Vascular Endothelial Growth Factor-A and Islet Vascularization Are Necessary in Developing, but Not Adult, Pancreatic Islets

Mouse models were generated by breeding male hemizygous transgenic Pdx1-Cre mice (Mouse Genome Informatics [MGI] nomenclature: Tg(Pdx1-cre)^89.1Dam) (14) or Pdx1^PB-CreER mice (MGI nomenclature: Tg(Pdx1-cre/ERT)^1Mga) (15) with female mice expressing a conditional VEGF-A allele (Vegfa^fl/fl; MGI nomenclature: Vegfa^tm2Gne) (16). In all experiments, Cre-positive mice were compared with their respective age-matched Vegfa^fl/fl littermates. PCR genotyping was performed on tail biopsies with primers described (14,16,17). Before all terminal procedures, mice were anesthetized with a solution of 90 mg/kg ketamine and 10 mg/kg xylazine (Henry Schein, Melville, NY). Animal studies were approved by the Institutional Animal Care and Use Committee at Vanderbilt University Medical Center. Tm (cat. no. T5648; Sigma) administration was performed as previously described (18). Tm-treated mice were housed separately to prevent cross-contamination (19).

Islets were isolated by collagenase P digestion of the pancreas (20) and handpicked under microscopic guidance to nearly 100% purity. Quantification of islet VEGF-A production was performed by ELISA (R&D Systems, Minneapolis, MN) (1). Quantitative RT-PCR analysis on total cellular RNA from isolated islets was performed using Rn18s as housekeeping gene and the primer-probe sets (Applied Biosystems, Foster City, CA) listed in Supplementary Table 1 (21,22).

Embryonic, neonatal, and adult pancreata were processed and imaged as previously described (12,20). Cryosections and whole mounts were labeled by immunohistochemistry as previously described (20) using antibodies and working dilutions listed in Supplementary Table 2. Transmission electron microscopy (TEM) was performed on pancreata from Vegfa^fl/fl and Pdx1^PB-CreER;Vegfa^fl/fl mice 1 month after Tm treatment as previously described (1).

Glucose tolerance testing (2 g/kg of body wt i.p.) was performed after a 14- to 16-h fast as previously described (1). Islet function was studied in a dynamic cell perifusion system (1,23). Insulin concentration in perifusate fractions and insulin content in pancreatic extracts were determined by radioimmunoassay (cat. no. RI-13K; Millipore) (24). Hyperglycemic clamps were performed in conscious chronically catheterized (carotid artery and jugular vein) mice by the Vanderbilt Mouse Metabolic and Phenotyping Center according to established protocols (25).

In adult pancreatic sections, analyses of vessel density and size were performed with MetaMorph software (Universal Imaging, Downington, PA) as previously described (1). In whole mount embryonic pancreata, the ratio of platelet endothelial cell adhesion molecule (PECAM-1⁺) vessel area to area of Pdx1⁺ pancreatic epithelium was calculated on every third optical section of z-stacks obtained by confocal microscopy and averaged for each pancreas.

Pancreatic β-cell area measurements were based on whole-section imaging with either a tiling (Olympus BX41) or scanning (ScanScope FL; Aperio Technologies, Vista, CA) approach. For β-cell area, four pancreatic sections (at least 200 μm apart) were immunolabeled for insulin/amylase or alternatively islet hormones. Sections were counterstained with the nuclear label DAPI. Analysis of β-cell area was performed using MetaMorph and ImageScope software (Aperio). β-Cell proliferation was assessed by either pH3 or BrdU labeling (2,000–3,000 β-cells analyzed/tissue). BrdU (Sigma, St. Louis, MO) was administered at 0.8 mg/mL in drinking water for 7 days prior to tissue collection. For these analyses, n ≥ 3 mice were used per group.

Statistical analyses were performed with Prism software (GraphPad Software, La Jolla, CA). Comparisons between two groups were assessed by Student t test. Data are reported as means ± SEM. Statistical significance was assigned when P < 0.05.

To investigate the role of VEGF-A and ECs in adult islet function, we used complementary genetic Cre-loxP approaches to temporally inactivate VEGF-A during pancreas and islet development (Pdx1-Cre) (14) or in adult islets (Pdx1^PB-CreER) (15) (Fig. 1A). For studies of VEGF-A inactivation in adult islets, we determined a Tm dose effectively reducing VEGF-A expression with minimal toxicity (3 × 1 mg Tm), which resulted in a 91.4% reduction in islet VEGF-A secretion in vitro (18) (Supplementary Fig. 1).

Compared with pancreatic acinar tissue, islet endocrine cells from adult vehicle-treated and Tm-treated Vegfa^fl/fl mice and from vehicle-treated Pdx1^PB-CreER;Vegfa^fl/fl mice expressed high levels of VEGF-A (Fig. 1 and data not shown). In contrast, reduced VEGF-A immunoreactivity was observed throughout the pancreas of Pdx1-Cre;Vegfa^fl/fl mice (Fig. 1C). Reduced VEGF-A immunoreactivity was also observed in β-cells from adult Pdx1^PB-CreER;Vegfa^fl/fl mice both 1 and 3 months after Tm treatment, though VEGF-A was still expressed by α-cells (Fig. 1D and data not shown). Isolated islets from adult Pdx1-Cre;Vegfa^fl/fl mice and Tm-treated Pdx1^PB-CreER;Vegfa^fl/fl mice showed a significant reduction in Vegfa and Kdr (VEGF-A receptor 2) mRNA compared with their respective Vegfa^fl/fl controls (Fig. 1E and F). There were no compensatory changes in other VEGF family members (Fig. 1E and F and data not shown).

To investigate effects of VEGF-A inactivation in the developing pancreas and adult islets, we assessed the pancreatic and islet vasculature by immunohistochemistry and morphometry (Fig. 2A–C). Morphometric analysis of islets from adult Pdx1-Cre;Vegfa^fl/fl mice showed a 72% reduction in islet vessel density (Fig. 2D). The islet vessel density was reduced 53% in Tm-treated Pdx1^PB-CreER;Vegfa^fl/fl mice (Fig. 2F and data not shown). Similarly, the islet vessel size/branching was reduced 73% in adult Pdx1-Cre;Vegfa^fl/fl mice and 45% in Tm-treated Pdx1^PB-CreER;Vegfa^fl/fl mice (Fig. 2E and G). Similar vascular changes in Pdx1^PB-CreER;Vegfa^fl/fl islet were also observed 1 month after Tm treatment (data not shown). Exocrine vessel density and size were also reduced in Pdx1-Cre;Vegfa^fl/fl mice (Fig. 2H and I). In contrast, exocrine vessel density was unchanged in Pdx1^PB-CreER;Vegfa^fl/fl mice, though exocrine vessel size was slightly reduced (Fig. 2J and K).

We next examined how reduced VEGF-A expression affected capillary permeability in adult Pdx1^PB-CreER;Vegfa^fl/fl islets by TEM. As previously described (1,5), ECs lining capillaries in Tm-treated Vegfa^fl/fl controls contained highly fenestrated endothelium (Fig. 2L). The islet capillaries in Tm-treated Pdx1^PB-CreER;Vegfa^fl/fl mice displayed a wide range in the degree of fenestrated endothelium. Fenestrations were clearly less prevalent (open arrowheads in Fig. 2M and Supplementary Fig. 2A–D), although they still could be readily found in some capillaries of VEGF-A–deficient islets. In contrast to VEGF-A inactivation during development (1,5), the abundance of caveolae in intraislet capillaries of Tm-treated Pdx1^PB-CreER;Vegfa^fl/fl mice was more variable between individual capillaries (Supplementary Fig. 2A–D).

Early pancreas-wide inactivation of VEGF-A not only led to severely reduced islet vascularization but also resulted in a 43% reduction in overall pancreatic tissue mass (Fig. 3A). This change in pancreatic mass was accompanied by a 42% reduction in pancreatic insulin content (Fig. 3C) and a 36% reduction in the islet area (even when normalized to the pancreatic area [Fig. 3E]). In contrast, pancreas size, pancreatic insulin content, and β-cell area per pancreas were unchanged in Tm-treated adult Pdx1^PB-CreER;Vegfa^fl/fl mice (Fig. 3B, D, and F). Although the timing of VEGF-A inactivation in Pdx1-Cre;Vegfa^fl/fl and Tm-treated Pdx1^PB-CreER;Vegfa^fl/fl mice differentially affected β-cell mass, islet morphology was indistinguishable between the two mutants and their respective controls (Fig. 3G–I).

We showed previously that islet vasculature is important for insulin delivery into the peripheral circulation (1). For examination of the impact of reduced adult islet vasculature on islet function and glucose homeostasis, mice from each line were challenged with a bolus of glucose. Six-month-old male Pdx1-Cre;Vegfa^fl/fl mice, which have greatly reduced islet vascular supply and β-cell mass, showed severe glucose intolerance (Fig. 4A). In contrast, Tm-treated Pdx1^PB-CreER;Vegfa^fl/fl males had only mildly impaired glucose tolerance both 1 month and 3 months after VEGF-A inactivation (at ages 5 and 7 months [Fig. 4B–D]). In female mice, this phenotype only manifested 3 months after VEGF-A inactivation and was much less pronounced (Supplementary Fig. 3).

For further investigation of the milder changes in glucose tolerance in Pdx1^PB-CreER;Vegfa^fl/fl mice, an in vivo analysis of insulin secretion by a hyperglycemic clamp experiment was performed. It demonstrated that insulin release into systemic circulation was slightly delayed in mice with hypovascularized Pdx1^PB-CreER;Vegfa^fl/fl islets compared with Vegfa^fl/fl controls (Supplementary Fig. 4). These data were consistent with the greater peak glucose and subsequently delayed glucose clearance during the glucose tolerance test of Pdx1^PB-CreER;Vegfa^fl/fl mice (Fig. 4B–D).

Next, we sought to determine whether changes in islet vascularization had direct effects on β-cells. Reduced oxygenation, as a consequence of reduced islet vascularization, could lead to increased expression of hypoxia-inducible factor (Hif)1α and altered glucose sensing in β-cells (26,27). In Pdx1-Cre;Vegfa^fl/fl islets with 72% capillary loss, Hif1α showed only a modest β-cell nuclear enrichment, and islets maintained normal Gck and Glut2 expression (Supplementary Fig. 5), but Hif1α was undetectable in β-cells of Tm-induced Pdx1^PB-CreER;Vegfa^fl/fl islets (which maintained 50% of their vascular supply), suggesting that islets have a relatively high capacity to adapt to reduced vascularization.

Islets from both Pdx1-Cre;Vegfa^fl/fl and Tm-treated Pdx1^PB-CreER;Vegfa^fl/fl mice had normal Ins2 and Mafa gene expression, though Pdx1 mRNA levels in Pdx1-Cre;Vegfa^fl/fl islets were slightly reduced (Fig. 5A and B). Additionally, β-cells in both models of VEGF-A inactivation displayed normal nuclear expression of two principal regulators of insulin gene transcription, pancreatic and duodenal homeobox 1 (PDX1) and V-maf musculoaponeurotic fibrosarcoma oncogene homolog A (MAFA) (data not shown). β-Cells of Tm-treated Pdx1^PB-CreER;Vegfa^fl/fl mice had normal morphology by TEM with densely packed insulin secretory granules displaying their characteristic halos (Fig. 5C and D). Furthermore, a functional analysis of islets isolated from Tm-treated Pdx1^PB-CreER;Vegfa^fl/fl mice by an in vitro perifusion assay showed that reducing islet vasculature in adult islets does not affect β-cell insulin secretion (Fig. 5E).

Taken together, these data suggest that the insulin secretory defect in islets from both Pdx1-Cre;Vegfa^fl/fl and Tm-treated Pdx1^PB-CreER;Vegfa^fl/fl mice is mediated by impaired vascular delivery of insulin into the peripheral circulation. The defect in glucose clearance in Pdx1-Cre;Vegfa^fl/fl mice was most likely exacerbated due to the combined reduction in both islet vasculature and β-cell mass. A caveat of these studies is that Cre recombinase activity has been detected in the brain (and specifically within the hypothalamus) in both our Cre driver lines (28).

Prior studies using the Rip-Cre;Vegfa^fl/fl model to inactivate VEGF-A in developing islets showed that in the setting where islet vasculature is reduced by 50%, mice maintained normal pancreatic insulin content and β-cell mass (1). In contrast, the current studies with Pdx1-Cre;Vegfa^fl/fl mice demonstrate that the earlier and more extensive loss of VEGF-A and islet vasculature (72%) reduces pancreatic insulin content and β-cell mass but not Ins2 gene transcription. To address whether the reduction in β-cells was caused by reduced endocrine cell genesis, we first analyzed Pdx1-Cre;Vegfa^fl/fl pancreata at the onset of secondary transition (Fig. 6A–G). VEGF-A inactivation with Pdx1-Cre line resulted in a substantially reduced pancreatic vascularization at e14.5 (Fig. 6A–C). Although the density of PECAM-1⁺ vascular structures was significantly decreased in Pdx1-Cre;Vegfa^fl/fl pancreata compared with their controls, there was no difference in the organization of carboxypeptidase A1⁺ tip cells and SOX9⁺ trunk progenitors, suggesting no defect in tip-trunk compartmentalization. Furthermore, the density of neurogenin 3⁺ endocrine progenitors (Fig. 6F and G) and PAX6⁺ endocrine precursors (data not shown) per E-cadherin⁺ pancreatic epithelium was similar in Pdx1-Cre;Vegfa^fl/fl and Vegfa^fl/fl pancreata, suggesting that the process of endocrine cell specification is not perturbed by pancreatic hypovascularization.

At postnatal day 1 (P1), Pdx1-Cre;Vegfa^fl/fl islets were severely depleted of ECs (Fig. 6H and I) but had a normal mantle-core arrangement of endocrine cells (Fig. 6J and K). However, severe islet hypovascularization was associated with a 31% reduction in pancreatic insulin content (Fig. 6P) and a 50% reduction in β-cell proliferation (Fig. 5L, M, and Q). β-Cells in Pdx1-Cre;Vegfa^fl/fl islets continued to proliferate at lower rate throughout postnatal life, as measured by BrdU labeling in 2-month-old mice (Fig. 6N, O, and R). By contrast, β-cell proliferation was unchanged in 7-month-old Tm-treated Pdx1^PB-CreER;Vegfa^fl/fl mice compared to Vegfa^fl/fl controls (Supplementary Fig. 6). Collectively, these data indicate that pancreatic vasculature modulated by VEGF-A signaling is necessary for growth of both the exocrine and endocrine pancreas and that the extent and timing of VEGF-A inactivation differentially impact β-cell mass.

VEGF-A is critical for EC recruitment during islet development and is responsible for the highly vascularized state of pancreatic islets. Using complementary genetic models to temporally control VEGF-A inactivation, we show: 1) that loss of VEGF-A expression during early pancreas development impairs pancreatic and islet vascularization and leads to a sustained reduction in β-cell proliferation, resulting in reduced pancreatic and β-cell mass and impaired glucose clearance; 2) that loss of VEGF-A expression in adult β-cells reduces islet vascularization and the number of intraislet ECs; and 3) that despite a 50% reduction in adult islet vasculature, adult islets are surprisingly capable of maintaining islet morphology, gene expression, and β-cell mass with only minimal impairment in glucose homeostasis. These results indicate that VEGF-A and the islet vasculature it regulates have a different role during pancreatic and islet development than in the adult islet (Fig. 7) and that the adult islet and β-cell have considerable capacity to survive and function when blood flow, oxygenation, and intraislet ECs are markedly reduced.

The close relationship between islet cells and the vasculature begins in early pancreatic development, when ECs are required for the initiation of pancreas bud formation and insulin gene induction (9,10). Pancreatic vascularization is also important in differentiation and development of the exocrine pancreas (29). The assembly of the islet vasculature and the aggregation of developing islet cell clusters into an islet occur concomitantly in the later stages of development (E16-P1 in the mouse) (1). The data presented here provide new insights into these coordinated and integrated processes by showing that VEGF-A inactivation using a Pdx1-Cre leads to reduced β-cell mass at birth and that the β-cell mass is never recovered. These results are quite different from when VEGF-A is inactivated slightly later in development with an Ins-Cre where β-cell mass is not affected, but delivery of insulin into the systemic circulation is impaired by the reduced islet vascularization (1). The reduced β-cell mass in Pdx1-Cre;Vegfa^fl/fl mice is caused not by reduced islet endocrine cell genesis but, instead, by a reduction in β-cell proliferation. This reduced β-cell proliferation was noted at P1 and continued in adult islets. Since this reduction in β-cell proliferation was not seen when the Tm-based system was used to inactivate VEGF-A in adult β-cells, we conclude: 1) that either β-cells present in the developmental window between VEGF-A inactivation with the Pdx1-Cre and Ins-Cre respond differently to EC signals promoting proliferation or ECs during this brief developmental window produce different signals than those in later development and in the adult; 2) that if β-cells do not receive this proliferative stimulus from ECs during this developmental window, their subsequent ability to proliferate is compromised; and 3) that more extensive loss in islet vasculature with Pdx1-Cre vs. Ins-Cre leads to reduced β-cell proliferation. In addition, the early inactivation of VEGF-A with the Pdx1-Cre could affect developing islets by reducing the number of growth factor–producing ECs adjacent to developing islets and limiting delivery of nutrients through a reduced blood supply. Some of the current results with the Pdx1-Cre;Vegfa^fl/fl mice differ from a previous report (5), in which the pancreatic islet area was not reduced, but glucose tolerance was impaired. In the current report, more detailed analyses demonstrated that both pancreatic and β-cell mass were reduced, with β-cell mass being assessed by independent measurements of insulin content, β-cell area, and β-cell mass.

Because inactivation of VEGF-A during pancreas and islet development had such a major effect on islet β-cell mass and because islet oxygenation and blood flow are thought to be critical for insulin secretion, we predicted that inactivation of VEGF-A in adult β-cells would lead to reduced β-cell mass and markedly impaired insulin secretion. Although we were able to induce a substantial and prolonged reduction in islet VEGF-A production, the β-cell mass and function were remarkably similar to controls and the mice had only slight changes in glucose metabolism in vivo. In addition, islets from Tm-treated Pdx1^PB-CreER;Vegfa^fl/fl mice had no change in expression of the β-cell genes Ins2, Pdx1, or Mafa and no difference in pancreatic insulin content or β-cell area. We did not observe a reduction in pancreatic β-cell area or proliferation even 3 months after VEGF-A inactivation, suggesting that a 50% reduction of islet vasculature does not impair basal β-cell survival.

The main defect observed in Pdx1^PB-CreER;Vegfa^fl/fl mice was slightly impaired glucose tolerance after VEGF-A inactivation. This suggests that there is either a delay in the glucose stimulus reaching the β-cells or a delay in insulin release into the bloodstream. Because β-cells can secrete insulin into the interstitium and directly into capillaries (30), it is likely that the insulin produced in Tm-treated Pdx1^PB-CreER;Vegfa^fl/fl mice has a longer path to traverse before finding the bloodstream. Indeed, perifusion of islets from Tm-treated Pdx1^PB-CreER;Vegfa^fl/fl mice showed no alterations in glucose-stimulated insulin secretion in vitro, while the hyperglycemic clamp data demonstrated a slight delay in glucose-stimulated insulin secretion in vivo. Importantly, Pdx^PB-CreER;Vegfa^fl/fl mice were able to attain the insulin secretion seen in the control mice at the end of the clamp, consistent with the minimal difference in blood glucose levels at the end of the glucose tolerance test. Previous studies showed that adult Rip-Cre;Vegfa^fl/fl mice had more severe glucose intolerance and reduced insulin secretion, even though they display a reduction in intraislet vessel density similar to that in Tm-treated Pdx1^PB-CreER;Vegfa^fl/fl mice and normal β-cell mass (1). However, Rip-Cre;Vegfa^fl/fl mice also show dramatic changes in intraislet capillary ultrastructure, displaying few fenestrations and increased caveolae (1), similar to intraislet capillaries in mice treated with VEGF signaling inhibitors (13) and those reported by Lammert et al. (5). In contrast, Pdx1^PB-CreER;Vegfa^fl/fl mice display a mix in capillary ultrastructure, with the preservation of many fenestrations. It is possible that non–β endocrine cells and β-cells in which VEGF-A was not inactivated are able to partially compensate for the lack of VEGF-A production by β-cells in Pdx1^PB-CreER;Vegfa^fl/fl mice and that this lesser reduction in capillary permeability may account for the very mild glucose intolerance seen in this model. Our metabolic data contrast with data from mice treated with a VEGF receptor inhibitor, in which fasting blood glucose and glucose tolerance were improved (13). Although the inhibitor led to a reduction in pancreatic islet vascular density similar to that in our model, it also likely affected the vasculature of multiple other tissues.

These findings have implications for islet-related physiology and pathophysiology. For example, in insulin-resistant states such as obesity, pregnancy, and puberty, the β-cell must respond by increasing insulin biosynthesis and output, and one would predict that both the delivery of nutrients and oxygen and the efflux of insulin from the islet require adaptations in blood flow. The current results indicate that there is likely sufficient redundancy in islet blood flow so that islet vascularization would not limit the β-cell’s response to such challenges. We have noted that islet blood flow increases in hyperplastic islets, but this is the result of intraislet vessel dilatation and not an increase in the number of intraislet vessels (31). In these expanded islets with dilated vasculature, β-cells are a greater distance from a capillary and thus may have an islet/capillary relationship similar to that seen in islets with VEGF-A inactivated in adult mice.

Decreased vessel density and oxygen tension (32,33) in islet grafts may contribute to islet death or dysfunction in the early transplantation period. Since islets after transplantation are avascular, the reduction in islet blood flow would be greater than in Tm-induced reduction in islet vascularization, so efforts to hasten islet revascularization are still warranted. However, the current results do suggest that it may not be necessary for transplanted islets to achieve the highly vascularized state of islets within the pancreas. In fact, Olsson and Carlsson (34) reported recently that many islets in the pancreas are exposed to low oxygenation and are only recruited when necessary, so perhaps β-cells do not require a high oxygen tension and can function when vascularization is reduced. Thus, additional investigation is required before the relationship of islet vascularization and islet function in adult islets is clearly defined.

Finally, insufficient VEGF-A signaling to ECs and reduced EC signals to developing islet cells at critical stages of development not only reduced pancreatic and β-cell mass but also reduced basal β-cell proliferation in adult islets, suggesting that the embryonic effect persists. Thus, it is possible that reduced VEGF signaling during development could lead to a reduced ability to expand β-cell mass in response to challenges such as insulin resistance.

This work was supported by a Merit Review Award from the VA Research Service (BX000666), by National Institutes of Health (NIH) grants (DK69603, DK68764, DK89572, DK66636, DK63439, DK72473, and DK89538), by grants from JDRF International, by the Vanderbilt Mouse Metabolic Phenotyping Center (NIH grant DK59637), and by the Vanderbilt Diabetes Research and Training Center (DRTC) (DK20593). Islet isolation was performed in collaboration with the Vanderbilt Islet Procurement and Analysis Core (supported by the Vanderbilt DRTC). Tissue sections were acquired in part by the Vanderbilt Translational Pathology Shared Resource. Imaging was performed with support from the Vanderbilt University Medical Center Cell Imaging Shared Resource (NIH grants CA68485, DK20593, DK58404, HD15052, DK59637, and EY08126).

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

R.B.R. conceived and designed the experiments, performed the experiments, analyzed data, and wrote the manuscript. M.B. conceived and designed the experiments, analyzed data, and wrote the manuscript. A.S. performed the experiments. F.C.P. performed the experiments and reviewed and edited the manuscript. G.P. performed the experiments and reviewed and edited the manuscript. Q.C. and G.L.H. performed the experiments. J.K. conceived and designed the experiments and performed the experiments. C.S.T. performed the experiments and reviewed and edited the manuscript. C.D. conceived and designed the experiments, performed the experiments, analyzed data, and reviewed and edited the manuscript. O.P.M. conceived and designed the experiments, analyzed data, and reviewed and edited the manuscript. A.C.P. conceived and designed the experiments, analyzed data, and wrote the manuscript. A.C.P. 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.

Parts of this study were presented in abstract form at the 66th Scientific Sessions of the American Diabetes Association, Washington, D.C., 9–13 June 2006, and at the 69th Scientific Sessions of the American Diabetes Association, New Orleans, Louisiana, 5–9 June 2009.

The authors thank Guoqiang Gu and Douglas A. Melton (Harvard University) for the Pdx1-Cre mice, Maureen Gannon and Christopher V.E. Wright (Vanderbilt University) for the Pdx1^PB-CreER mice, and Napoleone Ferrara for the Vegfa^loxP mice. The authors are grateful to Christopher V.E. Wright, Roland Stein, and Guoqiang Gu for providing antibodies (Vanderbilt University).