Deciphering von Hippel-Lindau (VHL/Vhl)-Associated Pancreatic Manifestations by Inactivating Vhl in Specific Pancreatic Cell Populations

The von Hippel-Lindau (VHL) syndrome is a pleomorphic familial disease characterized by the development of highly vascularized tumors, such as hemangioblastomas of the central nervous system, pheochromocytomas, renal cell carcinomas, cysts and neuroendocrine tumors of the pancreas. Up to 75% of VHL patients are affected by VHL-associated pancreatic lesions; however, very few reports in the published literature have described the cellular origins and biological roles of VHL in the pancreas. Since homozygous loss of Vhl in mice resulted in embryonic lethality, this study aimed to characterize the functional significance of VHL in the pancreas by conditionally inactivating Vhl utilizing the Cre/LoxP system. Specifically, Vhl was inactivated in different pancreatic cell populations distinguished by their roles during embryonic organ development and their endocrine lineage commitment. With Cre recombinase expression directed by a glucagon promoter in α-cells or an insulin promoter in β-cells, we showed that deletion of Vhl is dispensable for normal functions of the endocrine pancreas. In addition, deficiency of VHL protein (pVHL) in terminally differentiated α-cells or β-cells is insufficient to induce pancreatic neuroendocrine tumorigenesis. Most significantly, we presented the first mouse model of VHL-associated pancreatic disease in mice lacking pVHL utilizing Pdx1-Cre transgenic mice to inactivate Vhl in pancreatic progenitor cells. The highly vascularized microcystic adenomas and hyperplastic islets that developed in Pdx1-Cre;Vhl f/f homozygous mice exhibited clinical features similar to VHL patients. Establishment of three different, cell-specific Vhl knockouts in the pancreas have allowed us to provide evidence suggesting that VHL is functionally important for postnatal ductal and exocrine pancreas, and that VHL-associated pancreatic lesions are likely to originate from progenitor cells, not mature endocrine cells. The novel model systems reported here will provide the basis for further functional and genetic studies to define molecular mechanisms involved in VHL-associated pancreatic diseases.


Introduction
The von Hippel-Lindau (VHL) syndrome is an autosomal, dominant inherited disorder caused by mutations in the VHL tumor suppressor gene. VHL patients are predisposed to develop highly vascular tumors in multiple organs, including hemangioblastomas of the retina and central nervous system (CNS), clear cell renal carcinomas, pheochromocytomas, cyst and neuroendocrine tumors in the pancreas [1]. This familial cancer syndrome is caused by germ-line mutations in the VHL gene, which was mapped to chromosome 3p25 by positional cloning [2]. Following Knudson's two-hit hypothesis, loss or inactivation of the remaining wildtype allele is associated with VHL tumorigenesis [3,4]. The spectrum of VHL tumors in affected families varies [5] and biochemical analysis of the VHL gene product has provided the molecular basis that explains the phenotype-genotype correlations evident in VHL disease [6,7,8,9].
At the molecular level, the von Hippel-Lindau protein (pVHL) is a critical factor in the oxygen sensing pathway. Under normoxic conditions, pVHL forms a multiprotein complex with E3 ubiquitin ligase that targets the a-subunits of hypoxia-inducible factor (HIFa) for degradation by the proteosome [10,11]. Under hypoxic conditions, HIFa subunits escape ubiquitin-mediated proteolysis, allowing HIFa to accumulate, translocate to the nucleus, and activate downstream targets. In subsets of VHL mutations, the lack of functional pVHL leads to accumulation of HIF, and results in the activation of HIFa target genes even in the presence of oxygen [12]. Enhanced transcription of a wide variety of HIFa target genes, such as vascular endothelial growth factor (VEGF), is thought to contribute to the highly vascular tumors that develop in VHL patients [13]. Independent of its function in regulating the HIFa pathway, pVHL also binds other cellular proteins [14,15,16] and promotes extracellular matrix assembly [16,17]. For example, pVHL has been shown to bind and stabilize microtubule structures [18], and regulate fibronectin deposition to maintain vascular integrity [19,20].
Expression of VHL has been reported in most tissues and cell types, but little is known about its role in normal development [2,21,22]. Homozygous inactivation of Vhl in mice resulted in embryonic lethality at embryonic stage (E) E10.5-E12.5 due to defects in placental vasculogenesis [23]. Thus, tissue-specific knockout of Vhl has been utilized to investigate the biological functions of pVHL. Inactivation of Vhl utilizing an albumin-cre in liver or a mosaic b-actin-cre resulted in hepatic vascular tumors [24,25], similar to the ones observed in Vhl heterozygous knockout animals. Unexpected roles of Vhl in spermatogenesis, thymus cell survival, and bone development have also been reported with conditional loss of pVHL [25,26,27]. Notably, inactivation of Vhl in renal epithelial cells led to development of tubular cysts that share morphologic and molecular characteristics with renal cysts found in VHL patients [28]. While genetic analysis of Vhl in mice have uncovered novel functions of pVHL in various tissues, much is yet to be learned about pVHL's role in organs normally affected in VHL patients, such as CNS, adrenals and pancreas.
In VHL patients, the most common manifestations affecting the pancreas are benign cysts and microcystic adenomas (MCA), which occur in 35-75% of VHL patients [29,30]. In addition, 12-17% of VHL patients develop pancreatic neuroendocrine (islet cell) tumors, which possess the malignant potential to develop metastasis [31,32]. However, the cellular origins and molecular mechanisms leading to these pancreatic lesions found in VHL patients are not known. Here we present three mouse models of VHL in the pancreas generated by conditionally inactivating Vhl in specific pancreatic cell populations. Not previously reported in the literature, we showed that pVHL is dispensable for the normal functions of terminally differentiated islet a-cells and b-cells in the endocrine pancreas. Whereas mice with Vhl deletion in endocrine pancreas showed normal survival, mice lacking pVHL in pancreatic progenitor cells that give arise to both exocrine and endocrine pancreas (Pdx1-Cre;Vhl f/f), exhibited significant postnatal death. Most significantly, the few surviving Pdx1-Cre;Vhl f/f mice developed MCA and islet hyperplasia, similar to those found in VHL patients. Taken together, these results provided evidence suggesting that pVHL is functionally important for postnatal exocrine pancreas, and that VHL-associated pancreatic lesions originate from progenitor cell lineages. The novel model systems reported here will provide the basis for further genetic studies to define molecular events involved in VHL-associated pancreatic diseases.

Loss of pVHL in a-cells is Dispensable and Insufficient to Induce Tumorigenesis
VHL-associated pancreatic neuroendocrine tumors (PNET) afflict 12-17% of VHL patients, and these tumors have been demonstrated to exhibit focal positivity for glucagon [33]. Thus, we speculated that these PNET might originate from glucagonpositive a-cells deficient of pVHL in the endocrine pancreas. To inactivate Vhl specifically in a-cells, we first generated an a-cell specific Cre transgenic line (Glu-Cre) utilizing a rat glucagon promoter sequence to direct Cre recombinase expression. By crossing a Glu-Cre transgenic line with the Z/AP reporter mice [34], we confirmed a-cell specific expression of Cre recombinase in Glu-Cre;Z/AP pancreas at 5 and 12 months of age. As expected, expression of Cre-recombinase localized in a-cells at the outer ring of endocrine islets, as indicated by purple alkaline phosphatase (AP) staining ( Figure 1A, right panels). No positive AP expression in pancreas was observed in age-matched control genotypes of Glu-Cre and Z/AP mice (data not shown).
To determine the functional significance of Vhl in a-cells, the Glu-Cre transgenic mice were next bred with mice in which exons 2 and 3 of the Vhl alleles are flanked by loxP sites (flox or f) [25]. Since variability in phenotype was documented in other tissuespecific knockouts of Vhl [24,25], we generated Vhl-deficient a-cells in A/J and C57BL/6 backgrounds. At the time of weaning (3-4 weeks after birth), we obtained the expected Mendelian frequencies for all genotypes in both background strains (Table 1). Mouse pancreases with Vhl deletion in a-cells (Glu-Cre;Vhl f/f, n = 12) appeared histological normal when compared to age-matched control genotypes (Glu-Cre, n = 8 and Vhl f/f, n = 6) from 10 months to 23 months of age (data not shown). No obvious physiological and behavioral phenotypes were observed for Glu-Cre;Vhl f/f mice. To confirm that Vhl alleles were indeed deleted in Glu-Cre;Vhl f/f mice, we isolated a-cells, b-cells and endothelial cells from pancreatic islets via flow cytometry, and obtained genomic DNA for each cell population. Genotyping PCR analysis showed the expected deletion of Vhl alleles only in glucagon-positive a-cells, but not in insulin-positive b-cells and lectin-positive endothelial cells in Glu-Cre;Vhl f/f animals at 27 months of age ( Figure 1B). Together, these results demonstrated that Vhl is not essential for mature a-cells, and that deficiency of Vhl in a-cells is insufficient to cause tumorigenesis in the endocrine pancreas.
Vhl is not essential for Normal Function of Differentiated Islet b-cell Since b-cells are the major cell type in the endocrine pancreas and some VHL PNETs are positive for insulin staining [33], we next hypothesized that Vhl deficiency in insulin-positive b-cells might lead to PNET development. To conditionally inactivate Vhl in b-cells, we crossed Rip-Cre transgenic mice [35] with mice in which exons 2 and 3 of the Vhl alleles are flanked by loxP sites (flox or f). Similar to our findings in the Glu-Cre;Vhl colony, mice from the Rip-Cre;Vhl colony were generated at the expected Mendelian frequencies for all genotypes regardless of the strain backgrounds ( Table 2). Histological analysis of Rip-Cre;Vhl f/f mouse pancreases (n = 4) did not show any abnormality or islet tumor formation when compared to age-matched control genotypes (Rip-Cre and Vhl f/f) at 15 months of age ( Figure S1A). Immunostaining of endocrine markers, such as insulin and glucagon, did not reveal distinct features unique to mice deficient for pVHL in bcells (data not shown). Determination of Vhl allele status in the Rip-Cre;Vhl f/f pancreas utilizing DNA isolated from microdissection further confirmed that Vhl deletion only occurred in the expected endocrine islets, but not in the surrounding exocrine pancreas ( Figure S1B).
Notably, the physical appearance of the Rip-Cre;Vhl f/f mice were proportionally smaller than their littermates at time of weaning. Indeed, data confirmed that Rip-Cre;Vhl f/f mice weighed significantly less than age-matched littermates at all time points analyzed (Figure 2A). Since Vhl has been shown to regulate glucose metabolism [36], we next measured fasting glucose levels in a cohort of Rip-Cre;Vhl mice to determine if an elevated basal glucose level was responsible for the small size seen in these mice.
However, our data demonstrated no significant differences in basal glucose levels between control genotypes and Rip-Cre;Vhl f/f mice at all time points measured ( Figure 2B). We then speculated that low expression of Cre-recombinase in the hypothalamus reported for Rip-Cre transgenic mice [35] affects hormone feedback loops regulated by the hypothalamus due to inactivation of Vhl. Among various hormones regulated by the hypothalamus, we chose to measure growth hormone since its deficiency is associated with lean body mass. Again, our data did not show significant differences in serum growth hormone levels between control genotypes and Rip-Cre;Vhl f/f mice at all time points tested ( Figure 2C). Therefore, it is unclear if the small size of Rip-Cre;Vhl f/f mice is due to the loss of Vhl in pancreatic islet b-cells or in the hypothalamus.

Vhl is Functionally Important for Postnatal Exocrine Pancreas
The most common manifestations of pancreatic VHL disease are benign cysts and microcystic adenoma, which occur in 35-75% of VHL patients [29,30]; however, it remains unclear whether these lesions arise from the endocrine or exocrine pancreas. To investigate whether loss of Vhl in the exocrine pancreas can lead to cyst development, we utilized pancreatic and duodenal homeobox 1 (Pdx1) transgenic mice in which cre recombinase expression was driven by the Pdx1 promoter [37]. By breeding Pdx1-Cre transgenic mice with Vhl floxed mice, we generated mice with Vhl deletion in pancreatic endocrine and exocrine cells as a result of Pdx1-Cre expression in pancreatic progenitor cells during embryogenesis [37]. In contrast to mice deficient in Vhl in a-cells or b-cells, we observed some lethality in Pdx1-Cre;Vhl f/f mice. At the time of weaning, the distribution of genotypes was significantly different from the expected values with fewer than expected Pdx1-Cre;Vhl f/f mice in all three different genetic backgrounds (Table 3). Statistically, observed lethality is more severe in the A/J and Balb/C genetic backgrounds than in the C57BL/6 background.
To determine if the lethality in Pdx1-Cre;Vhl f/f mice occurs embryonically or postnatally, Mendelian ratios of the six possible genotypes were observed immediately after birth and during different postnatal time intervals (Table 4). The Pdx1-Cre;Vhl f/f mice were born at the anticipated Mendelian ratio; however, these mice could not survive more than 5 days postnatally (P5). We further analyzed whole-mount histology of mouse pups (P1-P5) in an attempt to uncover the cause of death. However, pathologists blinded for genotypes could not distinguish Pdx1-Cre;Vhl f/f pups from pups of control genotypes ( Figure 3A). To confirm that the Vhl alleles were indeed deleted in the pancreas of Pdx1-Cre;Vhl f/f newborn pups, whole pup body DNA and pancreatic DNA were isolated by laser capture microdissection. Genotyping analysis demonstrated the expected Vhl deleted allele in the Pdx1-Cre;Vhl f/f pup pancreas, as well as the floxed Vhl allele in cells of nonpancreatic lineage, such as epithelial and endothelial cells ( Figure 3B, animal #3). Thus, whereas mice with homozygous inactivation of Vhl in a-cells or bcells (Glu-Cre;Vhl f/f and Rip-Cre;Vhl f/f, respectively) showed normal survival rates, Pdx1-Cre;Vhl f/f mice exhibited reduced postnatal survival that is likely to be associated with exocrine dysfunction due to the loss of Vhl in the exocrine pancreas.

Loss of Vhl in Exocrine Pancreas Recapitulates Cysts and Microcystic Adenoma (MCA) That Develop in VHL Patients
Since the postnatal lethality observed for Pdx1-Cre;Vhl f/f did not exhibit 100% penetrance, we were able to perform histological analysis of the pancreases of the few surviving Pdx1-Cre;Vhl f/f at adulthood in the C57BL/6 background. At 6-7 months of age, Pdx1-Cre;Vhl f/f mouse pancreases (n = 2) were indistinguishable from control genotypes (data not shown). However, at 16-18 months of age, an obvious loss of exocrine pancreas and fat replacement were observed in Pdx1-Cre;Vhl f/f mice (n = 3, Figure 4A). Pathologists blinded for mouse genotypes further identified the presence of cysts and MCA in all Pdx1-Cre;Vhl f/f mice lacking pVHL in the pancreas ( Figure 4A, panel c and f).
In VHL patients, the pancreatic cysts and MCA are highly vascularized and characterized by a layer of epithelial lining intermixed with endothelial cells forming capillaries around the cysts [38]. To investigate the similarity between the MCA developed in Pdx1-Cre;Vhl f/f pancreas and that found in VHL patients, we used a CD31 antibody to identify endothelial cells, and a cocktail of cytokeratin antibodies (MAK6) to identify the epithelial cell lining of MCA. Immuno-histochemical staining revealed that epithelial lining of the MCA was positive for MAK6 but not CD31, and that CD31 positive endothelial cells were intermixed within MCA ( Figure 4B). The extensive vasculature network of MCA was more evident in Pdx1-Cre;Vhl f/f mice injected with FITC-lectin prior to euthanizing ( Figure 4C, panels a and b). Utilizing immuno-fluorescent staining, we further demonstrated that cytokeratin-positive epithelial lining and lectin-positive vasculature did not co-localize, but instead were distinctive cellular structures displayed by MCA in Pdx1-Cre;Vhl f/f animals ( Figure 4C, panels c and d).
Within the pancreas tissue largely replaced by fat deposition, some islets in Pdx1-Cre;Vhl f/f animals appeared small and abnormally shaped without much exocrine acinar cells surrounding them ( Figure 4A, panel f), whereas some appeared hyperplastic ( Figure 5A). Regardless of the size, islets of Pdx1-Cre;Vhl f/f were characterized by disorganized, dilated and tortuous vascular networks within, as well as outside the islets ( Figure 5B). Since VHL neuroendocrine tumors are highly vascularized, it is possible that these islets lacking pVHL might be progressing toward developing pancreatic neuroendocrine tumors.
Modest Upregulation of Hif1a is Associated with Pancreatic Phenotypes Developed in Pdx1-Cre;Vhl f/f Mice pVHL deficiency is known to result in constitutive HIF asubunits stabilization and increased expression of HIF target genes [11]. To examine if the pancreatic phenotypes observed in Pdx1-Cre;Vhl f/f mice were associated with activation of Hif1a or Hif2a pathways, quantitative real-time PCR analysis was performed to detect Hif1a and Hif2a mRNA expression. As shown in Figure 6A, expression of Hif1a was upregulated in Pdx1-Cre;Vhl f/f pancreas (n = 3) when compared with control mice (n = 3). Expression of Hif2a and a Hif2a-preferred target gene (Vegf) [39] appeared similar between control and Pdx1-Cre;Vhl f/f mice ( Figure 6A, data not shown). While the upregulation of Hif1a mRNA in Pdx1-Cre;Vhl f/f pancreas was not statistically significant, we speculated that this might be the reason why a few of the Pdx1-Cre;Vhl f/f mice survived postnatal lethality. Consistent with the mRNA data, we further confirmed the trend that protein expression of Hif1a was upregulated in the pancreas of pVHL deficient mice utilizing a Hif1a ELISA assay ( Figure 6B).

Discussion
The major significance of this study is that generation of three new pancreatic mouse models of VHL has allowed us to address the cellular origins and biological roles of VHL during pancreatic development and tumorigenesis. In terms of VHL's role during pancreatic development, we demonstrated that deletion of Vhl in endocrine a-cells and b-cells which terminally differentiated around E12.5 [40], does not appear to affect normal functions of endocrine pancreas during embryogenesis, postnatal development, and in adulthood (Glu-Cre;Vhl f/f and Rip-Cre;Vhl f/f animals). In contrast, deletion of Vhl in pancreatic progenitor cells that give arise to ductal, exocrine and endocrine pancreas resulted in significant postnatal death in mice (Pdx1-Cre;Vhl f/f animals), even though pancreatic organogenesis was not affected during embryonic development. Together, this data suggest that VHL is functionally important in postnatal ductal and exocrine pancreas, and that pVHL is not essential for mature endocrine pancreatic cells. While this finding uncovers the novel aspect of VHL in postnatal ductal and exocrine pancreas, it remains to be determined whether HIF-dependent or HIF-independent pathways are involved.
In terms of VHL's role during pancreatic tumorigenesis, little is known about the cellular origins and molecular mechanisms related to VHL-associated pancreatic diseases, even though a large portion of VHL patients are affected by pancreatic manifestations, such as cysts, MCA and neuroendocrine tumors. By specifically inactivating Vhl in distinct pancreatic cell populations, we reported the first mouse model of VHL that recapitulates some of the clinical features found in the pancreas of VHL patients, and demonstrated that inactivation of Vhl in the endocrine pancreas is insufficient to initiate tumorigenesis. VHL-associated pancreatic lesions were only found in mice where Vhl was inactivated in pancreatic progenitor cells (Pdx1-Cre;Vhl f/f mice), but not in mice in which Vhl was inactivated in mature endocrine cells (Glu-Cre;Vhl f/f and Rip-Cre;Vhl f/f mice). Thus, our data supports the notion of a progenitor cell origin for these VHL-associated pancreatic lesions. In addition, it has been previously suggested by histopathological studies that MCA originated from pancreatic exocrine cells, such as the centroacinar cells [41,42] or ductal cells [43,44]. Moreover, none of the VHL patients (n.108) with PNET disease evaluated at our center had a functional neoplasm by hormone levels or symptoms [45]. Together, these data suggest that VHL-associated MCA and PNET originate from pancreatic ductal or exocrine progenitor cells, not differentiated endocrine cells. Analogous to our hypothesis, VHL-associated hemangioblastomas have been demonstrated to derive from embryologic multipotent cells [46]. However, further studies will be required to definitively determine if these VHL-associated pancreatic manifestations can be recapitulated when Vhl is specifically deleted in progenitor cells of ductal, exocrine or endocrine lineage.
Similar to pervious reports on Vhl knockout mice [24,25], we also observed variability in phenotypic penetrance of the Vhl deletion due to differences in mouse genetic backgrounds. The postnatal lethality in Pdx1-Cre;Vhl f/f mice was most severe in A/ J and Balb/C background strains, but not as severe in the C57BL/ 6 background strain. This finding suggests that strain-specific genetic modifiers may provide protection for the survival of these mice with pVHL deficiency in the pancreas. Notably, we showed that pancreatic phenotypes observed in Pdx1-Cre;Vhl f/f mice correlate with a modest upregulation of Hif1a, but not HIf2a. It is thus possible that these strain-specific genetic modifiers may function to interfere with VHL downstream signaling, and compensate for the loss of pVHL. This does not exclude the possibility that the pancreatic phenotypes observed in Pdx1-Cre;Vhl f/f mice resulted from mechanisms completely independent of HIF signaling pathways, such as VHL's role in regulating extracellular matrix assembly [47]. Identification of these modifier genes and further comparative functional analysis utilizing mice deficient in Hif1a and Hif2a will provide insight into the precise molecular mechanisms leading to the development of VHLassociated pancreatic disease.
An unexpected observation during our studies was the small size seen in mice deficient in pVHL in the islet b-cells. While our analyses excluded the idea that alterations in growth hormone and basal glucose levels led to this phenotype in Rip-Cre;Vhl f/f mice, it is described in the literature that pVHL regulates glucose metabolism in liver and kidney cells [10,36]. Establishment of our Rip-Cre;Vhl f/f homozygous mice now allows us to begin investigating whether or not pVHL also plays an important role in glucose sensing in pancreatic islet b-cells, such as via glucose transporter 1 (Glut1). However, phenotypes seen in Rip-Cre;Vhl f/f mice are confounded by the fact that Rip-Cre is expressed at a low level in the hypothalamus [35]. Thus, it would be critical to confirm and determine that deletion of Vhl indeed occurred in particular nuclei within the hypothalamus in Rip-Cre;Vhl f/f homozygous mice. Only when that information is available, will we be able to conclusively interpret the physiological data resulting from the loss of pVHL in pancreatic islet b-cells or in the hypothalamus.
In summary, to decipher the functional significance of pVHL in the pancreas during development and tumorigenesis, we conditionally inactivated Vhl in distinct pancreatic cell populations and reported the first mouse model of VHL that recapitulates clinical features found in the pancreas of VHL patients. Importantly, our data demonstrated that pVHL is functionally important for postnatal ductal and exocrine pancreas, and suggested that pancreatic progenitor cells, not mature endocrine cells, as the cell of origin for VHL-associated pancreatic lesions. The novel mouse model systems reported in this study will provide the foundation for further functional and genetic analysis to advance our understanding of VHL-associated pancreatic manifestations.

Ethics Statement
National Cancer Institute (NCI) and NCI-Frederick are accredited by AAALAC International and follows the Public Health Service Policy for the Care and Use of Laboratory Animals. Animal care was provided in accordance with the procedures outlined in the ''Guide for Care and Use of Laboratory Animals (National Research Council; 1996; National Academy Press; Washington DC). All animal experiments were conducted in accordance with NIH-approved protocols and guidelines.

Animals and Genotyping
Mice carrying the Vhl alleles flanked by loxP sites [25] were rederived in C57BL/6, A/J and Balb/C backgrounds before crossing with different Cre transgenic mice. The Pdx1-Cre transgenic mice were a kind gift from Dr. D. Melton [37], and the Rip-Cre (B6.Cg-Tg (Ins2-cre)25Mgn/J) transgenic mice were purchased from The Jackson Laboratory (stock number: 003573). The Glu-Cre transgenic mice were generated utilizing the pBKCMV/Glu-CreGH plasmid kindly provided by Dr. M. Magnuson. Briefly, the plasmid fragment containing 2.3 kb of rat glucagon promoter upstream of the Cre-recombinase was isolated by SalI and NotI digestion prior to injection to generate transgenic founder lines. Tissue expression of Glu-Cre was evaluated in offspring from crosses of the Glu-Cre transgenic mice with Z/AP reporter mice [34]. Detailed analysis of the Glu-Cre transgenic mice will be described elsewhere. Cre recombinase expression in tissues of Z/AP mice will delete the floxed LacZ expression cassette upstream of the hPLAP gene, allowing hPLAP expression and detection by standard staining techniques. Tissue sections were counterstained with nuclear fast red (Sigma, St. Louis, MO).

Animal Tissue Collection and Processing
The pancreatic tissue was processed for frozen histological analysis by embedding tissues in Tissue-Tek OCT freezing medium, and for formalin-fixed paraffin embedding (FFPE). Frozen (10-20 mm) and FFPE pancreas sections (5 mm) were routinely stained with Mayer's hematoxylin and eosin (H&E) for histopathological analysis. For all histological analysis, Cre;Vhl f/f mice were compared with age-matched controls genotypes of Cre, Vhl f/f, Vhl f/+ and Cre;Vhl f/+ mice. Only representative control genotypes are shown in figures. For whole mount mouse pup analysis, a sagittal incision was made to allow formalin fixation to penetrate internal organs. Mouse pups were divided into two parts via the sagittal incision prior to the paraffin embedding process.
FITC-lectin perfusion to visualize the blood vasculature was performed as described [49]. Briefly mice were injected intravenously via tail vein with 50 ug of FITC-labeled lectin (Lycopersicon esculentum, Vector Laboratories, Burlingame, CA), which was allowed to circulate for 3 minutes. Then mice were euthanized via cervical dislocation, and the pancreas was removed. All procedures involving mice were performed with approval by the NIH Institutional Animal Care and Use Committee.

Islet Isolation, Dissociation, and Flow Cytometry
Mice were perfused with FITC-lectin as described above prior to being euthanized. Pancreatic islets were isolated by standard techniques, with minor adaptations. Briefly, pancreata were inflated via bile duct cannulation and retrograde pancreatic duct injection of 3-4 ml of ice-cold collagenase type XI (1 mg/ml in HBSS, Sigma, St. Louis, MO). Following digestion (37uC, 14 min), pancreata were dispersed by gently aspirating through a 14G needle, then filtered through a metal strainer (0.8 mm). Pancreas suspensions were then subjected to buoyant density gradient centrifugation (14-15% Optiprep, Accurate Chemicals, Westbury, NY), followed by handpicking with great care to collect all visible islets.

Glucose Measurements
Mice underwent a 24 hour fast prior to collecting whole blood via a tail snip. Blood glucose was measured using a glucometer (Ascensia Contour, Bayer HealthCare, Mishawaka, IN).  For immuno-staining of FITC-lectin injected pancreas, frozen sections were briefly fixed in 4% paraformaldehyde, washed in PBS, and incubated in blocking buffer (5% normal goat serum/2.5% BSA in PBS). Primary antibodies guinea pig anti-swine insulin (1:500, DAKO, Carpinteria, CA) and rabbit anti-cytokeratin 18 (1:50, Santa Cruz Biotechnology, Santa Cruz, CA) were diluted in 0.56 blocking buffer. Sections were incubated with primary antibody overnight at 4uC in a humidified chamber followed by brief washes in PBS. Secondary antibodies Alexa Fluor 594 or 633-conjugated anti-guinea pig and Alexa Fluor 594 -conjugated anti-rabbit (1:200; Invitrogen, Carlsbad, CA) were applied to sections. After incubation at room temperature for one hour, the fluorescently stained sections were washed several times in PBS, cover slipped using mounting medium with DAPI (Vector Laboratories, Burlingame, CA), and visualized using either a Zeiss Axiovert fluorescence microscope (Carl Zeiss MicroImaging, Thornwood, NY).

RNA collection and Quantitative Real-Time PCR
Islet RNA was collected from frozen pancreas sections on slides by microdissection and extracted using the PicoPure RNA Isolation kit (Arcturus, Mountain View, CA). RNA concentration was determined by NanoDrop ND-1000 (NanoDrop, Wilmington, DE), and equal amounts (1.4 mg) of total RNA was used to generate cDNA. Reverse transcription and quantitative PCR was performed as described previously [50]. TaqMan primers and probes were purchased from Applied Biosystems (Foster City, CA): Gapdh (Mm99999915_g1), Hif1a (Mm00468878_m1), Hif2a/ Epas1 (Mm0438717_m1).

Hif1a ELISA
Total pancreatic protein lysate was harvested by sonicating pancreas on ice in RIPA buffer (Pierce, Rockford, IL) containing 16 Complete Protease Inhibitor Cocktail (Roche, Nutley, NJ). Protein concentration was determined with the Quick Start Bradford Protein Assay (Bio-Rad Laboratories, Hercules, CA). Mouse Hif1a protein was measured using Surveyor IC human/ mouse total Hif1a immunoassay according to manufacturer's protocols (R&D Systems, Minneapolis, MN).

Statistical Analysis
Statistical analysis was performed using GraphPad Prism v.5.01 and Microsoft Excel. The distributions of genotypes after interbreeding of Cre;Vhl f/+ mice were compared to the expected Mendelian distributions with Chi-square-test. Chi-square and pvalue (2 degrees of freedom) are indicated for each strain and each postnatal time point in Tables. Mouse body weight, serum growth hormone levels, and fasted blood glucose levels were evaluated with one-way ANOVA with multiple comparisons testing. Expression data of Hif1a and Hif2a were analyzed with Mann-Whitney (unpaired, nonparametric) t-test. A p value less than (,) 0.05 was considered statistically significant. Figure S1 A. H&E staining of representative Rip-Cre and Rip-Cre;Vhl f/f pancreas at 15 months of age. Islets are as indicated (Is), and images are taken at 1006. B. Genotyping PCR to determine Vhl allele status using genomic DNA isolated from