Ablation of 3-Phosphoinositide-Dependent Protein Kinase 1 (PDK1) in Vascular Endothelial Cells Enhances Insulin Sensitivity by Reducing Visceral Fat and Suppressing Angiogenesis

Abstract

The phosphatidylinositol 3-kinase signaling pathway in vascular endothelial cells is important for systemic angiogenesis and glucose metabolism. In this study, we addressed the precise role of the 3-phosphoinositide-dependent protein kinase 1 (PDK1)-regulated signaling network in endothelial cells in vivo, using vascular endothelial PDK1 knockout (VEPDK1KO) mice. Surprisingly, VEPDK1KO mice manifested enhanced glucose tolerance and whole-body insulin sensitivity due to suppression of their hepatic glucose production with no change in either peripheral glucose disposal or even impaired vascular endothelial function at 6 months of age. When mice were fed a standard diet at 6 months of age and a high-fat diet at 3 months of age, hypertrophy of epididymal adipose tissues was inhibited, adiponectin mRNA was significantly increased, and mRNA of MCP1, leptin, and TNFα was decreased in the white adipose tissue of VEPDK1KO mice in comparison with controls. Consequently, both the circulating adiponectin levels and the activity of hepatic AMP-activated protein kinase were significantly increased, subsequently enhancing whole-body insulin sensitivity and energy expenditure with increased hepatic fatty acid oxidation in VEPDK1KO mice. These results provide the first in vivo evidence that lowered angiogenesis through the deletion of PDK1 signaling not only interferes with the growth of adipose tissue but also induces increased energy expenditure due to amelioration of the adipocytokine profile. This demonstrates an unexpected role of PDK1 signaling in endothelial cells on the maintenance of proper glucose homeostasis through the regulation of adipocyte development.

A growing body of evidence has implicated the phosphatidylinositol 3-kinase (PI3K)/Akt signaling pathway, which is initiated by several growth factors, as a key player in energy and glucose homeostasis, and also in the normal development and function of many cells and tissues (1, 2). In vascular endothelial cells, PI3K/Akt signaling is a powerful regulator of cellular proliferation, viability and migration (3–5). Recent studies suggested that PI3K/Akt signaling plays a central role in angiogenesis by showing the defects in angiogenic sprouting and vascular remodeling and development in knockout mice lacking p110α or phosphatase and tensin homolog (PTEN) deleted from chromosomes specifically in vascular endothelial cells (3, 5–7). Vascular endothelial growth factor (VEGF) potently stimulates the PI3K/Akt pathway in endothelial cells and regulates angiogenesis in vitro and in vivo (3).

3-Phosphoinositide-dependent protein kinase-1 (PDK1) is a serine-threonine kinase that mediates downstream signaling of PI3K and regulates the activity of Akt and p70 ribosomal S6 kinase (S6K). Because PDK1 exclusively phosphorylates Akt at threonine (Thr) 308, PDK1 is thought to play a central role in the PI3K/Akt pathway in many cell types. Mice expressing PDK1 at approximately 10% of normal levels exhibit normal activation of Akt and S6K in response to insulin stimulation, despite significant reductions in body weight and organ volumes (8). A line of tissue-specific PDK1 knockout mice were generated, and either decreased organ mass and volume (9) or impaired glucose metabolism (10–12) was reported. These studies strongly suggest that the PDK1/A-kinase, cGMP-kinase, C-kinase kinase signaling pathway is crucial in an array of cellular processes and in the maintenance of normal organ function.

To address the role of the PDK1-regulated signaling network in endothelial cells, we generated mice with PDK1 deficiencies only in vascular endothelial cells [vascular endothelial PDK1 knockout (VEPDK1KO)] by the excision of floxed alleles of PDK1 with Cre expressed under the control of the Tie2 promoter (Tie2-Cre) (13) and investigated the physiological role of the endothelial PDK1 and PI3K signal for regulation of angiogenesis, blood pressure, and systemic metabolism in vivo.

Results

Generation of endothelial cell-specific PDK1-deficient mice

VEPDK1KO mice were born at the expected Mendelian frequency and littermate mice not expressing Tie2-Cre (PDK1^flox/floxTie2Cre^−/−) were used as the control mice throughout this study. Although primary cultures of mouse lung vascular endothelial cells (MLEC) of VEPDK1KO mice did not express the PDK1 protein, PDK1 expression in gastrocnemius muscle (Gastro), heart, aorta, liver, white adipose tissue (WAT), brown adipose tissue (BAT), and brain were comparable in the VEPDK1KO and control mice (Fig. 1A).

Fig. 1.

Generation of the VEPDK1KO. A, PDK1 expression in primary cultured endothelial cells isolated from mouse lung (MLEC) or Gastro, heart, aorta, liver, WAT, BAT, and brain tissues. B, VEGF-stimulated PDK1 signal in MLEC. C, Insulin-stimulated PDK1 signal in MLEC. All figures are representative images.

VEGF- and insulin-induced signaling in VEPDK1KO mice

VEGF stimulates growth, proliferation, and permeability of endothelial cells through the PI3K/PDK1/Akt and MAPK pathways (14). To determine whether lack of PDK1 expression in vascular endothelial cells impairs activation of Akt, primary cultures of MLEC were stimulated with VEGF and phosphorylation of Akt at Thr308, which is directly phosphorylated by PDK1, and at Ser473, which is independent of PDK1, and were investigated using phosphospecific antibodies. In the control mice, VEGF stimulated phosphorylation of Akt at both Thr308 and Ser473 in endothelial cells. In contrast, VEGF failed to induce phosphorylation of Akt at Thr308 by 58%, but not at Ser473, in MLEC from VEPDK1KO mice. VEGF-induced phosphorylation of S6K at Thr389 and endothelial nitric oxide synthase (eNOS) at Ser1177 also was inhibited by 98 and 86%, respectively, in VEPDK1KO mice, whereas that of ERK1/2 at Thr202/Tyr204 residues was preserved in VEPDK1KO mice (Fig. 1B). The amounts of Akt, S6K, eNOS, and ERK1/2 did not differ between the two genotypes. In terms of the insulin-mediated signaling, phosphorylation of Akt at Thr308 and of eNOS at Ser1177 was decreased by insulin in MLEC in VEPDK1KO mice, as expected. In contrast, phosphorylation of Akt at Ser473 and ERK1/2 at Thr202/Tyr204 was not affected (Fig. 1C). These results demonstrated that inactivation of PDK1 specifically impaired the PDK1/Akt, but not the Ras/ERK, pathway in vascular endothelial cells of VEPDK1KO mice.

Physiological effects of PDK1 inactivation in endothelial cells

Body weights of VEPDK1KO mice did not differ from those of control littermates at 3 and 6 months of age [Supplemental Table 1 (published on The Endocrine Society's Journals Online web site at http://mend.endojournals.org) and Fig. 2A], and survival up to 12 months of age was not different between VEPDK1KO and control mice. The concentrations of glucose, free fatty acids, and total cholesterol in the plasma of VEPDK1KO mice were comparable to those in control mice at 3 and 6 months of age. Plasma insulin and triacylglycerol (TG) concentrations were similar between the two genotypes at 3 months of age. In contrast, at 6 months, plasma insulin and TG concentrations were reduced in VEPDK1KO mice by approximately 37% (P < 0.05) and 36% (P < 0.05), respectively (Supplemental Table 1).

Fig. 2.

Growth curves and glucose metabolism in VEPDK1KO. A, Growth curves of VEPDK1KO (circles) and control (Con) (squares) mice fed a SD (filled symbols) or HFD (open symbols) (n = 12–15 per group). B, Laser Doppler blood flow of VEPDK1KO (black bars) and control (white bars) (n = 5–8 per group). C and D, Blood glucose after the GTT (C) and ITT (D) in VEPDK1KO mice (filled circles; n = 9) and control mice (open squares; n = 12) fed a SD at 3 months (left panel) or 6 months (right panel) of age. E, Blood glucose after the ITT (1.2 U/kg) for control (open squares; n = 8) and VEPDK1KO (filled circles; n =10) mice fed a HFD at 3 months (HFD3) of age. Data are recorded as the mean ± sem. *, P < 0.05 (ANOVA) vs. corresponding value for control mice fed a HFD. **, P < 0.05 (ANOVA) vs. corresponding value for control mice fed a SD.

Malformation of vascular structures was not detected in VEPDK1KO mice (data not shown). Systolic and diastolic blood pressure as well as heart rate at 3 and 6 months of age did not differ between VEPDK1KO and control mice. Nitric oxide metabolism levels also were similar between the two genotypes, despite ablation of VEGF-induced eNOS phosphorylation in endothelial cells from VEPDK1KO mice (Fig. 1B). Because phosphorylated eNOS plays a critical role during NO synthesis, by directly dilating vessel diameters and improving blood flow, we investigated whether the deletion of PDK1 in vascular endothelial cells could physiologically inhibit the insulin-stimulated elevation of blood flow. We measured blood flow of the lower thigh using a laser Doppler blood flow meter before and after insulin stimulation (Fig. 2B). As expected, insulin (10 mU/g) stimulated blood flow (4.0 ± 0.9 to 5.9 ± 1.4 ml/min · 100 g, P < 0.05) in control mice. Nevertheless, insulin injection failed to stimulate blood flow (4.3 ± 2.1 before and 4.0 ± 1. 2 ml/min · 100 g after, P < 0.05) in VEPDK1KO mice. These data suggested that the PDK1 and PI3K signal in vascular endothelial cells regulates blood flow after insulin injection.

Glucose metabolism in VEPDK1KO mice

Initially, we presumed that inhibition of the insulin signal by PDK1 ablation in vascular endothelial cells would aggravate systemic insulin sensitivity. However, VEPDK1KO at 6 months of age showed a significant decrease in plasma insulin and TG concentrations compared with controls. Then, we further evaluated the glucose metabolism of VEPDK1KO.

Although at 3 months of age, no significant difference was observed between VEPDK1KO and control mice in a glycemic profile after an ip glucose loading (1 g/kg) (Fig. 2C), contrary to our expectation, blood glucose concentrations were significantly lower in 6-month-old VEPDK1KO mice at 15 and 30 min (P < 0.05) after glucose loading compared with control mice (Fig. 2C). In an insulin tolerance test (ITT), blood glucose concentrations in VEPDK1KO mice were also reduced significantly throughout the test at 6 months of age. In contrast, at 3 months of age, there were no differences between VEPDK1KO and control mice (Fig. 2D). In a euglycemic-hyperinsulinemic clamp test at 6 months of age, the glucose infusion rate required to maintain euglycemia was significantly higher in VEPDK1KO than in control mice by 41%. In contrast, insulin-stimulated whole-body glucose uptake was not significantly different between the two genotypes (Fig. 3A).

Fig. 3.

Euglycemic-hyperinsulinemic clamp study and hepatic insulin signaling in VEPDK1KO. A, The glucose infusion rate (GIR) and insulin-stimulated whole-body glucose disappearance (Rd) in control (white bars) and VEPDK1KO (black bars) mice. B, Insulin-stimulated 2-deoxyglucose (2DG) uptake in cerebellum (Cerebel), Gastro, soleus, extensor digitorum longus (EDL), heart, BAT, eWAT, and sc WAT (sWAT) in control (white bars) and VEPDK1KO (black bars) mice. C, Percent suppression of basal hepatic glucose production in control (white bars) and VEPDK1KO (black bars) mice during the hyperinsulinemic-euglycemic clamp study. D, PEPCK and G6Pase gene expression in the liver after the hyperinsulinemic-euglycemic clamp. Con, Control. E, Immunoblot analysis of the tyrosine-phosphorylated β-subunit of insulin receptors (pIR-β), total insulin receptor β-subunit (IR-β), Akt phosphorylated at Ser473 (pAkt), and total Akt in total cell lysates isolated from the liver after an iv insulin load (10 U/kg) for 5 min. HFD3, HFD at 3 months; SD3, SD at 3 months. Data are recorded as the mean ± sem of eight animals per genotype. *, P < 0.05 vs. control littermates.

Insulin-stimulated glucose uptake in the cerebellum, skeletal muscle, heart, and BAT, as well as in epididymal and sc WAT, was not significantly different between the two groups (Fig. 3B). In contrast, insulin-induced suppression of hepatic glucose production was significantly greater, by approximately 2.4-fold, in VEPDK1KO than in control mice, although basal hepatic glucose output did not differ between genotypes (82 ± 14 in VEPDK1KO vs. 87 ± 14 μmol/kg · min in control; n = 8; P value not significant) (Fig. 3C). A major mechanism by which insulin inhibits hepatic glucose output is the regulation of expression of a number of gluconeogenic genes, such as phosphoenolpyruvate carboxykinase (PEPCK) and glucose-6-phosphatase (G6Pase) (15, 16). In VEPDK1KO mice, expression of both PEPCK and G6Pase was significantly reduced, by 37 and 32%, respectively (Fig. 3D). Hepatic expression of PEPCK and G6Pase genes in the basal (nonclamped) state was similar between the two genotypes (data not shown). These results suggest that the unanticipated enhancement of insulin sensitivity observed in VEPDK1KO mice at 6 months of age was due mostly to suppression of hepatic glucose output.

Effects of a high-fat diet (HFD) on insulin sensitivity in VEPDK1KO mice

To confirm the physiological impact of enhanced whole-body insulin sensitivity, the effects of a HFD were investigated. Surprisingly, VEPDK1KO mice were resistant to body weight gain compared with control mice (Fig. 2A). The difference in body weight between the two groups was significant at 3 months of age and gradually increased throughout the observation period. At 6 months of age, VEPDK1KO mice weighed approximately 16% less than control mice (P < 0.05).

Food consumption adjusted for body weight and activity did not differ between groups fed either a standard or a HFD (Fig. 4, A and B). Rectal temperature did not differ between groups (37.2 ± 0.1 in VEPDK1KO vs. 37.1 ± 0.2 C in control, n = 8 each). However, total O₂ consumption (Fig. 4C), CO₂ production (Fig. 4D), and respiratory exchange ratio (RER) measured by indirect calorimetry, was significantly increased during the dark period in VEPDK1KO fed a HFD at 3 months of age (Fig. 4E). These results strongly suggest that obese VEPDK1KO had enhanced systemic energy expenditure compared with obese controls.

Fig. 4.

Food intake, activity, and energy expenditure. A and B, Food intake (A) and activity (B) in control (white bars) and VEPDK1KO (black bars) mice fed a SD or a HFD at 3 months of age (n = 8 per group). C–E, Mean O₂ consumption (C), CO₂ production (D) and RER (E) of control (white bars and squares) and VEPDK1KO (black bars and circles) fed a SD (black bars) or a HFD (white bars) at 3 months of age (n = 8 per group). Con, Control; HFD3, HFD at 3 months; SD3, SD at 3 months. Data are recorded as the mean ± sem. *, P < 0.05 vs. control littermates.

We next examined insulin sensitivity of VEPDK1KO fed a HFD at 3 months of age. The ITT (1.2 U/kg), performed in 3-month-old mice consuming a HFD, clearly demonstrated that blood glucose concentrations in VEPDK1KO mice were significantly lower than in control mice (Fig. 2E). In contrast, whole-body insulin sensitivity in 3-month-old mice fed a standard diet (SD) did not differ between the two genotypes (Fig. 2D). In addition, the phosphorylation of Akt at Ser473 in the liver after iv administration of insulin was increased 2.2-fold in VEPDK1KO mice fed a HFD compared with control mice fed a HFD (Fig. 3E).

Physiological change of adipocytes in VEPDK1KO mice

To further assess parameters involved in body weight reduction, the proportion of the weight of the liver, Gastro, the heart, and WAT to total body weight, was determined in both groups fed either a standard or a HFD at the indicated age. When fed a, the relative tissue weight did not differ between groups at 3 months of age. Those of epididymal WAT (eWAT) in VEPDK1KO mice fed a SD at 6 months of age and a HFD at 3 months of age were significantly lower than control mice, regardless of the tissue weight of the liver, Gastro, and the heart, was similar in both groups (Fig. 5A). The weight ratio of the sc WAT also was lowered in VEPDK1KO fed a HFD at 3 months of age (Fig. 5A).

Fig. 5.

Reduced age-related and diet-induced obesity in VEPDK1KO mice. A, Proportion tissue weights relative to body weight in control (white bars) and VEPDK1KO (black bars) mice (n = 8 per group). B, CT images at the L5 level of control and VEPDK1KO mice (scale bar, 1 cm). Estimated amounts of total fat (TF), visceral fat (VF), and sc fat (SF) in the abdominal area of control (white bars) and VEPDK1KO (black bars) calculated based on the CT images (n = 8–9 per genotype). C, Mean size of adipocytes from eWAT of control (white bars) and VEPDK1KO (black bars) mice (n = 8 per group) (scale bar, 100 μm). D, Mean cell count of adipocytes from eWAT of control (white bars) and VEPDK1KO (black bars) mice (n = 8 per group). Con, Control; HFD3, HFD at 3 months; SD3, SD at 3 months. Data are recorded as the mean ± sem. *, P < 0.05 vs. control littermates.

Total and visceral fat areas were significantly reduced in VEPDK1KO mice compared with control mice at 3 months of age. In contrast, sc fat areas were not different between the two genotypes (Fig. 5B). The average size of adipocytes was decreased in VEPDK1KO mice fed a SD at 6 months of age and those at 3 months of age fed a HFD (Fig. 5C). On the other hand, the number of adipocytes was not altered in both lineages on a SD or on a HFD (Fig. 5D). These results suggest that the reduced WAT weight may be attributable to adipocytes that are smaller size but similar in number in VEPDK1KO fed a HFD.

Secretion and expression of adipokines in eWAT

First, we examined the profiles for secretion and expression of adipokines. The mRNA encoding adiponectin, leptin, TNFα, and monocyte chemoattractant protein 1 (MCP1) in eWAT did not differ between the two groups at 3 months of age when animals consumed a standard diet. However, mRNA expression of adiponectin was increased and that of leptin, TNFα, and monocyte chemottractant protein 1 (MCP1) was decreased at 6 months of age in VEPDK1KO mice fed a standard diet. These results were enhanced when 3-month-old mice were fed a HFD (Fig. 6A). Because of the interest in the relationship between insulin resistance and inflammation of adipocytes, we also determined the mRNA of markers for macrophages in eWAT. At 3 months of age, when fed a HFD, the expression of cluster of differentiation antigen 68 (CD68) and F4/8 mRNA was lower in VEPDK1KO mice than in control mice. A similar result was observed at 6 months of age with a SD (Fig. 6A).

Fig. 6.

Gene expression and adipokines in eWAT of VEPDK1KO mice. A, Amounts of the indicated mRNA in the eWAT of control (white bars) and VEPDK1KO (black bars) (n = 8 per group). B, Serum adiponectin in VEPDK1KO (black bars) and control mice (white bars) (n = 8 per group). C, Serum leptin, MCP-1, and VEGF in VEPDK1KO (black bars) and control mice (white bars) fed a HFD (n = 8 per group). Con, Control; HFD3, HFD at 3 months; SD3, SD at 3 months. Data are recorded as the mean ± sem. *, P < 0.05 vs. control littermates.

At 6 months of age on a SD and 3 months of age on a HFD, serum adiponectin concentrations were slightly but significantly increased in VEPDK1KO as compared with control mice (Fig. 6B). Serum leptin and MCP1 concentrations were reduced in VEPDK1KO compared with control mice, but the difference was not statistically significant (Fig. 6C). Because mature adipocytes secrete VEGF (17), serum VEGF concentrations also were determined. In mice fed a HFD for 3 months, VEGF concentrations were significantly lower in VEPDK1KO mice compared with control mice (40.0 ± 1.3 in VEPDK1KO vs. 48.7 ± 5.6 pg/ml in control; n = 8; P < 0.05) (Fig. 6C).

Hepatic AMP-activated protein kinase (AMPK) activity and fatty acid metabolism in liver and Gastro

AMPK is regulated by adiponectin signaling and might be chiefly involved in the molecular mechanisms underlying the enhancement of whole-body and hepatic insulin sensitivity in VEPDK1KO mice. Initially, we analyzed mRNA expression of adiponectin receptor 1 (AdipoR1) and AdipoR2. Unexpectedly, the expression of AdipoR1 and AdipoR2 was comparable in both groups in the indicated experimental conditions (Fig. 7C).

Fig. 7.

Hepatic AMPK signaling and fatty acid oxidation, gene expression of liver, Gastro, and BAT in VEPDK1KO mice. A, Representative immunoblot analysis of AMPK phosphorylated at Thr172 (pAMPK), total AMPK, ACC phosphorylated at Ser79 (pACC), and total ACC in the liver of VEPDK1KO and control mice. B, Hepatic triglyceride content of VEPDK1KO (black bars) and control (white bars) mice. C, E, and F, Amounts of indicated mRNA in liver (C), Gastro (E), and BAT (F) of VEPDK1KO (black bars) and control mice (white bars). D, Hepatic fatty acid oxidation of VEPDK1KO (black bars) and control (white bars) mice. Con, Control; HFD3, HFD at 3 months; SD3, SD at 3 months. Data are recorded as the mean ± sem. *, P < 0.05 vs. control mice.

Nevertheless, phosphorylation of AMPK and acetyl-coenzyme A (CoA) carboxylase (ACC), involved in fatty acid oxidation, was elevated approximately 2.0- and 6.6-fold, respectively, in VEPDK1KO mice fed a SD at 6 months of age and approximately 3.7- and 11.2-fold, respectively, in VEPDK1KO mice fed a HFD at 3 months of age (Fig. 7B), as we expected. In addition, hepatic TG content was significantly reduced in VEPDK1KO mice fed a HFD at 3 months of age, and mRNA expression of sterol regulatory element-binding protein 1c (SREBP1c), which is controlled by AMPK, was significantly reduced in VEPDK1KO mice when fed a SD at 6 months of age and when fed a HFD at 3 months of age (Fig. 7C). Furthermore, expression of genes related to fatty oxidation including acetyl-CoA oxidase (ACO), medium-chain acyl-CoA dehydrogenase (MCAD), carnitine palmitoyltransferase 1a (CPT1a) and those related to fatty acid synthesis including fatty acid synthase (FAS) and stearoyl-CoA desaturase 1 (SCD1) were determined to evaluate the mechanism of increased insulin sensitivity and enhanced AMPK signaling in the livers of VEPDK1KO mice. Expression of ACO, MCAD, and CPT1a, were up-regulated in the livers of VEPDK1KO on a SD at 6 months of age and on a HFD at 3 months of age. Expression of ACO and CPT1a in liver were also higher in VEPDK1KO mice fed a SD at 6 months of age. In addition, gene expression of hepatic FAS was higher in VEPDK1KO only when mice were fed a HFD. SCD1 expression did not differ between the groups (Fig. 7C). On the other hand, expression of genes related to fatty acid oxidation, lipogenesis, and energy expenditure in Gastro were not significantly altered in either group. Gene expression of uncoupling proteins in the liver (UCP2), Gastro (UCP2,3), and BAT (UCP1) also were unaltered between VEPDK1KO and control (Fig. 7, C, E, and F). To determine whether the increased expression of genes related to fatty acid oxidation in the liver of VEPDK1KO mice, which were fed a SD at 6 months of age and a HFD at 3 month of age, directly contribute to increased hepatic fatty acid oxidation, we measured β-oxidation of the liver by using ¹⁴C-labeled palmitate in VEPDK1KO and control mice. Incorporation of [¹⁴C]palmitate into acid-soluble metabolites was increased 109% in VEPDK1KO fed a SD at 6 months of age and 100% in VEPDK1KO fed a HFD at 3 months of age, compared with corresponding control mice (Fig. 7D). These results suggest that fatty acid oxidation in the liver of VEPDK1KO mice was elevated, consistent with the results described above for RT-PCR.

Angiogenesis in the WAT of VEPDK1KO mice

In endothelial cells, the PDK1/Akt signaling pathway mediates angiogenesis (18), and the VEGF signal is inhibited in VEPDK1KO mice (Fig. 1B). Therefore, it was hypothesized that dysregulation of angiogenesis in WAT might be involved in the mechanism by which PDK1 deficiency in endothelial cells alters adipocyte size (Fig. 5C). To determine whether angiogenesis in the VEPDK1KO mice was, in fact, defective, emerging endothelial cells were measured in the transplanted Matrigel containing either VEGF (50 ng/ml) or vehicle. The area of emergent proliferating endothelial cells, labeled with an anti-CD31 antibody in the Matrigel containing VEGF, was significantly increased compared with the VEGF-free Matrigel. On the other hand, endothelial cell area did not differ between VEGF-containing Matrigel and VEGF-free Matrigel transplanted in VEPDK1KO mice (Fig. 8A). These results demonstrate that the proliferation of endothelial cells in vivo was suppressed in VEPDK1KO mice compared with control mice.

Fig. 8.

Lowered angiogenesis in the Matrigel and each tissue in VEPDK1KO. A, The mean area of the CD31-positive cells in the transplanted Matrigel containing VEGF (50 ng/ml) or vehicle in VEPDK1KO (black bars) and control (white bars) (scale bar, 200 μm). B, The representative images and approximate weight of IB-4 lectin-positive cells in eWAT, liver, and Gastro in VEPDK1KO (black bars) and control (white bars) mice fed a HFD (scale bar, 200 μm). C, Expression of the PECAM1 and VCAM1 mRNA in the Gastro and liver of VEPDK1KO (black bars) and control mice (white bars) fed a HFD (n = 8 per group). Con, Control; HFD3, HFD at 3 months; SD3, SD at 3 months. Data are recorded as the mean ± sem. *, P < 0.05 vs. control mice.

The weight of total endothelial cells in bilateral eWAT was determined by measuring the area labeled with an family of five tetrameric type 1 isolecctin (IB-4) lectin, which is a marker of vascular endothelial cells. The proportion of lectin-positive area in WAT was significantly lower in VEPDK1KO than in control mice (Fig. 8B). On the other hand, weights of endothelial cells in Gastro, liver, and BAT (data not shown) did not differ between the groups (Fig. 8B). The mRNA encoding platelet-endothelial cell adhesion molecule 1 (PECAM1) and vascular cell adhesion molecule 1 (VCAM1) were down-regulated in the WAT of VEPDK1KO mice in concert with a reduction in the numbers of endothelial cells and capillary vessels (Fig. 6A). The approximate weight of total endothelial cells and the PECAM1 and VCAM1 gene expression in skeletal muscle and liver was not changed, as evaluated in VEPDK1KO on a HFD at 3 months of age (Fig. 8C). Together, these results suggest that the significant inhibition of angiogenesis in WAT (Fig. 8B), but not liver and skeletal muscle, might be due to a decrease in adipocyte size (Fig. 5C) in VEPDK1KO mice fed a SD at 6 months of age and a HFD at 3 months of age.

Angiogenesis per se is an important factor for the growth and expansion of each tissue. Several vascular growth factors, such as fibroblast growth factor, VEGF, and angiopoietins, comprehensively coordinate neovascularization through the activation of the PI3K/PDK1/Akt pathway in vascular endothelial cells. Although expression of VEGF in eWAT (Fig. 6A), liver (Fig. 7A), and Gastro (Fig. 7D) did not differ between both groups, that of hypoxia-induced factor 1α (HIF1α), a transcription factor that regulates angiogenic factors, was significantly up-regulated only in eWAT but not in the liver and Gastro (Fig. 7D) of VEPDK1KO fed a HFD.

Discussion

The results of the present study demonstrate that prevention of age-related and HFD-induced adipocyte hypertrophy due to suppression of angiogenesis in WAT, together with modulation of adipokine production, contributes to improved glucose metabolism in VEPDK1KO mice (Fig. 2C-E and 3A-D).

The PI3K/Akt pathway in endothelial cells regulates angiogenesis and arterial structure. Endothelial Akt transgenic mice showed increased NO production and prevention of apoptosis in aortic endothelial cells (6). In addition, heterogeneous deficiency of PTEN, a negative regulator of PI3K signaling specific to endothelial cells, increased angiogenesis in the matrigel and in the transplanted tumor tissues (3). Inhibition of angiogenesis in the matrigel in VEPDK1KO mice (Fig. 8A) is consistent with these reports, suggesting that PI3K/Akt/ signaling in endothelial cells is one of the principal regulators of angiogenesis in endothelial cells. Deletion of the PI3K signal in vascular endothelial cells was expected to deteriorate systemic insulin sensitivity, because the insulin signal was inhibited in vascular endothelial cells (Fig. 1C). Unexpectedly, VEPDK1KO at 3 months of age on a SD showed no change in glucose metabolism, but VEPDK1KO at 6 months of age on a SD or at 3 months of age on a HFD did not exhibit hypertrophy of eWAT and improvement in hepatic insulin sensitivity. The relationship between angiogenesis and glucose metabolism was indeed controversial. It was reported that insulin stimulates glucose uptake in cultured endothelial cells via the PI3K and MAPK signals in vitro (19). On the other hand, vascular endothelial insulin receptor knockout mice showed no remarkable change in insulin sensitivity under SD conditions (20). Recent reports have stated that endothelial insulin receptor substrate 2 (IRS2) in knockout mice (eIrs2KO) deteriorated systemic insulin sensitivity due to the suppression of glucose uptake in skeletal muscle but not in liver. Total body weight and adipose tissue weight were not altered in eIrs2KO (21). These findings suggested that insulin signaling in vascular endothelial cells per se might impair systemic insulin sensitivity, although other cytokines or growth factors, which could potentially stimulate the PI3K/PDK/Akt pathway in vascular endothelial cells, are also important for angiogenesis in growing tissues including eWAT. Discrepancy of glycemic metabolism between VEPDK1KO and eIrs2KO may be due to a difference in vascularization in adipose tissue.

The coordination between adipogenesis and angiogenesis during adipocyte development has been well documented both morphologically and cytochemically, and there is evidence indicating that development of adipocytes through adipogenesis is accompanied by vascular growth in adipose tissue. Treatment with TNP-470, a selective angiogenesis inhibitor, resulted in marked vascular remodeling concomitant with adipose tissue loss and weight reduction in mice (22). Furthermore, targeted induction of apoptosis in the vasculature of adipose tissue using prohibitin, a proapoptotic peptide, reversed obesity by ablation of WAT (23). These two studies reported the improvement of obesity using drugs or peptides. Our present results focused for the first time on the relationship between congenitally ablated angiogenesis and the prevention of obesity. Total body weight and eWAT weight ratio in VEPDK1KO on a HFD was significantly reduced compared with the control mice on a HFD (Figs. 2A and 5A). Significant reduction in eWAT weight was also observed in VEPDK1KO on a SD at 6 months of age; however, total body weight of VEPDK1KO on a SD was not altered at any age. The contribution of change in epididymal fat tissue weight might be too small to affect total body weight. Macrophage infiltration into cell clusters containing stromal cells, endothelial cells, macrophages, and microcapillaries, located either between adipocytes or peripheral to them, appear essential for adipocyte development. The macrophage has other roles in WAT. When the volume of adipose tissue must expand for the storage of redundant dietary energy, macrophages in adipose tissue are thought to accumulate smoothly and secrete VEGF for recruitment of new vascularization (24). In VEPDK1KO mice, F4/80 and CD68, markers of macrophages, were decreased in epididymal adipose tissue, which could mean that distribution of fat to adipose tissues is lower in VEPDK1KO. A further assessment is required to determine whether the deletion of PDK1 in vascular endothelial cells directly inhibits the accumulation of macrophage. Deletion of angiogenesis in WAT might not have directly influenced gene expression related to adipogenesis including PPAR and C/EBP. Delivery of lipid substrates with a vascular system in adipose tissue is necessary to grow lipid droplets in adipose cells. Thus, the results shown in VEPDK1KO mice strongly suggested that angiogenesis and adipogenesis are closely related and that the attenuation of vascular growth should result in the retardation of adipocyte growth and development. On the other hand, the number of adipocytes was unaltered in VEPDK1KO under conditions of SD for 3 and 6 months and HFD for 3 months. As reported in a previous paper, the number of adipocytes in C57BL/6 mice on a HFD was not increased until 30 wk of age (25). Because we fed our mice a HFD and followed up until 3 months (18 wk) of age, it is possible that the number of adipocytes might not change in VEPDK1KO under conditions of suppressed angiogenesis. Expression of PECAM1 and VCAM1 and total weight of endothelial cells were similar in liver and skeletal muscle, suggesting that angiogenesis of these tissues in VEPDK1KO mice was comparable with that in control mice (Fig. 8C). However, the mechanism underlying the difference in angiogenesis in each tissue observed in VEPDK1KO remains unclear. In VEPDK1KO fed a HFD, expression of HIF1α in eWAT was increased when compared with control mice (Fig. 6A) but was similar in liver or Gastro in both types of mice (Fig. 7, C and D). Furthermore, tissue weight of eWAT was dramatically increased (Fig. 5A), as suggested by the increased vascularization in these tissues (Fig. 8B). Weights of liver and skeletal muscle were not increased even when mice were fed a HFD (Fig. 5A). Based on these results, it is hypothesized that, with rapid growth, eWAT may be more sensitive to hypoxia than other tissues such as Gastro and liver. These mechanisms might provide a plausible explanation as to why deletion of PDK1 in vascular endothelial cells easily deteriorated angiogenesis in eWAT. However, additional studies are required to ascertain these results.

The obese VEPDK1KO mice did not gain body weight to the same extent as the control mice despite similar food intake between the two groups. Surprisingly, VEPDK1KO fed a SD at 6 months of age and fed a HFD at 3 months of age demonstrated increased energy expenditure compared with control mice (Fig. 4). Because VEPDK1KO on a SD at 3 months of age showed similar energy expenditure to control mice, PDK1 deletion of vascular endothelial cells might not influence systemic energy metabolism directly. Changes of energy metabolism observed in mildly obese VEPDK1KO should be a secondary effect, possibly related to inhibition of adipogenesis. These results are consistent with the previous paper reporting that energy expenditure was elevated in mice bearing lower angiogenesis by TNP-470 administration (22). Presumably, elevation of circulating adiponectin and increased O₂ consumption, CO₂ production, and RER during a dark period (Figs. 4 and 6B) in VEPDK1KO mice should be sufficient to improve whole-body insulin sensitivity. However, we also considered that the concentration of adiponectin measured in the peripheral blood does not at all reflect that in the portal veins, which is obviously more dynamic and more critical for hepatic metabolism. Thus, we suggest that portal adiponectin concentration would be higher in KO mice than in control mice, directly after a meal in particular. The unchanged peripheral glucose uptake rate in VEPDK1KO mice is essentially consistent with previous studies investigating the effects of elevated circulating adiponectin concentrations. Both the injection of recombinant adiponectin (26, 27) and the transgenic overexpression of adiponectin (28) improved hepatic insulin sensitivity without altering glucose uptake in peripheral tissues in a euglycemic-hyperinsulinemic clamp test. The insulin-sensitizing effects of adiponectin in skeletal muscle and liver have now been clearly established (26–30). AdipoR1 chiefly activates AMPK, which inhibits gluconeogenesis and stimulates fatty acid oxidation through SREBP1c in the liver, whereas AdipoR2 is associated with peroxisome proliferator-activated receptor activation (31). Presumably, the major mechanism underlying the inhibition of adipose tissue hypertrophy and enhanced insulin sensitivity in VEPDK1KO mice, which was characterized by suppression of hepatic glucose production and hepatic TG content in concert with reduced expression of hepatic PEPCK, G6Pase, and SREBP1c (Figs. 3D and 7C), may be activation of hepatic AMPK by elevated serum adiponectin concentration. As expected, phosphorylation of AMPK and ACC in the liver was increased in VEPDK1KO mice (Fig. 6B). The unaltered serum free fatty acid concentration shown in VEPDK1KO mice (Supplemental Table 1) is thought to be consistent with stimulated hepatic fatty acid consumption. In addition, expression of genes related to fatty acid oxidation was significantly increased in liver but not in skeletal muscle of VEPDK1KO on a HFD. Increased fatty acid β-oxidation in mitochondria leads to the production of two molecules of CO₂ because acetyl-CoA is used in the tricarboxylic acid cycle. These phenomena are consistent with higher O₂ consumption, CO₂ production, and RER, particularly during dark periods, as measured by indirect calorimetry in VEPDK1KO fed a HFD. Nevertheless, food intake and activity were unaltered in both groups (Fig. 4). In addition, the level of expression of uncoupling protein in BAT in VEPD1KO was not sufficient to repress obesity (Fig. 7). Taken together, we suggest that fatty acid, not delivered to adipose tissue, would be processed immediately in the liver in VEPDK1KO under conditions of HFD or SD for 6 months. However, further investigation would be required for a thorough understanding of the relationship between increased circulating adiponectin with hepatic AMPK activation and food consumption and energy expenditure.

In conclusion, the present observations in VEPDK1KO mice are consistent with the emerging concept of the close relationship between angiogenesis and adipogenesis and demonstrate the significance of PDK1 signaling in endothelial cells to maintain proper glucose homeostasis through regulation of adipocyte development.

Materials and Methods

Mice

VEPDK1KO were generated by breeding PDK1^flox/flox mice (32), which harbor a modified endogenous PDK1 gene in which exons 3 and 4 are flanked by loxP sites, with mice that express the Cre recombinase gene under the control of the Tie2 gene promoter (Tie2-Cre) (13). The heterozygous offspring of both the loxP-targeted PDK1 gene (PDK1^flox/+) and the Tie2-Cre transgene (PDK1^flox/+Tie2Cre^+/−) were then crossed with PDK1^flox/flox mice to generate mice homozygous for the Tie2-Cre transgene and for the PDK1 floxed allele (PDK1^flox/floxTie2Cre^+/−). PDK1^flox/floxTie2Cre^+/− and PDK1^flox/floxTie2Cre^−/− were interbred to generate male VEPDK1KO and control mice. These mice were all bred from the C57/BL6 mouse.

Animals were housed in a 12-h dark, 12-h light cycle at a controlled temperature and were allowed free access to water and a SD (Oriental Yeast Co., Tokyo, Japan). In the HFD experiments, mice were allowed free access to the HFD (30% fats, 33% proteins, 25% carbohydrates, and 12% vitamins and minerals by weight; Oriental Yeaset) from the age of 4 wk to the time of euthanasia. Food intake and O₂ consumption, CO₂ production, RER (Columbus, Columbus, OH) were measured daily for 1 wk at the indicated age. Activity levels were measured by activity sensors (FDM 300; Melquest, Toyama, Japan) Rectal temperature was taken by thermometer every other day from 21–22 wk of age.

All aspects of animal care and experiments were performed at the Laboratory Animal Center with the approval of the Animal Research Committee at the Kawasaki Medical School.

Blood pressure measurement and biochemical assays

Blood pressure was measured in conscious animals with a tail-cuff blood pressure monitor (Muromachi Kikai Co., Tokyo, Japan). Blood glucose was measured using a Free Style Kissei glucose meter (Kissei Pharmaceutical Co. Ltd., Tokyo, Japan). Plasma insulin (Morinaga, Tokyo, Japan), nitric oxide, adiponectin, leptin, MCP1, and VEGF (R&D Systems, Minneapolis, MN) were measured using ELISA kits. Serum free fatty acids, TG, and cholesterol were measured using assay kits (Wako, Osaka, Japan). Liver TG content was determined using a kit (Wako) according to the manufacturer's instructions.

Glucose metabolism in vivo

The glucose tolerance test was performed by ip injection of glucose (1 g d-glucose/kg) after an overnight fast. The ITT was performed by ip injection of human regular insulin (Novo Nordisk, Bagsvaerd, Denmark) after a 5-h fast. A 120-min euglycemic-hyperinsulinemic clamp was conducted after an overnight fast, while animals were conscious, as previously described (33). Briefly, a primed-continuous infusion of human regular insulin at a rate of 15 pmol/kg · min raised plasma insulin to within the physiological range (∼780 pmol/liter), and 40% glucose was periodically infused at variable rates to clamp plasma glucose at approximately 6.5 mmol/liter. [3-³H]Glucose (3.7 × 10⁵ Bq bolus, followed by 3.7 × 10³ Bq/min; GE Healthcare Biosciences Co., Piscataway, NJ) was infused throughout the clamp to estimate insulin-stimulated whole-body glucose flux. Insulin-simulated glucose uptake in individual tissues was estimated by administration of a bolus (3.7 × 10⁵ Bq) of 2-deoxy-d-[1-¹⁴C]glucose (GE Healthcare Biosciences). At the end of the study, tissues were collected and frozen in liquid nitrogen for subsequent analysis.

Measurement of fatty acid oxidation

Fatty acid oxidation was measured in fresh liver tissue using the palmitate oxidation method as previously described (34). The rates of palmitate oxidation were calculated from acid-soluble metabolites. Briefly, 150 mg liver was homogenized in Sugar-Tris-EDTA buffer containing 0.25 m sucrose, 10 mm Tris, 1 mm EDTA, 1 m trichostatin A, and 5 mm nicotinamide with half-maximal speed and was incubated with reaction buffer containing 100 mm sucrose, 10 mm Tris-HCl, 5 mm KH₂PO₄, 0.2 mm EDTA, 5 mm nicotinamide, 1 m trichostatin A, 0.3% fatty acid-free BSA, 80 mm KCl, 1 mm MgCl₂, 2 mml-carnitine, 0.1 m malate, 0.05 mm CoA, 0.5 mm palmitate, and 0.4 μCi [¹⁴C]palmitate (GE Healthcare Bio-sciences) for 1 h at 37 C. The reaction mixture was transferred into a tube containing 1 m HCl and a filter paper immersed with 1 m NaOH in the tube cap and was incubated for 1 h at 37 C. Finally, acid-soluble metabolites containing oxidized long-chain fatty acid was measured by liquid scintillation counter using scintillation cocktail.

Analysis of sc and visceral fat area by computed tomography (CT) scan

An abdominal transversal CT image at the level of the fifth lumbar vertebrae was analyzed for measurement of sc, visceral, and total fat areas (LaTheta; Aloka, Tokyo, Japan).

Preparation of MLEC

After 8- to 12-wk-old control or VEPDK1KO male mice were euthanized by cervical dislocation, the lungs were excised and digested in 0.1% type 1 collagenase (Worthington Biochemical Co., Lakewood, NJ) for 1 h, followed by filtration using a 140-μm sieve. The cell suspension was plated on endothelial cell medium, and MLEC were isolated by two-step immunoselection with antimouse CD16/CD32 and antimouse CD102 monoclonal antibodies (BD Biosciences Pharmingen, San Jose, CA) (35). To evaluate VEGF or insulin signals, MLEC in passage 2 or 3 of the primary culture were starved overnight and stimulated with either 50 ng/ml recombinant human VEGF (R&D Systems) or 100 ng/ml human insulin (Dako, Carpinteria CA) for 10 min.

Angiogenesis in vivo

Matrigel (BD Biosciences Pharmingen) containing VEGF (50 ng/ml) and heparin (100 U/dl) was injected into the right inguinal sc area of 8-wk-old female mice fed a normal chow diet. After 3 wk, frozen tissues were sliced every 30 μm, frozen sections were incubated with CD31 antibody (BD Biosciences Pharmingen), and areas of CD31-positive cells were enumerated manually. For immunofluorescent staining with IB-4 lectin (Invitrogen) of eWAT, liver, and Gastro tissues were stained with 20 μg/ml IB-4 lectin (Invitrogen) overnight at 4 C. The weight of total endothelial cells in each tissue was calculated by the ratio (percent) of the surface area stained with lectin in one visual field (×400) multiplied by the total weight (in milligrams) of eWAT, Gastro, and liver.

Laser Doppler blood flow meter

Control mice and VEPDK1KO were anesthetized with isoflurane, and blood flow was measured in the left lower thigh using a laser Doppler blood flow meter (Advance Co. Ltd, Kyoto, Japan) before and after insulin stimulation. Blood flow was measured before and 10 min after ip injection of insulin (10 mU/g).

Western blot

Tissue detergent extracts were prepared by homogenization in lysis buffer. After electrophoresis of each lysate (10–30 μg) on SDS-PAGE, proteins were transferred to a polyvinylidene difluoride membrane, which was blotted overnight with primary antibody at 4 C. After blotting with the appropriate secondary horseradish peroxidase-labeled antibodies for 1 h at room temperature, signals were detected using ECL Plus Western Blotting Detection Reagents (GE Healthcare Biosciences) by chemiluminescence. The anti-PDK1 antibody was purchased from Merck Biosciences (San Diego, CA). Other primary antibodies (phospho-Akt at Ser473, phospho-Akt at Thr308, Akt, phospho-S6K at Thr389, S6K, phospho-eNOS at Ser1177, eNOS, phospho-ERK1/2 at Thr202/Tyr204, ERK1/2, phospho-AMPKα at Thr172, AMPKα, phospho-ACC, and ACC) were purchased from Cell Signaling Technology (Danvers, MA). Insulin signaling was determined in liver isolated from mice deprived of food for 16 h and iv injected with 1 mU/g human regular insulin (Novo Nordisk) for 5 min.

RNA preparation and quantitative RT-PCR method

Total RNA was extracted using an RNeasy lipid tissue mini kit (QIAGEN, Valencia, CA) according to the manufacturer's instructions. cDNA was produced from mRNA (2 μg) using TaqMan reverse transcription reagents (Applied Biosystems, Foster City, CA). Quantitative RT-PCR was performed using a 7500 Real-Time PCR system (Applied Biosystems). The relative abundance of mRNA was calculated with 36B4 mRNA as the invariant control. The primers were purchased from Takara Bio Co. (Shiga, Japan).

Statistics

Results are expressed as mean ± sem. Differences were tested for statistical significance using Mann-Whitney's U test and/or ANOVA test.

Acknowledgments

We thank Fumiko Sakamoto and Miho Kobayashi for their assistance with genotyping, RT-PCR, and cell culture.

This work was supported by a Grant-in-Aid from the Japan Society for the Promotion of Science (to K.Ka. and K.Ko.; 18591008 and 18591007), by Research Project Grants from the Kawasaki Medical School (to K.Ka., K.Ko., M.H., and K.T.; 17-502, 17-504, 18-501, 19-501, 19-502, 19-504, 20-505, and 21-501) and Grant-in-Aid for Mitsubishi Pharma Research Foundation, by the Takeda Science Foundation (to K.Ko. and K.Ka.), and by the promotion of insulin research from Novo Nordisk Pharma, Ltd. (to M.H.).

Disclosure Summary: The authors have nothing to disclose.

Abbreviations

 
• ACC

  Acetyl-coenzyme A carboxylase

 
• AdipoR

  adiponectin receptor

 
• AMPK

  AMP-activated protein kinase

 
• BAT

  brown adipose tissue

 
• CoA

  coenzyme A

 
• CT

  computed tomography

 
• eNOS

  endothelial nitric oxide synthase

 
• eWAT

  epididymal WAT

 
• Gastro

  gastrocnemius muscle

 
• HFD

  high-fat diet

 
• ITT

  insulin tolerance test

 
• MCP1

  monocyte chemoattractant protein

 
• MLEC

  mouse lung vascular endothelial cells

 
• PDK1

  3-phosphoinositide-dependent protein kinase-1

 
• PI3K

  phosphatidylinositol 3-kinase

 
• RER

  respiratory exchange ratio

 
• SD

  standard diet

 
• S6K

  p70 ribosomal S6 kinase

 
• SREBP1c

  sterol regulatory element-binding protein 1c

 
• TG

  triacylglycerol

 
• VEGF

  vascular endothelial growth factor

 
• VEPDK1KO

  vascular endothelial-specific PDK1 knockout

 
• WAT

  white adipose tissue.