Metabolism Regulates Exposure of Pancreatic Islets to Circulating Molecules In Vivo Free

Islets of Langerhans in adult mice are primarily composed of insulin-producing β-cells (up to 90%), surrounded by other endocrine cell types (1), and serve to maintain whole-body glucose homeostasis. Islets are organized in clusters distributed preferentially along large blood vessels (2), and the resident endocrine cells are interwoven with a dense capillary network. Islet capillaries are denser, more tortuous, and more fenestrated than vessels of the surrounding exocrine tissue (3). In addition, relative to their total mass (∼1% of the pancreas), islets are highly perfused (∼10% of pancreatic blood flow) (4,5). Both these traits are critical to ensure adequate oxygen and nutrient supply, as well as fast sensing of circulating molecules and rapid hormone transport into the blood. Although β-cells integrate a range of signals from the circulation involved in insulin secretion and/or β-cell function, survival, and proliferation (glucagon-like peptide-1 [GLP-1] and prolactin/placental lactogen) (6–8), the access dynamics of circulating molecules to islets remain largely unexplored.

Although adult β-cells are terminally differentiated with low proliferative capacity, conditions of increased insulin demand such as pregnancy and obesity can profoundly alter β-cell function and/or mass (9). Failure to compensate for increased insulin demand contributes to gestational diabetes mellitus and type 2 diabetes (T2D) in genetically predisposed individuals (10). β-Cell adaption during gestation and high-fat diet (HFD) is widely studied (11–15), and vessel properties are modified during these highly proliferative phases. For instance, islet blood flow varies during acute and chronic blood glucose concentration changes (4,16–18), insulin resistance states (19), and pregnancy in the rat (20). Highlighting the importance of islet blood flow in β-cell function, ablation of endothelial insulin signaling impairs insulin release through effects on the pancreatic blood circulation (21). However, whether islet vascular permeability is also affected by altered metabolism remains uncharacterized. Understanding the changes in molecule access and distribution that accompany compensatory β-cell expansion may be key to developing novel antidiabetic therapies/strategies. Therefore, using in vivo imaging approaches, we examined whether alterations in islet vascular permeability are associated with β-cell adaptation to both physiological and pathological metabolic triggers.

All animal studies were approved by the Languedoc Roussillon Institutional Animal Care and Use Committee (CEEA-LR-1053). Virgin or G14.5 gestating 9–12-week-old female C57BL/6J mice were used. All experiments were performed on animals fed ad libitum. In some experiments, virgin RIPTVA-mCherry mice, produced at the National Institute for Medical Research, London, were used to facilitate localization of islets. The construct used to produce these mice contained a 0.7-kb rat insulin promoter (RIP) sequence upstream of the TVA-IRES-mCherry gene, followed by the 3′ untranslated region of the rat prolactin gene (hGH minigene was not used) (22,23), and was microinjected into the pronucleus of fertilized oocytes of superovulated (CBA/Ca × C57BL/10)F1 mice. The RIPTVA-mCherry mice obtained were backcrossed to C57BL/6J mice for more than eight generations. Mice were either placed on standard diet (SD), tryptophan (TRP)^+/− diets (0/0.18%), or HFD (63% calories from fat) (Safe Diets). In some cases, virgin animals fed 2 weeks with SD were used as controls for G14.5 and 2-week HFD–fed animals. Intraperitoneal glucose tolerance tests (IPGTTs) and glucose-stimulated insulin secretion tests were as previously described (21,24). For in vivo imaging, the pancreas was exteriorized by surgery. Animals were anesthetized by injection of ketamine/xylazine (0.1/0.02 mg/g) and temperature controlled as previously described (25,26). Respiration was controlled by tracheotomy. An incision was made in the peritoneum, and the pancreas was gently maneuvered onto a custom-made metallic stage covered with a 2-mm-thick layer of soft polymer (Bluesil). The pancreas was pinned to the polymer using four to five stainless steel minutien insect pins (tip = 0.0125 mm). To prevent desiccation, the tissue was superfused with NaCl 0.9% heated to 37°C. Imaging commenced 30–40 min postanesthesia.

Molecule extravasation was imaged using a multiphoton microscope (Zeiss 7MP) adapted with a long-working distance objective M Plan Apo NIR ×20, 0.4 NA (Mitutoyo), and multiphoton excitation and emitted fluorescence collection were as previously described (26). Surface islets were identified through mCherry localization or by light contrast. Fluorescein isothiocyanate (FITC)– or rhodamine-labeled dextrans (Sigma-Aldrich) (25 mg/mL in NaCl 0.9%, 100 μL/20 g body weight) were injected through a jugular catheter. In some experiments, 100 μL/20 g body weight of 50 mg/mL dextran was used. As this had no effect on measurements, data were pooled. Recordings in islets, distinguishable by their tortuous vessels, at depths of 15–40 µm below the surface, were started at the time of injection, and acquisition rate was set to 3.5 frames/s. Regions of interest (ROIs) in vessels and in the directly adjacent parenchyma were selected in each video (3–4-µm diameter circles). At least four videos from four different mice were analyzed per molecular weight (MW) and condition.

Tissue movement was corrected using a registration tool based on a subpixel translation obtained from a minimal image difference search, as previously described (25,26). In brief, videos were imported into ImageJ (National Institutes of Health), and the average distance of each frame was calculated versus a reference image over eight trials. A translation was then performed corresponding to the minimum of a parabolic interpolation before application to the entire image stack. To obtain the extravasation rate value p (µm/s) for each molecule, an automated routine programmed in MATLAB R2011a software was used (26). This is described mathematically as

∇Φ corresponds to the fluorescence gradient in µm⁻¹ and equals

where Φ_inside − Φ_outside is the measured difference between fluorescence intensities in the vessel and the parenchyma and is half the distance between the ROIs in µm. D corresponds to the diffusion coefficient in µm²/s, p is the extravasation rate in µm/s, and DΔΦ corresponds to the diffusion term. The ascending phase of the fluorescence profile obtained experimentally in the vessel ROI was used as input to the model. Then, the value of p for which the simulated profile of fluorescence intensity increase in the parenchyma over time best fitted experimental values was selected. The decreasing phase of the fluorescence profiles in selected ROIs was modeled using the Nelder-Mead Method optimization algorithm, consisting of a decreasing exponential function (MATLAB), before calculation of fluorescence decay rates (s⁻¹), corresponding to the time constant (1/τ) of the exponential fluorescence decay [f(x) = e^(−t/τ)].

2-NBDG (2-(N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)amino)-2-deoxyglucose) (Life Technologies) was injected either 5 mg/kg i.v. or 10 mg/kg i.m. in 100 or 40 µL of 0.9% NaCl, respectively. Although plasma volume increases during gestation in mice (∼26%) (27), this is in accordance with a similar increase in weight (∼24%), justifying a mg/kg correction. Furthermore, uptake rates were unchanged when 2-NBDG was injected above 4 mg/kg i.v. or 6.25 mg/kg i.m. Vessels were labeled with fluorescent dextran to facilitate islet localization. Multiphoton excitation was delivered at 890 nm. Each Z-plane was scanned every 15 s (120-µm stacks). Videos were stabilized using Huygens Essential (Scientific Volume Imaging). Fluorescence increase over time in ROIs was measured. After intravenous injection, the first 1.5 min of recordings was excluded from analysis, due to contamination of signal by molecules diffusing out of vessels. Background fluorescence in exocrine tissue was subtracted from signal measured in islets, and fluorescence measurements were normalized. Uptake of 2-NBDG was modeled using either a one-phase association curve (intravenous), or a variable slope sigmoidal curve (intramuscular). First derivatives of the modeled sigmoidal curves were generated using GraphPad Prism. Vmax of derivatives and inflection points of sigmoidal and one-phase association curves were compared.

Blood cell velocity was measured using an epifluorescence stereomicroscope (SteREO Discovery; Carl Zeiss), as previously described (25). Mice were anesthetized using ketamine/xylazine (0.1/0.02 mg/g i.p.). Plasma was labeled with FITC–150 kDa dextran. Fluorescence emission was captured using an ORCA Flash4.0 sCMOS (Hamamatsu). Acquisition rate was 150 frames/s, >1,000 frames per video were captured. Velocities were obtained by analyzing diagrams of the intensities along defined capillary paths (in ImageJ) (25). Five to six mice per condition and two to nine islets per mouse were analyzed.

Pancreata were fixed overnight in 4% paraformaldehyde before head-to-tail slicing on a Leica vibratome (100 µm) or were snap frozen in optimal cutting temperature and sectioned using a cryostat (20 µm). In some experiments, vessels were filled using gelatin (Sigma-Aldrich) labeled with FITC through intracardiac perfusion (25). Antibody labeling was as previously described (26). Images were acquired using a Zeiss LSM 780 confocal microscope. Images were analyzed using Imaris (Bitplane), Volocity (PerkinElmer), and ImageJ.

For quantifications, one to four slices were randomly selected from at least six animals per group, and all islets present were analyzed, corresponding to a minimum of 15 islets per mouse. A priori, this is sufficiently powered to detect a minimum 1.2-fold difference with an SD of 40%, a power of 0.9, and α = 0.05 (G*Power 3.1). β-Cell proliferation was measured on slices stained for Ki67 (rabbit, 1:200; CliniSciences), insulin (guinea pig, 1:400; Abcam), and DAPI (Sigma-Aldrich). The proportion of proliferative β-cells was obtained by dividing the number of Ki67⁺ nuclei by total number of nuclei of insulin⁺ cells in islets, as previously described (14). Vessel density in Z-stack images was assessed by calculating vessel (gelatin-FITC labeled or rat anti-CD31, 1:100; BD Pharmingen) and islet (insulin⁺ or mCherry) volume using Volocity. To determine β-cell size, pancreas slices were stained for insulin, E-cadherin (rat, 1:500; Takara), and DAPI. The cross-sectional area of E-cadherin insulin⁺ cells in which a cut nuclei was present was measured using ImageJ, as previously described (28).

Pancreata were collected and fixed in 4% paraformaldehyde and 2.5% glutaraldehyde at 4°C and sliced on a Leica vibratome. Unlabeled slices were treated as previously described (29). Images were acquired on a transmission electron microscope (Hitachi H-7100). Density of fenestrae per micrometer was obtained by counting fenestrae on available endothelium of individual vessels after subtraction of the perikaryal length. Internal fenestrae diameters and islet vessel lumen perimeters in vessels present in cross sections were manually measured using ImageJ.

Values are represented as means ± SEM, and tests were performed using GraphPad Prism. Normality was assessed using D’Agostino-Pearson test. Comparisons were made using either unpaired Student t test, or two-tailed Mann-Whitney U test, as appropriate. Multiple comparisons were made using one-way ANOVA followed by Bonferroni post hoc test or two-way ANOVA using Tukey multiple comparison test. P values were considered significant at P < 0.05, 0.01, and 0.001.

To study molecule access in islets, we developed a novel approach to image the pancreas directly in vivo in anesthetized mice using a two-photon microscope adapted with long working distance objectives (Fig. 1A and B) (26). Islet localization was initially facilitated by using mice on a C57BL/6J background expressing mCherry under the control of the RIP (Supplementary Fig. 1). As no difference with wild-type C57BL/6J mice was observed, data were pooled. Fluorescent dextran was used for extravasation rate measurements, since it avoids artifacts introduced by receptor binding and/or uptake by cells. Injection of FITC-labeled dextran (MW <70 kDa) into the jugular vein of virgin animals induced a steep increase followed by a rapid decrease in intraislet fluorescence, suggesting that β-cells are only briefly exposed to circulating molecules (Fig. 1C and D and Video 1). Extravasation rates from endocrine capillaries were significantly greater than from the surrounding exocrine vessels (∼5-fold for 4- and 10-kDa molecules and ∼10-fold for 20- and 40-kDa molecules) (Fig. 1E), as expected from the reported differences in fenestration density (3). In the islets, molecules <70 kDa diffused rapidly across capillaries (Fig. 1E and Video 1). By contrast, diffusion of 20–40-kDa molecules in the exocrine tissue was very slow, albeit detectable.

In addition to molecule access to pancreatic tissue, we also assessed molecule retention in the extracellular space subsequent to entry. To measure this, fluorescence exponential decay rates in tissue were calculated. In virgin animals, small molecules (∼4 kDa) diffused rapidly out of both exocrine and endocrine tissues, whereas for larger molecules (∼10 kDa), diffusion occurred more slowly out of the exocrine compartment (Fig. 1F). This finding supports the notion that endocrine cells only very briefly encounter extravasating molecules.

We next sought to investigate if islet vessel permeability was modified in response to changes in metabolic demand. First, gestation was chosen as it represents a normal physiological state of insulin resistance accompanied by changes in β-cell proliferation and function (11,30). As expected, islets from G14.5 animals were 2.5-fold larger than from virgins, and this was associated with β-cell hypertrophy/hyperproliferation, increased basal insulinemia, and unchanged glucose-stimulated insulin secretion (Fig. 2A–G and Supplementary Fig. 2). As evidenced by IPGTT, G14.5 mice were glucose intolerant (Fig. 2H and I). Second, to determine whether adaptive changes in the islet vasculature were specifically associated with pregnancy, or reflected a more generalized response to short-term metabolic stress, virgin female mice were fed HFD for 2 weeks. This paradigm has the advantage of preventing the hormonal changes occurring during gestation. As anticipated (13–15), animals fed HFD were heavier and more glucose intolerant than their standard chow–fed counterparts, whereas basal and glucose-stimulated insulin secretion were not significantly modified (Fig. 3A–E). Although an increase in islet size could not yet be observed at this stage (Fig. 3F and G), β-cells were hypertrophic (Fig. 3H), and an increase in β-cell proliferation was evident, as measured by the proportion of Ki67⁺ β-cells (twofold increase compared with SD-fed mice) (Fig. 3I and J) and reported in Mosser et al. (14).

At G14.5, although the general shape of the permeability curve and molecule size cutoff remained unchanged, molecule extravasation rates from the islet vasculature were significantly decreased (Fig. 4A and B and Video 2). This effect was particularly marked for 10- and 20-kDa molecules and was specific to islets, since no alteration could be detected in exocrine vessels (Fig. 4B). In addition, there was an increase in the rate of molecule diffusion out of the endocrine tissue compared with virgin animals (Fig. 4C). Similarly to G14.5, a significant decrease in molecule extravasation rate was present in islets of HFD-fed mice, and this was more pronounced for smaller molecule sizes (Fig. 4D and E and Video 3). In addition, fluorescence decay rates in the endocrine tissue of HFD-fed animals were comparable to those measured at G14.5 (Fig. 4F). Decreased retention would shorten the already short-lived peak of molecules in the tissue. Thus, β-cells unexpectedly experience less exposure to circulating molecules during acute metabolic demand induced by pregnancy and HFD.

Next, to investigate potential mechanisms involved in modifications to islet vessel permeability, relevant structural vascular parameters were assessed. At G14.5, there was a striking increase in islet volume occupied by capillaries (Fig. 5A), consistent with the reported increase in islet endothelial proliferation during gestation (30). By contrast to the effects of long-term high-fat feeding (Supplementary Fig. 3), and described in Dai et al. (19), changes in vessel density were absent after 2 weeks of HFD (Fig. 5B). As an increase in islet vessel volume can result from dilation and/or angiogenesis (19), we measured lumen perimeters in cross sections using electron microscopy. Mean islet vessel perimeter was significantly increased in gestating animals (Fig. 5C and D). Although most vessels were a similar size to virgin animals, some appeared dilated (Fig. 5D). By contrast, the reported long-term effects of HFD on vessel dilation (Supplementary Fig. 3) (19) were not evident after 2 weeks of treatment (Fig. 5C and E). Fenestration density and mean fenestrae diameter remained unchanged between both virgin and G14.5 animals, and SD- and HFD-fed animals (∼75 nm in both cases) (Fig. 5F–K).

A common feature of both gestation and HFD treatment is β-cell expansion (11,14). To test whether cell proliferation could drive modifications in vascular permeability, the compensatory increases in β-cell mass during gestation were prevented by feeding animals from G8.5 to G14.5 with a TRP-deficient diet (TRP⁻ diet), thereby inhibiting serotonin synthesis (31). Both virgin and gestating mice fed with TRP⁻ diet for 6 days lost weight (Fig. 6A), but only gestating mice became glucose intolerant (Fig. 6B and C). No increase in islet size or β-cell proliferation could be observed at G14.5 compared with virgin animals (Fig. 6D–G), as expected (31). However, extravasation rates of molecules were still reduced in gestating animals fed with TRP⁻ diet (Fig. 6H and I and Video 4). Remarkably, in both virgin and G14.5 TRP⁻ diet–fed mice, values obtained were similar to those of animals fed standard chow, despite the absence of gestational increases in islet size. However, islet size distribution and β-cell density in islets were altered at G14.5 (Supplementary Fig. 4A–C), and other changes induced by gestation were still present, including β-cell hypertrophy and increased islet volume occupied by vessels (Supplementary Fig. 4D–F). Fenestrae diameters and density remained unchanged (Supplementary Fig. 4G and H). These findings suggest that changes in islet vessel permeability during gestation are independent of β-cell proliferation.

Extravasation and retention of molecules may conceivably be modified by binding and uptake kinetics, questioning whether islets actually experience “lesser exposure” to circulating bioactive molecules. To examine this, glucose uptake was tracked directly in vivo in islets using 2-NBDG. This “trappable” fluorescent glucose analog was injected either intramuscularly to mimic an increase in molecule concentration in the periphery (e.g., glucose release and GLP-1 secretion) or intravenously to control for effects of peripheral glucose transport on islet fluorescence accumulation (Fig. 7 and Video 5 and Video 6). Islet fluorescence increase followed sigmoidal or one-phase association curves in response to intramuscular (Fig. 7A and B) or intravenous injection (Fig. 7C and D), respectively. Comparison of fitted curves versus virgin mice revealed, for both injection regimes, reduced 2-NBDG incorporation rates during gestation (G14.5) and HFD feeding, with effects noticeably more pronounced in the latter. Thus, metabolic stressors impinge upon substrate uptake, and this occurs in a direction that strengthens a role for permeability in islet function through delayed exposure to circulating molecules.

Although the method used to calculate the extravasation rate of the molecule is independent of vascular flow, the 2-NBDG binding rate could conceivably be affected by changes in islet blood flow velocity. To assess the relative importance of blood flow dynamics, we measured red blood cell velocity in vivo and found that it was comparable in both the endocrine and exocrine compartments in virgin, G14.5, and 2-week HFD–fed animals (Fig. 8A and B and Video 7). Consistent with that previously reported (18), exocrine and islet blood flow under hyperglycemia did not significantly differ (Fig. 8B and C). In addition, anesthesia-induced hyperglycemia (32) was similar under all conditions examined (Fig. 8C). Thus, the delayed 2-NBDG uptake rate during increased metabolic demand is unlikely to be a consequence of reduced islet blood flow.

T2D can be described as a failure of β-cells to compensate for peripheral insulin resistance, leading to glucose intolerance and health complications (33). Although most studies have understandably focused upon impaired β-cell function and insulin receptor signaling, the islet microenvironment may provide another route for defective insulin secretion during T2D. We therefore aimed to study the access of circulating molecules to islets under normal conditions, as well as during both physiological (pregnancy) and pathological (HFD) increased metabolic demands. Using in vivo two-photon microscopy applied to the pancreas of anesthetized mice, we found that β-cells fleetingly encounter circulating molecules, which rapidly pervade the tissue before clearance, and that both gestation and short-term HFD (2 weeks) induce decreases in islet vascular permeability and molecule retention. Vascular permeability remained reduced after prevention of gestation-induced β-cell mass expansion, suggesting that cell proliferation is unlikely to be a primary driver of islet vascular changes. Divergent effects of gestation and 2-week HFD were observed on islet vessel dilation, suggesting that different mechanisms may be involved in the regulation of microenvironment properties. Nonetheless, altered molecule access dynamics and lowered exposure to circulating molecules were unifying features of both paradigms. A summary of results is presented in Supplementary Table 1.

The circulation of molecules within the intercellular microenvironment is dependent on tissue properties and can greatly influence the sensing of blood-borne signals, in addition to downstream tissue responses (34). By tracking events in vivo in real time, we were able to determine that circulating molecules (<70 kDa) could rapidly diffuse through the highly fenestrated capillaries of islets, similar to that described in other fenestrated vascular beds in vivo (26). In the pancreas, fenestration density greatly influences molecule extravasation, as evidenced by the divergence in rates between exocrine and endocrine tissue, the former irrigated by less fenestrated capillaries (3). Given the density of vascularization in islets, and the direct proximity of each β-cell to a vessel, this result strongly suggests that β-cells are capable of almost immediately (<1 s) sensing a peak in circulating molecules. Incidentally, β-cells are capable of responding to stimuli delivered over similar time courses (35,36). This finding has important implications for the access of factors secreted in a pulsatile manner in distant body compartments and is involved in β-cell function and/or insulin secretion (e.g., GLP-1, 3.2 kDa; prolactin/placental lactogen, 22 kDa).

By finely regulating the distribution of incoming and outgoing factors through their storage/segregation, the tissue microenvironment plays an important role in molecule action. For example, various niches that favor cell proliferation have been identified, and this may be linked to particular properties of the tissue or presence of growth factors (37,38). In virgin animals, small molecules such as 4-kDa dextran diffused equally rapidly out of both exocrine and endocrine tissues, whereas larger molecules such as 10-kDa dextran diffused more rapidly out of the latter, suggesting that islets only perceive “flashes” of circulating molecules and that passive retention by the tissue is very limited. Such findings might partly explain why β-cells in vivo are less proliferative than some cell types in other tissues, where the microenvironment better supports the retention and buildup of mitotic factors (39).

Gestation is characterized by β-cell expansion and modifications in β-cell activity and hormonal balance (11). HFD feeding induces a similar adaptive response, albeit in a pregnancy hormone-independent context. Under both conditions, extravasation was unexpectedly reduced, suggesting that β-cells experience delayed exposure to circulating molecules during heightened insulin demand. This reduced extravasation likely results from a decrease in the transfer rate of molecules across the vascular membrane, rather than active removal from the tissue, i.e., through lymphatic drainage, since lymphatics are rarely associated with islets (40). Moreover, the molecule fluorescence profile was modeled in the ascending phase, where the extravasation rate has more pronounced effects than clearance. However, a contribution of the latter to the measures detailed here cannot be ruled out. This notwithstanding, alterations to molecule exposure might provide a mechanism to downregulate insulin signaling during gestation, helping to maintain the hyperglycemia required to support the energy requirements of fetal growth, and may partly explain the impaired insulin secretion detected in mice in vivo 1 week after commencing HFD (13–15), despite improved β-cell function in isolated islets (41). Importantly, this demonstrates that molecule diffusion is a dynamic process modulated by physiological state.

Increased islet vessel volume during gestation may be explained by both marked islet endothelial proliferation (30) and vessel dilation, as supported by the observed increase in the perimeters of a subset of islet vessels at G14.5. By contrast, changes in the islet vasculature during long-term HFD treatment depend solely on dilation, and not angiogenesis (19), supporting a role for divergent mechanisms in controlling adaptation to metabolic demand. We did not observe an increase in the islet vessel perimeter after 2 weeks of HFD feeding, suggesting that dilation is triggered during later stages of treatment (19). Although it was previously reported that fenestrae density may contribute to the acute regulation of molecule entry into tissues due to their rapid turnover/appearance (3), no change in mean fenestrae diameter or density was detected. We cannot, however, exclude a role for alterations in the composition of fenestrae diaphragms, which act as size-selective molecular sieves involved in the regulation of basal permeability (42,43).

It has been shown that the perivascular space can provide a niche for cell proliferation (44). Therefore, it is plausible that β-cell proliferation may directly affect islet capillary properties to favor this process. We therefore tested whether inhibition of β-cell proliferation during gestation, through suppression of serotonin synthesis (31), was able to modify vascular permeability in islets. We found that, besides proliferation, the structural changes occurring at G14.5 were unaffected, and permeability in islets was decreased compared with virgin mice fed with TRP⁻ diet, possibly through similar mechanisms to gestation under normal feeding. Therefore, neither β-cell proliferation nor increased serotonin levels appear to be instructive for decreased molecule exposure during gestation, despite the reported vasoactive properties of the latter monoamine (45). Alternative mechanisms may include prolactin/placental lactogen, whose circulating levels peak during gestation. Both these peptides may regulate vascular plasticity/permeability through signaling directly to endothelial cells (46,47), as well as increasing vascular endothelial growth factor production (30).

The modifications to vascular permeability measured using fluorescent dextran in vivo may not be relevant for biologically active molecules. Indeed, it could be argued that the natural affinity of islets for their substrate may maintain molecule binding/uptake, even in the face of altered permeability during metabolic demand. Suggesting that this is not the case, the 2-NBDG uptake rate decreased in islets of gestating animals, despite the reported increase in GLUT2 expression by G15 (48). Likewise, the 2-NBDG uptake rate was decreased after short-term high-fat feeding, probably reflecting changes in membrane trafficking of GLUT2 (49). Although changes in blood flow velocity, generally associated with alterations to capillary pressure (50), may affect 2-NBDG uptake rates, we were unable to detect any differences in red blood cell velocity in virgin, G14.5, and 2-week HFD–fed animals. Although the effects of acutely increased metabolic demand on vascular parameters in awake mice remain unknown, and an effect of anesthetic and/or blood pressure on tissue perfusion cannot be completely excluded, this demonstrates that the delayed uptake of 2-NBDG is unlikely a consequence of reduced islet blood flow. Thus, the data together suggest that vascular permeability likely contributes to islet (patho)physiology by modifying the exposure of β-cells to bioactive substances.

In summary, we show that β-cells in vivo are exposed to peaks of circulating molecules, and that islet vessel permeability and molecule diffusion are dynamic processes that can be influenced by the physiological state. Together with well-characterized molecular and cellular mechanisms, the islet vasculature may thus be targeted by T2D insults to precipitate insulin secretory failure.

Acknowledgments. The authors thank C. Lafont and Dr. M. Desarménien (Institute for Functional Genomics, Montpellier, France) for technical assistance with surgery, Dr. C. Cazevieille (Centre Régional d’Imagerie Cellulaire, Montpellier, France) for assistance with the electron microscope, Dr. M. Strom (National Institute for Medical Research) for help with transgenic mouse model production, Dr. X. Bonnefont (Institute for Functional Genomics) for useful comments on the manuscript, and all the animal facility staff at the Institute for Functional Genomics.

Funding. The authors were supported by grants from the INSERM, the CNRS, the University of Montpellier, the National Biophotonics and Imaging Platform of Ireland (NBIPI), IBiSA, Diabetes UK (RD Lawrence Fellowship; 12/0004431 to D.J.H.), the Medical Research Council (MR/N00275X/1 to D.J.H.), the European Foundation for the Study of Diabetes (EFSD/Novo Nordisk Rising Star Fellowship to D.J.H.), the Agence Nationale de la Recherche (ANR BETA-DYN JCJC13 to M.S.), and the Région Languedoc-Roussillon (IPAM).

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

Author Contributions. A.M. performed experiments and analyzed data. D.J.H. designed and performed experiments, analyzed data, and wrote the manuscript. P.F. and F.M. analyzed data. A.G. and G.E.-C. performed experiments. C.J.P., I.C.R., and P.L.T. produced transgenic mice. P.M. designed experiments and wrote the manuscript. M.S. designed and performed experiments, analyzed data, and wrote the manuscript. M.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.