Effect of Endothelium-Specific Insulin Resistance on Endothelial Function In Vivo

Endothelial cell–specific transgene overexpression was achieved using the murine Tie2 promoter and intronic enhancer as previously described (19). A Thr¹¹³⁴ mutation was introduced into cDNA encoding the human insulin receptor by site-directed mutagenesis (15). The plasmid pHHNS was a gift from Keith Channon (University of Oxford, Oxford, U.K.). After excision of the LacZ sequence with SbfI and MluI (19), the Thr¹¹³⁴ mutant human insulin receptor cDNA was cloned into pHHNS (Fig. 1A). Plasmid fidelity and orientation were confirmed using standard DNA sequencing and PCR analysis.

The transgene Tie2-hIR was excised with SalI, purified, and microinjected into fertilized eggs from superovulated C57BL/6 × CBA mice. Potential founders were screened by genotyping of genomic DNA from tail lysates using primers specific for the human insulin receptor (forward, 5′-GGT GGC AGC TTT CCC CAA CAC T; reverse, 5′-AGC CTT GGC TTC AGG CAT GGT C-3′).

Animals were backcrossed eight times onto a C57BL/6J background. All studies were comparisons of male ESMIRO mice aged 8–10 weeks and age- and sex-matched wild-type littermates. Morphological assessment was performed in 8- to 10-week-old male mice (20). Two lines of ESMIRO mice were generated, and experiments were performed in both lines to exclude a positional effect of transgene integration into the murine genome.

Total RNA was extracted from a variety of tissues using an RNeasy Mini kit (Qiagen). Equal quantities of RNA were reverse transcribed using Superscript II RT (Invitrogen) and random decamer oligonucleotides. To assess transgene expression, PCR was performed using primers specific for the human insulin receptor (forward, 5′-GGT GGC AGC TTT CCC CAA CAC T; reverse, 5′-AGC CTT GGC TTC AGG CAT GGT C-3′).

Real-time PCR analysis of endothelial NO synthase (eNOS), NOX2, NOX4, and β-actin expression was performed (eNOS forward, GGGAAAGCTGCAGGTATTTGAT, and reverse, CACTGTGATGGCTGAACGAAGA; Nox2 forward, ACTCCTTGGGTCAGCACTGG, and reverse, GTTCCTGTCCAGTTGTCTTCG; Nox4 forward, TGAACTACAGTGAAGATTTCCTTGAAC, and reverse, GACACCCGTCAGACCAGGAAT; and β-actin forward, CGTGAAAAGATGACCCAGATCA, and reverse, TGGTACGACCAGAGGCATACAG) (20). The standard curve method was used, and results are normalized to β-actin expression.

Thoracic aorta was excised and mounted vertically in OCT embedding matrix (CellPath, Powys, U.K.) before freezing in cold isopentane. Eight-micrometer transverse sections were fixed in paraformaldehyde (4% in PBS, 30 min). A human-specific anti-insulin receptor antibody (1 in 200, 30 min; E5844; Spring Bioscience, Fremont, CA) was used, which does not cross-react with the murine insulin receptor. A Vector M.O.M. immunodetection kit (Vector Laboratories, Burlingame, CA) reduced background staining of endogenous mouse immunoglobulins. Antibody staining was visualized using a secondary antibody conjugated to fluorophores (Alexa Fluor Dyes; Molecular Probes). Aortic sections were also probed with anti–von Willebrand factor antibody (Sigma-Aldrich, Gillingham, U.K.).

Fasted mice were anesthetized. An injection of either human recombinant insulin (5 units; Actrapid; NovoNordisk, Bagsvaerd, Denmark) or saline vehicle (0.9% NaCl) was administered via the inferior vena cava. After 5 min, the thoracic aorta was excised. Immunoblotting was performed on total aortic tissue homogenates using mouse monoclonal antibodies against total murine eNOS and phospho-eNOS (Ser¹¹⁷⁷) or a rabbit polyclonal antibody against β-actin (20).

Intraperitoneal glucose and insulin tolerance tests were performed in conscious fasted mice. Plasma insulin was measured by enzyme-linked immunoassay using mouse insulin standards. Fasting triglycerides and fasting free fatty acids were measured by colorimetry (20,21). Homeostasis model assessment of insulin resistance (HOMA-IR) was calculated from fasting insulin and glucose measurements.

Systolic blood pressure was measured using tail cuff plethysmography (20,21).

Vascular function was studied in aortic rings mounted in an organ bath as described previously (20,21). Cumulative dose responses to phenylepherine (1 nmol/l to 10 μmol/l) were measured. Phenylepherine constriction responses were reassessed after incubation with insulin (100 mU/ml; 2 h; Actrapid) (20). Relaxation responses to acetylcholine (ACh; 1 nmol/l to 10 μmol/l), A23187 (1 nmol/l to 1 μmol/l), and sodium nitroprusside (SNP; 0.1 nmol/l to 1 μmol/l) were then measured. ACh relaxation experiments were repeated after incubation with the SOD mimetic Mn(III)tetrakis(1-methyl-4-pyridyl) porphyrin pentachloride (MnTMPyP; 10 μmol/l; 30 min) (22).

Six mouse hearts were used for each preparation of coronary microvascular endothelial cells (CMECs) as described previously (23,24).

Three different methods were used to detect superoxide. First, in situ reactive oxygen species (ROS) generation by CMECs was assessed using dihydroethidium (DHE) fluorescence (2 μmol/l; 5 min) as previously reported (24,25). CMECs were grown to 70% confluence before exposure to DHE. Peak fluorescence intensity per cell was quantified microscopically using a computerized image analysis system (Improvision) from at least 30 cells per group; n = 6 mice per CMEC isolation.

Second, 5 μmol/l lucigenin-enhanced chemiluminescence was performed on homogenates of CMECs or aorta to assess NADPH-dependent superoxide production (22,23). In some experiments, one of the following agents was preincubated for 10 min: 20 mmol/l tiron, 10 μmol/l diphenyleneiodonium (DPI), or 100 μmol/l NG-nitro-l-arginine methyl ester (l-NAME).

Finally, functional evidence for the generation of superoxide was obtained by assessing the effect of an SOD mimetic (MnTMPyP) on aortic relaxation responses to ACh in the ESMIRO model (24).

Data are expressed as means ± SE. In vascular studies, concentration-response relationships were compared using two-way repeated-measures ANOVA. Half-maximal effective concentration (EC₅₀) and maximal response (Emax) were compared using a Student's t test. One-way ANOVA was used to analyze chemiluminescence data (Newman-Keuls multiple comparison test for post hoc analysis). Other variables were compared using Student's t test. P < 0.05 taken as statistically significant.

Endothelial cell–specific expression of mutant Thr¹¹³⁴ receptors was achieved using a Tie-2 promoter/intronic enhancer construct (26). Two lines of transgenic mice were generated. Semiquantitative RT-PCR performed on mRNA isolated from a range of organs from ESMIRO mice demonstrated that transgene expression was greater in endothelial-rich tissues such as lung and to a lesser extent kidney as opposed to liver, spleen, and heart (Fig. 1C). Expression of mutant human insulin receptor protein was confirmed by immunohistochemistry on sections of thoracic aorta using an antibody specific for the human insulin receptor (Fig. 1D). Positive staining was found at the luminal endothelial surface of ESMIRO mouse aorta but not wild type. The transgene protein colocalized with endothelial cells as labeled with the endothelial marker, von Willebrand factor (data not shown).

ESMIRO mice and their wild-type littermates appeared morphologically similar. No differences in fertility, lifespan, or behavior were observed. Whole-body weight, mean organ weights, and markers of adiposity were similar in ESMIRO and wild-type mice (data not shown).

It has been suggested that endothelial cell insulin resistance might promote global metabolic insulin resistance secondary to decreased insulin-stimulated blood flow to insulin-sensitive tissues (27–29). However, metabolic characterization of the ESMIRO mice did not support such a mechanism. ESMIRO and wild-type mice demonstrated similar fasting blood glucose levels (6.5 ± 0.2 and 6.7 ± 0.4 mmol/l, n = 8) and fasting serum insulin levels (0.34 ± 0.03 and 0.30 ± 0.03 ng/ml, n = 8) (Table 1). Glucose tolerance testing demonstrated no difference between the two groups (n = 11, P > 0.05; data not shown). Insulin tolerance testing suggested a trend toward improved insulin sensitivity in ESMIRO mice (n = 11, P = 0.05 by repeated-measures ANOVA; data not shown). Serum insulin levels rose equally in response to an intraperitoneal glucose challenge (ESMIRO, 0.78 ± 0.10 and wild type, 0.72 ± 0.06 ng/ml). HOMA-IRs were similar in ESMIRO and wild-type mice (2.1 ± 0.2 and 2.4 ± 0.6, respectively; Table 1). Fasting serum triglycerides (0.80 ± 0.08 and 0.98 ± 0.14 mmol/l, respectively) and free fatty acids (0.92 ± 0.1 and 1.0 ± 0.2 mmol/l, respectively) were similar in both groups.

It has been previously suggested that insulin resistance at the vascular level may lead to hypertension (30). However, ESMIRO and wild-type mice had similar systolic blood pressures (Table 1; 131 ± 4 and 129 ± 4 mmHg, n = 8).

Isolated aortic rings of ESMIRO mice and wild-type mice had similar constrictor responses to 40 mmol/l KCl (Emax 0.68 ± 0.04 and 0.73 ± 0.06 g, ESMIRO and wild type, respectively) and phenylepherine (Fig. 2A; ESMIRO, Emax 0.84 ± 0.06 g, EC₅₀ 227 ± 43 nmol/l; and wild-type, Emax 0.87 ± 0.05 g, EC₅₀ 242 ± 26 nmol/l).

Acute exposure to insulin blunts the vasoconstrictor response to phenylepherine through an endothelial- and NO-dependent mechanism (20). Consistent with this, exposure of wild-type aorta to insulin caused a significant blunting of maximal phenylepherine-induced constriction (compare Fig. 2A and B). However, this effect was lost in ESMIRO mice (Fig. 2B; phenylepherine plus insulin Emax 0.83 ± 0.07 and 0.59 ± 0.06 g, ESMIRO and wild type, respectively; P < 0.05).

The endothelial-dependent vasodilator effect of insulin involves Akt-mediated phosphorylation of eNOS on residue Ser¹¹⁷⁷, resulting in enhanced NO production (31). To investigate the effect of transgene expression on this effect of insulin on endothelial cells, we studied eNOS protein expression and Ser¹¹⁷⁷ eNOS phosphorylation in the thoracic aorta of ESMIRO and wild-type mice exposed to insulin in vivo. Basal levels of eNOS protein expression were similar in ESMIRO and wild-type mice (Fig. 2C and D). After administration of insulin, wild-type mice demonstrated a significant increase in the phospho-eNOS–to–eNOS ratio (1.00 vs. 1.78 ± 0.15, P < 0.05); this effect was absent in ESMIRO mice (0.63 ± 0.1 vs. 0.61 ± 0.1, NS). Therefore, overexpression of the mutant Thr1134 insulin receptor leads to impaired endothelial insulin signaling. A similar result was demonstrated in complementary experiments performed using lung tissue (data not shown).

Vasorelaxation to the agonist ACh was blunted in ESMIRO mice compared with wild type (Fig. 3A; Emax 68 ± 6 and 102 ± 6%, ESMIRO and wild type, respectively, P < 0.01; EC₅₀ 180 ± 73 and 114 ± 37 nmol/l, ESMIRO and wild type, respectively, NS). Similarly, vasodilator responses to the calcium ionophore A23187 were also blunted (Fig. 3B; Emax 51 ± 12 and 79 ± 8%, ESMIRO and wild type, respectively, P < 0.05; EC₅₀ 119 ± 29 and 39 ± 8 nmol/l, ESMIRO and wild type, respectively, P < 0.01). However, responses to SNP were similar in ESMIRO and wild-type mice (Fig. 3C; Emax 109 ± 1 and 110 ± 2%, ESMIRO and wild type, respectively, NS; EC₅₀ 23 ± 6 and 17 ± 2 nmol/l, ESMIRO and wild type, respectively, NS). These data demonstrate that the ESMIRO exhibits significant endothelial dysfunction.

A similar vascular phenotype was seen in a second line of ESMIRO mice (data not shown). These results exclude a positional effect after transgene integration into the murine genome.

A common mechanism of impaired endothelium-dependent vasodilatation is an increased production of ROS (32–35). We therefore assessed the effects of a SOD mimetic on ACh-induced relaxation in aortae. MnTMPyP restored vasorelaxation responses to ACh in ESMIRO aortae (Fig. 4A; Emax 83 ± 5 and 86 ± 7%, ESMIRO and wild type, respectively, NS), consistent with a role for increased aortic superoxide production in mediating the endothelial dysfunction seen in ESMIRO mice.

We assessed NADPH-dependent superoxide production by homogenates of thoracic aortae using lucigenin-enhanced chemiluminescence. Consistent with the functional findings in aortic vasodilatation experiments, ESMIRO aortae demonstrated increased superoxide production compared with wild type (Fig. 4B; 12.6.0 ± 0.5(×106) vs. 8.1 ± 0.4(×106) integrated light units, P < 0.001). ESMIRO and wild-type superoxide production were almost completely abolished by the antioxidant tiron (P < 0.001) or the flavoprotein inhibitor DPI (P < 0.001). The NO synthase inhibitor l-NAME had no effect on NADPH-dependent superoxide production in either group.

DHE fluorescence of cultured CMECs confirmed that ESMIRO mice have increased superoxide at the level of the endothelial cell compared with wild type (2,819 ± 151 and 1,439 ± 129, P < 0.001, n = 6 mice per group, Fig. 5A). This was confirmed by lucigenin-enhanced chemiluminescence performed on CMEC homogenates (Fig. 5B). As in experiments in whole aorta, NADPH-dependent superoxide production in CMECs was inhibited by tiron and DPI but not by l-NAME.

Consistent with the equal eNOS protein expression previously demonstrated in aorta (Fig. 2), there was no difference between groups in eNOS mRNA expression in CMECs (data not shown). However, a significant (approximately twofold) increase in mRNA expression of the NADPH oxidase isoforms Nox2 and Nox4 was seen in ESMIRO aortae (Fig. 6). Upregulation was even more pronounced in CMECs, especially that of Nox2.

The present study demonstrates a number of important findings that add to our understanding of the role of insulin resistance in endothelial cell homeostasis: 1) The induction of endothelial cell–specific insulin resistance leads to markedly reduced NO bioavailability in conduit vessels without significant alteration in global insulin sensitivity. 2) Possible mechanisms include reduced NO bioavailability secondary to increases in aortic and endothelial cell superoxide production. 3) Endothelial cell–specific insulin resistance does not alter systolic blood pressure. 4) Insulin action at the level of the endothelium is not a prerequisite for maintenance of normal blood glucose homeostasis.

To address the role of impaired insulin signaling specifically in the endothelium (as opposed to global impairment in all tissues), we generated a novel murine model of endothelial cell–specific insulin resistance by overexpressing a mutant human insulin receptor in the endothelium (15).

Insulin binding to a normal insulin receptor leads to activation of the tyrosine kinase and a rapid cascade of phosphorylation, leading to activation of multiple intracellular substrates (36). The function of these substrates is to reversibly bridge the activated insulin receptor to a variety of distal signaling molecules. As discussed above, emerging evidence supports eNOS as one of these distal signaling molecules in endothelial cells (37).

Recently, insulin receptors with a mutation where Ala¹¹³⁴ is replaced by Thr in the tyrosine kinase domain of the β-subunit have been found in humans (15). In an affected family, heterozygotes were severely insulin resistant. In in vitro transfection studies, insulin receptor processing, expression on the cell surface, and affinity of insulin binding to the mutant receptor were normal. The ability of insulin to stimulate receptor autophosphorylation and tyrosine kinase activity was, however, substantially impaired. Moreover, cells expressing Thr¹¹³⁴ insulin receptors had blunted insulin-stimulated glucose uptake (16). Moller and colleagues (17) went on to produce tissue-specific transgenic mice overexpressing Thr¹¹³⁴ insulin receptors in skeletal muscle. These mice were a model of nondiabetic insulin resistance and had reduced responses to insulin during insulin tolerance testing and elevated serum glucose and insulin levels when fed ad libitum.

ESMIRO mice expressed mutant human insulin receptors in the endothelium, as confirmed by immunohistochemistry. Functional impairment of endothelial insulin signaling was confirmed in two ways. First, the endothelium-dependent, NO-mediated vasorelaxant action of insulin (to reduce the vasoconstrictor response to phenylepherine) (20) was significantly impaired in aortic rings from ESMIRO mice compared with wild type. Second, and consistent with this finding, insulin-stimulated phosphorylation of eNOS at Ser¹¹⁷⁷ was also blunted in the aortae of ESMIRO mice. However, the ESMIRO had similar metabolic responses to intraperitoneal boluses of insulin and glucose when compared with wild-type littermates, indicating that there was no decline in global insulin sensitivity. Moreover, we detected no differences in body weight, adiposity, plasma lipids, fasting insulin, or HOMA. Despite no difference in metabolic homeostasis, ESMIRO mice had a striking defect in endothelium-dependent relaxation in aorta, attributable to an unfavorable imbalance between the production of NO and ROS. Therefore, there was a significant divergence between the impact of impaired endothelial insulin signaling on vascular function versus glucose homeostasis.

Insulin resistance and type 2 diabetes are now established as important risk factors for the development of cardiovascular atherosclerosis (3,8,38). The alarming increase in type 2 diabetes in children and young adults and the fact that a substantial proportion of patients with type 2 diabetes have coronary artery disease at presentation makes understanding the mechanisms underlying accelerated atherosclerosis in insulin resistance of particular importance. Compelling evidence supports endothelial cell dysfunction as a key early event in the pathogenesis of atherosclerosis (9).

The bioavailability of NO is dependent on the balance between its production by eNOS and its inactivation by ROS. Classical activation of eNOS (e.g., by ACh) involves a rise in intracellular Ca²⁺ and binding of Ca²⁺/calmodulin to the enzyme. Recently, a Ca²⁺-independent regulatory pathway for eNOS stimulated by shear stress or insulin has also been described (31). Both shear stress and insulin increase endothelial NO production via the activation of PI 3-kinase and protein kinase B (PKB/Akt), which phosphorylates eNOS on Ser¹¹⁷⁷ (37). In addition, insulin upregulates eNOS transcription in endothelial cells (39). In the present study, impaired endothelial insulin signaling was accompanied by reduced insulin-mediated vasodilation as might be expected. Interestingly, ACh- and calcium ionophore–stimulated vasodilation were also significantly impaired despite eNOS expression being unaffected. The mechanism underlying this observed endothelial dysfunction appears to be an increase in endothelial cell and aortic superoxide production. In a recent report, we demonstrated that mice with global haploinsufficiency of the insulin receptor (IRKO mice) have blunted insulin-dependent vasodilator responses in aortic rings but preserved vasorelaxant responses to ACh at a young age (20). However, as IRKO mice age, they develop blunting of ACh-mediated aortic relaxation secondary to increased superoxide production (24). The present report indicates a similar mechanism of endothelial dysfunction and suggests that sustained or severe perturbation of insulin signaling at the level of the endothelial cell may lead to increased superoxide generation and therefore a reduction in NO bioavailability.

Although we used a Tie2 promoter construct to direct endothelial cell–specific expression, there is now some evidence that Tie2 may also be expressed within a small subset (2–7%) of circulating blood monocytes (40). This raises the possibility that the increased superoxide demonstrated in whole ESMIRO aortae could result from circulating hematopoietic cells. However, this is unlikely in view of the small population of cells involved and the fact that endothelial dysfunction was confirmed in isolated aorta in the organ bath. Furthermore, both DHE fluorescence and lucigenin-enhanced chemiluminescence performed on isolated endothelial cells from ESMIRO mice confirmed increased NADPH-dependent superoxide production by these cells.

Interestingly, in contrast with the data from Kahn and colleagues (14) and the vascular endothelium insulin receptor knockout (VENIRKO) mouse, total eNOS expression was not downregulated in our model either at mRNA or protein level. There are a number of potential explanations for this difference. The VENIRKO mouse is a conditional knockout model generated using a Cre-lox approach, whereas the ESMIRO mouse presented in the current report demonstrates overexpression of mutant insulin receptors with a dominant-negative action. This mutant receptor has been demonstrated previously to bind insulin normally but to lack tyrosine kinase activity. These fundamental differences between the models may explain the differences observed. It is also unclear from the VENIRKO paper at which age the mice were studied. This may be important because we have recently demonstrated that mice with haploinsufficiency of the insulin receptor are prone to an age-related decline in endothelial function (24).

Three different methods were used to confirm the increased superoxide production in the ESMIRO mouse. Perhaps most notably, a SOD mimetic restored endothelial function through conversion of superoxide to hydrogen peroxide. The source of increased endothelial superoxide in the study does, however, remain to be fully defined. NADPH oxidases are recognized as major sources of ROS in the vasculature (41) and are dysregulated in models of type 2 diabetes, insulin resistance, and obesity (42–45). Several isoforms of these enzymes have been described previously (46), among which Nox2 and Nox4 have been implicated in increased oxidase activity in obesity (47,48) and in insulin signaling in nonendothelial tissues (49). An interesting finding in the current study was the increase in Nox2 and Nox4 mRNA levels in aortae and CMECs of ESMIRO mice, which would be in keeping with a role for NADPH oxidases in the increased superoxide generation. However, in the absence of available good selective inhibitors of these enzymes, definitive evidence of their involvement would require additional studies probably involving appropriate gene-modified models.

The mechanism of Nox2 and Nox4 upregulation in the endothelium requires further elucidation. Recent studies have suggested an association between insulin signaling via the PI 3-kinase pathway and Nox4 expression (49). It has been hypothesized that Nox4 activity may maintain insulin sensitivity. Therefore, in our model Nox4 may be upregulated to compensate for the impaired endothelial insulin signaling. Future studies would also clarify whether Nox2 and Nox4 are upregulated in aortic tissues other than the endothelium.

Finally, the ESMIRO demonstrates a reduced phospho-eNOS–to–eNOS ratio compared with wild type both after stimulation with insulin and at baseline. The latter difference was small. This does raise the possibility that reduced eNOS activity may contribute to the endothelial dysfunction in the ESMIRO irrespective of preserved eNOS expression. This requires further investigation.

Despite a substantial defect of insulin signaling in vascular endothelial cells, we found no demonstrable abnormality in the metabolic phenotype of ESMIRO mice. In fact, consistent with the findings in VENIRKO mice (14), there was a tendency for an improvement in insulin sensitivity during the later stages of insulin tolerance testing. Although a euglycemic insulin clamp may have revealed more subtle abnormalities in glucose homeostasis, our data and those from Kahn and colleagues (14) do not support a nonredundant role for insulin-induced endothelial NO release in blood glucose homeostasis.

The observed phenotype of ESMIRO mice indicates that abnormal endothelium-dependent vasodilatation in the context of insulin resistance may be attributable to impaired insulin signaling at the level of the endothelial cell per se, independent of other abnormalities that are the result of global insulin resistance. Therefore, the integrity of endothelial cell insulin signaling plays a key role in determining conduit vessel endothelial function and NO bioavailability in response to conventional (noninsulin receptor–mediated) agonists.

E.D. was a British Heart Foundation (BHF) clinical PhD fellow. P.C. was supported by Diabetes UK. V.E. was a BHF clinical PhD fellow. A.M.S. holds the BHF Chair of Cardiology at King's College London. M.K. was a BHF intermediate fellow.