Pancreatic Islet Production of Vascular Endothelial Growth Factor-A Is Essential for Islet Vascularization, Revascularization, and Function

Compound VEGF^fl/fl, Rip-Cre;VEGF^fl/wt, and Rip-Cre;VEGF^fl/fl mice were generated by crossing Rip-Cre transgenic mice on C57BL/6 background (30) with VEGF^fl/fl mice on a congenic background (31). Littermates were used in the studies. Tie1^LacZ/wt mice were previously characterized (32). Islet transplants were performed into the immunodeficient NOD-SCID mouse model from The Jackson Laboratory (http://jaxmice.jax.org/jaxmice-cgi/jaxmicedb.cgi?objtype=pricedetail&stock=001303). All animal studies were approved by the institutional animal care and use committee at the Vanderbilt University Medical Center.

The function of the islet/transplant vasculature was assessed by infusing fluorescein isothiocynate-conjugated tomato lectin (Lycopersicon Esculentum, 1 mg/ml; Vector Laboratories, Burlingame, CA) into the jugular vein of embryos or adult mice. Mouse embryos at e16.5 were infused with 3 μl lectin at an infusion rate of 10 μl/min using 30- to 31-gauge needle attached through the catheter to a 10-μl Hamilton syringe that is placed in an infusion pump (model 11 plus; Harvard Apparatus, Holliston, MA). Lectin was allowed to circulate for 5 min, after which time the embryos were killed and their digestive organs dissected using an Olympus SZX12 (Olympus, Tokyo, Japan) dissecting microscope attached to an Olympus DP12 digital camera. In adult mice and islets transplanted beneath the kidney capsule, 100 μl lectin was injected through a 28-gauge needle attached 0.3-ml syringe and allowed to circulate for 5 min before the animal was killed and dissected.

Collection of adult pancreata or kidneys bearing islet transplants was performed as described (26). Embryos were dissected in ice-cold 10 mmol/l PBS. Entire embryos (e13.5) or dissected digestive organs (e16.5) were fixed in freshly prepared 2% paraformaldehyde (Electron Microscopy Sciences)/10 mmol/l PBS for 30 min on ice. Following fixation, the tissues were washed four times with 10 mmol/l PBS over a period of 1 h and then equilibrated in 18% sucrose/10 mmol/l PBS overnight at 4°C. The tissues were cryopreserved in optimum cutting temperature compound (VWR Scientific Products, Willard, OH) at −80°C. Immunohistochemistry on 10-, 30-, or 60-μm cryosections was performed as described (26). β-Galactosidase activity was detected on 10-μm cryosections as described (26). Antibodies to the following antigens were used: insulin, glucagon, somatostatin, PECAM-1, and VEGFR2 (26); pancreatic polypeptide (1:2,000; Bachem, San Carlos, CA); VEGF (1:200; Lab Vision Corporation, Fremont, CA); VEGF-A (1:250; R&D Systems, Minneapolis, MN); angiopoietin (Ang)-1 (1:200; Alpha Diagnostic International, San Antonio, TX); Tie2 (1:200; Santa Cruz Biotechnology, Santa Cruz, CA); ephrin-A1 (1:1,000; Amgen, Thousand Oaks, CA); and EphB4 (1:500; R&D Systems). The antigens were visualized using appropriate secondary antibodies conjugated with Cy2, Cy3, and Cy5 fluorophors from Jackson ImmunoResearch Laboratories (West Grove, PA). Secondary antibodies were used at concentrations recommended by the manufacturer. Digital images of the 10-μm cryosections were acquired with a MagnaFire digital camera (Optronics, Goleta, CA) connected to an Olympus BX-41 fluorescence microscope (20× and 40× magnification) or using a Zeiss LSM510 META laser scanning confocal microscope (40× magnification; Carl Zeiss, Jena, Germany). Sections, 30- and 60-μm thick, were subjected to optical sectioning, and digital images were 3D reconstructed using MetaMorph version 6.1 software (Universal Imaging, Downington, PA). To systematically examine vascularization of islets and exocrine tissue, sections collected from three different depths of the tissue block spaced by 250 μm were examined. In case of islet grafts, the sections were spaced by 150 μm. Using MetaMorph version 6.1 software (Universal Imaging), integrated morphometry analysis was applied to 30–40 islets per tissue block (or comparable areas of exocrine tissue or graft) to calculate vessel density and area per vessel. Vessels were counted using a technique described by Weidner and colleagues (33,34).

Pancreas was fixed by vascular perfusion with 2.5% glutaraldehyde (Electron Microscopy Sciences)/0.1 mol/l sodium cacodylate. After islets were located on a semithin section, 70-nm sections were placed on slot grids, stained with uranyl acetate and lead citrate, and analyzed by transmission electron microscopy (Philips CM-12 electron microscope). For immunoelectron microscopy, pancreas was perfused with 0.5% glutaraldehyde plus 3.0% paraformaldehyde in 0.1 mol/l phosphate buffer followed by 1 h immersion fixation in the same fixative. The tissues were then postfixed for 4 h in 3.0% paraformaldehyde in phosphate buffer, washed, dehydrated, and embedded in LR white resin (Electron Microscopy Sciences). Sections 70-nm thick were mounted onto nickel grids and incubated with guinea pig anti-insulin (Linco) followed by donkey anti–guinea pig antibody-12-nm colloidal gold (Jackson ImmunoResearch) and counterstained with aqueous uranyl acetate and lead citrate.

Size-matched islets were cultured in eight-well chamber sides containing 50–70 islets/well in 470 μl RPMI-1640 media (the same as above) for 48 h at 37°C. VEGF-A production by isolated islets was measured by mouse VEGF-A–specific enzyme-linked immunosorbent assay (R&D Systems).

Unpaired t test and one-way ANOVA with Newman-Keuls multiple comparison test were used to compare outcomes in mice of different genotypes. Data are expressed as means ± SE.

Prior information about the mouse pancreatic vasculature was rather limited; therefore, we first quantified vascularization in endocrine and exocrine compartments of the pancreas (online appendix Fig. 1 [available at http://diabetes.diabetesjournals.org]). Our results indicate that both vessel density and vessel size/branching are as much as twofold greater in islets than in the surrounding exocrine tissue, which is in agreement with previously reported data on the rat pancreatic vasculature (2,3). To investigate molecular factors that might be involved in much higher vascularization in the islets compared with exocrine tissue, we examined histological sections of adult mouse pancreas. Interestingly, we found that the members of three angiogenic factor families such as VEGFs, Angs, and ephrins are expressed at much greater levels in pancreatic islet cells than in surrounding exocrine or ductal cells (Fig. 1).

While expression of VEGF-A was described previously in islets (16–18), it was unclear which islet cell types express this angiogenic factor. Colocalization of VEGF-A with insulin in adult mouse pancreas demonstrated that all insulin⁺ cells express VEGF-A (Fig. 1A). Since we noted that there were also VEGF-A⁺/insulin⁻ non–β-cells around the islet perimeter, we costained for glucagon and demonstrated that α-cells also express VEGF-A. (Fig. 1B). VEGFR2, the receptor that largely mediates the actions of VEGF-A as an endothelial cell mitogen and vascular permeability factor, was expressed in the microvasculature of the islets, exocrine tissues, and in periductal capillary plexus (Fig. 1C–F). It is noteworthy that the level of VEGFR2 was consistently greater in the microvessels of islets compared with exocrine capillaries, as reflected by fluorescence intensity. VEGFR2 expression was significantly downregulated in larger pancreatic arterial and venous vessels (see PECAM-1⁺/VEGFR2⁻ vessels in Fig. 1E–J).

Expression of Ang-1 and Tie receptors in the pancreas has not been previously reported. In sharp contrast to VEGF-A, Ang-1 was detected only in β-cells (Fig. 1K) but not in α-cells (Fig. 1L). β-Cell–specific expression of Ang-1 was also found in human islets (online appendix Fig. 2) in spite of the difference in mouse and human islet architecture (37). The expression pattern of Tie2 receptors was substantially different than what we found for VEGFR2. While VEGFR2 expression was restricted to microvessels, Tie2 receptor was broadly expressed in endothelial cells of all vessel types and especially was noted at high levels in arteries and veins (Fig. 1M and N). Using a Tie1^lacZ/wt knock-in mouse model (32), the Tie1 receptor expression pattern was unique compared with VEGFR2 and Tie2 receptors in that it was detected in microvessels of islets and exocrine tissues (Fig. 1O and P), periductal capillary plexus, and arteries but appeared to be downregulated in veins (Fig. 1Q and R).

To determine whether members of the ephrin family of angiogenic factors are expressed in the pancreas, we analyzed expression of ephrin-A1 ligand, which was previously reported to be expressed by islet tumors in the RIP-Tag model (38), and found that normal β-cells and non–β-cells express ephrin-A1 (Fig. 1S–U). We then examined the expression of ephrin receptors. Since ephrin-B2 ligand and its EphB4 receptor are specifically localized to arterial and venous endothelial cells in other tissues (39), respectively, we sought to utilize EphB4 expression and hoped to identify the venous pole of the islets. Intra-islet capillaries were negative for EphB4 expression. However, as these vessels became postcapillary venules at the islet periphery, EphB4 was highly expressed (at the same levels as seen in large pancreatic veins) (Fig. 1V–X).

Thus, we propose that the greater vascularity and distinct phenotypic characteristics of endothelium (such as permeability) within the adult pancreatic islets compared with exocrine tissue are the result of angiogenic ligands (VEGFA, Ang-1, and ephrin-A1) being enriched in islet endocrine cells and differential expression of the VEGFR2, Tie1, Tie2, and EphB4 receptors in the pancreatic vasculature.

We envisioned two possible models of how pancreatic islets become so intensely vascularized. In the first model, pancreatic islets, which form by the migration and coalescence of developing endocrine cells, would secrete angiogenic factors and then become penetrated by vessels “diving” into the islet after islet assembly in late embryogenesis. Such a system has been described for coronary artery development (40). In the second model, islet vasculogenesis/angiogenesis would occur simultaneously with the migration and coalescence of developing endocrine cells during islet assembly throughout embryogenesis. To examine which model was operational, we assessed the expression of angiogenic ligands and receptors in the developing mouse pancreas. Enriched expression of angiogenic factors was typical not only for adult islet cells but was also found in endocrine cells within developing islet cell clusters. Immunohistochemical analysis of the embryonic mouse pancreas showed that VEGF-A and Ang-1 were expressed by developing endocrine cells as early as e13.5 (Fig. 2A–G). In contrast to adult islet cells, both embryonic glucagon⁺ and insulin⁺ cells expressed Ang-1 (data not shown).

At this early time point, endocrine cells were already in a close association with VEGFR2⁺ and Tie2⁺ endothelial cells (Fig. 2D and H). Therefore, we were interested to know whether complete islet assembly was a prerequisite for the endocrine cells to be exposed to the blood flow with its delivery of oxygen and nutrients. At e16.5, which coincides with the beginning of islet morphogenesis, we used intravital lectin labeling (41) to mark functional vessels in the embryonic pancreas (Fig. 2I–N). Figure 2K–N demonstrate that coalescing endocrine cells are adjacent to blood-perfused lectin⁺ capillaries, thus indicating that blood flow to the pancreas and endocrine cells in particular precedes the assembly of mature islets. In addition, previous studies showed that islet mass expands postnatally several-fold (42), and interestingly, we found that this period is also marked by a significant vascular remodeling within the islets (online appendix Fig. 3). Based on these observations, we propose a model of pancreatic islet vascularization where development of islet vasculature and establishment of islet blood flow occur concomitantly and in coordination with islet formation (Fig. 2O). Critical components of this islet vascularization model include: 1) production of angiogenic factors by early developing islet cells, 2) recruitment of endothelial cells and their association with developing islet cell clusters, 3) establishment of blood flow to small endocrine cell clusters before complete islet assembly, and 4) coordinated assembly of islet cell types and the islet vascular structures in late embryogenesis.

To address the role of vascularization in islet function, we inactivated one of the islet angiogenic factors, VEGF-A, in a β-cell–specific manner using the Cre-loxP system. In Rip-Cre;VEGF^fl/fl islets, VEGF-A inactivation led to a 70% loss in VEGF-A production (Fig. 3A), which was consistent with β-cell abundance/islet and our finding that VEGF-A is also expressed by non–β-cells (Fig. 1A and B). Reduced VEGF-A expression by β-cells in Rip-Cre;VEGF^fl/wt and Rip-Cre;VEGF^fl/fl islets did not impact islet morphology (Fig. 3B). However, islet vasculature was altered in VEGF-A dose-dependent manner (Fig. 3C and D). 3D reconstructed islet vasculature (Fig. 3D) and data in Fig. 3E and F show that vessels in VEGF-A–deficient islets were not only fewer in number compared with wild-type islets but also reduced in size/branching. In Rip-Cre;VEGF^fl/fl islets, it also appeared that vessel density was much lower in the islet core than around the islet perimeter, where VEGF-A is still expressed by non–β-cells (Fig. 3D, bottom panel). These vascular abnormalities were confined to the islets, as both microvascular density and vessel size were normal in exocrine pancreas (see Fig. 3 legend).

To elucidate the relationship between VEGF-A, islet vascularization, and islet function, we subjected Rip-Cre;VEGF^fl/wt and Rip-Cre;VEGF^fl/fl mice to intraperitoneal glucose tolerance testing and compared them with their wild-type littermates. Rip-Cre;VEGF^fl/wt and Rip-Cre; VEGF^fl/fl mice had normal fasting blood glucose levels, but they cleared the blood glucose at a considerably slower rate than VEGF^fl/fl mice (Fig. 4A and B). This delayed glucose clearance in Rip-Cre;VEGF^fl/wt and Rip-Cre;VEGF^fl/fl mice coincided with an ∼40% reduction in plasma insulin levels measured 15 min after glucose administation (Fig. 4C). In contrast to a recent report (43), we found no effect of the Rip-Cre transgene alone on the glucose clearance (data not shown). To investigate the mechanism by which Rip-Cre;VEGF^fl/wt and Rip-Cre;VEGF^fl/fl mice had lower plasma insulin levels than their wild-type littermates, we examined their insulin expression and islet mass. Figure 4D shows that the decreased insulin secretion in VEGF-A mutant mice was not caused by reduced pancreatic insulin content. Islet mass was also very similar in Rip-Cre;VEGF^fl/fl (2.20 ± 0.28 mg, n = 3) compared with wild-type (1.72 ± 0.27 mg, n = 3) mice, as were the relative populations of the different endocrine cell types (data not shown).

To investigate why plasma insulin levels were decreased in Rip-Cre;VEGF^fl/wt and Rip-Cre;VEGF^fl/fl mice during glucose challenge, we tested β-cell secretory capacity. Isolated islets from control, Rip-Cre;VEGF^fl/wt, and Rip-Cre;VEGF^fl/fl mice were placed in a cell perifusion system and challenged with β-cell secretagogues. The insulin secretory profile of islets with reduced VEGF-A was very similar to that of wild-type islets (Fig. 5A), suggesting that β-cell secretory capacity was not altered by reduced islet VEGF-A levels.

Since insulin content, islet morphology, islet mass, and β-cell insulin secretion were normal in VEGF-A mutant islets, we hypothesized that vascular abnormalities in Rip-Cre;VEGF^fl/wt and Rip-Cre;VEGF^fl/fl islets result in reduced insulin levels in the vascular system and that this leads to perturbation of glucose homeostasis. To test this hypothesis, we took advantage of the in situ pancreas perfusion technique (35). Unlike the cell perifusion system, where secretagogues reach islet cells via diffusion, pancreas perfusion delivers the stimulus to the islet cells utilizing the native vasculature of the islets and exocrine pancreas and thus mirrors how secretagogues reach islets cells in vivo. Using the perfusion system for vascular delivery of β-cell secretagogues, we measured insulin secretory output in wild-type and Rip-Cre;VEGF^fl/fl mice. In sharp contrast to the uniform insulin response by wild-type and mutant isolated islets in the diffusion-dependent perifusion system (Fig. 5A), vascular delivery of β-cell stimuli resulted in a significant attenuation of insulin secretion in Rip-Cre;VEGF^fl/fl mice (Fig. 5B).

To further investigate the islet vasculature, we examined the ultrastructure of intra-islet capillaries by electron microscopy. Using this approach, we found that intra-islet endothelial cells from Rip-Cre;VEGF^fl/fl mice had a dramatic loss of fenestrations and contained an abundant amount of vesiculo-vacuolar organelles consistent with caveolae (Fig. 5C and online appendix Fig. 4). Both VEGF¹²⁰ and VEGF¹⁶⁴ isoforms, which are produced by islets (20), have been shown to induce endothelial fenestrations in vitro (44). This process is first marked by rapid mobilization of caveolae, which under prolonged VEGF-A treatment progress to fenestrations (45). Therefore, it appears that in Rip-Cre;VEGF^fl/wt and Rip-Cre;VEGF^fl/fl islets insufficient VEGF-A levels arrested formation of the endothelial fenestrations at caveolae stage. These results are consistent with in vitro studies indicating a role for VEGF-A in caveolae development and maintenance (45) and similar to the findings of Lammert et al. (22) when they deleted VEGF-A using a Pdx1-Cre approach. Thus, the impaired insulin delivery into the vascular system in Rip-Cre;VEGF^fl/wt and Rip-Cre;VEGF^fl/fl appears to result from not only reduced vessel density and size (Fig. 3E and F) but also reduced vessel permeability as a result of the ultrastructural changes in the intra-islet endothelial cells. Our results with in vitro/in situ and in vivo insulin secretion in Rip-Cre;VEGF^fl/fl mice indicate that not only normal β-cell secretory potential and insulin stores but also normal islet vasculature are essential to achieve appropriate insulin levels in the vascular system in response to physiologic stimuli.

While normal islets are highly vascularized (online appendix Fig. 1A–C) and contain a fenestrated endothelium (Fig. 5C, left panel), islet isolation procedures sever vascular connections of islets and thus during the first few days after transplantation, until islets revascularize, they are avascular and are dependent on diffusion of nutrients and oxygen for function and survival. We postulated that islet-derived angiogenic factors like VEGF-A are fundamental to recruitment of endothelial cells from the transplant recipient and mobilization of intra-islet endothelial cells. Previous studies indicated that increased VEGF-A expression in islets may accelerate islet revascularization and improve transplanted islet survival, but results have been conflicting (46–48). Furthermore, no studies have addressed the role of physiologic production of angiogenic factors by islets in the revascularization process. To address these questions, we asked how reduced production of VEGF-A in β-cells affects the revascularization of transplanted islets. Two hundred VEGF^fl/fl or Rip-Cre;VEGF^fl/fl islets were transplanted beneath the renal capsule of NOD-SCID mice (Fig. 6). Interestingly, at early time points, the level of revascularization and pattern of new vessel formation were very similar in both types of grafts. At 7 days posttransplantation, vessel density reached ∼50% of that in wild-type islet grafts at 1 month and new vessels were mainly found around the graft perimeter (Fig. 6A–D). This vascular pattern was prevalent in Rip-Cre;VEGF^fl/fl islet transplants even at 1 month posttransplantation (Fig. 6G and H), while vasculature was homogeneously distributed across the wild-type islet grafts (Fig. 6E and F). When revascularization was complete at 1 month, vessel density in the endocrine compartment of Rip-Cre;VEGF^fl/fl islet grafts was twofold lower than in wild-type islet transplants (Fig. 6I) and was similar to the vessel density of Rip-Cre;VEGF^fl/fl islets in the native pancreas (Fig. 3E). In both types of transplants, capillary size expressed as an area/vessel (Fig. 6J) was similar to that in exocrine pancreas (online appendix Fig. 1C). These data indicate that islet production of VEGF-A is required not only for normal vasculature of native islets in the pancreas but also for the vessel recruitment and revascularization of transplanted islets.

While endocrine and exocrine cells of the pancreas arise from the pancreatic duodenal homeobox-1–expressing progenitors, pancreatic islets become more vascularized than exocrine tissue. Furthermore, the microvessels in these two pancreatic compartments have a different capillary size, tortuousity, and vascular permeability. Assembly of a complex structure, such as vascularized islet, requires coordinated interactions of islet cells and endothelial cells. There is increasing evidence that during pancreas development, endothelial cells produce an instructive signal that induces pancreatic endocrine differentiation and morphogenesis (13,14). Recent studies by Edsbagge et al. (15) indicate that in addition to endothelial-endocrine cell signaling, functional blood vessels are required in subsequent steps of pancreatic development. The data presented here suggest that endocrine cell–endothelial cell communication is reciprocal, since early differentiating endocrine cells produce angiogenic factors including VEGF-A and Ang-1. Our studies utilizing intravital labeling with endothelium-binding lectin provide the first evidence that development of islet vasculature occurs concomitantly with islet morphogenesis, as coalescing endocrine cells are adjacent to blood-perfused capillaries before being assembled into the final structure of mature islet. Nevertheless, it remains to be determined how early in development islet arterioles and venules are specified and when islet microvasculature becomes fenestrated. The islet capillary network undergoes considerable remodeling in the early postnatal period, which coincides with a rapid increase in islet mass after birth, as recently demonstrated by Georgia and Bhushan (42).

Consistent with the higher vascularity in adult islets compared with exocrine tissue, angiogenic factors remain enriched in the endocrine cells into adulthood. It is intriguing, however, that the pattern of their expression is not uniform throughout the islet. We found that unlike VEGF-A and ephrin-A1, Ang-1 was expressed only in the β-cells. This unique pattern of Ang-1 expression could be associated with regulation of branching in arteriole(s) entering the islets. It is also possible that Ang-1 expression in β-cells counteracts VEGF effects on vascular permeability (49) or promotes interactions of endothelial cells with β-cells, as surrounding support cells, and extracellular matrix (50). Our data provide several lines of evidence pointing to distinct phenotypic characteristics of islet endothelium, which have also been recognized previously in other vascular beds (51,52). For example, VEGFR2 is highly expressed in intra-islet capillaries but downregulated in microvasculature of exocrine tissue and even more significantly in islet arterioles and venules. On the other hand, islet arterioles and venules express Ang-1 receptor, Tie2, at much higher levels than intra-islet capillaries. Additionally, as intra-islet capillaries become specified into postcapillary venules around the islet periphery, they have increased expression of the venous marker EphB4.

Although islets express several angiogenic factors (described in this study and by others [16–18,20]), our genetic data indicate that VEGF-A plays a unique role in islet vascularization. β-Cell–specific inactivation of VEGF-A led to substantial loss in islet vessel density, vessel size, and vascular permeability, while vasculature in exocrine tissue remained intact. It appeared that islet vascularization is quite sensitive to the level of VEGF-A expression. VEGF-A heterozygosity in β-cells (VEGF-A also remains expressed in non–β-cells) caused vascular abnormalities in the islets, suggesting haploinsufficiency. Thus, this approach allowed us to investigate how islet-specific vascular alternations relate to islet function, which could not be accomplished previously by global inactivation of VEGF-A in the pancreas (22).

In spite of normal pancreatic insulin content and β-cell mass, mice with reduced VEGF-A expression in β-cells had impaired GSIS, a phenotype similar to some defects in β-cell gene expression at the level of transcription factors, enzymes, or ion channels. Using vascular or diffusion delivery of β-cell secretagogues into the islets, we were able to show that insulin deficiency in these mice is not a result of β-cell dysfunction but rather caused by abnormalities in the islet vasculature. The decreased insulin secretory output in response to β-cell stimulus delivered through vascular route could be attributed to defects in several events that are difficult to discern. Reduced vessel density and vessel size likely cause less flow into the VEGF-A mutant islets, which will decrease overall flux of the stimulus into the islets and could lead to a lower or delayed β-cell response. It is also possible that islet vascular alternations could hamper exit of exocytosed insulin from the islets.

Previous studies (53,54) suggest that following β-cell exocytosis, insulin can take two separate pathways: either enter the extracellular space or directly enter intra-islet capillaries through fenestration in their endothelium. In normal islets, the majority of the β-cells are apposed to capillaries (55). Our 3D reconstructions showed, that in VEGF-A mutant islets, capillary support of β-cells is substantially reduced, and moreover, it is possible that the changes in the endothelial cell ultrastructure affect insulin transport across vascular endothelial cell barrier. The role of caveolae in transendothelial transport of plasma proteins is well-established (56), and a study by Bendayan (54) suggested that insulin can be internalized by caveolae at least in the endothelial cells of acinar parenchyma. Even if such caveolae-mediated transendothelial transport of insulin occurs in VEGF-A mutant islets, it is not known how rapidity and efficiency of transcytosis compares with the insulin transport through endothelial fenestrae.

Another possible explanation for attenuated insulin secretion in Rip-Cre;VEGF^fl/fl mice during vascular delivery of β-cell stimuli is that disruption of vascular cells in the islets perturbs cell-cell communication and information circuits that coordinate overall islet insulin secretory output. It is noteworthy that Rip-Cre;VEGF^fl/wt mice lost only 25% of islet vessel density, yet their glucose clearance was similar to Rip-Cre;VEGF^fl/fl mice. For example, there is growing evidence that in the brain, vascular endothelial cells actively participate in communication of calcium signals or calcium signal propagation, a feature that until recently was attributed only to glial cells, neurons, or astrocytes (57). Astrocyte–endothelial cell cocultures demonstrated that calcium signals could be communicated bidirectionally utilizing both gap junctions and ATP-mediated paracrine signaling through purinergic receptors (58,59). The components of both these signaling pathways are present in the β-cells and intra-islet endothelial cells (60,61). While there is evidence that β-cells communicate the calcium signal through gap junctions (62) and extracellular messengers (ATP) (63), it must be determined whether such signal communication also exists between β-cells and intra-islet endothelial cells. Additionally, Patel et al. (64) demonstrated that in intact liver, calcium signaling is coordinated by endothelial-derived nitric oxide and such coordination is lost following isolation of hepatocytes. Similar to isolated hepatocytes, in isolated islets subjected to cell perifusion studies, the close relationship between β-cells and endothelial cells is also disrupted since the intra-islet capillary network collapses rapidly during culture due to endothelial cell death and/or dedifferentiation (data not shown) (28). Taken together, our data indicate that β-cells and islet microvasculature participate cooperatively in the maintenance of glucose homeostasis. Factors modulating VEGF-A expression, perhaps glucose or islet-enriched transcription factors, may influence islet vascularity and, consequently, insulin levels in the systemic circulation. Thus, abnormalities in islet vascularization may underlie a new mechanism for impaired islet insulin output associated with diabetes.

The observation that reduced islet vascularization causes a pre-diabetic phenotype similar to some defects in β-cell gene expression has not only implications for islet development and function but also for revascularization of transplanted islets. Whether the revascularization of transplanted islets and vascularization of islets during development share similar mechanisms is not known, but our data provide evidence that the two processes share VEGF-A signaling. Even though the transplants of VEGF-A–deficient islets became partially revascularized, the blood vessels were mostly found around the graft perimeter. This suggests that reduced production of VEGF-A by β-cells not only reduces the number of intra-islet endothelial cells participating in graft vessel formation (26) but also limits recruitment of host endothelial cell and their invasion into the graft. On the other hand, vessel size was similar in transplants of wild-type and VEGF-A mutant islets and comparable with that in exocrine tissue, thus suggesting that this parameter in the transplant can be regulated in a VEGF-A–independent manner. Our observation that the vessel size was reduced to a similar extent in both types of transplants points out that even wild-type islet grafts have an inferior vascular supply compared with the islets in the native pancreas. Further studies will determine how formation of islet graft capillaries and their permeability is modulated by upregulation of VEGF-A.

This study was supported by a Merit Review Award from the VA Research Service, research grants from the National Institutes of Health (NIH; DK68764, DK63439, DK62641, and DK59637), and the Juvenile Diabetes Research Foundation International, the Vanderbilt Mouse Metabolic Phenotyping Center (DK59637), and the Vanderbilt Diabetes Research and Training Center (NIH DK20593). Electron microscopy was carried out in the Vanderbilt University Research Electron Microscopy Core of the cell imaging resource. This resource is partially supported by NIH grants CA68485, DK20593, and DK58404.

We thank Napoleone Ferrara for the VEGF^fl/fl mice.