Islet vascularization is regulated by primary endothelial cilia via VEGF-A dependent signaling

Accumulating evidence point to a role for primary cilia in endothelial cell function. Islet vascularization is an important determinant of islet function and glucose homeostasis. We have previously shown that β-cell cilia directly regulate insulin secretion. However, it is unclear whether primary cilia are also implicated in islet vascularization and thus contribute to glucose homeostasis. Objective To characterize the role of primary cilia in islet vascularization. Methods and Results At four weeks, Bbs4−/− islets show markedly lower intra-islet capillary density with enlarged diameters. We transplanted islets into the anterior chamber (ACE) of mouse eyes for longitudinal and non-invasive in vivo monitoring of vascular morphology. Bbs4−/− islets exhibited significantly delayed re-vascularization and enlarged vessels during engraftment. Similar vascular phenotypes were observed in two other ciliopathy models. By shifting the relative contributions of host versus donor endothelial cells in islet revascularization, we found that primary cilia on endothelial cells is essential for this process. Electron microscopy analysis further revealed a lack of fenestration in engrafted Bbs4−/− islets, partially impairing vascular permeability and glucose delivery to β-cells. Finally, we identified that Vascular endothelial cell growth factor A (VEGF-A)/VEGF receptor 2 (VEGFR2) signalling is involved in islet vascularization, islet function and vascular fenestration. In vitro silencing of two different ciliary genes in endothelial cells disrupts VEGF-A/ VEGFR2 internalization and phospho-activation of downstream signalling components. Consequently, key features of angiogenesis including proliferation, migration and tube formation are attenuated in BBS4 silenced endothelial cells. Conclusions Endothelial cell primary cilia regulate islet vascularization and vascular barrier function via VEGF-A/ VEGFR2 signaling pathway. Islet vascularization is impaired in four weeks old Bbs4−/− mice. Long-time monitoring of re-vascularization of WT and Bbs4−/− islets recapitulates the phenotype and demonstrates a role for cilia in islet vascularization and vascular barrier function. VEGF-A/ VEGFR2-dependent signalling is regulated by endothelial primary cilia.


Introduction
Organ growth, development and function are processes that are critically linked to the development of functional blood supply. Tissues rely on functional blood vessels for efficient delivery of oxygen or nutrients such as glucose. In addition, blood vessels are important to establish microenvironments or "niches" required for stem or progenitor cell maintenance among others 1 . Moreover, mis-regulated angiogenesis plays a role in numerous diseases, including but not limited to, diabetic complications, cancer progression and metastasis 2 . In the pancreas, the exocrine portion of the organ is organized by branching ducts while the endocrine islets of Langerhans are scattered throughout the exocrine tissue and interconnected by blood vessels 3 . Pancreatic bud formation and blood vessel formation are initiated at the same time; however, although both endocrine and exocrine pancreatic tissue are derived from the same progenitors, vessel density in the exocrine portion of the organ is considerably lower compared to intra-islet vessel density. This suggests that separate factors play a role in endocrine pancreatic vascularization. In addition, islet re-vascularization has been suggested to be a critical determinant in graft survival following islet transplantation in the treatment of Type 1 Diabetes 4-6 .
One of the main players in islet vascularization is vascular endothelial growth factor A (VEGF-A)/ VEGF receptor 2 dependent signaling; VEGF-A is secreted by islet endocrine cells (both α -and β -cells) and interacts with VEGFR2 receptor on microvessels in the periphery to recruit blood vessels to the islets 5,7 . Constitutive, targeted deletion of Vegfa from β -cells revealed a critical role for VEGF-A dependent signals in blood vessel formation, maintenance and function 5 . Mice were severely glucose intolerant but showed normal islet function in isolated islets, suggesting that either nutrients (such as glucose) did not get delivered efficiently to the β -cells or insulin disposal into the blood stream was significantly hampered. In contrast, removal of Vegfa from mature β -cells had a less severe impact on glucose tolerance although intra-islet capillary density was reduced by 50% 8 ; this could suggest that, after islet maturation, local VEGF-A/ VEGFR2 dependent signaling is less important for the maintenance of intra-islet capillaries and insulin disposal compared to during development.
Primary cilia are present on roughly eighty percent of the cells in the adult body plan 9 .
Among others, endothelial cells are ciliated and endothelial cilia have been implicated in flow-sensing and vascular hypertension, intracranial blood vessel formation, and atherosclerosis prevention [10][11][12] . Recent studies also unveiled a novel role of primary cilium in preventing vascular regression 13 . In metabolically active organs, they play a role in sensing metabolic signals and energy homeostasis 14 15 . In addition, we have shown that β -cell cilia play a role in insulin signaling, insulin secretion and glucose homeostasis 16 (Volta et al, in revision). Endothelial cells in both central and peripheral organs of insulin action have been implicated in insulin resistance 17 and diabetes etiology 18 . Here, we present one of the first studies addressing the role of endothelial cilia in islet vascularization and re-vascularization of transplanted islets.

Intra-islet capillary density is reduced in Bbs4 -/pancreata
To test if primary cilia play a role in pancreatic islet vascularization, we characterized intraislet capillaries in pancreatic cryosections of two-month old Bbs4 -/mice (N=8). Platelet endothelial cell adhesion molecule (PECAM-1) immunofluorescence as a marker of endothelial cells revealed 35% reduced intra-islet capillary density compared to wildtype (wt) littermate controls (Fig. 1A, P=0.0019). To calculate intra-islet capillary density, the relative PECAM-1 positive volume was normalized to insulin positive islet volume (Fig. 1B). In addition, the average vessel diameter was increased by 18% compared to that in wt controls 6 ( Fig. 1C, P<0.0001). Because tissue integrity is often compromised in cryopreserved samples, we corroborated our results by whole mount staining and imaging of freshly isolated islets of two-month old mice ( Fig. 1D-1F). In good agreement with our previous observations, capillary density was reduced by 33% and vessel diameter increased by 18% in islets of Bbs4 -/mice (P=0.0044 and 0.0067 respectively). Of note, we observed no change in pericyte coverage of intra-islet capillaries based on Neuron-glial 2 (NG2) immunofluorescence (Suppl. Fig. 1A, 1B, P=0.7045). In four-month old Bbs4 -/mice, however, the difference in intra-islet capillary density and diameter in Bbs4 -/mice approximated that of wt littermates (Supp. Fig. 1C-1E, P=0.0676 and 0.0736 respectively).
These dynamics implicate that primary cilia and centrosomal/ basal body integrity play a role during pancreas and islet development. Importantly, there was no detectable difference in vascularization of the exocrine portion of the pancreas at four months of age (Suppl. Fig. 1F).
Therefore, Bbs4 function seems to be mostly restricted to the endocrine pancreas and not relevant for the exocrine compartment.

Bbs4 -/endothelial cells exhibits blunted angiogenic response during islet re-vascularization
To test if cilia play a role in intra-islet capillary formation during engraftment of transplanted tissue as well as during development, we transplanted murine islets into the anterior chamber of the eye (ACE) 19 . This approach allows for longitudinal and non-invasive in vivo monitoring of islet engraftment and re-vascularization. We and others have previously shown that endothelial cells disappear over prolonged periods of cultivation of isolated islets.
Immediately after isolation, endothelial cells maintain the intra-islet capillary network. After two days in culture, endothelial cell clusters remain. Seven days post isolation endothelial cells have disappeared from the islet (Suppl. Fig. 2A). Thus, when transplanted shortly after isolation, a significant number of endothelial cells from the donor survives and contributes to revascularization of the islet graft 4,20,21 . Importantly, both islet cells and endothelial cells are ciliated. Therefore, to determine whether the role of primary ciliary/ centrosomal function in islet endothelial or endocrine cells is underlying the reduction in intra-islet capillary density, we capitalized on the temporal differences in the ratio of donor or recipient endothelial cells in intra-islet capillaries (Suppl. Fig. 2B).
In one experimental setting, islets isolated from Bbs4 -/and wt littermate controls were cultivated for two days before transplantation into the ACE of wt recipients (Fig. 1G, Suppl. injection of fluorescently labeled dextran once a week for a total of twelve weeks (Fig. 1G).
The percentage of dextran-related fluorescence in total islet volume (based on backscatter signal 22 ) was used as a measure of islet vascularity, and re-vascularization rate of each islet was calculated by normalizing its dextran-related volume at each individual week to the volume at twelve weeks. Two weeks after transplantation, Bbs4 -/islets showed significantly less re-vascularization compared to wt controls (Fig. 1H). Until four weeks after transplantation, re-vascularization remains lower than that of wt islets (Fig. 1H). Twelve weeks post transplantation, intra-islet vascular density measured as dextran related fluorescence intensity relative to islet volume is similar in both Bbs4 -/and wt islets (Fig. 1I).
Morphologically, the newly formed blood vessels vary among the two different genotypes: the vessel diameter is greater in Bbs4 -/islets, suggesting that there are fewer, wider islet capillaries in these islets compared to the wt controls during the first four weeks of engraftment ( Fig. 1J), similar to the results obtained from cryosections of young Bbs4 -/animals.
8 In the other experimental setting, islets were kept in culture for seven days after isolation before transplanting to the ACE of wt recipients; this treatment ablates endothelial cells in the donor tissue and favors almost exclusively revascularization by the recipients' vascular system (Suppl. Fig 2B, lower panel). We did not observe significant differences in re-vascularization rate over time or islet vascular density at the twelve-week endpoint ( Fig.   2A-2C). In addition, vessel morphology and diameter were similar in Bbs4 -/and wt control islets (Fig. 2D). Because Bbs4 -/endothelial cells were lost over the seven days cultivation period, our findings suggest that primary cilia on endothelial cells are required for the angiogenic response and thus islet re-vascularization.
To verify our conclusion, we reversed the transplantation strategy by transplanting wt islets to Bbs4 -/recipients either directly after isolation or after keeping them in culture for two and seven days respectively (Suppl. Fig. 3). To avoid confounding effects of glucose intolerance on islet engraftment, we transplanted islets into the eyes of four months old Bbs4 -/mice that were obese but not diabetic yet (Suppl. Fig. 3A-3C). Based on the apparent role of endothelial primary cilia in the angiogenic response, we expected stronger impairment in re-vascularization compared to wt recipients. Indeed, we observed significantly slower revascularization in this transplantation scheme and the rates correlated with the time islets were kept in culture (Suppl. Fig. 3D, 3E). The revascularization rate of wt islets transplanted to the ACE of Bbs4 -/mice seven days post-isolation was significantly lower than that of wt islets two days post-isolation (P=0.0147), and both were significantly lower than that of wt islets transplanted to wt ACE (P=0.0111). Interestingly, intra-islet vessel density after completion of re-vascularization was not significantly lower than that of wt islets transplanted into wt recipients at the twelve week endpoint (Suppl. Fig. 3F). The diameter of intra-islet capillaries, however, was markedly wider in wt islets transplanted to Bbs4 -/recipients than the other two groups after seven days in culture (Suppl. Fig. 3G). Average diameters of newly formed capillaries in wt and Bbs4-/-islet grafts in wt recipients.

Primary cilia of endothelial cells regulate islet re-vascularization
Because Bbs4 protein might have additional roles in cellular processes unrelated to the basal body, we used two additional mutant mouse models. Pitchfork (Pifo -/-) mice have dysfunctional cilia that cannot be properly disassembled 22 . We determined intra-islet vascular density in three-month old Pifo -/mice; PECAM-1 staining revealed 20% reduction in intraislet vascular density and 30% increase in vessel diameter in Pifo -/mice compared to littermate controls, similar to our observation in Bbs4 -/islets (Suppl. This generates a β -cell specific, inducible ciliary knockout mouse (Ift88 loxP/loxP ; Pdx1-CreER) with an efficiency of 50%, that we will refer to as β ICKO (pronounced BICKO) from here on (Volta et al., in revision) (Suppl. Fig. 5A). Recombination was induced between postnatal day 25 and 35 and we quantified intra-islet vascular density in Tx-treated β ICKO mice six weeks after induction; we found no observable difference between induced β ICKO mice and oiltreated controls (Suppl. Fig. 5B, 5C). In addition, we isolated β ICKO islets two months after Tamoxifen treatment and transplanted them into wt mice after overnight culture. We did not observe differences in relative re-vascularization rate between wt and Ift88 Δ/Δ ; Pdx1-CreER islets (Suppl. Fig. 5D-5G). Overall, our observations strongly suggest that cilia in endothelial cells regulate re-vascularization during islet engraftment.

Intra-islet vasculature plays a role in glucose metabolism
Besides a significant higher vascular density than exocrine pancreas, another hallmark of intra-islet capillary is fenestrae as interfaces for substance exchange between islets and the blood stream 23 . Thus, we further examined the barrier function of Bbs4 -/islets four months after engraftment, when the intra-islet vascular density is already comparable with wt. As previously reported, in isolated islets or Bbs4 -/depleted MIN6m9 cells, we did not observe a difference in β-cell glucose uptake or glucose metabolism 16  we observed an immediate increase of fluorescence intensity outside the islet graft followed by a plateau that was established at comparable times after dextran addition due to the dynamic aqueous humor flow. Importantly, both the initial rate and the plateau were decreased in Bbs4 -/islet grafts (Fig. 3E, P=0.0437). Simulation of this process also revealed different kinetics of dye leakage (Fig. 3F), suggesting that the delivery of medium sized molecules is less efficient in Bbs4 -/intra-islet capillaries. Nutrients such as glucose or fatty acids are small molecules often relying on active transport across cell membranes. But in the case of highly fenestrated capillaries, solutes including small molecules, ions and hormones pass freely through the pores between the blood and abluminal surface. To test if glucose delivery to β -cells were affected by the ultrastructural changes in islet capillaries, we injected fluorescently labeled glucose analog (2-NBDG) into animals transplanted with Bbs4 -/or wt islets. 2-NBDG was rapidly transported into β -cells from intra-islet capillaries and simultaneously leaked into the surrounding aqueous humor from iris vessels. We followed glucose uptake by measuring fluorescence intensity originating from β -cells and normalized it to the fluorescence intensity originating from the aqueous humor over time (Fig. 3G).
Importantly, wt β -cells internalized significantly more 2-NBDG than Bbs4 -/β -cells under the same period of time (Fig. 3H, P=0.0418). This result points to a critical role of ciliary/ basal body proteins in maintaining islet capillary fenestration with implications for the delivery of glucose and potentially other nutrients to islet cells. To investigate the physiological relevance of our finding, we utilized our transplantation model in combination with Bbs4 -/or wt islets to better understand how reduced intra-islet capillary density and permeability may affect islet output and glucose tolerance. It was previously shown that 100 islets (diameter between 150 and 200 μ m) transplanted into the ACE of chemically induced diabetic mice were sufficient to revert blood glucose levels to normoglycemic levels; this amount of pancreatic islets is defined as marginal islet mass 24 . C57/B6 J mice were treated with strepotozotocin and subsequently transplanted with marginal islet mass of wt or Bbs4 -/islets.
Four weeks after transplantation, mice transplanted with Bbs4 -/islets were significantly glucose intolerant compared to those mice transplanted with wt islets (Fig. 3I). Eight weeks after transplantation, mice showed impaired glucose tolerance and, 12 weeks posttransplantation, only mildly impaired glucose handling compared to those transplanted with wt islets (Fig. 3J, 3K). Overall, the phenotype is ameliorated over time, which is consistent with our previous observations in re-vascularization ( Fig. 1, 2), suggesting that a lack of capillaries in Bbs4 -/islets indeed negatively affects whole body glucose homeostasis.
Glucose handling is mildly impaired by the end of 12 weeks, which is probably caused by impaired glucose delivery to β -cells and a combination of defective 1 st phase insulin secretion 16 .  6A). We found no observable change in Tie1, Tie2, endothelial Nitric oxide synthase eNOS or Notch ligand Delta-4 (Dll4), while VEGFR2, Ephrin A1 and B2 were downregulated by 20% (Fig. 4A).
Angiogenesis is not exclusively regulated on the transcriptional level but also relies on a complex network of signaling cascades. Therefore, we tested if VEGFR2-dependent phospho-activation of downstream signaling molecules, including Phospholipase Cγ1 (PLCγ1), Protein Kinase B (Akt) and Mitogen-activated protein kinase (MAPK 42/44) were affected. We used HDMEC cells stably transfected with shRNA targeting BBS4 and tested before, seven and fifteen minutes after VEGF-A addition (Fig. 4B-4D, Suppl Fig. 6A, 6B). In control cells (scrambled), VEGFR2 Tyr 1175 phosphorylation is markedly increased at seven minutes (4.3 fold) and sustained until 15 minutes (3.4 fold) after stimulation (Fig. 4B, 4D).
Binding of VEGF-A to receptors triggers their endocytosis instantly, which is the initial step for subsequent intracellular signaling, and mainly regulated by EphrinB2 28-30 .
Several studies have shown that VEGFR2 does not signal from the plasma membrane but from other intracellular compartments, mainly early endosomes and the endosomal sorting complex required for transport (ESCRT) 30,31 . The observed attenuation of VEGFR2 downstream signaling may stem from dysregulation of these initial steps. Therefore, we examined the uptake of biotinylated VEGF-A in BBS4-depleted HDMECs. After 30 min incubation, BBS4-depleted cells had only internalized 50 % VEGF-A compared to controls ( Fig. 4E, 4F). We also tested VEGF-A uptake dynamics in a pancreatic endothelial cell line,

MS-1 cells. Similar to our observations in HDMEC, control MS-1 cells efficiently and rapidly internalized VEGF-A, while by comparison, Ift88-depleted MS-1 cells internalized
less VEGF-A after 30 min (58% compared to controls, Suppl Fig. 6F-6H). On the other hand, we also tested if VEGFR2 internalization was impaired in primary human endothelial cells depleted of BBS4 (Fig. 4G). Plasma membrane bound VEGFR2 was evident prior to VEGF-A stimulation in both control and BBS4-depleted HDMEC cells. After ten min incubation with VEGF-A, a significant number of VEGFR2 was internalized in control cells, while 25% more VEGFR2 remained on the cell in BBS4-silenced cells, which indicates decreased internalization of VEGFR2 (Fig. 4H) and may be explained by our previous finding of lowered EphrinB2 levels (Fig. 4A, 4B). These results suggest that the VEGFR2 internalization upon ligand binding is regulated by ciliary and basal body proteins.
Furthermore, we examined the relative expression of the same set of angiogenic factors in isolated islets and found most expression levels unchanged, while VEGF-A expression was reduced by 20% in Bbs4 -/islets compared to wt controls. Von-Hippel-Landau gene expression, a known ciliary protein mutated in Von Hippel-Landau disease and a known regulator of Hypoxia induced factor 1a, was also unchanged (Suppl. Fig. 6I). VEGF-A gene expression was also slightly decreased in β ICKO islets (Suppl. Fig. 6J). To test if this reduction was physiologically relevant, we tested VEGF-A secretion in Bbs4 -/and β ICKO islets. Loss of Bbs4 or Ift88 function respectively decreased VEGF-A secretion by 33% and 28% (Suppl. Fig. 6K, P=0.0063 and Suppl. Fig. 6L, P=0.007), which could be explained by the reduction in VEGF-A gene expression. Altogether, these results consistently suggest that the intra-islet VEGF-A signaling pathway is impaired in Bbs4 -/islets.

VEGF-A signaling triggers a complex array of biological processes including
endothelial cell proliferation and migration, two characteristic processes of angiogenesis. We A scratch assay revealed that BBS4-depleted cells covered only 73% of the distance of control cells in the same time frame (Fig. 4K, P<0.0001), similar to previous reports of ciliary involvement in cell migration 13 . Matrigel tube formation is a common in vitro assay for angiogenesis and we found that Bbs4-depleted cells formed less organized tubules and a subset of cells was not incorporated into tubular structures at all (Fig. 4L). Total tubular length was 80% in Bbs4-depleted compared to control cells (Fig. 4M, P<0.0001)). In addition, tubular structures formed by Bbs4-depleted cells established significantly fewer branching points than control cells (Fig. 4N, 75%, P<0.0001). These independent lines of evidence -reduced endothelial cell proliferation, migration and tube formation, and impaired vessel fenestration -all point to a role for cilia and basal body proteins in VEGF-A/ VEGFR2-dependent regulation of endothelial cell function. with previous observations. We conclude that the observed phenotypes of reduced intra-islet capillary density and impaired fenestration of intra-islet micro-vessels are caused by defective VEGF-A/ VEGFR2 dependent signaling in endothelial cells.
Intra-islet capillary networks provide endocrine cells with nutrients, oxygen and growth factors which support the development and maturation of islets. In addition, the microvasculature constantly adapts its barrier properties to the dynamic metabolic rates 32 .
Finely tuned interplay between nutrient delivery to islet endocrine cells, endocrine cell function and insulin disposal into the blood stream is essential for the maintainance of glucose homeostasis 5,33 . Whereas intra-islet capillary density normalizes over time in Bbs4 -/islets, vessel function remains impaired when cilia/ basal function is compromised more than four months after transplantation. Blood vessel fenestration is significantly reduced in Bbs4 -/islets. Fenestrae are key features of islet capillaries that enable efficient nutrient delivery from and insulin disposal into blood vessels. In vivo 2-NBDG diffusion is blunted and leakage of 40 KDa dextran diminished when Bbs4 -/is depleted from intra-islet capillaries. With respect to glucose handling, decreased vessel permeability in Bbs4 -/islets could impose an additional barrier on insulin disposal, adding to the observed defects in 1 st phase insulin. We thus suggest that defective delivery of nutrients such as glucose to β -cells and the disposal of insulin from β -cells into the blood stream may both contribute to impaired glucose handling of Bbs4 -/mice 16 .
Functional imaging of micro-vessels remains challenging and is largely limited to easily accessible tissues such as cornea and skin 34,35 . Transplantation of islets into the ACE not only renders established intra-islet capillaries more accessible for standard imaging techniques; it also provides the opportunity to follow dynamic processes such as islet revascularization under physiological as well as pathological conditions. This technique could also prove to be a useful resource for further investigation into blood vessel function in other tissues when transplanted into the ACE.
Finally, re-vascularization and establishing a functional interface between islet cells and blood supply are not only important in the context of islet transplantation but will likely prove vital for all approaches to β -cell replacement therapy. Moreover, taking into account the scarcity of organ donations worldwide, it is of paramount importance to optimize the outcomes of organ transplants. Revascularization rates in liver transplants correlate with long term graft survival, renal function and early allograft dysfunction 36

Animal models
Experimental procedures involving live animals were carried out in accordance with animal welfare regulations and with approval of the Regierung Oberbayern (Az 55.  (Volta et al.).

Islet isolation and transplantation
Islets were isolated by cannulation of the common bile duct of donor animals and infusion of 2.5ml collagenase P, 1.0 mg/ml in HBSS containing 25mM HEPES and 0.2% BSA (Sigma-Aldrich, USA). Inflated pancreata were dissected out and digested at 37°C for 10 min, and cultured in RPMI Medium 1640 (11mM D-glucose) supplemented with 10% fetal bovine serum, 100 IU/ml penicillin, 100 μ g/ml streptomycin and 2 mM L-Glutamine (all from Thermo Fisher Scientific, USA). Recipient mice were anesthetized with isoflurane (Baxter, USA) and fixed with a custom-made head holder (Narishige, Japan) as described previously 19 . A small incision was made in the cornea with a 25 G needle and a glass cannula containing islets was inserted through the opening into the anterior chamber of the eye.

In vivo imaging of islet grafts
Imaging was performed between one to twelve weeks post transplantation by confocal microscopy. Images of islet grafts were obtained by a Leica SP5 system with 25× objective

Construction of lentiviral vectors
We used a commercial lentiviral system from Sigma-Aldrich for designing and construct

Statistical tests
All results are presented as either mean ± SEM or shown as box-and-whisker plots (running from minimal to maximal values). Two-way ANOVA test was used for time course analysis, and two-tailed student t-test was used to assess statistical significance for other experiments, with a p value < 0.05 considered to indicate significance.