb of the 5′-untranslated region and the first exon and intron of the Vegfb gene in front of a luciferase reporter construct. Plasmids encoding PGC-1α and/or the transcription factors ERR-α, NRF1, or PPARγ were then cotransfected with the luciferase reporter construct in COS-1 cells (Fig. 1A). The results show that only coexpression of PGC-1α and ERR-α induced significant transactivation of the reporter construct. To identify putative ERR-α binding sites in the Vegfb promoter, the murine genomic sequence was analyzed by the ECR Browser, Mulan, and multiTF algorithms (9,12). The analyses identified two ERR-α consensus DNA binding sites (13) located in the 5′promoter region (−315 to −310) and in the first intron of the Vegfb gene (+340 to +345) and one putative binding site (−571 to −566) (Fig. 1B). To identify the ERR-α binding sites responsible for the transactivation of the Vegfb promoter, new reporter constructs were generated where each ERR-α binding site was mutated by a single bp substitution. No luciferase activity was detected in the absence of PGC-1α and ERR-α, showing that the Vegfb construct per se did not induce luciferase activity (Fig. 1B, upper bar). Cotransfection of PGC-1α and ERR-α with WT or modified luciferase constructs revealed that mutating the −571 to −566 binding site significantly decreased luciferase activity compared with control reporter constructs (Fig. 1B). Mutations in the two other ERR-α DNA binding sites did not decrease luciferase activity. However, additional cryptic ERR-α binding sites most likely exist between −0.5 and 0 kb in the Vegfb promoter and in the first intron of the Vegfb gene because reporter constructs of both these genomic regions alone also displayed high luciferase activity (Fig. 1B).

We crossed MPGC-1αTG mice (6) with Vegfb-deficient animals (3) to ultimately create MPGC-1αTG//Vegfb^−/− mice. qPCR analysis demonstrated a sevenfold increase in muscular expression of Ppargc1a transcripts in MPGC-1αTG mice compared with WT controls (Fig. 1C). Similar analyses showed six- and threefold upregulation of the expression of Vegfb and Esrra, respectively (Fig. 1C). MPGC-1αTG mice also showed an approximate twofold upregulation of the previously reported PGC-1α target genes Vegfa; transcription factor A, mitochondrial (Tfam); NADH dehydrogenase (ubiquione) 1 α subcomplex 5 (Ndufa5); NADH dehydrogenase (ubiquione) 1 β subcomplex 10 (Ndufb10); and cytochrome C, somatic (Cycs) (Fig. 1C and D). Ablation of Vegfb in MPGC-1αTG mice did not affect the expression of any of these genes.

We analyzed muscular expression of FATPs and several other FA handling proteins previously reported to be regulated by PGC-1α (14). The expression of fatty acid translocase (Cd36), Fatp1, and Fatp4 was upregulated in MPGC-1αTG mice compared with WT mice (Fig. 1E). Fatp3, the vascular-specific target of VEGF-B signaling, was downregulated in the MPGC-1αTG//Vegfb^−/− mice compared with the WT and MPGC-1αTG animals (Fig. 1E). Taken together, PGC-1α and ERR-α regulate Vegfb expression, which requires at least one intact ERR-α binding site located in the promoter region of Vegfb. In vivo, PGC-1α controls Vegfb expression in parallel with the expression of OXPHOS genes and Vegfa. The expression of the OXPHOS genes was not altered by Vegfb deletion, whereas Fatp3 was downregulated only in the MPGC-1αTG//Vegfb^−/− mice.

Muscle sections from WT, MPGC-1αTG, MPGC-1αTG//Vegfb^−/−, and Vegfb^−/− mice were stained with ORO to visualize accumulation of neutral lipids. Quantifications showed a fivefold increase in lipid accumulation in MPGC-1αTG compared with WT mice (Fig. 2A). In contrast, MPGC-1αTG//Vegfb^−/− mice had almost normalized levels of lipid accumulation (Fig. 2A). The lowest level of lipid accumulation was detected in Vegfb^−/− mice, in line with previous results (1).

Intracellular lipids can accumulate either through increased FA uptake, decreased β-oxidation, or increased de novo FA synthesis. Therefore, muscular expression of genes encoding critical enzymes in β-oxidation and de novo FA synthesis were measured by qPCR analysis (Fig. 2B). The transcript levels of regulatory genes in β-oxidation and de novo FA synthesis were upregulated in MPGC-1αTG compared with WT mice (Fig. 2B), which is in line with previous reports (7). No further increase in the expression of any of the genes was detected in MPGC-1αTG//Vegfb^−/− mice (Fig. 2B), and in line with previous studies, no difference in expression levels of these genes was detected between WT and Vegfb^−/− mice (1).

Vegfb^−/− mice have a shift in nutrient utilization toward increased use of glucose and decreased use of FAs compared with WT animals (1). However, on chow diet, no differences in postprandial blood glucose levels or glucose tolerance were detected among WT, MPGC-1αTG, MPGC-1αTG//Vegfb^−/−, and Vegfb^−/− mice (Fig. 2C and D). A minor increase in the muscular expression of Glut4 (gene name Slc2a4) but not of Glut1 (gene name Slc2a1) was detected in MPGC-1αTG//Vegfb^−/− mice compared with MPGC-1αTG mice (Fig. 2E). PGC-1α/ERR-α has been reported to regulate glucose utilization by controlling the expression of several intracellular glucose handling enzymes (15,16); therefore, we assessed muscular expression of these transcripts in WT, MPGC-1αTG, MPGC-1αTG//Vegfb^−/−, and Vegfb^−/− mice (Fig. 2F). Muscular pyruvate dehydrogenase kinase 4 (Pdk4) expression was downregulated, with ∼60% in MPGC-1αTG//Vegfb^−/− and Vegfb^−/− mice compared with MPGC-1αTG mice. A minor increase in citrate synthase (Cs) levels (25%) were detected in MPGC-1αTG//Vegfb^−/− compared with MPGC-1αTG mice. In summary, no major differences in glucose tolerance, blood glucose levels, or expression of glucose transporters were found between WT and MPGC-1αTG, MPGC-1αTG//Vegfb^−/−, and Vegfb^−/− mice on chow diet.

To characterize the impact of VEGF-B–driven muscular lipid accumulation of the diabetic phenotype in HFD-fed MPGC-1αTG mice, transcriptional profiling of Vegfb and genes involved in lipid uptake and oxidation were analyzed in chow- and HFD-fed animals. Vegfb expression was induced fourfold by HFD feeding in WT mice compared with chow-fed mice and twofold in HFD-fed MPGC-1αTG mice compared with chow-fed MPGC-1αTG mice (Fig. 3A). HFD feeding did not increase Esrra levels in any of the mouse strains compared with the respective genotype fed the chow diet (Fig. 3A). However, HFD feeding increased muscular expression of genes encoding FA handling proteins in both WT and MPGC-1αTG mice (Fig. 3A). Deletion of Vegfb in the HFD-fed MPGC-1αTG mice significantly prevented the upregulation of Fatp4 (Fig. 3A). On the contrary, the expression of mitochondrial genes was reduced upon HFD feeding in WT, as previously reported (2), and also in MPGC-1αTG mice (Fig. 3B). Deletion of Vegfb in the HFD-fed MPGC-1αTG preserved the expression of several OXPHOS genes compared with MPGC-1αTG mice (Fig. 3B). Thus, transcriptional analysis suggested that HFD feeding induced an imbalance between lipid uptake and lipid oxidation, and reducing VEGF-B levels may prevent this, leading to metabolic normalization. To analyze whether ablation of Vegfb reduced muscular FA uptake during HFD, muscular tissue sections from the various genotypes were stained with ORO. Muscular lipid accumulation was increased eightfold in HFD-fed MPGC-1αTG mice compared with WT HFD-fed mice but only fourfold in MPGC-1αTG//Vegfb^−/− mice (Fig. 4A). Hence, deletion of Vegfb alone in MPGC-1αTG mice lowered muscular lipid accumulation by 80%. In accordance with our previous findings (2), the lowest levels of lipid accumulation was seen in the HFD-fed Vegfb^−/− mice. The decreased muscular lipid accumulation in HFD-fed MPGC-1αTG//Vegfb^−/− mice was not due to a compensatory increase in β-oxidation or decrease in de novo FA synthesis because transcript levels of the regulatory genes in these pathways were not altered (Fig. 4B). We also assessed whether genetic ablation of Vegfb in HFD-fed MPGC-1αTG mice directly affected the capacity of β-oxidation by incubating isolated skeletal muscle preparations with ¹⁴C-oleic acid and measured formed ¹⁴CO₂ (Fig. 4C). However, no differences in the levels of released ¹⁴CO₂ were observed between genotypes.

The transcriptional alterations of the FATPs were modest in relation to the drastic reduction in lipid accumulation in the HFD-fed MPGC-1αTG//Vegfb^−/− mice, and we used immunohistochemistry to analyze protein expression of FATP3 and FATP4 in chow-fed WT, HFD-fed WT, MPGC-1αTG, MPGC-1αTG//Vegfb^−/−, and Vegfb^−/− mice (Fig. 4D). CD31, a vessel-specific marker protein, was used to identify vascular localization of the FATPs. The expression of both FATP3 and FATP4 was strongly upregulated in HFD-fed MPGC-1αTG mice compared with HFD-fed WT mice. Strikingly, the protein expression of both FATP3 and FATP4 were downregulated by ∼70% and 90% in HFD-fed MPGC-1αTG//Vegfb^−/− and HFD-fed MPGC-1αTG mice, respectively. In summary, HFD feeding induced the expression of Vegfb and FA handling protein transcripts in both WT and MPGC-1αTG mice, and muscular lipid deposition was increased accordingly. Of note, muscular lipid accumulation was dramatically reduced in the MPGC-1αTG//Vegfb^−/− mice, which was caused by lowered FA uptake through decreased expression of FATP3 and FATP4.

High intramyocellular lipid content in both humans and experimental animal models have been shown to be associated with insulin resistance (17,18). Specifically, both diacylglycerols (DAGs) derived from lipid droplets and ceramides have been shown to affect insulin signaling and/or insulin-mediated muscular glucose uptake (19). Therefore, we used ultrafast LC-MS–based lipidomics to quantify the levels of triglycerides (TGs), DAGs, and ceramides in muscle biopsy specimens from chow-fed WT mice and HFD-fed WT, MPGC-1αTG, MPGC-1αTG//Vegfb^−/−, and Vegfb^−/− mice (Fig. 5). Area under the curve (AUC) analysis of all lipid species showed that MPGC-1αTG//Vegfb^−/− and Vegfb^−/− mice had ∼35% and 60% lower muscular TG content, respectively, compared with MPGC-1αTG mice (Fig. 5A and B). Similar analysis showed that muscular levels of DAGs and ceramides were reduced by ∼35% and 25% in MPGC-1αTG//Vegfb^−/− mice and by ∼70% and 35% in Vegfb^−/− mice compared with MPGC-1αTG mice (Fig. 5C–F). To analyze whether VEGF-B controls the degradation of TGs to DAGs or de novo DAG synthesis, we measured muscular expression levels of adipose triglyceride lipase (Atgl) and phospholipid phosphatase 1 (Plpp1) (Fig. 5G). Expression of Atgl was downregulated in both HFD-fed MPGC-1αTG//Vegfb^−/− and Vegfb^−/− mice, suggesting that lower amounts of TGs are available for degradation. No differences were found between genotypes in expression of Plpp1 or in the expression of transcripts coding for key enzymes for ceramide formation (Fig. 5G and H).

We analyzed whether deletion of Vegfb could reverse insulin resistance in HFD-fed MPGC-1αTG mice (Fig. 6). Postprandial blood glucose levels were normalized in HFD-fed MPGC-1αTG//Vegfb^−/− and Vegfb^−/− animals (Fig. 6A). As previously reported, Vegfb^−/− mice had an increased body weight compared with WT littermates upon HFD feeding (Fig. 6A) (2). However, no differences in body weights were detected among WT, MPGC-1αTG, and MPGC-1αTG//Vegfb^−/− mice (Fig. 6A). HFD-fed WT and MPGC-1αTG mice but not MPGC-1αTG//Vegfb^−/− mice developed hyperinsulinemia, whereas HFD-fed Vegfb^−/− mice displayed moderately increased plasma insulin levels (Fig. 6B). The HOMA-IR index mirrored the plasma insulin levels (Fig. 6B). qPCR analysis showed that MPGC-1αTG//Vegfb^−/− mice had higher Glut4 expression (Fig. 6C), lower Pdk4 expression (Fig. 6D), and higher Cs expression (Fig. 6D) compared with MPGC-1αTG mice, suggesting higher glucose utilization (20). In line with reduced muscular lipid accumulation and increased expression of glucose handling enzymes, the HFD-fed MPGC-1αTG//Vegfb^−/− and Vegfb^−/− mice showed improved glucose tolerance and insulin sensitivity compared with HFD-fed WT and MPGC-1αTG mice (Fig. 6E and F).

To visualize how insulin resistance correlated to VEGF-B signaling and muscular lipid accumulation, we plotted muscular lipid content and measurements of insulin resistance in individual HFD-fed mice (Fig. 7A). HFD-fed MPGC-1αTG were characterized by high muscular lipid content and insulin resistance. Reducing Vegfb^−/− levels in HFD-fed MPGC-1αTG mice strongly affected lipid accumulation and increased insulin sensitivity. In summary, deletion of Vegfb can completely prevent insulin resistance in HFD-fed MPGC-1αTG mice.

In addition to insulin resistance, dyslipidemia is a key feature of type 2 diabetes. Hence, the levels of several plasma lipid species were analyzed in HFD-fed mice (Fig. 7B and C). The results showed that HFD feeding leads to increased levels of plasma TGs in WT and MPGC-1αTG mice. In contrast, TG levels were decreased in HFD-fed MPGC-1αTG//Vegfb^−/− and Vegfb^−/− mice compared with both WT and MPGC-1αTG animals (Fig. 7B). LDL and VLDL cholesterol were also decreased in MPGC-1αTG//Vegfb^−/− mice compared with MPGC-1αTG mice, whereas no difference in HDL cholesterol was detected (Fig. 7B). Similar levels of nonesterified fatty acids (NEFAs) and ketone bodies (KBs) were also detected in MPGC-1αTG and MPGC-1αTG//Vegfb^−/− mice (Fig. 7C). In summary, the metabolic tolerance tests showed that genetic deletion of Vegfb in HFD-fed MPGC-1αTG mice is sufficient to prevent hyperglycemia and hyperinsulinemia, reduce insulin resistance, and improve dyslipidemia, thus rescuing the aberrant phenotype of this previously enigmatic mouse model.

In this study, we demonstrate that Vegfb is a downstream target of the PGC-1α/ERR-α signaling pathway. PGC-1α is induced by physiological stimuli, including exercise, fasting, and cold temperature (4). Of note, mRNA levels of Vegfb are also increased by nutrient deprivation (Supplementary Fig. 1) and during exercise, which is shown in muscles biopsy specimens from mice and human subjects (21,22). The coregulation of PGC-1α and VEGF-B ensures that lipid uptake from the circulation is coordinated with tissue lipid oxidation (Fig. 7D). PGC-1α/ERR-α have also been shown to induce VEGF-A levels and angiogenesis in skeletal muscle in vivo during starvation (8,23–25). Thus, PGC-1α and ERR-α coordinate VEGF-A and VEGF-B expression levels, mitochondrial biogenesis, and lipid oxidation not only by controlling vessel growth but also by regulating lipid uptake.

Given the impact of PGC-1α as the main activator of mitochondrial biogenesis and energy metabolism, efforts have been made to elucidate the role of PGC-1α in metabolic conditions such as insulin resistance and type 2 diabetes. The muscular expression of Ppargc1a and OXPHOS genes is reduced in patients with type 2 diabetes, suggesting that low PGC-1α activity could be the underlying mechanism of type 2 diabetes (26,27). Therefore, it was surprising and paradoxical that HFD-fed MPGC-1αTG mice developed insulin resistance and type 2 diabetes (7). In this study, we have unraveled the mechanism underlying the diabetic phenotype reported for the HFD-fed MPGC-1αTG mice (7). We show that muscular overexpression of PGC-1α upregulates Vegfb levels, driving FA uptake and tissue accumulation and, as a consequence, insulin resistance. Inactivation of Vegfb in HFD-fed MPGC-1αTG mice reduced ectopic lipid accumulation and the development of type 2 diabetes by reduced protein levels of endothelial FATPs. Hence, the HFD-induced insulin resistance in MPGC-1αTG mice first described by Choi et al. (7) is most likely caused by a PGC-1α–dependent upregulation of Vegfb expression.

Ablation of Vegfb in MPGC-1αTG mice did not affect plasma levels of NEFAs or KBs, despite reduction of muscular lipid accumulation. The current data agree with that of Choi et al. (7), which showed that HFD-fed MPGC-1αTG mice had lower plasma levels of KB, despite increased muscular FA inflow. We have previously shown that reducing VEGF-B levels in db/db mice reduced both muscular lipid accumulation and plasma levels of NEFAs and KBs (2). Furthermore, lipid infusions in unchallenged Vegfb^−/− mice redirected lipids from the peripheral tissues to the adipose tissue (AT) (1). PGC-1α is described as a regulator of several myokines shown to induce alterations in the AT (22). Therefore, the lack of differences in plasma NEFA levels between MPGC-1αTG and MPGC-1αTG//Vegfb^−/− could be caused either by the effect of PGC-1α on AT tissue or by VEGF-B–dependent shunting of lipids to the AT.

Efforts have been made to try to link the VEGF-B signaling pathway to insulin resistance in humans, but it was reported that serum VEGF-B levels did not differ between patients with type 2 diabetes and healthy control subjects (28). However, a major confounding factor was that the trial enrolled patients taking oral antidiabetic drugs, such as metformin and thiazolidinediones. The authors showed that thiazolidinediones reduced circulating levels of VEGF-B. Similarly, in a more recent clinical trial of metformin treatment in women with polycystic ovary syndrome and insulin resistance, metformin was shown to reduce plasma VEGF-B levels (29). This trial enrolled only patients who had not received antidiabetic drugs within the past 3 months before the metformin treatment was initiated. Measurements of serum VEGF-B levels before metformin treatment showed that VEGF-B levels were increased in the subjects with insulin resistance and correlated positively with insulin resistance. In addition, gene association studies have linked the VEGF-B signaling pathway to insulin resistance in humans because a sequence variant in Fatp4 was identified to be associated with the metabolic syndrome and insulin resistance (30). Moreover, serum VEGF-B levels as well as AT Vegfb expression have been shown to be upregulated in patients who are obese (31,32). Taken together, several human studies have shown interesting links among VEGF-B levels, obesity, and insulin resistance.

In summary, the current study identifies Vegfb as a downstream target of the PGC-1α/ERR-α signaling pathway and a mediator of the PGC-1α–induced adverse effects upon HFD feeding. We provide evidence that tissue metabolism is intimately linked to the functional features of the vasculature and highlights the importance of PGC-1α as a major metabolic sensor/regulator to coordinate respiratory capacity with vessel growth through VEGF-A and lipid uptake through VEGF-B.

Acknowledgments. The authors thank Janet Rossant (Departments of Molecular Genetics and Obstetrics and Gynaecology, University of Toronto) and Guo-Hua Fong (Department of Cell Biology and Center for Vascular Biology, University of Connecticut Health Center) for the gift of the MPGC-1αTG mice; Thomas Perlmann (Department of Cell and Molecular Biology, Karolinska Institutet) for the gifts of pCMX-ERR-α, pcDNA3.1/myc-HisA-PGC-1α-FLAG, pTK-LUC, pCMX-β-Gal plasmids; and the Swedish Metabolomics Centre (Umeå University; www.swedishmetabolomicscentre.se) for help with lipid analysis. The authors also thank Sofia Wittgren and Karin Pettersson (Department of Medical Biochemistry and Biophysics, Karolinska Institutet) for mouse care and mouse genotyping.

Funding. This study was supported by the Wilhelm och Else Stockmanns Stiftelse (C.E.H.), Swedish Heart-Lung Foundation (U.E.), Novo Nordisk Foundation (U.E.), Swedish Cancer Foundation (U.E.), Swedish Research Council (U.E.), Torsten Söderbergs Stiftelse (U.E.), Ragnar Söderbergs Stiftelse (U.E.), Ludwig Institute for Cancer Research (U.E.), CSL Ltd., Melbourne, Australia (U.E.), and Karolinska Institutet.

Duality of Interest. A.M., U.E., and A.F. are shareholders in a company within the diabetes field. This study was supported by CSL Ltd., Melbourne, Australia (U.E.). U.E. is a consultant for CSL Ltd., Melbourne, Australia. This does not alter the author’s adherence to all policies of Diabetes. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. A.M. and A.F. contributed to the study design and research, data analysis, and writing and review of the manuscript. I.P. performed expression, immunohistochemistry, and lipid analyses and contributed to the review of the manuscript. X.W. performed the in vitro analysis and contributed to the in vivo experiments and review of the manuscript. C.E.H. initiated the project and contributed to the review of the manuscript. U.E. designed the research and contributed to the review of the manuscript. U.E. and A.F. are the guarantors of this work and, as such, had full access to all data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at 2013 Keystone Symposia on Molecular and Cellular Biology Diabetes—New Insights into Mechanism of Disease and its Treatment, Keystone, CO, 27 January–1 February 2013.