Enhanced expression of VEGF-A in β cells increases endothelial cell number but impairs islet morphogenesis and β cell proliferation

Abstract

There is a reciprocal interaction between pancreatic islet cells and vascular endothelial cells (EC) in which EC-derived signals promote islet cell differentiation and islet development while islet cell-derived angiogenic factors promote EC recruitment and extensive islet vascularization. To examine the role of angiogenic factors in the coordinated development of islets and their associated vessels, we used a “tet-on” inducible system (mice expressing rat insulin promoter-reverse tetracycline activator transgene and a tet-operon-angiogenic factor transgene) to increase the β cell production of vascular endothelial growth factor-A (VEGF-A), angiopoietin-1 (Ang1), or angiopoietin-2 (Ang2) during islet cell differentiation and islet development. In VEGF-A overexpressing embryos, ECs began to accumulate around epithelial tubes residing in the central region of the developing pancreas (associated with endocrine cells) as early as embryonic day 12.5 (E12.5) and increased dramatically by E16.5. While α and β cells formed islet cell clusters in control embryos at E16.5, the increased EC population perturbed endocrine cell differentiation and islet cell clustering in VEGF-A overexpressing embryos. With continued overexpression of VEGF-A, α and β cells became scattered, remained adjacent to ductal structures, and never coalesced into islets, resulting in a reduction in β cell proliferation and β cell mass at postnatal day 1. A similar impact on islet morphology was observed when VEGF-A was overexpressed in β cells during the postnatal period. In contrast, increased expression of Ang1 or Ang2 in β cells in developing or adult islets did not alter islet differentiation, development, or morphology, but altered islet EC ultrastructure. These data indicate that 1) increased EC number does not promote, but actually impairs β cell proliferation and islet formation; 2) the level of VEGF-A production by islet endocrine cells is critical for islet vascularization during development and postnatally; 3) Angiopoietin-Tie2 signaling in endothelial cells does not have a crucial role in the development or maintenance of islet vascularization.

Introduction

Islets represent approximately 1–2% of pancreatic mass, but receive 6–20% of total pancreatic blood flow and up to 20 times more blood flow than pancreatic acinar tissue (Lifson et al., 1980; Lifson et al., 1985; Nyman et al., 2008). This intense vascularization allows the β cell to quickly sense and respond to changes in blood glucose by secreting insulin, a critical regulator of whole organismal metabolism. However, the molecular events and signals involved in islet vascularization and its relationship with islet development are incompletely understood.

Several growth factor families, most notably vascular endothelial growth factors, angiopoietins, basic fibroblast growth factor, and ephrins, and their receptors are known to play a role in vasculogenesis/angiogenesis (Carmeliet, 2003; Cines et al., 1998; Yancopoulos et al., 2000), but our understanding of their role in islet vascularization is incomplete. VEGF-A facilitates EC proliferation, migration and recruitment, and generally participates in the early phase of blood vessel formation by vasculogenesis or angiogenesis (Ferrara et al., 2003; Gale and Yancopoulos, 1999). Angiopoietins, including the counteracting molecules, Ang1 and Ang2, activate intracellular signaling pathways that result in EC migration, tube formation, vascular remodeling and maturation (Ramsauer and D’Amore, 2002). Increased Ang1 expression in the mouse ear also stabilizes the blood vessel leakage in response to increased VEGF-A expression (Thurston et al., 1999; Yancopoulos et al., 2000). Ang2 inhibits Ang1 signaling by competitive binding to the Tie2 receptor, and the interaction between Tie2 receptor and Ang2 allows blood vessels to regress when VEGF-A levels are low, and stimulates neovascularization in the presence of increased levels of VEGF-A (Lobov et al., 2002). Moreover, Ang1 and Ang2 can modulate cellular survival pathways in the absence of Tie2 through interaction with integrins (Carlson et al., 2001).

VEGF-A is the predominant growth factor expressed by islet endocrine cells (Bergers et al., 2000; Brissova et al., 2006; Christofori et al., 1995; Inoue et al., 2002). Mouse models of pancreas-specific gene deletion have found that VEGF-A is critical in maintaining normal intra-islet EC number, morphology, and fenestrations, and thus the vessel permeability within islets (Brissova et al., 2006; Lammert et al., 2003). β Cell-specific ablation of VEGF-A results in reduced and abnormal islet vascularity, reduced insulin output, and a pre-diabetic phenotype (Brissova et al., 2006; Lammert et al., 2003). In addition to a role in islet function, islet vascularization is important following islet transplantation, an experimental therapy for type1 diabetes. Intra-islet VEGF-A production also plays an important role in islet revascularization (Brissova et al., 2004; Brissova and Powers, 2008; Brissova et al., 2006; Nyqvist et al., 2005). Efforts that improve islet revascularization after transplantation appear to improve islet survival and function. For example, islets with increased VEGF-A expression had improved islet revascularization and more rapidly reversed diabetes (Lai et al., 2005; Sigrist et al., 2003; Su et al., 2007), but islet grafts with β cell-specific VEGF-A ablation had a slower rate of revascularization (Brissova et al., 2006). Increased Ang1 expression in transplanted islets provided cytoprotection, enhanced engraftment, and reduced islet apoptosis (Su et al., 2007).

As part of the reciprocal relationship between blood vessels and islet cells, ECs located near pancreatic and islet progenitor cells provide critical signals for islet cell differentiation (Cleaver and Melton, 2003; Edsbagge et al., 2005; Jacquemin et al., 2006; Lammert et al., 2001). For example, inhibition of embryonic aortic EC formation prevented pancreatic β cell insulin expression and differentiation, while co-culture of mouse pancreatic endoderm with aorta or umbilical artery promoted insulin expression (Lammert et al., 2001). Increased VEGF-A expression in the pancreas (using a Pdx1-VEGF-A transgene) led to a hypervascularized pancreas, increased islet number and islet area, and ectopic insulin expression in the stomach, suggesting that increased EC-derived signals promote islet cell proliferation and mass (Lammert et al., 2001).

To better understand this reciprocal signaling between pancreatic islet ECs and neighboring endocrine cells, we tested the effects of increased β cell-derived angiogenic factor expression on ECs during islet development and vascularization. We hypothesized that 1) overexpressed VEGF-A or Ang1 specifically in β cells would enhance islet-endothelial cell signaling, increase islet vascularization, and increase β cell proliferation and mass; 2) Increased expression of Ang2, an antagonist of Ang1, by β cells would reduce islet vascularization and impair β cell differentiation. To test these hypotheses, we used an inducible transgenic approach to overexpress VEGF-A, Ang1, or Ang2 in β cells and examined islet vascularization, development, and function. Surprisingly, increased expression of VEGF-A, but not Ang1 or Ang2, increased EC numbers but impaired β cell growth and islet morphology. Ang1 or Ang2 overexpression did not dramatically alter EC number but led to abnormal ultrastructure of ECs and impaired islet function.

Materials and methods

Mouse models and genotyping

An inducible “tet-on” (tetracycline-regulated) transgenic system was employed to allow temporal and dosage change of angiogenic factor expression by adjusting the time and the amount of inducer-Doxycycline (Dox), a tetracycline derivative. The transactivator mouse line contained a transgene in which the rat insulin promoter (Rip) drives β cell specific expression of reverse tetracycline activator (rtTA) (Milo-Landesman et al., 2001). The responder lines used the tet-O-CMV promoter to drive the expression of transgenes (VEGF-A, Ang1, or Ang2) (Bureau et al., 2006; Haninec et al., 2006; Ohno-Matsui et al., 2002). The Ang1 responder line included a bicistronic transgenic cassette hAng1-IRES-lacZ that produced human Ang1 and LacZ under the control of the tetracycline response element (Haninec et al., 2006). The Ang2 responder line had a c-Myc tag engineered to the C-terminal of Ang2 transgene (Bureau et al., 2006). VEGF-A, Ang1, or Ang2 overexpression mouse lines (Rip-rtTA;tet-O-VEGF-A; RIP-rtTA;tet-O-Ang1; Rip-rtTA;tet-O-Ang2) were generated by breeding Rip-rtTA^tg/tg (C57BL/6/CBA background) with tet-O-VEGF-A^tg/wt (C57BL/6 background), or tet-O-Ang1^tg/wt(CD-1 background) or tet-O-Ang2^tg/wt (CD-1 background). The transgene expression was activated after Dox (2mg/ml) was provided in light-protected drinking water containing 1% Splenda® (zero-calorie sucralose as a sweetener) during the following time periods: embryonic day (E) 5.5 to E10.5, E5.5 to E11.5, E5.5 to E12.5, E5.5 to E14.5, E5.5 to E16.5, E5.5 to P1, E5.5 to P7, E5.5 to P14, E5.5 to P28, E5.5 to 8 week, P1 to P7, P21 to P28, or 8–12 weeks of age (Figure 1A). Drinking water containing Splenda® with or without Dox was changed every two days. Inducible whole-body knockout of Ang1 was generated using ROSA-rtTA;tet-O-Cre;Ang1^fl/fl mouse lines with Dox (2 mg/ml) administered from E13.5 to P28, as described previously (Jeansson et al., 2011). Using tail genomic DNA, mice were genotyped by PCR with the following primer sequences: 5′ Ang1-primer-ATA GGA ACC AGC CTC CTC TCT; 3′ Ang1-primer-AAG GAC ACT GTT GTT GGT GGT A; 5′ Ang2 primer-GGG AAT GAG GCT TAC TCA TTG; 3′ Ang2 primer-GCT GGT CGG ATC ATC ATG GT; 5′ VEGF-A primer-TCG AGT AGG CGT GTA CGG; 3′ VEGF-A primer-GCA GCA GCC CCC GCA TCG; 5′ rtTA primer-GTG AAG TGG GTC CGC GTA CAG; and 3′ rtTA primer-GTA CTC GTC AAT TCC AAG GGC ATC G (Bureau et al., 2006; Ohno-Matsui et al., 2002; Ward et al., 2004). Animal studies were conducted with the approval of the Institutional Animal Care and the Use Committee at Vanderbilt University Medical Center.

Figure 1. Overexpression of VEGF-A in early development affects islet morphology and vascularization whereas overexpression of Ang1 and Ang2 has little effect.

(A) Mice were treated with Dox to induce angiogenic factor expression for the time periods shown by the arrows. (B–I) Immunostained sections of P1 stage pancreatic tissue from mice treated with Dox from E5.5 to P1. (B) Rip-rtTA; tet-O-Ang1 mice, insulin (Ins, blue) and Ang1 (red); (C) Rip-rtTA; tet-O-Ang1 mice, insulin (Ins, blue) and PECAM1 (green). (D) Rip-rtTA; tet-O-Ang2 mice, insulin (Ins, blue) and c-Myc (red; the Ang2 transgene has a Myc tag). (E) Rip-rtTA; tet-O-Ang2 mice, insulin (Ins, blue) and PECAM1 (green). (F) Rip-rtTA; tet-O-VEGF-A mice, insulin (Ins, blue) and VEGF-A (red); (G) Rip-rtTA; tet-O-VEGF-A, insulin (blue) and PECAM1 (green). (H) Rip-rtTA mice, insulin (Ins, blue) and PECAM1 (green). (I) Pancreatic insulin content at P1 in Rip-rtTA mice and Rip-rtTA;tet-O-VEGF-A mice treated with Dox from E5.5 to P1. Insulin content: Rip-rtTA, 528.5±8.1 ng/pancreas (n=5 mice); Rip-rtTA; tet-O-VEGF-A, 372.1± 23.9 ng/pancreas (n=8 mice); ***; p<0.001. (J) Pancreatic glucagon content at P1 stage from mice treated with Dox from E5.5 to P1; Glucagon content: Rip-rtTA, 22.1±1.7 ng/pancreas (n=5 mice); Rip-rtTA; tet-O-VEGF-A, 15.4±2.1 ng/pancreas (n=8 mice); *; p<0.05. (K) Percentage of β cell area per total pancreatic area; Rip-rtTA, 1.22±0.14% (n=3 mice); Rip-rtTA;tet-O-VEGF-A, 0.70±0.10% (n=3 mice); **; p<0.05. Scale bar in panel B represents 50 μm and also corresponds to panels D–H.

Pancreatic tissue collection

Mice were anesthetized by intraperitoneal injection of a 20 mg/kg xylazine and 80 mg/kg ketamine mixture. Postnatal pancreata were dissected and fixed in 4% paraformaldehyde (Electron Microscopy Sciences, Cat. #15710) in 100 mmol/l PBS for 90 min on ice (Brissova et al., 2006). Embryonic mouse pancreata were dissected and fixed in 4% paraformaldehyde in 10 mmol/l PBS for 4 hours on ice. After fixation, tissues were washed 4 times in 10 mmol/l PBS for 30 minutes, and then equilibrated overnight at 4°C in 30% sucrose in 10 mmol/l PBS. Then the tissues were preserved in Tissue-Tek Optimal Cutting Temperature compound (Sakura Finetek, Cat. #4583) and Tissue-Tek Cryomold Standard (Sakura Finetek, Cat. #4557) at −80°C and sectioned at 5 μm or 10 μm on a Leica CM 3050 S cryostat.

Immunofluorescence and immunohistochemistry

Tissue cryosections were air-dried at room temperature for 5 minutes and then post-fixed in 1% paraformaldehyde/10 mmol/l PBS for 10 minutes. Sections were washed for 15 minutes in PBS with three changes, permeabilized in 0.2% Triton X-100 for 10 minutes (room temperature), and then washed again in PBS with three changes for 15 minutes. The sections were blocked with 5% normal donkey serum for 90 minutes and then incubated with the primary antibody diluted in 10 mmol/l PBS containing 1% BSA and 0.1% Triton x-100 overnight at 4°C. These primary antibodies were used for detecting antigens in tissue: polyclonal guinea pig anti-swine insulin (DakoCytomation), 1:500; rabbit anti-glucagon (Cell Signaling), 1:5000; rat anti-mouse CD31 (BD Biosciences), 1:100; goat anti-VEGF-A (R&D systems), 1:200; goat anti-hAng1 (R&D systems), 1:1000; biotinylated dolichos biflorus agglutinin (DBA) (Vector laboratories), 1:200; rabbit anti-myc (GenScript), 1:100; rabbit anti-Ki67 (Abcam), 1:500; rabbit anti-MafA (a generous gift from Roland Stein, Vanderbilt University, TN, USA), 1:1000; rabbit anti-Pdx1 (a generous gift from Christopher Wright, Vanderbilt University, TN, USA),1:2000; rabbit anti-Glut2 (alpha Dignostic Intl), 1:100; Goat anti-Ngn3 (a generous gift from Guoqiang Gu, Vanderbilt University, TN, USA) 1:500; rabbit anti-Nkx6.1 (antibody core of the Beta Cell Biology Consortium, Nashville, TN, USA) 1:800; rabbit anti-amylase (Sigma), 1:100; Goat anti-Cpa1 (R&D systems) 1:250; rabbit anti-Sox9 (Millipore) 1:2000; rabbit anti-Pax6 (Covance) 1:500; mouse anti-E-Cadherin (BD Biosciences) 1:500; sheep anti-somatostatin (American Research Products, Inc.) 1:500. Following incubation with the primary antibody, sections were washed in 0.1% Triton X-100/PBS with three changes for 15 minutes, and then incubated with secondary antibodies diluted in 10 mmol/l PBS containing 1% BSA and 0.1% Triton X-100 for 1 hour at room temperature. Secondary antibodies were conjugated with Cy2, Cy3, and Cy5 at 1:200, 1:500, and 1:200 dilutions, respectively (Brissova et al., 2004). Images were acquired using an Olympus BX-41 fluorescence microscope (Olympus America Inc. Center Valley, PA, USA), a Zeiss LSM510 META laser scanning confocal microscope (Carl Zeiss Microimaging, Thornwood, NY, USA), or Scan Scope FL (Aperio Technologies, Inc., Vista, CA, USA). To examine the vascularization of islets, 5 or 10 μm sections were collected 250 μm apart in adult pancreata, and continually collected in embryos, at P1, or P7 stages. Vessel density and area per vessel of 30–40 islets per pancreatic tissue block were analyzed using MetaMorph version 7.1 software (Universal Imaging) (Brissova et al., 2006; Weidner, 1995; Weidner et al., 1991).

Transferase-mediated dUTP nick-end labeling (TUNEL) assay

Apoptotic cells in pancreatic tissue were detected by terminal deoxynucleotidyl transferase-mediated dUTP nick-end labeling (TUNEL) using ApopTag In Situ Apoptosis Detection Kits (Chemicon international, Cat. #S7165). Pancreatic tissues were collected and fixed as described above. Tissue sections were post-fixed with 1% paraformaldehyde (Electron Microscopy Sciences, Cat. #15710) for 10 minutes at room temperature, permeabilized with 0.5% Triton X-100 for 15 minutes on ice, washed three times with 10 mmol/l PBS, pH 7.4, and incubated for 1 hour at 37°C with TUNEL working strength TdT enzyme reaction mixture. After washing three times with 10 mmol/l PBS, tissue sections were incubated at room temperature with anti-digoxigenin conjugate (rhodamine) for 30 min. TUNEL-positive nuclei were visualized with an Olympus BX41 fluorescence microscope using standard fluorescein excitation and emission filters.

Transmission electron microscopy (TEM)

Pancreatic tissues were dissected, quickly rinsed in rinse buffer (0.1M sodium cacodylate/1% CaCl₂), minced into 0.5 mm pieces, and fixed in Karnovsky’s fixative buffer (2% paraformaldehyde/2.5% glutaraldehyde made in 0.1 M sodium cacodylate/1% CaCl₂, pH 7.4) at room temperature for 1 hour, and then kept in fixative buffer overnight at 4°C. After post-fixation with 1% osmium tetroxide, tissues were dehydrated through this series of ethanol concentrations: 25% ethanol, 10 minutes; 50% ethanol, 10 minutes; 70% ethanol, 10 minutes; two changes of 95% ethanol, 10 minutes each; three changes of 100% ethanol, 5 minutes/each. Dehydrated tissues were then embedded in Spurr resin for electron microscopy. To locate the endocrine cell area of interest for future ultra-thin sectioning, semi-thin sections (500 nm) were stained with toluidine blue and analyzed by light microscopy. Ultrathin sections (60–70 nm) were collected on electron microscopy grids and stained with uranyl acetate and lead citrate. Images were captured with an AMT digital camera system using a Phillips CM-12 Transmission Electron Microscope at an operating voltage of 80 KeV (Brissova et al., 2006).

Islet Isolation

Rip-rtTA;tet-O-Ang1, Rip-rtTA;tet-O-Ang2, and Rip-rtTA mice were anesthetized by intraperitoneal injection of a mixture of ketamine and xylazine (80 and 20 mg/kg, respectively). Pancreata were injected through the bile duct with 3.5 ml of 0.6 mg/ml collagenase P in Hanks Balanced Salt Solution (HBSS) and digested in 6.3 ml of 0.6 mg/ml collagenase in HBSS for 8 min at 37°C using a wrist-action shaker. Islets were purified by hand-picking under microscopic guidance (Brissova et al., 2006).

Islet perifusion

Isolated islets from Rip-rtTA;tet-O-Ang1, Rip-rtTA;tet-O-Ang2, and Rip-rtTA mice induced with Dox from E5.5 to 2 months of age were tested for insulin secretion in the dynamic cell perifusion system (Brissova et al., 2006). Fifty islet equivalents were placed in a chamber and washed under baseline media for 30 minutes prior to the experiment. Islets were then perifused for 9 minutes with 5.6 mM glucose (Sigma, D16), followed by 9 minutes with 16.7 mM glucose, 21 minutes with 5.6 mM glucose, 9 minutes with 16.7 mM glucose and 100 μM isobutylmethylxanthine (Sigma, I5879-1G), and 21 minutes 5.6 mM glucose. The effluent fractions were collected at 3-minute intervals using an automatic fraction collector. The insulin content of each fraction was detected by radioimmunoassay (RI-13K, Millipore) (Brissova et al., 2006).

Enzyme-linked immunosorbent assay (ELISA) for Ang1 and Ang2

Freshly isolated islets were cultured in an eight-well slide chamber (50–70 islets/well in 470 μl RPMI-1640 media) for 48 hours at 37°C. Ang1 and Ang2 production by isolated islets was measured using human Ang1- or Ang2-specific enzyme-linked immunosorbent assay (R&D Systems) (Brissova et al., 2006).

Glucose Tolerance Testing

Eight and 12-week-old mice, fasted for 16 hours, were injected intraperitoneally with filter-sterilized glucose (2 g/kg body weight) prepared in 10 mmol/l PBS. Blood glucose concentration was measured in tail vein blood using a glucometer at six time points following injection: 0, 15, 30, 60, 90, and 120 minutes (Brissova et al., 2006).

Measurement of pancreatic insulin and glucagon content

Pancreata from mice at P1 and P7 were dissected in ice-cold 10 mmol/l PBS and placed in 1 ml vial containing 0.5 ml of acid-alcohol. Tissues were extracted for 48 hours at 4°C using a nutator. The amount of insulin or glucagon in extract was measured by radioimmunoassay (insulin, RI-13K; glucagon, GL-32K; Millipore) as previously described (Brissova et al., 2002).

Statistical analysis

Data were expressed as mean ± standard error of the mean. Group comparisons were performed using unpaired t-test or one-way ANOVA with Newman-Keuls multiple comparison test. The level of significance was set at p < 0.05 (*; p < 0.05, **; p < 0.01, and ***; p < 0.001).

Results

β Cell-specific overexpression of angiogenic factors induced during islet development alters islet vascularization

To investigate the effects of VEGF-A, Ang1, or Ang2 overexpression on islet development, we treated Rip-rtTA;tet-O-VEGF-A, Rip-rtTA;tet-O-Ang1, Rip-rtTA;tet-O-Ang2, and their littermate controls (Rip-rtTA) with Dox continuously from embryonic day 5.5 (E5.5) to postnatal day 1 (P1) (Figure 1A). Dox treatment increased hAng1 expression in β cells of Rip-rtTA;tet-O-Ang1 mice, increased Ang2 as reflected by myc expression in β cells of Rip-rtTA;tet-O-Ang2 mice, and increased VEGF-A expression in β cells of Rip-rtTA;tet-O-VEGF-A mice (Figure 1B, D, F). In contrast, we did not detect hAng1 expression, myc expression, or increased VEGF-A expression in β cells of Rip-rtTA mice (data not shown).

β Cell-specific VEGF-A overexpression dramatically increased islet vascularization, and altered islet morphology (Figure 1F, G and Supplemental Figure 1A–D). In addition, pancreatic insulin and glucagon content in newborn Rip-rtTA;tet-O-VEGF-A mice was reduced by 29% and 30%, respectively compared to controls (Figure 1I, J). β Cell-specific overexpression of VEGF-A also reduced β cell area (Figure 1K). While insulin content may not always correspond to β cell number (Olsson and Carlsson, 2011; Weir and Bonner-Weir, 2011), these results suggest that the reduced insulin content reflects decreased β cell number. In contrast, mice overexpressing Ang1 or Ang2 had normal islet morphology and vascularization (Figure 1B–E). Neither Ang1 nor Ang2 overexpression altered pancreatic insulin content at P1 (Supplemental Figure 1E, F). These data indicate that overexpression of VEGF-A, but not Ang1 or Ang2, affects islet development.

β Cell-specific overexpression of VEGF-A affects β cell proliferation

Reduced β cell number in Rip-rtTA;tet-O-VEGF-A neonatal pancreas could result from defects in endocrine cell differentiation and/or β cell turnover. To determine whether β cell-specific overexpression of VEGF-A affected endocrine progenitor cell development, we treated Rip-rtTA;tet-O-VEGF-A and Rip-rtTA mice with Dox continuously from E5.5 to when pancreatic tissues were collected at E10.5, E11.5, E12.5 and E14.5. Biological effects of VEGF-A overexpression were monitored by EC labeling in a whole mount of developing pancreas and optical sectioning through the entire pancreatic tissue. At E10.5, ECs surrounded pancreatic buds but there was no difference in tissue vascularization between Rip-rtTA;tet-O-VEGF-A embryos and controls (Supplemental Figure 2A, B). Increased number of ECs surrounding insulin⁺ cells was first detected in Rip-rtTA;tet-O-VEGF-A pancreas at E12.5 (Supplemental Figure 2C, D) and the density of PECAM1⁺ vascular structures adjacent to Pdx1^HI cell clusters increased progressively at E14.5 and E16.6 (Supplemental Figure 2E–H). The trunk regions in Rip-rtTA;tet-O-VEGF-A pancreas became hypervascularized at the onset of secondary transition (Supplemental Figure 2E, F) and harbored fewer Ngn3⁺ endocrine progenitors than Rip-rtTA controls (Figure 2A–C). However, we did not note any change in the expression pattern of Sox9 or the tip cell marker Cpa1 (Figure 2D, E). Thus, these results reveal that overexpression of VEGF-A in β cells inhibits the development of pancreatic endocrine cells.

Figure 2. β cell-specific overexpression of VEGF-A increases endothelial cell proliferation but decreases number of Ngn3⁺ endocrine progenitors and β cell proliferation.

(A and B) Pancreatic sections of Rip-rtTA and RIP-rtTA;tet-O-VEGF-A mice at E14.5 (Dox treatment from E5.5 to E14.5) were immunostained for E-cadherin (green), and Ngn3 (red). (C) Percentage of Ngn3⁺ cells/E-cadherin⁺ cells at E14.5: Control Rip-rtTA, 16.4±1.0 (n=3 mice); Rip-rtTA; tet-O-VEGF-A, 12.1±1.0 (n=3 mice); *; p<0.05.

(D and E) Pancreatic sections of Rip-rtTA and RiP-rtTA;tet-O-VEGF-A mice at E14.5 (Dox treatment from E5.5 to E14.5) were immunostained for Cpa1 (green), and Sox9 (red). (F and G) Pancreatic sections of Rip-rtTA and RIP-rtTA;tet-O-VEGF-A mice at E16.5 stage (Dox treatment from E5.5 to E16.5) were immunostained for insulin (Ins, green), TUNEL positive cells (red) and counterstained with DAPI (blue). (H and I) Pancreatic sections of Rip-rtTA and RIP-rtTA;tet-O-VEGF mice at P1 (Dox treatment from E5.5 to P1) were immunostained for insulin (Ins, blue), Ki67 (red) and PECAM1 (green). (J) Rate of β-cell proliferation at E16.5 and P1 (Dox treatment from E5.5 to E16.5 or P1). Rip-rtTA (E16.5), 4.0±1.0% (n=3 mice); Rip-rtTA;tet-O-VEGF-A (E16.5), 0.08±0.05% (n=3 mice); Rip-rtTA (P1), 11.3±1.7% (n=4 mice); Rip-rtTA;tet-O-VEGF-A (P1), 2.4±0.5% (n=4 mice); **; p<0.01 (K) Rate of endothelial cell proliferation at P1 (Dox treatment from E5.5 to P1). Rip-rtTA, 14.4±3.8% (n=4 mice); Rip-rtTA;tet-O-VEGF-A, 88.2±6.1% (n=4 mice); ***; p<0.001. Insets in A–I show magnification of boxed area in corresponding panels. Scale bars in panels A–E represent 50 μm. Scale bar in panel F represents 50 μm and also corresponds to panels G–I.

To further explore whether reduced β cell number in Rip-rtTA;tet-O-VEGF-A neonatal pancreas was caused by other processes than impaired endocrine cell differentiation, we measured β cell proliferation and apoptosis at E16.5 and P1. VEGF-A overexpression did not increase β cell apoptosis at P1, as measured by the TUNEL assay (Figure 2F, G). In contrast, β cell proliferation was decreased significantly in VEGF-A overexpressing mice at E16.5 and P1, while ECs adjacent to the developing islet clusters underwent robust proliferation (Figure 2H–K). Additionally, the fraction of β cells overexpressing VEGF-A decreased over time (E16.5, P1, and P7) (Supplemental Figure 1G). These data indicate that β cell-specific VEGF-A overexpression increased EC proliferation, but reduced β cells proliferation, led to reduced β cell mass at birth.

β Cell-specific overexpression of VEGF-A affects islet formation during embryonic and postnatal development

To examine how overexpression of VEGF-A affected islet development and altered islet formation, we treated Rip-rtTA;tet-O-VEGF-A and Rip-rtTA mice with Dox continuously from E5.5 to when pancreatic tissues were collected at E16.5, P1, and P7 (Figure 1A). At E16.5, α and β cells formed relatively large clusters in control Rip-rtTA embryos, but in Rip-rtTA;tet-O-VEGF-A mice, α and β cells were more scattered and remained adjacent to dolichos biflorus agglutinin-positive (DBA⁺) ductal structures (Figure 3A, B, and Supplemental Figure 3A–D). Co-labeling for insulin, glucagon and E-cadherin at E16.5 showed that defective α and β cell clustering in Rip-rtTA;tet-O-VEGF-A embryos was not related to islet cell delamination (Supplemental Figure 3E–H). The expression pattern of endocrine marker Pax6 was similar in Rip-rtTA and Rip-rtTA;tet-O-VEGF-A mice (Supplemental Figure 3E–H). At birth, the islet clusters in Rip-rtTA;tet-O-VEGF-A mice were distinctly smaller compared to controls (Figure 3C, D) and the normal islet formation was even more perturbed at P7 (Figure 3E, F and Supplemental Figure 4). PECAM1 labeling revealed that layers of ECs accumulated around DBA⁺ epithelial structures residing in the central region of the developing pancreas (associated with endocrine cells) at E16.5 (Figure 3G, H), and this EC accumulation continued at birth and postnatally (Figure 3I–L, and Supplemental Figure 4). We noted that at P7 presumptive islet areas were largely replaced by endothelial cells (Supplemental Figure 4A, B) without any significant effects on exocrine tissue (Supplemental Figure 4E). Furthermore, the number α and β cells associated with DBA⁺ ducts in Rip-rtTA;tet-O-VEGF-A mice at P7 was several fold higher compared to controls (Supplemental Figure 4C–G). These data suggest that overexpression of VEGF-A in β cells affects endocrine cell clustering and further perturbs islet formation.

Figure 3. Overexpression of VEGF-A during islet development dramatically increases islet vascularization, but decreases β cell number.

Pancreatic sections of Rip-rtTA and Rip-rtTA;tet-O-VEGF-A mice were immunostained at different developmental stages (Dox treatment from E5.5 to E16.5, P1, or P7). (A–F) Pancreatic sections of Rip-rtTA and Rip-rtTA;tet-O-VEGF-A mice at E16.5, P1 and P7 labeled for insulin (Ins, blue), glucagon (Glu, green), and DBA ductal marker (red). (D–F) Pancreatic sections of Rip-rtTA;tet-O-VEGF-A at E16.5, P1 and P7 stages: Insulin (Ins, blue), Glucagon (Glu, green), and DBA ductal marker (red). (G–L) Pancreatic sections of Rip-rtTA and Rip-rtTA;tet-O-VEGF-A mice at E16.5, P1 and P7 were labeled for insulin (Ins, blue), PECAM1 (green), and DBA ductal marker (red). Arrows in panel B, D and F point to β cell associated with DBA⁺ ducts. Scale bar in panel L represents 50 μm and also corresponds to panels A–J.

We next investigated consequences of continuous VEGF-A overexpression in β cells during the postnatal period. Rip-rtTA;tet-O-VEGF-A and Rip-rtTA mice were treated with Dox starting from E5.5 until P28. VEGF-A overexpressing Rip-rtTA;tet-O-VEGF-A mice were viable, had impaired glucose tolerance, severely altered islet morphology, but normal fasting glucose levels at P28 (Figure 4A, B, and Supplemental Figure 5). The vast majority of β cells in Rip-rtTA;tet-O-VEGF-A mice at this time point were adjacent to or residing in ductal epithelium expressing high levels of E-cadherin (Supplemental Figure 5C–F). Although β cells overexpressing VEGF-A declined over time (Supplemental Figure 1G), they were still present in the P28 pancreas (Figure 4E–G). The expression of endocrine cell differentiation markers such as Pdx1 and Nkx6.1 in VEGF-A overexpressing β cells was similar to those with basal VEGF-A expression in Rip-rtTA;tet-O-VEGF-A and Rip-rtTA mice (Figure 4E–H), but VEGF-A overexpressing β cells had substantially reduced expression of MafA (Figure 4I, J).

Figure 4. Continuous and short-term overexpression of VEGF-A in β cells during postnatal period leads to altered islet morphology.

(A, B) Pancreatic sections of Rip-rtTA and Rip-rtTA;tet-O-VEGF mice treated with Dox from E5.5 to P28; insulin (Ins, blue); PECAM1 (green); glucagon (Glu, red). (C) Pancreatic sections of Rip-rtTA;tet-O-VEGF mice treated with Dox from P1 to P7 were labeled for insulin (Ins, blue), PECAM1 (green), and glucagon (Glu, red) (D) Pancreatic sections of Rip-rtTA;tet-O-VEGF mice treated with Dox from P21 to P28 were labeled for insulin (Ins, blue), PECAM1 (green), and glucagon (Glu, red). (E–J) Pancreatic sections of Rip-rtTA and Rip-rtTA;tet-O-VEGF mice treated with Dox from E5.5 to P28. (E, F) Insulin (Ins, green), Pdx1 (red), and VEGF-A (blue); (G, H) Insulin (Ins, green), Nkx6.1 (red), and VEGF-A (blue); (I, J) Insulin (Ins, green), MafA (red), and VEGF-A (blue). Insets in E–J show magnification of boxed area in corresponding panels. In panels F, H, and J, the left and right insets represent β cells with basal VEGF-A expression and VEGF-A overexpressing β cells, respectively. Scale bar in panel A represents 50 μm and corresponds to panels B–J.

To address whether the different timing of VEGF-A expression is important for postnatal islet development, we treated Rip-rtTA and Rip-rtTA;tet-O-VEGF-A mice with Dox at two different postnatal stages, from P1 to P7 or from P21 to P28. Increased VEGF-A production by β cells at either stage resulted in altered islet morphology with α and β cell clusters surrounded by an increased number of endothelial cells and a reduced β cell number. (Figure 4C, D). However, the islet morphology was less perturbed during this brief period of VEGF-A overexpression compared to prolonged administration of Dox from E5.5 to P7 or P28. These results indicate that the tight control of VEGF-A expression is required for both embryonic and postnatal islet development.

β Cell-specific overexpression of VEGF-A affects endothelial cell ultrastructure

VEGF-A is a potent modulator of vascular permeability. To investigate the effects of VEGF-A overexpression on EC ultrastructure during islet development, we utilized transmission electron microscopy (TEM) to examine islet morphology at P1. In Rip-rtTA controls, ECs formed a thin monolayer around a functional capillary lumen and were surrounded by endocrine cells (Figure 5A). In contrast, islet-associated ECs in Rip-rtTA;tet-O-VEGF-A mice were tightly packed, forming multiple layers surrounding endocrine cell clusters (Figure 5B), and the ductal cells that were associated with β cells (Supplemental Figure 6A). In contrast to the tight EC junctions forming a complete capillary lumen in control mice, gaps between ECs were noted in islets from Rip-rtTA;tet-O-VEGF-A mice, and red blood cells were sometimes found outside the capillary lumen (Supplemental Figure 6B), suggesting vessel leakiness in Rip-rtTA;tet-O-VEGF-A mice. EC fenestrations were not found at the P1 stage in either Rip-rtTA;tet-O-VEGF-A or Rip-rtTA control mice, indicating that this specialization of intra-islet ECs develops postnatally. However, VEGF-A overexpression in Rip-rtTA;tet-O-VEGF-A mice led to a marked increase in caveolae at the luminal surface of ECs (Figure 5C, D). These data suggest that the level of VEGF-A affects caveolae development, vessel formation, and proper islet development.

Figure 5. β cell-specific overexpression of VEGF-A alters the ultrastructure of endothelial cells.

Transmission electron microscopy images show endocrine cells with adjacent endothelial cells. (A) Rip-rtTA mice; scale bar, 2 μm. (B) Rip-rtTA;tet-O-VEGF-A mice; scale bar, 2 μm. Boxed areas in panel A and B are magnified in C and D, respectively. (C) Rip-rtTA mice; scale bar, 500 nm. (D) Rip-rtTA;tet-O-VEGF-A mice; scale bar, 500 nm. Open arrowheads in C and D point to caveolae. BM, basement membrane; LM, vessel lumen; EC, endothelial cell; RBC, red blood cell. α, α cell; β, β cell; δ, δ cell.

β Cell-specific overexpression of Ang1 or Ang2 does not affect islet morphology, but affects intra-islet endothelial cell ultrastructure and islet function

To dissect the role of Tie2 signaling in islet vascularization, we overexpressed Ang1 and Ang2 in β cells not only during embryonic development, but also postnatally and in adult mice. Dox induction increased the number of β cells expressing Ang1 (Figures 6A–C) and the amount of the secreted Ang1 protein (Figure 6D). Islet morphology, islet vessel area, and islet density were unchanged in Rip-rtTA;tet-O-Ang-1 mice treated with Dox from either from E5.5 to 8 weeks (Figure 6E–H) or from 8 to 12 weeks (Figure 6I–L). In addition, Ang1 overexpression did not affect islet function in female Rip-rtTA;tet-o-Ang-1 mice (treated with Dox from E5.5 to 8 weeks or from 8 to12 weeks), or in adult male Rip-rtTA;tet-O-Ang-1 mice (treated with Dox from 8 to12 weeks), as assessed by intraperitoneal glucose tolerance testing (GTT) (Supplemental Figure 7A, B, D, E). Young male mice treated with Dox from E5.5 to 8 weeks showed slightly impaired glucose tolerance (Supplemental Figure 7A). This data suggested that β cell specific overexpression of Ang1 slightly impairs islet insulin secretion. The notion that angiopoietin-Tie2 signaling in endothelial cells does not play a crucial role in islet vascularization is further supported by recent studies of Jeansson et al. demonstrating that mice with Ang1 deletion after E13.5 did not have a vascular phenotype (Jeansson et al., 2011). Using this Rosa-rtTA;tet-O-Cre;Ang1^fl/fl system, Ang1 inactivation did not affect islet morphology, vascularization, or function (Supplemental Figure 7F–H).

Figure 6. Overexpression of Ang1 in the adult pancreas has little effect on islet morphology, vessel area, and vessel density.

(A, B) Pancreatic sections of Rip-rtTA;tet-O-Ang1 mice were immunostained with anti-insulin (Ins, green) and anti-β-galactosidase (β-gal, red). The Ang1 transgene has a bicistronic cassette that encodes Ang1 and the β-gal reporter. (A) Mice without Dox. (B) Mice treated with Dox from E5.5 to 8 weeks. (C) The percentage of β-gal⁺ β cells in No Dox, 10±2% (n=3 mice) and Dox group, 72±2% (n=3 mice); ***; p<0.001. (D) Islets from No Dox and Dox group were isolated and cultured for 48 hours. The amount of secreted Ang1 in the culture media was quantified by ELISA: No Dox, 0.67±0.03 pg/islet/48h (n=3 mice); Dox, 4.8±1.0 pg/islet/48h (n=3 mice); *; p<0.05. (E, F) Immunostaining of pancreatic sections using anti-insulin (Ins, green) and anti-PECAM1 (red) in Rip-rtTA;tet-O-Ang1 mice in the absence or presence of Dox treatment. (G) Islet vessel density was calculated by dividing the total number of vessels over the total islet area: Control Rip-rtTA, 1,682±71 vessels/mm² (n=3 mice); Rip-rtTA; tet-O-Ang1, 1,570±43 vessels/mm² (n=3 mice); p>0.05. (H) Area/Vessel in islets was calculated by dividing the total area of islet vessels over the number of islet vessels: Control Rip-rtTA, 65.5±3.8 μm² (n=3 mice); Rip-rtTA; tet-O-Ang1, 66.6±3.0 μm² (n=3 mice), p>0.05. (I, J) Pancreatic sections of Rip-rtTA and Rip-rtTA;tet-O-hAng1 mice were immunostained with anti-insulin (Ins, green) and anti-PECAM1 (red). Both were treated with Dox from 8 weeks to 12 weeks of age. (K) Islet vessel density in Rip-rtTA, 1,626±34 vessels/mm² (n=3 mice) and Rip-rtTA;tet-O-Ang-1, 1,643±38 vessels/mm² (n=3 mice); p>0.05. (L)Area/Vessel in Rip-rtTA, 77±3 μm² (n=3 mice) and Rip-rtTA;tet-O-Ang-1, 78±3 μm² (n=3 mice); p>0.05.

To further investigate Tie2 signaling, we overexpressed a myc-tagged Ang2, the antagonist of Ang1, in β cells (Figure 1C). Rip-rtTA;tet-O-Ang2 mice treated with Dox from E5.5 to 8 weeks had an increased number of Ang2-overexpressing β cells and increased amount of secreted Ang2 protein compared with untreated Rip-rtTA;tet-O-Ang2 mice (Figure 7A–D). Islets from Rip-rtTA;tet-O-Ang2 mice treated with Dox from E5.5 to 8 weeks of age had normal morphology, but a slight increase in vessel density and vessel area (Figure 7E–H). No changes in islet vasculature were detected in adult Rip-rtTA;tet-O-Ang-2 mice treated with Dox from 8 to 12 weeks of age (Figure 7I–L). Only male Rip-rtTA;tet-O-Ang2 with Dox from E5.5 to 8 weeks of age displayed slightly impaired glucose clearance (Supplemental Figure 8) and increased islet vessel area and density (Figure 6G, H). This data suggests that β cell specific overexpression of Ang2 only slightly impairs insulin secretory output and slightly alters islet vascularization.

Figure 7. Overexpression of Ang-2 in the adult pancreas has little effect on islet morphology but slightly enhances vessel area and density.

(A and B) Pancreatic sections of Rip-rtTA;tet-O-Ang2 mice were immunostained with anti-insulin (Ins, green) and anti-myc (red) as Ang2 has a c-myc tag. (A) Mice without Dox; (B) Mice treated with Dox from E5.5 to 8 weeks of age. (C) The percentage of myc positive β cells in No Dox, 35±1% (n= 3 mice) and Dox, 80±1% (n=3 mice); ***; p<0.001. (D) Islets from both groups were isolated and cultured for 48 hours. The amount of secreted Ang2 in the culture media was quantified by ELISA in No Dox, 116±20 pg/islet/48h (n=3 mice) and Dox, 1,580±582 pg/islet/48h (n=3 mice); *; p< 0.05. (E, F) Immunostaining of pancreatic sections using anti-insulin (Ins, green) and anti-PECAM-1 (red) in Rip-rtTA;tet-O-Ang2 mice in the absence or presence of Dox treatment. (G) Islet vessel density: No Dox, 1,526±24 vessels/mm² (n=3 mice); Dox, 1,660±60 vessels/mm² (n=3 mice); *; p<0.05. (H) Area/vessel in No Dox, 119±4 μm² (n=3 mice) and Dox, 140±7 μm² (n=3 mice); **; p<0.01. (I, J). Pancreatic sections of Rip-rtTA and Rip-rtTA;tet-O-Ang2 mice were immunostained with anti-insulin (Ins, green) and anti-PECAM1 (red). Both were treated with Dox from 8 weeks to 12 weeks of age. (K) Islet vessel density: Rip-rtTA, 1,283±55 vessels/mm² (n=3 mice); Rip-rtTA;tet-O-Ang2, 1,353±42 vessels/mm² (n=3 mice); p>0.05. (L) Area/vessel: Rip-rtTA, 116±8 μm² (n=3 mice); Rip-rtTA;tet-O-Ang2, 102±6 μm² (n=3 mice); p>0.05.

To determine whether overexpression of Ang1 or Ang2 affected islet function due to changes in β cell insulin secretion, we performed islet perifusion assays. Islets from Rip-rtTA;tet-O-Ang1, Rip-rtTA;tet-O-Ang-2, and Rip-rtTA mice treated with Dox from E5.5 to 8 weeks of age had a similar insulin secretory profile in response to 16.7 mM glucose and 16.7 mM glucose + 100 μM IBMX as control islets (Supplemental figure 7C and 8C). These results suggested that β cell-specific overexpression of Ang1 or Ang2 did not affect β cell insulin secretion per se but the slight in vivo changes in insulin secretory output resulting in delayed glucose clearance might be associated with alterations in intra-islet EC morphology.

To investigate the effect of Ang1 or Ang2 overexpression on ECs, we examined the ultrastructure of intra-islet ECs using transmission electron microscopy. Islet endothelial cells from Rip-rtTA;tet-O-Ang-1 or Rip-rtTA;tet-O-Ang-2 treated with Dox from E5.5 to 8 week of age had a loss of fenestrations and an increased amount of caveolae (Figure 8A–C). This data suggest that β cell specific overexpression of Ang1 or Ang2 arrested formation of endothelial cell fenestrations (mostly likely at the caveolae stage); this may reduce vessel permeability leading to slightly impaired insulin delivery into the peripheral circulation.

Figure 8. β cell-specific overexpression of Ang1 or Ang2 alters endothelial cell ultrastructure.

Transmission electron microscopy images of mice treated with Dox from E5.5 to 8 weeks. (A) Rip-rtTA mice, scale bar, 500 μm. (B) Rip-rtTA;tet-O-Ang-1 mice, scale bar, 500 μm. (C) Rip-rtTA;tet-O-Ang-2 mice, scale bar, 500 μm. Closed arrowheads in A point to fenestrations. Open arrowheads in B and C point to caveolae. LM, vessel lumen; RBC, red blood cell.

Discussion

To better understand the reciprocal relationship between islet endocrine cells and nearby endothelial cells and the coordinated development of the islet and its vasculature, we investigated the effects of β cell-specific overexpression of the angiogenic factors VEGF-A, Ang1, and Ang2 during critical times of islet development. We found that VEGF-A overexpression during islet development: 1) increased endothelial cell proliferation and vascularization near insulin-positive cells, 2) reduced β cell proliferation and mass leading to reduced insulin and glucagon content, and 3) altered islet morphology. In VEGF-A overexpressing islet areas, endocrine cells were dispersed and an increased number of α and β cells remained associated with developing pancreatic ducts. Islet formation progressively worsened with continuous overexpression of VEGF-A by β cells from early development throughout postnatal period (E12.5-P28) and perturbed islet morphology was seen even during a brief period of postnatal VEGF-A overexpression. In contrast, overexpression of Ang1 or Ang2 and global inactivation of Ang1 did not affect islet mass or morphology, but subtle alterations in intra-islet endothelial cell fenestrations, and slightly impaired insulin delivery were observed in Ang1 or Ang2 overexpressing islets. These results indicate that a tight level of VEGF-A expression in islet endocrine cells is critical for regulation of islet development and formation of normal islet capillary network during embryonic development and postnatal life, while angiopoietin-Tie2 signaling plays a minor role in maintenance of normal vascular permeability of intra-islet ECs.

These results and those from other groups, suggest an expanded model of pancreatic islet vascularization and islet endocrine-EC interactions (Figure 9). During normal pancreatic development, VEGF-A is expressed in trunk epithelial cells that are associated with endothelial cells (Pierreux et al., 2010). Neither VEGF receptor 1 nor VEGF receptor 2 are expressed by β cells (Brissova et al., 2006). Most endothelial cells surrounding pancreatic epithelial cells are not part of perfused vessels, but some near glucagon-positive cells are perfused as early as E10.5 (Pierreux et al., 2010; Shah et al., 2011), suggesting functional vessels are important for endocrine cell development. In our VEGF-A overexpression model, insulin-expressing cells with increased VEGF-A expression attract more endothelial cells as early as E12.5 and signaling from this increased number of endothelial cells appears to alter subsequent endocrine cell genesis as indicated by reduced number of Ngn3⁺ cells in the trunk domain at E14.5. One explanation for this impaired endocrine cell genesis may be impaired endocrine cell differentiation. Around E16.5, endothelial cells are normally organized into a single layer surrounding the epithelial cells and endocrine cell clusters, but VEGF-A overexpression greatly increased the number of endothelial cells, with multiple layers surrounding the endocrine cell clusters and ductal epithelium that still contained insulin-positive cells (Figure 9). These increased endothelial cells may not form functional vessels, and may limit endocrine cell differentiation. At P7, many α and β cells in the VEGF-A overexpressing pancreas fail to migrate away from ducts to form highly vascularized islets (Figure 9).

Figure 9. Model of pancreatic islet vascularization in VEGF-A overexpressing mice.

(A) In the normal pancreas, around E14.5, endothelial cells form a single layer close to the trunk with endocrine cells. Around E16.5, endocrine cells begin to cluster near the ductal epithelium that is surrounded with single layer vessels in the normal pancreas. By P7, in normal pancreas, endocrine cell clusters migrate away from ducts to form highly vascularized islets. (B) In the pancreas with VEGF-A overexpressing β cells, around E14.5, an increased number of endothelial cells is attracted to β cells along trunk domain. Around E16.5, endocrine cells form smaller clusters near the ductal epithelium but are surrounded with multiple layers of endothelial cells. By P7, in pancreas with VEGF-A overexpressing β cells, endocrine cell clusters and ducts are surrounded with multiple layers of endothelial cells that limit proper islet formation. Endocrine cells are adjacent to the ducts with some endocrine cells remaining in the ductal epithelium.

Multiple mechanisms may mediate the process how an increased EC mass impairs islet cell differentiation, islet formation, and reduces β cell mass. When VEGF-A was upregulated in other tissues such as ear and heart, this led to leaky vasculature and myocardial thinning (Miquerol et al., 2000; Thurston et al., 1999). A recent study showed that higher oxygen tension enhanced endocrine cell differentiation (Fraker et al., 2009). Endothelial cells alone were insufficient to induce organ development, and signals from blood flow and oxygen may be necessary (Magenheim et al., 2011). Edsbagge and colleagues also showed earlier that functional blood vessels were required for dorsal pancreas development (Edsbagge et al., 2005). Oxygen is important for normal β cell development and function, with hypoxia leading to decreased insulin content and islet formation (Cantley et al., 2009; Cheng et al., 2010; Fraker et al., 2007; Fraker et al., 2009; Olsson and Carlsson, 2011; Shah et al., 2011). Furthermore, mice with β cell specific hypoxia-inducible factor 1α (HIF-1α) inactivation have impaired β cell function (Cheng et al., 2010). Likely, hypoxia stabilizes HIF-1α and allows interactions with the Notch pathway (Cejudo-Martin and Johnson, 2005; Cheng et al., 2010; Diez et al., 2007; Edsbagge et al., 2005; Heinis et al., 2010; Pear and Simon, 2005; Sainson and Harris, 2006). Up-regulation of the Notch pathway may prevent or limit endocrine progenitor cell differentiation, since down-regulation of the Notch pathway activates pancreatic endocrine progenitor differentiation (Apelqvist et al., 1999; Jensen et al., 2000; Kaern et al., 2005; Lammert et al., 2000; Magenheim et al., 2011; Murtaugh et al., 2003). Our TEM data suggests that a somewhat leaky islet vessel system with an increased caveolae at the luminal surface of ECs and red blood cells extravasation exists in the pancreas of VEGF-A overexpressing mice. Immature and unstable vessels may not supply sufficient blood flow to developing islet cells, and further lead to hypoxia and affect endocrine cell development. Therefore, in our VEGF-A overexpression mice, low oxygenation resulting from dysfunctional vessels may up-regulate the Notch pathway, suppress endocrine cell differentiation, and further reduce neogenesis of β cells, leading to a reduced number of β cells.

Another mechanism that explains the reduced β cell mass in Rip-rtTA;tet-O-VEGF system is reduced β cell proliferation. This may be the result of the increased endothelial cells mass secreting factors that alter the β cell microenvironment, or by directly inhibiting β cell proliferation. For example, hepatocyte growth factor (HGF), which is secreted by ECs, increases β cell proliferation (Eberhard et al., 2010; Fiaschi-Taesch et al., 2008; Garcia-Ocana et al., 2003; Garcia-Ocana et al., 2000; Garcia-Ocana et al., 2001; Johansson et al., 2006; Lopez-Talavera et al., 2004; Mellado-Gil et al., 2011), and laminins, EC-derived extracellular matrix molecules, promote β cell differentiation and proliferation (Jiang et al., 1999; Nikolova et al., 2006; Virtanen et al., 2008). Such cell-matrix interactions are also known to promote β cell survival (Daoud et al., 2010; Weber et al., 2008). Therefore, VEGF-A overexpression may directly or indirectly alter the level of HGF, laminin, or the extracellular environment of differentiating β cells. Meanwhile, we also noted that an increased number of α and β cells were still associated with ductal structures in the VEGF-A overexpression model. Although there was no effect on α and β cell delamination in VEGF-A overexpressing pancreas at earlier stage (E16.5), and ductal cells may give rise to new β cells to overcome decreased β cell number (Bonner-Weir et al., 2010; Bonner-Weir et al., 2004; Butler et al., 2010; Huang et al., 2002; Martin-Pagola et al., 2008; Rooman and Bouwens, 2004; Tokui et al., 2006; Weaver et al., 1985; Xu et al., 1999), it is also possible that the increased number of endothelial cells may physically limit and obstruct endocrine cell delamination and migration away from the ducts at later stages (P7, P28). These results suggest that VEGF-A overexpression impairs the late, but not the early stage, of delamination. Furthermore, the increased endothelial cells may prevent β cell and endocrine cell interaction leading to absent or poorly formed endocrine cell clusters and islets.

Our data indicates that reduced β cell mass results from reduced endocrine cell progenitors, and decreased β cell proliferation, and not from increased β cell death. We did not identify apoptotic β cells indicative of β cell death at P1, but these cells may be difficult to detect, since the process of the apoptosis is short and the apoptotic cells disappear quickly in vivo (Elmore, 2007). In addition, signaling from endothelial cells appears to also affect α cell development in the endocrine cell clusters since glucagon content is also decreased. Consistent with our findings was the recent report that sphingosine-1-phosphate receptor 1 deficiency was noted to lead to endothelial cell hyperplasia and this limited pancreas development (Sand et al., 2011). Furthermore, a recent study by Magenheim et al. showed that VEGF-A overexpression throughout pancreatic epithelium in Pdx1-tTA;tet-O-VEGF-A transgenic mice led to pancreatic hypervascularization that reduced endocrine and exocrine cell differentiation (Magenheim et al., 2011). Consistent with this report, in Rip-tTA;tet-O-VEGF-A mice, β cell-specific VEGF-A overexpression caused hypervascularization along the trunk domain of the developing pancreas and decreased the number of Ngn3⁺ endocrine progenitor cells. However, exocrine tissue development progressed normally in this system most likely because short-range effects of VEGF-A¹⁶⁴ isoform on ECs were insufficient to influence vascularization around distal tip cells that give rise to acinar cells (Carmeliet and Tessier-Lavigne, 2005).

Our findings of reduced β cell number and disrupted islet formation following VEGF-A overexpression in the developing pancreas differ from prior reports. Pancreas-specific VEGF-A transgenic overexpression using a fragment of the pdx1 promoter resulted in a hypervascularized pancreas, islet hyperplasia, a 3-fold increase in islet number and area, and a 7-fold decrease in exocrine area when examined at 2 months of age (Lammert et al., 2001). In another transgenic mouse line (Rip-VEGF-A) in which the rat insulin promoter (Rip) was used to increase β cell expression of VEGF-A (Gannon et al., 2002), an increase in islet-derived VEGF-A did not alter blood vessel density, islet morphology or endocrine cell number. When islets of Rip-VEGF-A mice were transplanted into diabetic recipient mice, they had improved islet revascularization and survival (Lai et al., 2005). In another study, donor islets transduced with adenovirus expressing VEGF-A and transplanted into diabetic recipients had improved vascularization, but no change in β cell number (Zhang et al., 2004). The differing results between these studies and the current report may relate to the timing, level, and pattern of VEGF-A expression, and give new insight into the interaction of EC and islet endocrine cells. For example, the pdx1 promoter transgene (Lammert et al., 2001) likely results in lower, but a developmentally earlier and different level of VEGF-A than studies using the insulin promoter (Gannon et al., 2002; Lai et al., 2005) or the current study. The pdx1 promoter would result in VEGF-A expression throughout the pancreas, duodenum, and portions of the stomach during embryonic stages and then it would become restricted to most islet cells and some exocrine cells beginning at birth (Guz et al., 1995; Lammert et al., 2001). In contrast, VEGF-A is specifically expressed in the β cell in our Rip-rtTA;tet-O-VEGF-A model, and in the other studies using the Rip promoter (Gannon et al., 2002; Lai et al., 2005). Furthermore, the pdx1 promoter would lead to an earlier increase in VEGF-A than the insulin promoter (~E8.5 versus E11.5). Perhaps an earlier increase in VEGF-A impacts ECs differently or perhaps endocrine progenitor cells or developing endocrine cells respond differently to earlier signals from endothelial cells. A critical difference in our study is that increased VEGF-A expression was induced while in the other reports VEGF-A was constitutively expressed, possibly leading to adaptive or compensatory responses that prevented the major alterations we noted. The results from a recent report using an inducible VEGF-A system supports our findings (Magenheim et al., 2011).

Angiopoietins, especially Ang1, participate in the process of blood vessel maturation and remodeling following initiation of vasculogenesis or angiogenesis by VEGF-A. When Ang1 or Ang2 was overexpressed in the liver, enlarged and disorganized blood vessels were formed (Bureau et al., 2006; Haninec et al., 2006). Similarly, Ang1 when overexpressed in the skin led to enlarged vessels and increased vascularity (Suri et al., 1998; Thurston et al., 1999). Furthermore, cardiac-specific overexpression of Ang1 demonstrated a pivotal role of Ang1 in the embryonic development of the epicardium and coronary vasculature, and complete ablation of the coronary artery was observed in some of the most severely affected embryos (Ward et al., 2004). Our studies used the same tet-O-Ang1 or tet-O-Ang2 transgenes as these prior studies and we noted increased levels of secreted Ang1 and Ang2. (Bureau et al., 2006; Haninec et al., 2006; Ward et al., 2004). In our Ang1- and Ang2- overexpressing models, intra-islet vessels had ECs with less fenestration and more caveolae. Since vessel permeability is influenced by caveolae and fenestrae (Feng et al., 1996; Vasile et al., 1999) and EC fenestration is important for normal pancreatic islet function (Brissova et al., 2006; Lammert et al., 2003), this change in vessel ultrastructure may slightly alter vessel permeability. This may alter glucose entry or insulin exit from the islet or into the systemic circulation and possibly explain the mildly impaired glucose tolerance in Ang1- and Ang2-overexpressing male mice that usually become more readily glucose intolerant than female mice. A recent study by Jeansson and colleagues provides an important new insight into Ang1 requirements in early and later development (Jeansson et al., 2011) by demonstrating that mice with cardiac-specific knockout of Ang1 had early vascular abnormalities, but mice with Ang1 deletion at any time point after E13.5 had a normal vascular phenotype (Jeansson et al., 2011). In this report, we showed that mice with Ang1 deletion after E13.5 had normal islet morphology, vascularization, and normal glucose tolerance. Those data support our findings that angiopoietin-Tie2 signaling in endothelial cells does not play a critical role in islet vascularization. Since Ang2 and 3 are still present and a report noted that Tie2 signaling is required beyond E12.5 (Jones et al., 2001), one cannot completely exclude a role for Tie2 signaling.

Taken together, our study shows that increased EC number does not promote but impairs β cell proliferation and islet formation, and that the precise control of VEGF-A production by developing islet cells is important for normal islet development and vascularization. In contrast, Angiopoietin-Tie2 signaling in endothelial cells does not appear to have a major role in the development or maintenance of islet vascularization.

Supplementary Material

01

Supplemental Figure 1. β cell-specific VEGF-A overexpression in early development affects β cell formation and islet morphology but mice overexpressing Ang1 or Ang2 have normal insulin content at birth. (A–D) Pancreatic sections of Rip-rtTA and Rip-rtTA;tet-O-VEGF mice treated with Dox from E5.5 to P1. (A, B) Insulin (Ins, blue), and somatostatin (Som, red). (C, D) Insulin (Ins, blue), glucagon (Glu, red), and PECAM1 (green). (E, F) Pancreatic insulin content at P1 in Rip-rtTA;tet-O-Ang1, Rip-rtTA;tet-O-Ang2 and Rip-rtTA mice treated with Dox from E5.5 to P1 (n=3 mice/group). (G) Percentage of VEGF-A overexpressing β cell per total β cells at E16.5, P1 and P7 (Dox treatment from E5.5 to E16.5, P1, or P7); E16.5, 87.8±5.6% (n=3 mice); P1, 42.4±3.7% (n=3 mice); P7, 22.6±4.2% (n=3 mice). **; p<0.01, ***; p<0.001.

Supplemental Figure 2. VEGF-A overexpressing β cells start recruiting more ECs at E12.5. Pancreatic sections of Rip-rtTA (A, C, E, G) and Rip-rtTA;tet-O-VEGF-A mice (B, D, F, H) were immunostained with insulin (Ins, red), PECAM1 (green), and Pdx1 (blue) at different developmental stages (Dox treatment from E5.5 to E10.5, E12.5, or E14.5, or E16.5). Arrows in D, F and H point to EC accumulation. Scale bars in A–H represent 50 μm.

Supplemental Figure 3. β cell-specific overexpression of VEGF-A alters islet morphology but does not affect α and β cell delamination at E16.5. Pancreatic sections of Rip-rtTA and Rip-rtTA;tet-O-VEGF-A mice were treated with Dox from E5.5 to E16.5. (A–D) Islet morphology visualized by labeling for insulin (Ins, blue) and PECAM1 (green). (E, F) β Cell delamination visualized by staining for insulin (Ins, green), Pax6 (red), E-cadherin (blue). (G, H) α Cell delamination visualized by staining for glucagon (Glu, green), Pax6 (red), E-cadherin (blue). Scale bar in A represents 50 μm, and also corresponds to panels B–D. Scale bar in E represents 50 μm and also corresponds to panels F–H.

Supplemental Figure 4. β cell-specific overexpression of VEGF-A during islet development dramatically increases α and β cell association with DBA⁺ ducts, islet hypervascularization without changes in exocrine tissue. Pancreatic sections of Rip-rtTA and Rip-rtTA;tet-O-VEGF mice treated with Dox from E5.5 to P7. (A, B) Rip-rtTA and Rip-rtTA;tet-O-VEGF-A tissues were stained for insulin (Ins, red), PECAM1 (green), and amylase (purple). (C, D) Rip-rtTA and Rip-rtTA;tet-O-VEGF-A tissues were stained for insulin (Ins, red), DBA (green), and amylase (purple). (E) Acinar cell area expressed as fraction of pancreatic tissue: Rip-rtTA, 89.0%±1.0 (n=3 mice); Rip-rtTA;tet-O-VEGF-A, 88.1%±1.7 (n=3 mice); p>0.05. (F) Percentage of β cell associated with duct: Rip-rtTA, 1.5±0.3% (n=3 mice); Rip-rtTA;tet-O-VEGF-A, 12.9±2.2% (n=3 mice); **; p<0.01. (G) Percentage of α cell associated with ducts. Rip-rtTA, 1.3±0.5% (n=3 mice); Rip-rtTA;tet-O-VEGF-A, 5.4±0.7% (n=3 mice); *; p<0.05. Bar in A represents 100 μm and also corresponds to panels B–D.

Supplemental Figure 5. Continuous overexpression of VEGF-A in β cell during postnatal period leads to progressive alteration in islet morphology and glucose intolerance. Pancreatic sections of Rip-rtTA and Rip-rtTA;tet-O-VEGF-A mice treated with Dox from E5.5 to P28. (A, B) Insulin (Ins, green), DBA (red), and Dapi (blue). (C, D) Insulin (Ins, green), DBA (red). (E, F) Insulin (Ins, green), E-cadherin (red) (G) Following intraperitoneal injection of glucose, blood glucose was measured in Rip-rtTA and Rip-rtTA;tet-O-VEGF-A mice at the age of 28 days, with Dox treatment from E5.5 to 28 days; ●; Rip-rtTA, ▲; Rip-rtTA;tet-O-VEGF-A. *; p<0.05. Bar in A represents 100 μm and also corresponds to panel B. Bar in C represents 50 μm and also corresponds to panels D–F.

Supplement Figure 6. Transmission electron microscopy of pancreata with VEGF-A overexpressing β cells at P1. Transmission electron microscopy images show endocrine cells in Rip-rtTA; tet-O-VEGF-A pancreas adjacent to endothelial cells. L, vessel lumen; DL, ductal lumen; EC, endothelial cells; β, β cell; RBC, red blood cell. In panel B, a black arrow points to RBC outside the capillary lumen, a black arrowhead points to RBC inside the capillary lumen, and white arrows point to the gap between ECs. Scale bar in A and B is 2 μm.

Supplemental Figure 7. Glucose tolerance and insulin secretion in mice with altered overexpression of Ang1. Following intraperitoneal injection of glucose, blood glucose was measured in male (A) and female (B) mice at the age of 8 weeks, with or without Dox treatment from E5.5 to 8 weeks; or in male (D) and female (E) mice at the age of 12 weeks, with or without Dox treatment from 8 weeks to 12 weeks of age. ▲; Rip-rtTA;tet-O-Ang-1, with Dox; ●; Rip-rtTA;tet-O-Ang-1, without Dox; ■; control Rip-rtTA, with Dox, n=3–7 mice/group. (C) Insulin secretion in islets isolated from Rip-rtTA and Rip-rtTA;tet-O-Ang1 mice was analyzed in response to 16.7 mM glucose and 16.7 mM glucose + 100 μM isobutylmethylxanthine (IBMX). The integrated response to 16.7 mM glucose was 15.2 ± 2.0 ng insulin in Rip-rtTA mice vs. 5.8 ± 2.3 ng insulin in Rip-rtTA;tet-O-Ang1 mice (n=4 mice/group, p=0.8536). The integrated response to 16.7 mM glucose + 100 μM IBMX was 74.7 ± 6.5 ng insulin in Rip-rtTA mice vs. 103.0 ± 11.3 ng insulin in Rip-rtTA;tet-O-Ang1 mice (n=4 mice/group, p=0.0715). (F, G) Pancreatic sections from Rosa-rtTA and Rosa-rtTA;tet-O-Cre;Ang1^fl/fl mice treated with Dox from E13.5 to P28. Insulin (Ins, blue), Glucagon (Glu, red), Pecam1 (green). (H) Following intraperitoneal injection of glucose, blood glucose was measured in Rosa-rtTA and Rosa-rtTA;tet-O-Cre; Ang1^fl/fl mice at the age of 28 days, with Dox treatment from E13.5 to 28 days, ▲; Rosa-rtTA;tet-O-Cre; Ang1^fl/fl and ●; Rosa-rtTA (n=3 mice/group, P>0.05).

Supplemental Figure 8. Glucose tolerance and insulin secretion in Ang2-overexpressing mice. Following intraperitoneal injection of glucose, blood glucose was measured in male (A) and female (B) mice at the age of 8 weeks, with or without Dox treatment from E5.5 to 8 weeks; or in male (D) and female (E) mice at the age of 12 weeks, with or without Dox treatment from 8 weeks to 12 weeks of the age. ▲, Rip-rtTA;tet-O-Ang-2, with Dox; ●, Rip-rtTA;tet-O-Ang-2, without Dox, ■, control Rip-rtTA, with Dox, n=3–7 mice/group. (C) Insulin secretion of islets isolated from Rip-rtTA and Rip-rtTA;tet-O-Ang2 mice was analyzed by stimulation with glucose and isobutylmethylxanthine (IBMX). The integrated response to 16.7 mM glucose was 15.1 ± 2.2 ng insulin in Rip-rtTA mice vs. 17.9 ± 1.2 ng insulin in Rip-rtTA;tet-O-Ang1 mice (n=4 mice/group, p=0.3036). The integrated response to 100 μM IBMX in the presence of 16.7 mM glucose was 104 ± 25 ng insulin in Rip-rtTA mice vs.120 ± 7 ng insulin in Rip-rtTA;tet-O-Ang1 mice (n=4 mice/group, p=0.5577).

Highlights.

• Islet cell and endothelial cell interactions are critical for islet development.

• VEGF-A from developing endocrine cells is critical for islet vascularization.

• Increased VEGF-A increases endothelial cells, but reduces β cell mass.

• Increased VEGF-A impairs β cell proliferation and migration and islet formation.

• Angiopoietin-tie2 signaling does not have a crucial role in islet vascularization.

Abbreviations

EC

Endothelial Cells

VEGF-A

Vascular Endothelial Growth Factor-A

Ang1

Angiopoietin-1

Ang2

Angiopoietin-2

Dox

Doxycycline

Rip-rtTA

Rat Insulin Promoter-reverse tetracycline activator

Pdx-1

pancreatic-duodenal homeobox factor-1

GTT

Glucose Tolerance Test

PBS

Phosphate Buffered Saline