Insulin Downregulates the Transcriptional Coregulator CITED2, an Inhibitor of Proangiogenic Function in Endothelial Cells

Coronary artery and peripheral vascular disease are major complications of diabetes. In these atherothrombotic diseases, angiogenesis is an important component in the complex response to acute and chronic ischemia. Diabetes is characterized by an impaired ability of hypoxic tissues to improve perfusion through angiogenesis (1). This impairment is caused by reduced activation of hypoxia-inducible factor (HIF), changes in growth factor signaling, abnormalities in inflammation accompanying the response to ischemia, and changes in mobilization and function of bone marrow–derived proangiogenic cells (1). There is obvious potential benefit from improving perfusion without, or in addition to, surgical intervention. However, despite efficacy of several strategies to improve angiogenesis in animal models of diabetes (1), there is no established clinical pharmacological approach to improve angiogenesis in diabetes.

Insulin resistance is a central feature of type 2 diabetes (2). Insulin can promote angiogenesis by upregulating expression of vascular endothelial growth factor (VEGF) (3,4). In addition, insulin can act directly on vascular endothelial cells to promote angiogenesis. Knockout of the insulin receptor in endothelial cells in a mouse model of oxygen-induced retinopathy caused a pronounced reduction of retinal neovascularization, with the number of preretinal cells reduced by 57% (5). This reduction of retinal angiogenesis is similar in magnitude to the effect of anti-VEGF therapy in the same mouse model (6). Therefore, insulin resistance in endothelial cells may contribute to the impaired neovascular response to ischemia in obesity and type 2 diabetes.

Knowledge about the downstream effectors of the proangiogenic action of insulin on vascular endothelium is mostly limited to regulation of nitric oxide (NO) (7). Increased synthesis of NO (8–11) is one of the most established actions of insulin in endothelial cells, and NO is a critical mediator of angiogenesis (12). Insulin increases endothelial-derived NO through phosphorylation of endothelial NO synthase (eNOS) by Akt and by increasing expression of eNOS (2,9). Consistent with this mechanism, the bioavailability of endothelial-derived NO is decreased in patients with diabetes (13,14).

Insulin also acts as a potent inhibitor of FoxO transcription factors in many cell types, and endothelial cell FoxO has an angiostatic role in postnatal vascular homeostasis and angiogenesis (15,16). We therefore decided to investigate the ability of insulin to regulate 21 direct FoxO targets previously implicated in angiogenesis. These candidates were identified in adult mice harboring an inducible Cre transgene and “floxed” genes for the three major FoxO isoforms, FoxO1, 3, and 4 (16). The triple knockout mice developed hemangiomas in the uterus, liver, and skeletal muscle. The authors of this study hypothesized that genes that were differentially expressed in endothelial cells from hemangioma-prone and hemangioma-resistant vascular beds were driving proangiogenic function (16). Using gene array analysis, gene expression was measured in endothelial cells cultured from hemangioma-prone liver and compared with gene expression in endothelial cells from lung, which was not susceptible to the development of hemangioma. The promoter regions of genes differentially regulated in liver compared with lung endothelial cells were then searched for FoxO target sequences. In the majority of genes selected this way, chromatin immunoprecipitation confirmed that FoxO bound to their promoter region (16). These results prompt the question of how these direct FoxO target genes are regulated by physiological stimuli and disease.

We screened these candidates to determine if they were regulated by insulin in endothelial cells in a FoxO1-dependent manner. With this approach, we identified 10 novel insulin-regulated genes that may modulate angiogenesis in the context of insulin resistance. To confirm the validity of this concept, we further studied one of these genes, CBP/p300-interacting transactivator with ED-rich tail 2 (CITED2), a known inhibitor of HIF-1α transactivation (17). In the following sections, we describe how CITED2 expression in endothelial cells is downregulated by insulin and increased in insulin resistance and type 2 diabetes in animals and humans. We demonstrate that CITED2 inhibits endothelial cell HIF transactivation and propose that inhibition of CITED2 may improve angiogenesis in type 2 diabetes. In addition, the novel insulin-regulated gene program described here holds potential for further exploration of treatment targets in vascular complications of diabetes.

Additional methods are described in the Supplementary Data.

Mice with a floxed mutation in the Foxo1 gene (16) were provided by Dr. Domenico Accili (Columbia University). Conditional knockout mice with gene deletion in vascular endothelial cells were generated by breeding mice with a Cre transgene under control of the VE-cadherin (Cdh5) promoter (stock number 006137; The Jackson Laboratory) (18) with Insr^lox/lox mice (19) or Cited2^lox/lox mice (20). db/db mice and age-matched db/+ mice from the same colony were purchased from The Jackson Laboratory (stock number 000697).

Primary mouse lung, liver, or myocardial endothelial cells were isolated and cultured as described previously (21). MS1 endothelial cells were purchased from the American Type Culture Collection and cultured in DMEM containing 5% FBS. Human umbilical vein endothelial cells (HUVECs) (Lonza) were cultured in EGM-2 medium (Lonza). HUVECs were used between passages two and five.

Human mammary artery tissue was harvested from patients with type 2 diabetes or control subjects without diabetes undergoing coronary artery bypass grafting. Tissue left over from preparing the arterial graft was collected in the operating room and periadventitial tissue quickly removed. Parts of the cleaned vessel were either fixed in formalin or flash frozen. Immuhistochemistry of human arteries was performed on sections of paraffin-embedded tissue using a rabbit polyclonal antibody against CITED2 (catalog no. 7067; ProSci) after antigen retrieval by boiling in citrate buffer.

Adenoviral constructs expressing mouse wild-type FoxO1 or a constitutively active T24A/S253D/S316A FoxO1 triple mutant were provided by Dr. Domenico Accili (22).

Real-time PCR and Western blotting were performed as previously described (21). Primer sequences are shown in Supplementary Table 1. CITED2 Western blotting of mouse and human cells and tissue was performed using a polyclonal rat CITED2 antibody (catalog no. MAB5005; R&D Systems). In experiments where samples were prepared in several batches, the control conditions were set at 100% in individual experiments and therefore have no error estimates.

To overexpress CITED2, HUVECs were transfected with a plasmid containing the full-length cDNA of the human CITED2 gene, a gift from Dr. Shoumo Bhattacharya (17) (plasmid no. 48184; Addgene). The backbone vector pcDNA 3.1 (Invitrogen) was used as a control. A plasmid expressing hypoxia-regulated element (HRE)–luciferase was a gift from Navdeep Chandel (23) (plasmid no. 26731; Addgene). For knockdown of CITED2, a small interfering (siRNA) with the sense sequence 5′-UGC AGA AGC UCA ACA ACC AdTdT-3′ was used. An siRNA targeting a nonmammalian gene (eGFP) was used as a control.

Flow cytometry of endothelial cells in culture was done after trypsin treatment to produce single-cell suspensions, stained with antibodies, and analyzed on an LSRII instrument (BD Biosciences). For FACS, the ventricles of the heart or both gastrocnemius muscles were minced with scalpels and then dissociated for 1 h in DMEM containing 2 mg/mL (560 units/mL) collagenase I (Worthington), 3 mg/mL Dispase (Roche), and 0.2 mg/mL (400 Kunitz units/mL) DNase I (Millipore). Filtered cells were stained with antibodies and sorted on a Moflo Legacy (Cytomation, currently Beckman Coulter) or BD FACSAria II SORP (BD Biosciences). Additional detail is described in the Supplementary Data.

HUVECs were transfected with CITED2 siRNA or plasmid as described above. After 48 h, cells were trypsin treated and 5,000 cells were plated on Matrigel. Tube formation was photographed through a phase-contrast microscope 4 h later. Images from randomly chosen microscopic fields were selected for analysis and analyzed by Angiogenesis Analyzer for ImageJ (http://image.bio.methods.free.fr/) by an observer unaware of culture conditions. The software automatically analyzes components of endothelial cell tubular networks (Supplementary Fig. 5).

Femoral artery ligation was performed as described previously (24) with modifications as explained in the Supplementary Data.

All protocols for animal use and euthanasia were reviewed and approved by the Animal Care Committee of the Joslin Diabetes Center and were in accordance with National Institutes of Health guidelines. The study of discarded arterial tissue from patients undergoing coronary artery bypass surgery was approved by the regional ethical committee in Denmark (Region Syddanmark, ID S-20100044) and a protocol was submitted to the Danish Data Protection Agency (Datatilsynet) as required by Danish law.

Comparisons were made using paired Student t test for the cell culture studies and unpaired Student t test or Wilcoxon rank-sum test for the clinical data and animal studies, with P < 0.05 considered statistically significant. In text and graphs, data are presented as the mean ± SEM.

Our aim was to discover novel insulin-regulated genes in endothelial cells with a role in angiogenesis. In the study by Paik et al. (16) described above, results were derived from analysis of endothelial cells from FoxO1/3/4 knockout mice. However, the described proangiogenic phenotype was prominent in FoxO1 knockout mice, whereas knockout of FoxO3, FoxO4, or both in combination did not result in susceptibility to hemangioma formation. Therefore, we reasoned that genes altered by FoxO1 knockout alone were more likely to have potent angiogenic function.

Endothelial cells were isolated from livers of Foxo1^lox/lox mice by two rounds of immunomagnetic isolation. The cultures were then infected with an adenovirus expressing Cre recombinase or a control adenovirus. The proportion of CD31⁺ cells was 97.0 ± 0.4 and 98.2 ± 0.3%, respectively, in recombined and control cultures (Supplementary Fig. 1A). This demonstrated a high purity of endothelial cells given that CD31⁺ cells costained for other endothelial markers, including VE-cadherin and VEGF receptor 2 (KDR) (Supplementary Fig. 1B and C). Cre-mediated recombination resulted in complete deletion of FoxO1 at the protein level without changes in FoxO3 or FoxO4 protein expression (Fig. 1A). FoxO1 knockout significantly changed expression of Adm, Cited2, Ctgf, Fbn1, Hmga2, Klf6, Noct (also known as Ccrn4l), Pbx1, Spry2, and Tcf4 (Fig. 1C). These candidates were further studied as described below. Among the remaining genes, Ccnd1, Meis1, Mrc1, Sdpr, Selm, and Tsc22d1 were not regulated (Fig. 1C), whereas Bmper, Id1, Pcolce, and Rab34 were regulated in the opposite direction of the change previously reported in cells cultured from Foxo1/3/4 knockout mice (16). Nrep (known as D0h4s114 or P311) was not detected.

In FoxO1-knockout cells, insulin receptor substrate-1 (IRS-1) and IRS-2 mRNA decreased by 73 ± 7 and 57 ± 2% (data not shown), and IRS-1 and IRS-2 protein decreased by 38 ± 12 and 75 ± 6% (Fig. 1B). These results are similar to previously published results describing that IRS-1 and IRS-2 are decreased in aortic endothelial cells from mice with endothelial-specific deletion of FoxO1/3/4 (25). Consequently, a knockout approach would not be useful to prove involvement of FoxO1 in insulin-stimulated signaling. We therefore used adenoviral vectors to express wild-type FoxO1, a constitutively active FoxO1 mutant or a control virus. It seemed important to study a wide time range of exposure to insulin as we expected the effect to be transient and the expression kinetics to differ for each gene. Because it was not feasible to expand primary mouse endothelial cells to include all control conditions at several time points, we used MS1 endothelial cells, which we have previously shown are sensitive to insulin (21). Cultures were serum starved overnight and treated with 10 nmol/L insulin for 2, 4, 8, or 16 h. mRNA expression results are shown in Fig. 2 for the time point with the largest change in the nontransfected condition.

Among the 10 genes selected, all were regulated by insulin (Fig. 2). We considered typical FoxO-dependent regulation to exist if insulin and FoxO caused reciprocal gene regulation and if insulin counteracted a change induced by wild-type FoxO1 overexpression but not one induced by overexpression of the constitutively active FoxO1 mutant. The expression of Adm, Ctgf, and Cited2 was downregulated by insulin in a typical FoxO-dependent manner (Fig. 2A–C). Klf6, Spry2, and Tcf4 were downregulated by insulin but were not changed by FoxO overexpression (Fig. 2E–G). Pbx1 was upregulated by insulin but was not changed by FoxO1 overexpression (Fig. 2I). Insulin also increased Noct, but this regulation may not be mediated through FoxO1 because Noct was increased by both FoxO1 expression and insulin treatment and the mutant FoxO1 did not prevent insulin regulation (Fig. 2H). Fbn1 and Hmga2 were regulated by insulin and ectopic FoxO1 expression in opposite directions, as expected (Fig. 2D and J), but these changes were not concordant with regulation after FoxO1 knockout in liver endothelial cells (Fig. 1C). Selection of genes for further study based on the criteria above is outlined in Supplementary Fig. 1.

Among the genes with typical regulation, Ctgf was downregulated by 80 ± 1% after 2 h, Cited2 by 60 ± 1% after 2 h, and Adm by 54 ± 2% after 16 h (Fig. 2A–C). Ctgf encodes connective tissue growth factor (CTGF), a secreted protein that is a key effector of transforming growth factor-β signaling (26,27). Cited2 encodes a transcriptional coregulator discussed extensively below, and Adm encodes adrenomedullin, a circulating vasodilator peptide hormone that is highly expressed in endothelial cells and vascular smooth muscle cells (28,29).

Ctgf, Cited2, and Adm were all expressed in endothelial cells freshly isolated from gastrocnemius muscle from C57BL/6 mice, although endothelial cell levels of Ctgf were relatively low compared with nonendothelial cells (Supplementary Fig. 2). CTGF and adrenomedullin are secreted by a number of cell types, and the importance of endothelial-derived CTGF and adrenomedullin relative to nonendothelial sources is uncertain. In contrast, the function of CITED2 as a transcriptional coregulator is cell autonomous. In addition, HIF is a key regulator of endothelial biology (30–33) and CITED2 has been shown to regulate HIF in nonvascular cells (17). As such, insulin-regulation of CITED2 in endothelial cells is likely important for endothelial function. Therefore, we proceeded by studying CITED2 to test our paradigm that genes regulated by insulin through FoxO are likely to be dysregulated in type 2 diabetes where vascular insulin signaling is impaired.

Insulin downregulated CITED2 mRNA with a nadir at 2 h in time-course experiments (Fig. 3A). The effect of insulin was potent and concentration dependent, with reductions in CITED2 mRNA of 31 ± 12, 55 ± 6, and 59 ± 7% by 1, 10, and 100 nmol/L insulin for 2 h (Fig. 3B). At the protein level, the maximal effect of insulin occurred after 8 h, where CITED2 was downregulated by 54 ± 13% (Fig. 3C and D).

The insulin receptor antagonist S961 efficiently blocked insulin-induced phosphorylation of Akt (Supplementary Fig. 3A) and prevented reduction of CITED2 protein after insulin treatment (Fig. 3E and F). Insulin-like growth factor-1 (IGF1) at 10 nmol/L reduced CITED2 mRNA by 48 ± 5%, with no additional effect at 100 nmol/L (Fig. 3G). After RNAi-mediated downregulation of IGF1 receptor (IGF1R) protein to 91 ± 4% of control levels (Fig. 3H), insulin reduced CITED2 mRNA as efficiently as in control cells (Fig. 3I). These results demonstrate that although both insulin and IGF1 can reduce CITED2, the effect of insulin is mediated by the insulin receptor as opposed to the IGF1R. Treating endothelial cells with the PI3K inhibitor LY294002 or the Akt inhibitor MK2206 blocked insulin-stimulated downregulation of CITED2 mRNA (Fig. 3J–K). Therefore, insulin-stimulated downregulation of CITED2 is dependent on the insulin receptor, PI3K, Akt, and FoxO1. Exposure to high glucose concentrations for 78 h did not change the magnitude of insulin-induced downreguation of CITED2 compared with an osmotic control or low glucose concentrations (Supplementary Fig. 3B and C).

Next, we examined how CITED2 was regulated in endothelial cells in a vascular bed relevant to the vascular complications of diabetes. We isolated mouse myocardial endothelial cells by immunomagnetic isolation. In myocardial endothelial cells from control mice, insulin decreased CITED2 protein by 48 ± 9%, but in mice with endothelial-specific knockout of the insulin receptor, insulin had no effect (Fig. 4A and B).

Endothelial cell insulin signaling through the PI3K/Akt pathway is impaired in animal models of obesity-associated insulin resistance (34,35). We therefore measured the expression of CITED2 mRNA in freshly isolated endothelial cells from mice with diet-induced obesity and from db/db mice. Cardiac ventricles were enzymatically dissociated and single-cell suspensions separated by FACS. This sorting resulted in efficient isolation of endothelial cell markers; in CD45⁻ CD31⁺ cells from the heart of db/+ control mice, the endothelial cell markers Pecam1 and Kdr were 35.5 ± 1.3-fold and 50.8 ± 0.6-fold higher, respectively, than in CD45⁻ CD31⁻ cells (Supplementary Fig. 4A and B). In CD45⁻ CD31⁺ cardiac cells isolated from mice on a high-fat diet, CITED2 mRNA was increased to 159 ± 6% of the level in mice on a low-fat diet (Fig. 4C). Similarly, in CD45⁻ CD31⁺ cardiac cells from db/db mice, CITED2 mRNA was increased to 187 ± 19% of the level in db/+ mice (Fig. 4D). There was no difference in CITED2 mRNA in CD45⁻ CD31⁻ cells in mice on a low-fat versus a high-fat diet or in db/db versus db/+ mice (Fig. 4C and D).

We next measured expression of CITED2 in vascular tissue from obese patients with type 2 diabetes compared with age-matched patients without obesity or diabetes. Insulin signaling through the PI3K/Akt pathway is impaired in patients with metabolic syndrome with insufficient compensation by hyperinsulinemia (2). We therefore expected increased expression of CITED2 in these patients. We collected mammary arteries from five obese patients with diabetes and four patients without obesity or diabetes undergoing coronary artery bypass surgery (Table 1). Patients in both groups had an average age of 61 years (Table 1). BMI was 35.9 ± 3.4 kg/m² in the group of patients with diabetes and 26.2 ± 0.9 kg/m² in control subjects. Lipid profiles were similar between both groups, consistent with the fact that all patients in both groups were treated with statins. There was a nonsignificant trend toward higher systolic and diastolic blood pressure in the group without diabetes despite the fact that all patients were treated with antihypertensive therapy. Immunohistochemistry of CITED2 revealed weak staining throughout the vascular wall but intense staining in nuclei of endothelial cells (Supplementary Fig. 4C). In tissue lysate, CITED2 was 3.8-fold higher in obese patients with diabetes than in the group without diabetes (P = 0.003 by Student t test; P = 0.04 by Wilcoxon rank-sum test) (Fig. 4E and F). These data, in a limited sample of patients, suggest that CITED2 is upregulated in vascular tissue in patients with obesity and type 2 diabetes and is consistent with the previous demonstration that patients with type 2 diabetes exhibit endothelial cell insulin resistance (13,14). Importantly, this difference is maintained despite near optimal control of hemoglobin A_1c (HbA_1c), blood pressure, and plasma lipids in patients with diabetes and suggests a potential therapeutic benefit of CITED2 inhibition.

Having demonstrated that CITED2 is upregulated in endothelial cells in diabetes, we next sought to examine the significance of these changes by characterizing the effects of CITED2 in endothelial cells. We used HUVECs for these experiments because human endothelial cells perform better in tube formation assays than primary rodent endothelial cells (36). As in MS1 endothelial cells (Fig. 3) and primary mouse myocardical endothelial cells (Fig. 4A and B), insulin decreased CITED2 protein in HUVECs (Fig. 5A). After 8 h of insulin treatment, CITED2 was 44% of control levels (Fig. 5B). HUVECs were transfected with siRNA targeting CITED2, which produced efficient knockdown with 95% reduction of CITED2 protein at 50 nmol/L siRNA (Fig. 5C). Compared with HUVECs transfected with a control siRNA, HUVECs with CITED2 knockdown increased several measures of tube formation on Matrigel (Fig. 5D and E and Supplementary Fig. 5). We then labeled HUVECs with EdU in order to determine the rate of DNA synthesis, a measure of cell proliferation (Fig. 5F and G). Knockdown of CITED2 increased EdU incorporation by 2.1 ± 0.3-fold (Fig. 5G). We therefore conclude that CITED2 negatively affects tube formation and suppresses proliferation of endothelial cells.

CITED2 has been shown to regulate HIF transactivation in a number of cell types (17), but this effect has not been described in any vascular cells, including endothelial cells. Demonstration of this specific phenomenon is important because CITED2 can both activate and inhibit transcription factors by restricting the availability of CBP/p300 or by acting as a molecular scaffold recruiting these coactivators to the transcription factor complex. This effect must be measured by HIF transactivation rather than DNA binding. We therefore used a luciferase reporter with a promoter containing hypoxia-regulated elements (HREs). Expression of this reporter in HUVECs showed that overexpression of CITED2 (Fig. 6A) decreased HRE-related activity by 59 ± 4% (Fig. 6B). Therefore, CITED2 impairs HIF transactivation in endothelial cells and inhibits endothelial cell proangiogenic function.

To establish that CITED2 regulates HIF in endothelial cells in vivo, we generated mice with cre-mediated recombination of the Cited2 gene (20) after cross-breeding with Cdh5-Cre transgenic mice. In primary lung endothelial cells cultured from Tg(Cre) Cited2^lox/lox mice, CITED2 protein was reduced by 70 ± 7% compared with endothelial cells cultured from Cited2^lox/lox control mice (Fig. 6B and C). We studied these animals during hind limb ischemia after ligation of the femoral artery on one side.

We measured endothelin-1 (Edn1) mRNA in muscle lysate. Endothelin-1 is a HIF target gene (37) and is involved in angiogenesis (38). It is also expressed preferentially in endothelial cells, which we validated by measuring expression in nonischemic gastrocnemius muscle enzymatically dissociated and sorted by FACS. Edn1 was 24.2 ± 3.2-fold higher in CD31⁺ cells than in CD31⁻ cells (Fig. 6D). Therefore Edn1 serves as a model HIF target gene specifically expressed in endothelial cells that can be used to assay changes in endothelial HIF activity in whole tissue lysate. HIF-1α is among the most short-lived proteins with a half-life of a few minutes after restoration of normoxia after hypoxia (39), and HIF target genes are similarly short-lived with Edn1 mRNA half-life being less than half an hour (33). We therefore measured Edn1 in lysate from flash-frozen muscle. Edn1 was increased in ischemic hind limb gastrocnemius muscle from C57BL/6 mice 24 h after femoral artery ligation (Supplementary Fig. 6). Expression subsequently decreased at 2, 3, and 7 days after surgery (Supplementary Fig. 6). In Cited2^lox/lox control mice, Edn1 increased by 2.7 ± 0.4-fold in the ischemic compared with the nonischemic leg (Fig. 6E). However, in Tg(Cre) Cited2^lox/lox mice, there was a 4.1 ± 0.4-fold induction in Edn1 (P = 0.02, compared with induction in Cited2^lox/lox mice) (Fig. 6E). This result indicates that CITED2 suppresses endothelial cell HIF transactivation in hypoxia and supports the concept that induction of CITED2 may contribute to dysfunction of HIF in type 2 diabetes.

Diabetes remains a risk factor for cardiovascular disease (40) despite control of plasma glucose, lowering of LDL cholesterol, and treatment of hypertension (41). Therefore, there is a need to reduce cardiovascular risk by targeting other factors that are dysregulated in diabetes. Insulin resistance in vascular endothelial cells is one such target. For example, we have previously shown that loss of insulin signaling in endothelial cells increases atherosclerosis in apolipoprotein E–null mice by up to threefold without changes in glucose tolerance, plasma lipids, or blood pressure (21). Insulin resistance in endothelial cells may also impair angiogenesis (5) and contribute to cardiovascular risk independent of atherosclerosis. However, little is known of the proangiogenic effector mechanisms that are activated by insulin in endothelial cells. Genes with angiogenic or angiostatic action that are regulated by insulin in endothelial cells are likely dysregulated in conditions with insulin resistance and may contribute to the impaired response to hypoxia. The identification of such genes can provide novel targets for improvement of angiogenesis in obesity and type 2 diabetes.

With the results presented here, we have identified novel insulin-regulated genes in endothelial cells. They were selected by a combined genetic and functional filter. Importantly, they were differentially regulated in endothelial cells by deletion of FoxO in tissues prone to hemangioma formation in FoxO-knockout animals (16). Therefore, they have a high likelihood of being mechanistically involved in endothelial cell homeostasis and angiogenesis. We validated this notion by demonstrating that one of these genes, CITED2, is downregulated by insulin in primary mouse and human endothelial cells and upregulated in cardiac endothelial cells isolated from obese animals and in vascular tissue from patients with obesity and type 2 diabetes. Furthermore, in keeping with predicted antiangiogenic function, CITED2 limits HIF transactivation in endothelial cells and negatively regulates endothelial cell tube formation and proliferation.

Although we did not study the other nine insulin-regulated genes further, we have provided a valuable resource for further study of the mechanisms leading from insulin resistance in endothelial cells to dysfunctional angiogenesis. These genes were not previously reported to be regulated by insulin, although one of them, the transcription factor Klf6, was shown to be upregulated by IGF1 (42). Several of the genes have previously been associated with angiogenic function, including Ctgf (26,27), Adm (28,29), Pbx1 (43), Spry2 (44), and Tcf4 (45). However, a direct role for Fbn1, Hmga2, Klf6, and Noct in angiogenesis has not been demonstrated.

Our rationale for further exploring the role of CITED2 was that one of the best-described functions of CITED2 is negative regulation of HIF-1α (17). Expression of CITED2 is induced by HIF-1α. In turn, CITED2 inhibits HIF-1α transactivation by competing with the transactivation domain of HIF-1α for both p300 and CREB binding protein (CBP) (46). Activation of endothelial cell HIF is critical for angiogenesis. Deletion of HIF-1β (Arnt) (30) or targeting both HIF-α isoforms (31) specifically in endothelial cells results in failure of vascular development and embryonic lethality. HIF-α isoforms have overlapping specificity but distinct function, and both HIF-1α (32) and HIF-2α (33) are necessary for normal angiogenesis in adulthood. Insulin may function as a proangiogenic factor through negative regulation of CITED2, thereby releasing a regulatory brake on HIF-dependent transcription.

Consistent with the ability of insulin to suppress CITED2 expression, this gene is upregulated in animal models and humans with insulin resistance. Expression of CITED2 is increased in liver of mice with diet-induced obesity or with the db/db mutation (47). More importantly, in gene array analysis of skeletal muscle biopsies, CITED2 was one of the most downregulated genes during a euglycemic-hyperinsulinemic clamp in healthy volunteers (48). This was confirmed in an independent, unbiased screen of insulin-stimulated gene expression in skeletal muscle (49). In contrast, CITED2 was much less reduced during clamp in patients with type 2 diabetes than in healthy control subjects (48). Downregulation of hepatic CITED2 in mice limits gluconeogenesis and improves glucose tolerance (47). It is therefore possible that inhibition of CITED2 may improve both insulin sensitivity and angiogenesis. Given recent assertions that therapies for type 2 diabetes should be tested for efficacy in modifying cardiovascular morbidity or mortality alongside their glucoregulatory actions (50), CITED2 is an attractive therapeutic target.

With our study, we have identified a group of endothelial genes that are controlled by insulin through FoxO1 and are likely dysregulated in insulin resistance. One of these genes, CITED2, is upregulated in vascular tissue in patients with obesity and type 2 diabetes and suppresses HIF transactivation and proangiogenic function in endothelial cells. These results advance our understanding of insulin action on endothelial cells and points to new therapeutic approaches to improve angiogenesis in patients with obesity and type 2 diabetes.

Acknowledgments. The authors thank Dr. C. Ronald Kahn of the Joslin Diabetes Center for providing Foxo1^lox/lox mice from a colony maintained by his laboratory.

Funding. X.W. received grants from the National Natural Science Foundation of China (81370936) and the Shanghai Pujiang Program (16PJ1401700). S.M.L. was supported by a Non-Clinical Open Funding Programme from the Diabetes Research and Wellness Foundation. D.S. was supported by a research grant from the Danish Diabetes Academy, which in turn receives funding from the Novo Nordisk Foundation. B.T.O. was supported by National Institute of Diabetes and Digestive and Kidney Diseases grant K08DK100543. L.M.R. was supported by grants from the Danish Council for Independent Research and the Novo Nordisk Foundation. C.R.-M. is supported by National Cancer Institute grant R21CA185196 and a Pilot and Feasibility grant from the Diabetic Complications Consortium supported by National Institute of Diabetes and Digestive and Kidney Diseases grant U24 DK076169.

The content of this article is solely the responsibility of the authors and does not necessarily represent the official views of the funding agencies.

Duality of Interest. T.R. was an employee of Novo Nordisk and was supported by the STAR Programme. C.R.-M. receives research support from Novo Nordisk through the STAR Programme.

Author Contributions. X.W. and S.M.L. performed most of the experiments and edited the manuscript. T.R., H.A., D.S., and N.D. contributed to experiments and edited the manuscript. B.T.O. bred Foxo1^lox/lox mice and edited the manuscript. S.R.P., H.C.B., and L.M.R. contributed to collection of human arterial tissue and discussion and editing of the manuscript. S.L.D. generated Cited2 floxed mice. M.T.W. edited the manuscript. C.R.-M. oversaw experiments and drafted the manuscript. C.R.-M. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Some of these results have been presented in abstract form at the 75th Scientific Sessions of the American Diabetes Association, Boston, MA, 5–9 June 2015, and the 76th Scientific Sessions of the American Diabetes Association, New Orleans, LA, 10–14 June 2016 (abstract number 1882-P and 1737-P, respectively).