Maintenance of Hepatic Nuclear Factor 6 in Postnatal Islets Impairs Terminal Differentiation and Function of β-Cells

pdx1^PBHnf6 transgenic mice have been previously described (15) and maintained for >12 generations on a mixed (B6D2) background. For embryonic analyses, the morning of the vaginal plug was considered e0.5.

Construction of the RIPrGLUT2 transgene has been previously described (21). Expression was assayed by immunohistochemistry using a species-specific anti-rat antibody (21). One transgenic line, chosen for its high level of transgene expression in pancreatic β-cells and its similarity to previously described transgenic lines, was used for all experiments.

All mouse studies were performed in accordance with the Vanderbilt Institutional Animal Care and Use Committee guidelines under the supervision of the Division of Animal Care.

Digestive organs or isolated pancreata from embryonic and adult stages were dissected in PBS and fixed immediately in 4% paraformaldehyde (4°C, 45–60 min). Tissues were dehydrated in an ethanol series, embedded in paraffin, and sectioned at 5 μm.

Indirect protein localization was obtained using species-specific donkey Cy2- or Cy3-conjugated secondary antibodies (1/500; Jackson Immunoresearch) and the following primary antibodies incubated overnight at 4°C: guinea pig anti-insulin (1/1,000; Linco), rabbit anti-mouse GLUT2 (1/500) (22), rabbit anti-rat GLUT2 (1/500) (21), and rabbit anti-MafA (1/1,000; Bethyl Labs). Detection of GLUT2 and MafA required antigen retrieval in 10 mmol/l citrate buffer (pH 6.0) or 10 mmol/l Tris-EGTA buffer (pH 9.0), respectively. YoPro1 (Molecular Probes) was used to visualize nuclei. Images were captured on an Olympus BX41 research microscope or a Zeiss LSM 510 confocal microscope. Image tonal range and color balance were minimally adjusted via histogram using Adobe Photoshop.

Pancreata from perinatal (postnatal day [P] 3) and adult (6- to 8-week-old) mice were fixed in 2.5% glutaraldehyde in 0.1 mol/l sodium cacodylate buffer (SCB) for 1 h and washed in SCB. Specimens were postfixed in 1% osmium tetroxide in SCB and washed in SCB. Tissues were dehydrated, infiltrated with propylene oxide:Spurr resin, and embedded by standard techniques for sectioning. Ultrathin sections were stained with 2.5% uranyl acetate and 2.5% lead acetate to provide positive contrast. Sections were imaged on a Phillips CM-12 transmission electron microscope at 80 keV.

Intraperitoneal glucose tolerance tests (IPGTTs) were performed as previously described (15). Briefly, 4- to 8-week-old mice fasted for 16 h were given an intraperitoneal injection of filter-sterilized glucose in PBS (2.0 mg dextrose/g body wt). Glucose concentration was measured in tail vein blood using the Freestyle glucose meter and test strips (Therasense) before injection (time 0) and 15, 30, 60, 90, and 120 min after injection.

Islets were hand-isolated from collagenase P-digested pancreata dissected from 4- to 6-week-old mice essentially as described previously (23). A parallel, four-column apparatus controlled by peristalsis pumps (1.0 ml/min), immersed in a circulating 37°C water bath, and plumbed into four fraction collectors (3.0 min/fraction) was used to perifuse isolated islets. After overnight incubation, 20–40 islets from one mouse were loaded per column and perifused with low (2.8 mmol/l) glucose in perifusion media (38.1 mmol/l sodium bicarbonate, 4.0 mmol/l l-glutamine, 1.0 mmol/l sodium pyruvate, 0.5% phenol red, 5.0 mmol/l HEPES, and 0.1% BSA in 1.0 l Dulbecco’s modified Eagle’s medium without glucose, pH 7.4) for a 30-min equilibration period (baseline). Secretagogues included high glucose (16.8 mmol/l), isobutylmethylxanthine (IBMX) (100 μmol/l in high glucose), arginine (20 mmol/l in high glucose), tolbutamide (300 μmol/l in low glucose), and potassium chloride (20 mmol/l in low glucose).

Pancreata from perinatal and adult mice were dissected, weighed, and homogenized in acid alcohol for extraction of insulin (23). Insulin content from perifusate fractions or from acid alcohol-extracted pancreas was measured by solid-phase radioimmunoassay (¹²⁵I-labeled insulin; Diagnostic Products) for mouse anti-insulin (MP Biomedical). Average insulin concentration was calculated as a function of total pancreatic wet weight.

Pancreata were dissected at P1 and placed into RNAlater (Ambion). Total pancreatic RNA was extracted using RNAqueous Phenol-free RNA Isolation kit (Ambion) according to the manufacturer’s instructions. Spectrophotometric and fluorometric methods were combined to determine RNA integrity and concentration. cDNA synthesis was performed on individual wild-type (n = 3) and pdx1^PBHnf6 (n = 4) RNA extracts using TaqMan reverse transcription reagents (Applied Biosystems) according to manufacturer’s instructions.

Pancreatic mRNA expression was analyzed by quantitative RT-PCR. Reactions were carried out in duplicate using iQ SYBR Green Supermix (Bio-Rad) according to the manufacturer’s instructions at an annealing temperature of 64°C. Data were collected using an iCycler iQ Real-Time PCR Detection System and software (Bio-Rad). Primers optimized by melt curve analysis were generated as follows: ins-1 (forward, ccctgcttgccctctgg; reverse, tgctgtttgacaaaagcctgg), ins-2 (forward, acccatgtcccgccgt; reverse, acccagctccagttgtgcc), and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (forward, agctcactggcatggcttccg; reverse, gcctgcttcaccaccttcttgatg). All values for individual pancreatic transcripts were normalized to total pancreatic mRNA as measured by GAPDH expression.

We previously created pdx1^PBHnf6 transgenic mice expressing Hnf6 under control of a pancreatic endocrine enhancer from the pdx1 gene (15). These mice express Hnf6 primarily in β-cells during postnatal stages, when expression of the endogenous gene is downregulated (6,7,15). pdx1^PBHnf6 transgenic mice showed hyperglycemia at weaning and severe glucose intolerance after a glucose challenge (15). Although insulin protein was detectable by immunohistology, transgenic mice did not display the typical biphasic insulin response during IPGTTs. Islets contained increased numbers of glucagon⁺ cells and showed a mixed islet morphology (Fig. 1A and D); however, plasma glucagon concentration was not statistically different from wild type (15). pdx1^PBHnf6 islets also had a severe impairment in expression of GLUT2, a target of Hnf6 transcriptional regulation (Fig. 1E) (16). We hypothesized that downregulation of GLUT2 decreased import of glucose into β-cells, resulting in the inability of pdx1^PBHnf6 transgenic mice to secrete insulin in response to increased glucose.

To test this hypothesis, we generated mice in which expression of rat GLUT2 is driven by the rat insulin promoter (RIPrGLUT2) (21) in an attempt to rescue GLUT2 expression in pdx1^PBHnf6 mice. We obtained five RIPrGLUT2 transgenic lines, one of which showed rGLUT2 expression in 60–80% of β-cells at P1 (data not shown). Adult bigenic (pdx1^PBHnf6;RIPrGLUT2) mice showed levels of rGLUT2 similar to those detected in monogenic RIPrGLUT2 pancreata (Fig. 1C and F). To determine whether bigenic mice were able to efficiently remove glucose from the bloodstream, we compared their glucose tolerance profiles with those of wild-type and pdx1^PBHnf6 mice during IPGTTs (Fig. 2). Mice were tested at 4–6 weeks (data not shown) or 8–10 weeks of age with identical results. No difference in glucose tolerance was observed between wild-type and RIPrGLUT2 transgenic mice. Although bigenic mice tended to have an improved fasting blood glucose level, this difference was not statistically significant. No differences in glucose tolerance were detected between pdx1^PBHnf6 and bigenic mice. Thus, restoration of GLUT2 to β-cells was not sufficient to rescue the diabetic phenotype of pdx1^PBHnf6 mice.

One explanation for decreased GSIS from transgenic islets is that transgenic islets simply produce less total insulin than their wild-type littermates. We extracted total insulin protein from wild-type and pdx1^PBHnf6 pancreata at perinatal and adult stages (Fig. 3) and found that transgenic mice had ∼35% less total pancreatic insulin at both ages. Quantitative RT-PCR analysis of P1 pdx1^PBHnf6 pancreata also showed a significant decrease in insulin mRNA. Insulin-1 transcripts were 26-fold lower in transgenic samples than in wild type, and insulin-2 was undetectable in transgenic pancreata (P < 0.05). The decrease in total pancreatic insulin in pdx1^PBHnf6 pancreata correlates with a 40% decrease in overall adult body mass (15) and a small but significant decrease in β-cell area at P1 (L.C., E.T., M.G., unpublished data). The 35% decrease in pancreatic insulin content does not fully explain, however, why pdx1^PBHnf6 mice are unable to secrete sufficient insulin to maintain glucose homeostatsis in response to glucose challenge.

To obtain mechanistic insight into how GSIS is altered in pdx1^PBHnf6 transgenic β-cells, we used a perifusion system to test the ability of transgenic islets to secrete insulin in response to various secretagogues (Fig. 4). In concordance with our in vivo IPGTT results, pdx1^PBHnf6 islets released ∼85% less insulin than wild-type controls in response to high glucose (Fig. 4A). Bolstering the glucose response with the phosphodiesterase inhibitor IBMX, which increases the cytoplasmic concentration of cAMP (24), only slightly improved insulin secretion in transgenic islets (to ∼70% less than wild-type islets; Fig. 4A). Tolbutamide, a sulfonylurea that promotes insulin secretion by inhibiting the action of the SUR-1 subunit of the K⁺_ATP channel (25), likewise failed to stimulate significant insulin secretion (Fig. 4A). In wild-type mice, addition of either arginine (Fig. 4A) or KCl (Fig. 4B), both general membrane depolarizers (23), produced an increase in insulin concentration in the perifusate comparable with that produced by high glucose. Transgenic islets were unable to significantly increase insulin secretion in response to either secretagogue. These results show that, consistent with in vivo measurements of GSIS (15), isolated pdx1^PBHnf6 islets have a severe inability to secrete insulin in response to a variety of stimuli.

Because islet perifusion studies suggested that there was a defect in a distal component(s) of the insulin secretory pathway in transgenic islets, we performed an ultrastructural analysis of endocrine cells in wild-type and transgenic mice. Endocrine cells were easily distinguished from surrounding acinar cells by the presence of characteristic cytoplasmic vesicles, each ∼0.25–0.45 μm in diameter and filled with an electron-dense granule. Somatostatin-secreting δ-cells, identified by their compact granulated vesicles (Fig. 5A and B), appeared morphologically normal in both wild-type and transgenic islets. In striking contrast, β-cells were difficult to identify in pdx1^PBHnf6 endocrine tissue, mostly attributable to a reduced number of secretory granules (compare Fig. 5A with B and D) and an increase in protein-filled lumenal structures (Fig. 5B), identified as Golgi cisternae in higher magnification images (data not shown). Nuclei in pdx1^PBHnf6 endocrine cells appeared healthy, with no hallmarks of apoptosis or necrosis.

Analysis of the cellular ultrastructure of both wild-type and pdx1^PBHnf6 pancreata revealed differences in intra-islet capillary morphology. Cross sections through wild-type capillaries showed small endothelial cell protrusions that were occasionally filled with electron-clear vesicles resembling caveolae (Fig. 5C, arrowheads). Conversely, pdx1^PBHnf6 capillaries showed increased endothelial surface area and a dramatic increase in caveolae (Fig. 5D, arrowheads). Fenestrations in transgenic capillaries also appeared to be decreased, consistent with the accumulation of caveolae. These changes in endothelial cell morphology may compound the insulin secretion defect in transgenic mice (26) but do not appear to be the main cause for alterations in glucose homeostasis, because transgenic islets show a dramatic decrease in insulin secretion even in a perifusion system.

The damaging effects of prolonged hyperglycemia on β-cell function have been well described (rev. in 2). Because pdx1^PBHnf6 transgenic mice are already hyperglycemic during adult stages, we analyzed β-cells from control and transgenic pancreata (Fig. 5E and F) at P3. At this stage, pdx1^PBHnf6 animals are still normoglycemic (data not shown) but already show decreased insulin mRNA and protein (Fig. 3). Two classes of dense core granules were easily identified in wild-type β-cells (27): immature, in which granules appear lighter and less dense, and mature, in which granules are dark and compact (Fig. 5E). Previous studies have suggested that granules mature by a process in which the acidic environment of the granule promotes proper folding and crystallization of insulin (27,28). Perinatal transgenic β-cells displayed mature granules but appear to have fewer immature granules (4 vs. 8 per 3.0- × 3.0-μm² field in wild-type β-cells; Fig. 5F). Golgi apparati in pdx1^PBHnf6 β-cells are also more distended compared with wild-type β-cells, reminiscent of the adult phenotype. Lysosomal structures were frequently identified in transgenic β-cells, suggesting increased scavenging of excess intracellular protein. Taken together with our adult studies, these results are consistent with a defect in pdx1^PBHnf6 β-cells in intracellular insulin dense core granule formation and/or trafficking, leading to the accumulation of protein primarily in the Golgi cisternae.

β-Cell–specific expression of insulin involves the action of several transcription factors, including Pdx1 and RIPE3b1/MafA (17,18,29,30). pdx1 is a direct transcriptional target of Hnf6 (31). Analysis of Pdx1 immunoreactivity in pdx1^PBHnf6 endocrine cells revealed no detectable changes in adult islets (15), early embryonic epithelium (e13.5; data not shown), or late gestation–stage endocrine clusters (e18.5; data not shown).

To date, MafA is the only transcription factor identified that is expressed exclusively in mature β-cells (19). MafA has recently been found to be important for proper GSIS (20). MafA-null mutant mice bear a striking similarity to pdx1^PBHnf6 transgenic mice with hyperglycemia at weaning, impaired GSIS in response to multiple secretagogues, including KCl (in vivo and in isolated islets), and altered islet composition and morphology.

Based on the crucial role of MafA in β-cell function, we hypothesized that misregulation of MafA in pdx1^PBHnf6 transgenic mice may contribute to defects in insulin production and/or secretion. In adult wild-type islets (Fig. 6C), MafA protein was restricted to the nuclei of β-cells; however, MafA protein was nearly absent from transgenic islets (Fig. 6F). Because MafA expression has been previously reported to be influenced by hyperglycemia (32), we analyzed MafA expression in pdx1^PBHnf6 islets before weaning, when transgenic pups are still under the influence of maternally produced insulin and are not hyperglycemic. At e18.5, MafA can be detected in insulin-producing cells in wild-type pancreata (Fig. 6A) but is undetectable in transgenic pancreata (Fig. 6D). Similarly, whereas wild-type islets express MafA in most β-cells at P1 (Fig. 6B), pdx1^PBHnf6 islets do not express levels of MafA detectable by immunohistology (Fig. 6E). Thus, overexpression of Hnf6 in transgenic mice is accompanied by the downregulation of MafA before the onset of hyperglycemia.

Hnf6 is initially expressed during embryogenesis in a broad endodermal domain consisting mainly of the hepatic and pancreatic epithelial anlagen, but is downregulated specifically in islet endocrine tissue at late gestation such that adult mouse islets no longer express Hnf6 (6,7). We previously showed that failure to downregulate Hnf6 resulted in altered islet morphology and islet composition and impaired GSIS, leading to diabetes (15). We now report that pdx1^PBHnf6 transgenic β-cells have impaired GSIS, most likely attributable to defects in intracellular trafficking and packaging of insulin granules. Transgenic β-cells also display a dramatic decrease in expression of MafA, consistent with an immature β-cell phenotype. Thus, our data suggest that maintaining Hnf6 expression in postnatal endocrine cells prevents terminal differentiation of insulin-producing cells into mature, glucose-responsive β-cells.

Our initial hypothesis, that the loss of GLUT2 in pdx1^PBHnf6 islets was the major contributing factor for the loss of GSIS, seems unlikely in light of the results of the RIPrGLUT2 genetic rescue experiments (Fig. 1). Although GLUT2 is necessary for uptake of glucose into the β-cell, its K_m is greater than that of glucokinase; GLUT2 is estimated to transport glucose about three times faster than glucokinase phosphorylates glucose (33,34). It is thus much more likely that multiple inputs into the insulin secretory pathway, including changes in islet-enriched transcription factors and other proteins necessary for insulin biosynthesis, are responsible for the nearly complete loss of GSIS in pdx1^PBHnf6 transgenic mice. This is consistent with our perifusion results: each of the secretagogues tested failed to elicit an increase in insulin secretion (Fig. 4). Because these secretagogues have been shown to potentiate insulin secretion by different cellular mechanisms, it is possible that the loss of GLUT2 compounds defects in more distal components of the insulin secretory pathway.

Ultrastructural analysis of pdx1^PBHnf6 β-cells, even before the onset of hyperglycemia (Fig. 5), revealed defects consistent with improper intracellular trafficking of insulin vesicles. Specifically, perinatal transgenic β-cells appeared to have fewer immature granules and distended Golgi cisternae (Fig. 5F). Accumulation of insulin protein within intracellular compartments may grow worse over time, because adult β-cells have a nearly complete lack of insulin granules and have greatly distended endoplasmic reticulum and Golgi cisternae. Although global defects in posttranslational packaging and membrane localization of proteins, including insulin and GLUT2, may contribute to loss of GSIS, our data showing appropriate membrane localization of rGLUT2 in pdx1^PBHnf6;RIPrGLUT2 transgenic mice argue against defective GLUT2 trafficking as being a major contributor to the diabetic phenotype of these mice (Fig. 1). Furthermore, it is unlikely that the alterations seen in β-cell ultrastructure in these animals are solely attributed to prolonged hyperglycemia, despite the fact that pdx1^PBHnf6 transgenic animals do become overtly hyperglycemic shortly after weaning (15) (E.T., M.G., unpublished observations). Decreased insulin secretion (35) and changes in gene expression, including loss of β-cell–specific transcription factors (36), occur in models of chronic hyperglycemia and are collectively referred to the “glucose toxicity” (rev. in 2). In pdx1^PBHnf6 islets, loss of MafA occurs before the onset of hyperglycemia (Fig. 6), as does the loss of GLUT2 (15), and the ultrastructural changes in β-cells. Moreover, expression of the β-cell–enriched factors Pdx1 (this study) and Nkx6.1 (15) appears normal in transgenic islets. Therefore, the phenotypic alterations in GSIS that we have described are more likely to be caused by the cumulative effects of loss of GLUT2 and MafA and the improper packaging of granules necessary for regulated insulin secretion.

Recent studies have indicated a role for endoplasmic reticulum (ER) stress as a causative agent in the progression of type 2 diabetes (rev. in 37). For example, mice null for PKR-like ER kinase (perk^−/−), an eIF2a kinase important in repressing protein synthesis upon accumulation of unfolded proteins in the ER, exhibited defects in insulin secretion and hyperglycemia (38,39). Similarly, conditional inactivation of the Wolfram gene (wfs1) specifically in β-cells resulted in decreased GSIS, increased expression of the ER stress gene BiP (hspa5/GRP78), and ER distension (40). In contrast, our analysis of pdx1^PBHnf6 transgenic β-cells was unable to discern any of the hallmarks of ER stress. Electron microscopy revealed abnormalities in the Golgi compartment of perinatal β-cells but did not appear to have the significantly distended ER compartments documented in either perk^−/− (38) or wfs1^−/− (40) β-cells. Microarray analysis of e18.5 and P1 pdx1^PBHnf6 pancreata also failed to detect significant changes in the ER stress targets BiP, CHOP, XBP-1, or ATF-6 (L.C., E.T., M.G, unpublished data).

Previous research has shown that not only are certain transcription factors necessary for the establishment of endocrine cell identity, but the timing and sequence of expression of these factors is also critical. For example, transient factors such as ngn3 must be downregulated to produce the proper number and distribution of functional endocrine cell types (41,42). Our data likewise show that sustained Hnf6 expression in β-cells is incompatible with terminal differentiation and acquisition of a mature fate. The cumulative data on Hnf6 suggest that it acts during embryonic stages to initiate the expression of factors required for endocrine differentiation but then must be downregulated to achieve a mature β-cell fate and establish GSIS. Although hnf6 mRNA expression has been shown to be downregulated in islets late in pancreatogenesis (6,7), a full characterization of the timing of Hnf6 downregulation in late embryonic endocrine cell types and the molecular mechanisms that regulate Hnf6 expression in vivo is still needed. Further studies will also be necessary to further refine the downstream targets and mechanism of action of Hnf6 in both the development and function of pancreatic endocrine cells.

This study was supported by National Institutes of Health (NIH) Training Grant in Developmental Biology (5-T32-HD05702 (to E.T.); Juvenile Diabetes Research Foundation (JDRF) Postdoctoral Fellowship 3-2003-57P (to I.A.); NIH Grant DK42502 (to R.S.); Merit Review Award from the VA Research Service and NIH grants DK68764, DK63439, and DK62641 (to A.C.P.); NIH Grant DK065131 and JDRF Career Development Award 2-2002-583 (to M.G.).

We would like to thank members of the Gannon Lab for helpful discussions and critical review of this manuscript, as well as Venus Childress and David Lowe for excellent assistance with animal husbandry and genotyping. Additional thanks to Dr. Zhongyi Chen, Dr. Jay Jerome, and Dorinda Vardell for their expert technical assistance. This research was made possible through use of the following Vanderbilt shared resources: Transgenic Mouse and ES Cell Core, Mouse Metabolic Phenotyping Center, Islet Isolation Core, and Shared Imaging Resource (NIH grants CA68485, DK20593, DK58404, DK59637, HD15052, and EY08126).