Vascular Endothelial Growth Factor–Mediated Islet Hypervascularization and Inflammation Contribute to Progressive Reduction of β-Cell Mass

C57Bl6/SJL transgenic mice expressing murine VEGF₁₆₄ (provided by P.A. D’Amore, Boston, MA) under the control of the rat insulin promoter-I (RIP-I) were obtained by embryo pronuclear microinjection. Both VEGF^low and VEGF^high transgenic mice were born at the expected frequencies and fertile, and they did not develop tumors when old (>14 months). Two-month-old C57Bl6/SJL mice were used for adeno-associated viral (AAV) vector treatment. Diabetes was assessed by measuring tail vein blood glucose levels using a Glucometer Elite analyzer (Bayer). Mice were fed ad libitum with a standard chow diet (Basal Purified Diet with 12% energy from fat) or an HFD (Basal Purified Diet with 60% energy from fat, TD.88137; Harlan Teklad). In all experiments, male mice were age-matched. Animal care and experimental procedures were approved by the Ethics Committee in Animal and Human Experimentation of the Universitat Autònoma de Barcelona.

VEGF levels in serum and islet homogenates were measured by ELISA (Calbiochem, Nothingam, U.K.). Insulin levels in serum were determined by radioimmunoassay (Linco, St. Charles, MO).

Pancreata were fixed for 24 h in formalin, embedded in paraffin, and sectioned. For immunohistochemical detection of insulin, glucagon, VEGF-A, CD31, Mac-2, and glucose transporter (GLUT)-2, pancreatic sections were incubated overnight at 4°C with a guinea pig anti-insulin antibody (Sigma Chemical, St. Louis, MO), a rabbit antiglucagon antibody (ICN Biomedicals Inc.), a mouse anti–VEGF-A antibody (Santa Cruz Biochemicals), a rat anti-mouse CD31 antibody (BD Biosciences), a rabbit anti–Mac-2 antibody (Cedarlane Laboratories), or a rabbit anti–GLUT-2 antibody (Chemicon International, Temecula, CA). Secondary antibodies were: tetramethyl rhodamine isothiocyanate-conjugated goat anti-guinea pig (Molecular Probes, Leiden, the Netherlands); biotinylated goat anti-rabbit (Molecular Probes); biotinylated donkey anti-goat (Santa Cruz Biotechnology); biotinylated rabbit anti-rat (DakoCytomation, Glostrup, Denmark); biotinylated goat anti-rabbit (Molecular Probes); biotinylated horse anti-mouse (Vector Laboratories, Burlingame, CA); streptavidin-conjugated Alexa 488 (Molecular Probes); or streptavidin-conjugated Alexa 568 (Molecular Probes). For anti-mouse antibody detection, M.O.M kit (Vector Laboratories) was used. Nuclei were counterstained with Hoechst (Sigma-Aldrich). Bright field sections were counterstained with hematoxylin.

To visualize vessels within the islets, a dextran–fluorescein isothiocyanate (FITC) solution (Sigma-Aldrich; 50 mg/mL PBS) was administered by tail vein injection. After 10 min, animals were sacrificed, and pancreata were fixed for 12–24 h in formalin, embedded in paraffin, and sectioned.

For morphometrical analysis of β-cell mass, immunohistochemical detection of insulin was performed in three (2 to 3 µm) sections, separated by 200 µm in each animal. The percentage of β-cell area in the pancreas was calculated by dividing the area of all insulin-positive cells in one section by the total area of this section and multiplying the result by 100. The β-cell mass was calculated by multiplying the pancreas weight by the percent of β-cell area. For measurement of β-cell proliferation, replicating β-cells were identified by double Ki67/insulin immunostaining. Apoptotic β-cells were determined by double transferase-mediated dUTP nick-end labeling (TUNEL)/insulin immunostaining. At least 5,000 β-cells per pancreas were counted. Four animals per group were used in each analysis. For morphometrical analysis of islet vascularization, FITC-positive area and insulin-positive area were determined in pancreatic sections from FITC-conjugated dextran-injected animals. Macrophage infiltration was measured as the percentage of Mac-2–positive area per islet area. Islet vascularization and macrophage infiltration were quantified by measuring islets from three different sections per mouse, with four animals per group. A microscope (Eclipse E800; Nikon, Tokyo, Japan) connected to an image analyzer was used. For laser-scanning confocal analysis, a TCS SP2 microscope (Leica Microsystems, Heidelberg, Germany) was used.

Pancreatic islets were isolated by intraductal injection of collagenase solution (1 mg/mL collagenase P; Roche Molecular Biochemicals, Mannheim, Germany) and digestion in Hanks’ balanced salt solution (HBSS) containing 1% BSA and 5 mmol/L glucose for 10 min at 37°C. Cold HBSS (with 1% BSA and 5 mmol/L glucose) was added to stop the digestion. The tissue suspension was washed twice with cold HBSS. Islets were handpicked.

For quantitative RT-PCR analysis, total RNA was extracted from isolated islets using Tripure Isolation Reagent (Roche Molecular Biochemicals) and RNeasy Mini Kits (Qiagen, Hilden, Germany). Total RNA (1 μg) was reverse-transcribed for 1 h at 37°C using the Omniscript Reverse transcriptase kit (D-40724; Qiagen). Quantitative PCR (qPCR) was performed in a SmartCycler II (Cepheid, Sunnyvale, CA) using the Quantitect SYBR green kit (Qiagen).

For gene expression analysis, the following primers were used: interleukin-1β (IL-1β) forward, 5′-CAACCAACAAGTGATATTCTCCATG-3′ and reverse, 5′-CATCCACACTCTCCAGCTGCA-3′; tumor necrosis factor-α (TNF-α) forward, 5′-CATCTTCTCAAAATTCGAGTGACAA-3′ and reverse 5′-TGGGAGTAGACAAGGTACAACCC-3′; CCL-2 forward, 5′-CCCAATGAGTAGGCTGGAGA-3′ and reverse, 5′-TCTGGACCCATTCCTTCTTG-3′; IL-6 forward, 5′-AGTTGCCTTCTTGGGACTGA-3′ and reverse, 5′-TCCACGATTTCCCAGAGAAC-3′; RBS forward, 5′-ATTCGCTGCACGAACTGCG-3′ and reverse, 5′-CAGCAGGTCTGAATCGTGGT-3′; and GLUT-2 forward, 5′-CTGGAGCCCTCTTGATGGGA-3′ and reverse, 5′-CCAGTCCTGAAATTAGCCCAC-3′.

Isolated islets were homogenized in protein lysis buffer after being washed with PBS. Proteins (50–100 µg) were separated by 10% SDS-PAGE, transferred to polyvinylidene difluoride membranes, and probed with primary antibodies against VEGF (Abcam), E-cadherin (Santa Cruz Biotechnology), and β-actin (Abcam) overnight at 4°C. Detection was performed using horseradish peroxidase-labeled anti-goat immunoglobulin G or horseradish peroxidase-labeled anti-rabbit immunoglobulin G (DakoCytomation) and ECL Plus Western Blotting Detection Reagent (Amersham Biosciences). Data were obtained from three pools of islets and four mice per pool.

Vectors were generated by triple transfection of human embryonic kidney 293 cells. Cells were cotransfected with a plasmid carrying the expression cassette, a helper plasmid carrying the AAV rep2 and cap9 genes, and a plasmid carrying the adenovirus helper functions (provided by K.A. High, Children’s Hospital of Philadelphia). The transgene used in these experiments was murine VEGF₁₆₄ under the control of RIP-I promoter followed by woodchuck hepatitis virus posttranscriptional regulatory element sequence to increase expression. AAV null vectors were produced using pAAV-MCS plasmid (Stratagene, La Jolla, CA). Vectors were purified with an optimized method using cesium chloride gradients (27), dialyzed, filtered, titred by qPCR, and stored at −80°C until use.

Mice were anesthetized with an intraperitoneal injection of ketamine (100 mg/kg) and xylazine (10 mg/kg). A microclamp was placed on the bile duct caudal to the liver to prevent vector distribution. A 30-gauge needle was used to advance retrogradially through the sphincter of Oddi into the common bile duct. A volume of 100 µL of AAV vectors diluted in PBS was injected into the duct. A dose of 10¹² vector genomes per mouse was used.

All values are expressed as the means ± SEM. Differences between groups were compared by Student t test. A P value <0.05 was considered statistically significant.

Islet vascularization and VEGF expression were determined in 6-month-old C57Bl6 mice fed with an HFD for 4 months. HFD-fed mice developed insulin resistance, as they were normoglycemic (chow, 136 ± 13.5 vs. HFD, 143 ± 17.5 mg/dL) and highly hyperinsulinemic (chow, 2.3 ± 0.4 vs. HFD, 8.1 ± 1.7 ng/mL) and showed altered insulin tolerance (Supplementary Fig. 1). In contrast to mice fed a chow diet, islets from HFD-fed mice were larger (chow, 290 ± 65 vs. HFD, 435 ± 110 µm²; P < 0.05), indicating compensatory islet hyperplasia. Islets from these mice showed hypervascularization with thicker basal membranes after staining for the endothelial cell marker CD31 and the basement membrane marker collagen IV (Fig. 1A and B). To investigate whether intraislet vasculature was functional, a FITC-conjugated dextran solution was intravenously injected in these mice. An increased vessel area/islet area ratio (∼25%) was observed in HFD-fed mice, indicating higher islet vascularization (Fig. 1C and D). This suggests that in hyperplasic islets, angiogenesis might have increased to perfuse new β-cells. This was parallel to the increased VEGF content (about twofold) in HFD islets (Fig. 1E and F). Because VEGF can induce collagen IV synthesis by endothelial cells (28), VEGF upregulation was probably responsible for the increased amount of endothelial ECM of these islets (Fig. 1A). Because fibrosis and VEGF upregulation are also usually found during inflammatory processes (11), pancreatic sections were stained for the macrophage cell marker Mac-2, and an increase in macrophage infiltration in the islets of HFD-induced insulin resistant mice was observed (Fig. 1G and H). These results suggest that insulin resistance induces compensatory islet hyperplasia together with increased islet VEGF expression, angiogenesis, and macrophage recruitment.

To elucidate the role of VEGF in islet angiogenesis and β-cell function, transgenic mice overexpressing VEGF in β-cells under the control of the RIP-I were generated. Two transgenic lines, VEGF^low and VEGF^high, with a 2.7- and 17-fold increase in VEGF expression (Fig. 2A and B), respectively, were obtained. VEGF^low mice showed an increase in VEGF comparable to that observed in insulin-resistant mice (Fig. 1F) and rats (19). VEGF serum concentrations remained unchanged (wild-type, 221 ± 27.8; VEGF^low, 202 ± 29.3; and VEGF^high, 189 ± 25.1 ng/dL), suggesting an autocrine/paracrine effect of VEGF in the pancreas. Compared with wild-type, 2-month-old VEGF^low and VEGF^high islets displayed hypervascularization and basal membrane thickening, as shown by both CD31 and collagen IV immunostaining (Fig. 2C). Intravascular injection of FITC-conjugated dextran confirmed the increased vessel area/islet area ratio of functional capillaries (Fig. 2C and D).

Islets from 2-month-old VEGF^low and VEGF^high mice showed altered distribution of α and β-cells (Fig. 3A). Although the basal membrane of islet endothelial cells provides signals that promote β-cell proliferation (15) and transgenic islets were hypervascularized, the β-cell mass was similar in wild-type, VEGF^low, and VEGF^high mice (Fig. 3B). This was consistent with normal rates of β-cell replication and apoptosis (Fig. 3C–F). This altered islet morphology was parallel to decreased levels of E-cadherin (Fig. 3F and G), a key protein in maintaining islet endocrine cell contacts (29,30).

Two-month-old wild-type, VEGF^low, and VEGF^high mice presented similar glycemia, insulinemia (Fig. 4A and B), and glucose tolerance (Fig. 4C). In addition, comparable in vivo insulin secretion profiles were noted in all groups (Fig. 4D). This was consistent with normal GLUT-2 expression in VEGF^low islets (Fig. 4E and F). However, VEGF^high islets showed a decrease in GLUT-2 expression without a major change in insulin secretion (Fig. 4E and F). Therefore, although transgenic islets were disorganized and hypervascularized, glucose homeostasis in young transgenic mice was nevertheless unaltered, indicating normal islet function, in agreement with the normal β-cell mass.

VEGF^low mice remained normoglycemic up to 12 months of age, when they developed mild hyperglycemia and hypoinsulinemia (Fig. 4A and B), altered glucose tolerance, and decreased glucose-stimulated insulin release (Fig. 5A and B), but normal insulin sensitivity (data not shown). In contrast, VEGF^high mice, which expressed higher VEGF levels, developed glucose intolerance and decreased insulin secretion after a glucose load earlier, at 5 to 6 months of age (Fig. 5C and D). Moreover, 8-month-old VEGF^high mice displayed overt hyperglycemia, hypoinsulinemia (Fig. 4A and B), and glucose intolerance (Fig. 5E).

Morphometric analysis of pancreatic sections from VEGF^low mice showed a trend toward decreased β-cell mass only at the age of 12 months (Fig. 5F and G), although this did not reach statistical significance (P = 0.059). In contrast, 8-month-old VEGF^high mice displayed a marked loss of β-cell mass (∼80%) (Fig. 5F and G). In VEGF^high pancreata, insulin- and glucagon-positive cells were scattered and surrounded by abundant collagen IV-positive staining (Fig. 5F).

In addition to promoting angiogenesis, VEGF is involved in macrophage recruitment (13). Mac-2–positive cells were detected in islets from 2- to 12-month-old VEGF mice (Fig. 6A and B). Macrophage infiltration progressively increased as animals aged and in parallel with VEGF expression, being higher in VEGF^high than in VEGF^low islets (Fig. 6A and B). These results suggest that chronic overexpression of VEGF progressively increased macrophage recruitment. Consistent with this, a statistically significant increase in the expression of CCL-2 (monocyte chemoattractant protein-1), IL-1β, and TNF-α was observed in islets of 12-month-old VEGF^low mice and 5-month-old VEGF^high mice, although a tendency toward increased expression was already noted in 2-month-old mice (Fig. 6C–E). Thus, the development of hyperglycemia and decrease in β-cell mass were concomitant with progressive macrophage infiltration, cytokine production, and fibrosis.

To avoid potentially undesirable effects of VEGF overexpression during embryonic development and to verify a direct effect of VEGF in endocrine pancreas inflammation, AAV vectors were used to overexpress VEGF in islets of adult mice. Intraductal delivery of AAV vectors of serotype 9 (AAV9) leads to highly efficient and long-term transduction of β-cells and exocrine pancreas (31). To restrict transgene expression to β-cells, the RIP-I promoter was used. Thus, AAV9 vectors carrying a RIP-I/VEGF transgene were intraductally injected into 2-month-old mice, and glucose homeostasis and islet morphology were studied within 2 months after AAV injection. Compared with AAV9-null–treated mice, AAV9-VEGF–treated animals showed about a twofold increase (AAV9-VEGF: 2.2 ± 0.8-fold increase vs. AAV9-null) in islet content of VEGF, similar to the increases detected in islets from HFD-fed mice, VEGF^low transgenic mice, or Zucker rats (19). This value was, however, variable among mice paralleling differential efficiency of transduction. Despite the heterogeneity, all AAV9-VEGF–treated mice showed between 30 and 80% hypervascularized and disorganized islets, with increased collagen IV accumulation (Fig. 7A). These alterations were already observed 10 days after vector administration (Fig. 7A) and were also present at 40 days (Supplementary Fig. 2). Islet inflammation, determined by Mac-2 immunostaining, was increased in AAV9-VEGF–treated mice at 10 (Fig. 7A and B) and 40 days (Supplementary Fig. 2) after AAV injection. Moreover, increased VEGF expression in β-cells induced a progressive increase in glycemia (Fig. 7C), along with glucose intolerance (Fig. 7D). Thus, these results confirm that VEGF overexpression in islets of adult mice can induce hypervascularization, islet disorganization, and inflammation that result in impaired glucose homeostasis.

In the preliminary stages of T2D, expansion of β-cell mass is a key adaptive response to compensate for insulin resistance (2). During this period, islet vasculature also needs to expand to perfuse the newly formed β-cells. The development of vessels in adult organisms (angiogenesis) occurs through a multistep process requiring VEGF and other soluble factors (32). In this study, we found that islets from obese, insulin-resistant, HFD-fed mice displayed increased vascularization that paralleled increased VEGF levels. Similar observations were described in insulin-resistant prediabetic Zucker rats (32), in which islets of these rats showed increased vascularization and enhanced VEGF secretion compared with non–insulin-resistant rats. This may indicate that VEGF upregulation represents a compensatory response to support angiogenesis during β-cell expansion. To discern the role of increased VEGF and islet vascularization during T2D development, VEGF^low and VEGF^high transgenic mice were generated. Similar to HFD-fed mice, VEGF^low and VEGF^high mice displayed islet hypervascularization with irregular capillaries and thicker basal membranes. These vascular alterations resembled those observed in islets from HFD-fed mice and were consistent with pathologic angiogenesis. Similarly, it has been shown in tumors that prolonged deregulated expression of VEGF leads to aberrant immature vessels that often invade the surrounding tissue (33,34). In contrast, normal vasculature spreads by physiologic angiogenesis, requiring proper regulation of the dose and timing of VEGF production (11,32).

Islets from transgenic VEGF^low and VEGF^high mice showed thickening of the vessel basal membrane that resulted from increased collagen IV accumulation, which in VEGF^high islets disrupted islet structure and was parallel to β-cell loss and development of hyperglycemia. In islets from GK rats, ECM deposition develops from intra- and peri-islet vessels (21,35). The islet ECM is mainly synthesized by endothelial cells, because β-cells do not have an ECM (25,26). Moreover, VEGF is able to induce collagen IV synthesis in cultured endothelial cells (28), which suggests that in islets from VEGF transgenic mice, a VEGF-mediated activation of endothelial cells causes overproduction of ECM components and fibrosis. Similarly, islet fibrosis has also been described in islets of patients with T2D and in various spontaneous rodent models of T2D, such as db/db mice and Zucker and GK rats, in which it is involved in islet destruction (21,36–38). Thus, during T2D, chronic VEGF upregulation in islets may be responsible for ECM accumulation in surrounding islet vessels, leading to fibrosis.

Two-month-old VEGF^low and VEGF^high mice also displayed disorganized islets with α-cells scattered in the islet core. The Ca²⁺-dependent cell adhesion molecule E-cadherin, which is highly expressed in pancreatic islets, mediates β-cell-to-β-cell contacts, controlling islet architecture and function (29,30). The expression of E-cadherin was reduced in VEGF^low and VEGF^high islets, consistent with decreased numbers of direct β-cell-to-β-cell contacts and islet disorganization. It is widely accepted that proper insulin secretion requires the coordinated function of the β-cells that form an islet (29). Moreover, islet disorganization is considered a hallmark of islet dysfunction (39). However, despite altered islet architecture, 2-month-old VEGF^low and VEGF^high mice displayed normal in vivo insulin release and glucose tolerance. Therefore, these transgenic mice indicate that abnormal islet morphology is not directly responsible for impaired β-cell function. Because the alteration of insulin secretion developed in parallel with the severity of inflammation, our results suggest that islet inflammation, rather than islet disorganization alone, was probably the proximate cause of islet failure.

Although the precise mechanism or more likely multiple mechanisms underlying decreased β-cell mass and function in T2D remain to be elucidated, islets from patients with T2D display an inflammatory process characterized by the presence of immune cell infiltration, inflammatory cytokines, apoptotic cells, and, eventually, fibrosis. VEGF transgenic mice partially recapitulate those processes that occur specifically in the islets without the development of insulin resistance. VEGF is a powerful chemoattractant for macrophages and other leukocytes at the sites of angiogenesis, where these cells play an important role in vessel formation and tissue remodeling (13,40). Islets from HFD-fed mice showed infiltrating macrophages, comparable to those described in islets from several type 2 diabetic animal models and human patients (7,38,41,42). Increased numbers of macrophages are detectable very early in islets, before the onset of diabetes (41). Normoglycemic insulin-resistant HFD-fed mice showed islet macrophage infiltration concomitant with VEGF upregulation. Similarly, islets from both VEGF^low and VEGF^high mice were progressively infiltrated by macrophages. Moreover, Mac-2⁺ insulitis grade and VEGF upregulation level in VEGF^low mice were similar to those observed in prediabetic HFD-fed mice. Finally, VEGF transgenic islets showed macrophage infiltration even before developing impaired glucose homeostasis, suggesting that infiltration is not a consequence of β-cell death. Therefore, this study suggests that macrophage infiltration may already be induced during islet hyperplasia and play a role in initiating and/or accelerating β-cell failure. This hypothesis is supported by the AAV-mediated overexpression of VEGF in islets of adult mice, because, as early as 10 days after vector administration, mice already showed islet disorganization and inflammation, which subsequently progressed to hyperglycemia and glucose intolerance.

VEGF-expressing islets showed macrophage infiltration and increased expression of IL-1β, CCL-2, and TNF-α. A significant increased in cytokine expression was detected at a slightly delayed time compared with macrophage infiltration, but both inflammatory processes increased with age and VEGF levels. Macrophages are able to secrete a broad range of cytokines. In RIP-CCL2 transgenic mice, myeloid cell infiltration of islets is able to induce β-cell death and diabetes (43). Moreover, it has been shown that high glucose concentrations can induce β-cell expression of several cytokines, such as IL-1β and CCL-2 (44–46). Thus, during development of T2D, VEGF and inflammatory cytokines may be secreted and attract macrophages, and, in turn, macrophages can produce higher levels of proinflammatory cytokines. These processes probably stimulate each other in a positive-feedback manner, and, together with other processes such as endoplasmic reticulum (ER) stress, may contribute to β-cell failure and death. During T2D, higher insulin requirements and high circulating levels of free fatty acids and glucose can lead to the activation of the unfolding protein response and ER stress within the β-cells (47). Moreover, several studies have proposed that proinflammatory cytokines are also able to induce β-cell apoptosis by means of ER stress (47–49). Because VEGF overexpression in islets from our transgenic mice resulted in enhanced cytokine production, this may also contribute to β-cell loss through ER stress.

It is of interest that although previous reports demonstrated that decreased VEGF expression in β-cells resulted in insufficient vascularization leading to impaired insulin secretion and glucose intolerance (14), our results suggest that excessive islet VEGF production also triggers architectural abnormalities and impaired β-cell function. Thus, VEGF levels need to be finely and tightly regulated to avoid development of undesirable hypervascularization, fibrosis, and inflammation.

Finally, although no direct links between angiogenic factor genetic variants and diabetes have been reported, allelic variants of VEGF (50) and the VEGF-inducer transcription factor hypoxia inducible factor-1α have been described in T2D patients (51). Moreover, polymorphisms in the VEGF-A gene have been reported in patients with micro and macrovascular diabetic secondary complications (52,53). These data further suggest that dysregulation of VEGF could be involved in the pathogenesis of T2D.

In summary, the current study demonstrates a crucial role for islet vascularization and inflammation in the development of β-cell failure, suggesting that macrophage recruitment and sustained islet cytokine production may result from increased angiogenesis and/or VEGF upregulation and that this inflammatory process may contribute to connect insulin resistance with β-cell death and T2D.

This work was supported by grants from Ministerio de Ciencia e Innovación, Plan Nacional I+D+I (SAF2008-00962) and Generalitat de Catalunya (2009 SGR-224), and Spain and European Commission (EUGENE2, LSHM-CT-2004-512013; CLINIGENE, LSHB-CT-2006-018933). J.A., V.J., C.M., and A.R. were recipients of predoctoral fellowships from Ministerio de Educación and Cultura y Deporte and A.S. from Direcció General de Recerca, Generalitat de Catalunya.

No potential conflicts of interest relevant to this article were reported.

J.A., E.A., V.J., and F.B. designed experiments. J.A., E.A., V.J., A.C., C.M., A.S., S.T., M.O., A.R., M.M., and A.P. generated reagents and performed experiments. J.A. and F.B. wrote and edited the manuscript. F.B. 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.

The authors thank C.J. Mann, Universitat Pompeu Fabra, S. Franckhauser, Universitat Autònoma Barcelona, and J. Storch, Rutgers University, for helpful discussion.