IPF1/PDX1 Deficiency and ␤-Cell Dysfunction in Psammomys obesus, an Animal With Gil Leibowitz,1 Sarah Ferber,2 Åsa Apelqvist,3 Helena Edlund,3 David J. Gross,1 Erol Cerasi,1 Danielle Melloul,1 and Nurit Kaiser1

The homeodomain IPF1/PDX1 is stimulatory effect of glucose on required in ␤-cells for efficient expression of insulin, (4–6). Studies in animal models of type 2 diabetes suggest glucose transporter 2, and prohormone convertases 1/3 that the activities of IPF1/PDX1 are reduced in hyper- and 2. Psammomys obesus, a model of diet-responsive glycemia or hyperlipidemia, thus impairing insulin gene type 2 diabetes, shows markedly depleted insulin stores expression and insulin production (7–9). In humans, non- when given a high-energy (HE) diet. Despite hyper- sense mutations in the Ipf1/Pdx1 gene were identified in glycemia, insulin mRNA levels initially remained un- changed and then decreased gradually to 15% of the patients with maturity-onset diabetes of the young type 4 basal level by 3 weeks. Moreover, insulin gene expres- (10), and missense mutations of the Ipf1/Pdx1 gene were sion was not increased when isolated P. obesus islets observed in several patients with sporadic type 2 diabetes were exposed to elevated glucose concentrations. Con- (11). In addition, mutations in the Ipf1/Pdx1 gene in sistent with these observations, no functional Ipf1/Pdx1 siblings of diabetic patients are associated with decreased gene product was detected in islets of newborn or adult glucose-stimulated insulin secretion and predisposition for P. obesus using immunostaining, Western blot, DNA diabetes (12). Taken together, these findings provide evi- binding, and reverse transcriptase–polymerase chain re- action analyses. Other ␤-cell transcription factors (e.g., dence for an essential role of IPF1/PDX1 in maintaining ISL-1, Nkx2.2, and Nkx6.1) were expressed in P. obesus normal ␤-cell function and glucose homeostasis in both islets, and the DNA binding activity of the insulin tran- mice and humans. In addition, IPF1/PDX1 is required for scription factors RIPE3b1-Act and IEF1 was intact. the formation of the in mice and humans; ho- Ipf1/Pdx1 gene transfer to isolated P. obesus islets nor- mozygous null mutations in the Ipf1/Pdx1 gene results in malized the defect in glucose-stimulated insulin gene pancreas agenesis (13,14). Thus, IPF1/PDX1 is important expression and prevented the rapid depletion of insulin ␤ content after exposure to high glucose. Taken together, both for pancreas development and for -cell differentia- these results suggest that the inability of P. obesus is- tion and function. lets to adapt to dietary overload, with depletion of in- Although the relative contributions of insulin resistance sulin content as a consequence, results from IPF1/PDX1 and ␤-cell dysfunction to the development of diabetes are deficiency. However, because not all animals become being debated, considerable evidence supports a dominat- hyperglycemic on HE diet, additional factors may be ing role for deficient ␤-cell function at all stages of the important for the development of diabetes in this ani- disease (15,16). The gerbil Psammomys obesus shows in- mal model. Diabetes 50:1799–1806, 2001 sulin resistance and develops diet-induced obesity-linked diabetes, initially associated with hyperinsulinemia, and gradually progressing to hypoinsulinemia and severe hy- lucose is the main physiological regulator of perglycemia (17). Because of the obvious similarities, it the ␤-cell and controls insulin production by serves as a convenient model for human type 2 diabetes. both stimulating insulin gene transcription and To explore whether impaired IPF1/PDX1 expression and/ Gstabilizing its mRNA (rev. in [1]). Genetic anal- or function is the underlying cause for ␤-cell dysfunction yses in mice have revealed that IPF1/PDX1 is required for in P. obesus exposed to high-energy (HE) diet, we have the regulation of insulin gene expression and for ensur- monitored the expression of Ipf1/Pdx1 and the functions ing normal expression of key components in the glucose of its gene product in the islets of P. obesus. A functional sensing and insulin processing machinery required for Ipf1/Pdx1 gene product could not be detected in pancre- maintenance of normoglycemia (2,3). Indeed, IPF1/PDX1 atic islets of adult or newborn P. obesus, suggesting that is one of the main transcription factors mediating the the absence of IPF1/PDX1 activity is a primary genetic event responsible for the failure of ␤-cells to cope with an From the 1Department of Endocrinology and Metabolism, Hebrew University- increased secretory demand. Consistent with this idea, Hadassah Medical Center, Jerusalem; the 2Institute of Endocrinology, Sheba Medical Center, Tel Hashomer, Israel; and the 3Department of Microbiology, forced expression of Ipf1/Pdx1 in P. obesus islets corrects University of Umeå, Umeå, Sweden. the defect in glucose-regulated insulin gene expression Address correspondence and reprint requests to Gil Leibowitz, MD, Depart- and prevents the rapid depletion of insulin content after ment of Endocrinology and Metabolism, Hadassah University Hospital, P.O. Box 12000, 91120 Jerusalem, Israel. E-mail: [email protected]. exposure to high glucose. Together, these results further Received for publication 23 February 2001 and accepted in revised form emphasize the central role played by a functional IPF1/ 11 May 2001. ␤ HE, high-energy; IRI, immunoreactive insulin; LE, low-energy; MOI, multi- PDX1 gene product to ensure proper -cell function and plicity of infection; PCR, polymerase chain reaction; RT, reverse transcriptase. normoglycemia.

DIABETES, VOL. 50, AUGUST 2001 1799 PDX1 DEFICIENCY AND ␤-CELL DYSFUNCTION

TABLE 1 Characteristics of P. obesus during development of HE diet–induced diabetes HE diet (day) 04 7 1421 Glucose (mmol/l) 5.8 Ϯ 0.2 18.1 Ϯ 1.1* 16.4 Ϯ 1.3* 18.7 Ϯ 1.9* 23.8 Ϯ 1.6* Triglycerides (mmol/l) 0.48 Ϯ 0.04 0.83 Ϯ 0.06* 0.72 Ϯ 0.09‡ 1.03 Ϯ 0.13* 11.89 Ϯ 4.03‡ Serum IRI (pmol/l) 924 Ϯ 172 5,395 Ϯ 893* 4,686 Ϯ 688* 4,693 Ϯ 909* 4,244 Ϯ 765* Pancreatic IRI (nmol/g tissue) 56.0 Ϯ 13.8 10.0 Ϯ 6.1† 8.5 Ϯ 2.6* 4.1 Ϯ 1.5‡ 11.7 Ϯ 5.3‡ Results are means Ϯ SE. *P Ͻ 0.001; †P Ͻ 0.05; ‡P Ͻ 0.01 relative to day 0 control.

RESEARCH DESIGN AND METHODS with EcoR⌱, separated on a 1% agarose gel, transferred to a nylon membrane, ␣ 32 Animals. The P. obesus of the Hebrew University colony and SD rats were and UV cross-linked. The blots were probed with ( - P) dCTP-labeled both obtained from Harlan (Jerusalem, Israel). After weaning at 3 weeks, P. full-length rat Ipf1/Pdx1 cDNA at 65°C in Rapid-hyb buffer solution (Amer- obesus were maintained on low-energy (LE) diet (2.38 kcal/g; Koffolk, Petach- sham, U.K.). The blots were then washed with 2ϫ sodium chloride–sodium Tikva, Israel) and SD rats were given standard laboratory diet until studied. citrate containing 0.1% SDS. To induce diabetes (random blood glucose Ͼ8.3 mmol/l), P. obesus were Reverse transcriptase–polymerase chain reaction analysis. Total islet switched from LE to HE diet (2.93 kcal/g, Weizmann Institute for Science, RNA was extracted using RNAzol B (Tel-Test, Friendswood, TX). For poly- Rehovot, Israel), on which they were maintained until studied. Animals were merase chain reaction (PCR) analysis, total RNA was reverse-transcribed anesthetized with ketamine hydrochloride (Ketalar, Park-Davis), and exsan- using AMV reverse transcriptase (RT) (Promega, Madison, WI). The resulting guinated by cardiac puncture. Blood glucose concentrations were determined cDNAs were amplified by PCR using oligonucleotides complementary to by Accutrend Sensor (Roche Diagnostics, Brussels, Belgium). The collected sequences in the insulin gene: 5Ј-TTGTCAACCAGCACCTTTGT-3Ј and 5Ј- serum was frozen at Ϫ20°C for further analysis of the immunoreactive insulin GTTGCAGTAGTTCTCCAGCT-3Ј for P. obesus and 5Ј-CCTGCCCAGGCTTTT- (IRI) and triglyceride concentrations (GPO-Trinder kit; Sigma, St. Louis, MO). GTCA-3Ј and 5Ј-GGTGCAGCACTGATCCACAATG-3Ј for rat insulin I. Primers Islet IRI content was determined by RIA after extraction with 0.18N HCl in 70% were designed to cross an intron and amplified fragments of 257-bp and 208-bp ethanol for 24 h at 4°C. Islets were isolated by collagenase digestion of the coding sequences from P. obesus and rat insulin genes. 18S rRNA (Quantum- pancreas and handpicked under a binocular as described (18). The islets were RNA kit; Ambion, Austin, TX) was used as an internal control. Polymerization harvested for and RNA determinations or cultured either in suspen- reaction was performed in a 25-␮l reaction volume containing 2.5 ␮l cDNA (25 sion or as monolayer patches in RPMI-1640 medium containing different ng RNA equivalents), 80 ␮mol/l cold dNTPs (dATP, dCTP, dGTP, dTTP), 2.5 glucose concentrations, as previously described (19). ␮Ci (␣-32P)dCTP, 100 pmol/l of appropriate oligonucleotide primers, and 1.5 Immunohistochemistry. from newborn (24–48 h postbirth) and units of Taq polymerase (MBI Fermentas, Amherst, NY). PCR amplification adult P. obesus and rats were fixed in 4% paraformaldehyde and either conditions were as follows: 5 min at 94°C followed by 14 cycles (30 s each) of embedded in paraffin when studied at the Hebrew University-Hadassah 94°C, 60°C, and 72°C. The amplimers were separated on a 6% polyacrylamide Medical Center or kept at 4°C in 30% sucrose (in 0.1 mol/l phosphate buffer, pH gel in Tris borate EDTA buffer, the gel was dried, and the incorporated ␮ 7.4) until sent to the University of Umeå. Sections (5 m) processed at the (␣-32P)dCTP was measured by PhosphorImager. The number of cycles and the Hebrew University–Hadassah Medical Center were deparaffinized, rehydrated, final reaction conditions were adjusted to the exponential range phase of the and heated in 10 mmol/l citrate buffer (pH 6.0) at 92°C for 10 min in a amplification curve for each product. We verified that the amount of each PCR microwave oven for antigen retrieval. After cooling at room temperature, the product in a reaction increased linearly with the amount of starting cDNA. For sections were washed three times with phosphate-buffered saline, and endog- quantitation of insulin mRNA in islets of prediabetic and diabetic P. obesus, enous peroxidase was blocked by a 15-min incubation with 3% H O at room 2 2 the ratio of insulin to 18S rRNA band intensity was determined for each temperature. For IPF1/PDX1 immunohistochemistry, sections were incubated reaction. The homeodomain region of Ipf1/Pdx1 was amplified by PCR from for 16 h at 4°C with polyclonal antibodies raised against the COOH-terminal P. obesus, rat, and human islets as well as ␤-cell lines derived from different domain of the human IPF1/PDX1 or the NH -terminal domain of the Xenopus 2 species using oligonucleotides complementary to conserved sequences in the homolog (XlHbox8) of IPF1/PDX1 (generously provided by Dr. C.B. Wright, Ipf1/Pdx1 gene: HD-1: 5Ј-ACCAAAGCTCACGCGTGGAAA-3Ј and 5Ј-TGATGT- Nashville, TN). Detection was with a streptavidin-biotin-peroxidase complex GTCTCTCGGTCAAGTT-3Ј; and HD-2: 5Ј-GAGCTGGAGAAGGAATTCTTATT- developed with aminoethylcarbazol (Zymed, San Francisco, CA). Subse- Ј Ј Ј quently, sections were treated with avidin/biotin–blocking solutions (Avidin/ 3 and 5 -TT(C/T)TGGAACCAGAT(C/T)TTGAT-3 . Biotin blocking kit; Zymed) followed by a 60-min incubation at 37°C with HD-1 primers cross an intron and amplify a 196-bp fragment from the guinea pig anti-porcine insulin antibody diluted 1:100 (Sigma, St Louis, MO) cDNA, whereas HD-2 primers do not cross an intron and they amplify a 94-bp ␤ and detection with streptavidin-biotin-alkaline phosphatase complex devel- fragment. -Actin and gene 3 (Hox3) were used as internal controls oped with 5-bromo-4-chloro-indolyl phosphate/nitro blue tetrazolium liquid for cDNA and genomic DNA PCR amplification. Amplification conditions were substrate (Sigma). Pancreases at the University of Umeå were frozen in liquid as follows: 5 min at 94°C followed by 35 cycles (1 min each) of 94°C, 60°C, and nitrogen and then sectioned and processed, as previously described (3). The 72°C. primary antibodies used were rabbit anti-mouse IPF1/PDX1 (3), rabbit anti– Viral vectors and islet transduction. Recombinant adenoviruses were ISL-1 (3), rabbit anti-Nkx6.1 (3), rabbit anti-Nkx2.2 (3), rabbit anti–GLUT-2 (3), constructed by cloning the gene of interest into a pACCMV.pLpA plasmid, rabbit anti-glucagon (Linco, St. Charles, MO), and guinea pig anti-insulin followed by cotransfection of 293 cells with the E1-deleted adenovirus (Linco). The secondary antibodies used were fluorescein anti–guinea pig plasmid PJM17 (21). Stocks of the viruses were prepared by transduction of (Jackson ImmunoResearch, West Grove, PA) and Cy3 anti-rabbit (Jackson). 293 cells (cultured in 145/20 plates) with recombinant adenoviruses at a At the University of Umeå, specimens were analyzed using an epifluorescence multiplicity of infection (MOI) of 10 for 90 min. Cells and media were collected microscope (Zeiss Axioplan), and at the Hebrew University-Hadassah Medical 48 h later, then cells were pelleted by centrifugation at 800g and the virus in Center, specimens were analyzed using a Nikon Eclipse E600 microscope. the supernatant fraction was precipitated overnight in 20% polyethylene glycol Western blot and gel electrophoretic mobility shift assays. Nuclear and (PEG 8000; Sigma) and 2.5 mol/l NaCl at 4°C. The virus-containing medium whole-cell extracts of P. obesus islets, rat islets, and ␤-cell lines (INS-1 and was centrifuged at 10,000g, and the pellet was resuspended in physiological HIT) were prepared and analyzed for IPF1/PDX1 using standard Western blot saline. The titer of the virus was determined using the plaque assay in 293 analysis with several antibodies against IPF1/PDX1. DNA-binding reactions cells. for IPF1/PDX1 and IEF1 detection were performed as previously described Freshly isolated P. obesus islets kept in suspension were transduced with (4,20). Double-stranded oligonucleotides spanning the human insulin se- adenoviral vectors expressing mouse Ipf1/Pdx1 (AdCMV-Pdx1)or␤-galacto- quence from Ϫ206 to Ϫ227, rat II insulin from Ϫ101 to Ϫ126, and rat I insulin sidase (AdCMV-␤-gal) at different MOIs. The efficiency of transduction was from Ϫ108 to Ϫ116 were end-labeled by a fill-in reaction and used as probes determined by ␤-galactosidase and insulin staining of AdCMV-␤-gal–infected for the detection of IPF1/PDX1, RIPE3b1-Act, and IEF1 binding complexes, islets plated on extracellular matrix–coated dishes 72 h postinfection and respectively. Radioactive bands were quantified on a PhosphorImager (Fujix assayed 5–7 days later upon formation of islet monolayer patches (19). BAS 1000; Fuji Film, Tokyo, Japan). Expression of Ipf1/Pdx1 in transduced islets was assessed by Western blot Genomic DNA Southern blot. Genomic DNA was prepared from P. obesus analysis. The response of the insulin gene to glucose stimulation was studied liver and various cell lines derived from different species and was digested in suspension cultures of P. obesus islets exposed to normal (3.3 mmol/l) or

1800 DIABETES, VOL. 50, AUGUST 2001 G. LEIBOWITZ AND ASSOCIATES

content (5.95 Ϯ 0.78 vs. 1.78 Ϯ 0.51 pmol/islet in 3.3 and 22.2 mmol/l glucose, respectively; P Ͻ 0.005 by two-tailed Student’s t test), whereas in SD rats, there was only a 35% decrease (18.6 Ϯ 2.1 vs. 12.0 Ϯ 0.7 pmol/islet in 3.3 and 22.2 mmol/l glucose, respectively; P ϭ 0.058). These re- sults indicate that the production of insulin in response to glucose is perturbed. Defective glucose-regulated insulin gene expression in P. obesus. To examine whether reduced insulin pro- duction resulted from defects in insulin gene expression, we assessed by relative quantitative RT-PCR the level of islet insulin mRNA during the development of diabetes in vivo and after exposure of isolated islets to a high glucose concentration in vitro. In vivo, insulin mRNA levels re- mained unchanged during the first 7 days of HE diet, despite hyperglycemia (Fig. 1). Insulin mRNA then gradu- ally decreased, and by day 21 it reached 15% of basal values. Failure to increase insulin gene expression during hyperglycemia could result from unresponsiveness of the

FIG. 1. Insulin mRNA content in islets of P. obesus during development of diabetes. A: RNA was extracted from P. obesus islets isolated from animals kept on HE diet for up to 21 days. Islet RNA was reverse- transcribed, and the cDNA was amplified by PCR (14 cycles of the following, at 30 s each step: denaturation at 94°C, annealing at 60°C, and extension at 72°C) using primers complementary to the 5؅ and 3؅ ends of the P. obesus insulin gene. 18S rRNA was used as an internal control. Each column represents the mean of 3–5 different batches of P. obesus islets. B: Insulin–to–18S rRNA ratio over time. *P < 0.05 and **P < 0.005 relative to day 0 on HE diet. high (22.2 mmol/l) glucose concentrations for 48 h. Transduced islets were analyzed for insulin content and mRNA levels by relative quantitative PCR. Data presentation and statistical analysis. Data are expressed as the means Ϯ SE. Statistical differences between groups were determined using the two-tailed unpaired Student’s t test. A P value of Ͻ0.05 was considered significant.

RESULTS Insulin deficiency in diabetic P. obesus. Diabetes- prone P. obesus show a rapid increase in blood glucose levels after transition from LE to HE diet, from a mean of FIG. 2. Insulin mRNA in P. obesus islets exposed to high glucose in 5.8 mmol/l at outset to 18.1 and 23.8 mmol/l by days 4 and vitro. A: Relative quantitative PCR for insulin mRNA and 18S rRNA from rat islets cultured at 3.3, 5.5, and 22.2 mmol/l glucose and P. 21, respectively (Table 1). Although serum IRI was in- obesus islets cultured at 3.3 and 22.2 mmol/l glucose for 21 h. Total creased on day 4 of the HE diet and remained high RNA was extracted from ϳ500 islets of each group and reverse- transcribed. Equal amounts of cDNA were subjected to PCR using throughout the study, pancreatic IRI content rapidly de- primers complementary to the 5؅ and 3؅ ends of the P. obesus and rat clined to 18% of control levels on day 4 of the HE diet, I insulin genes. Fragments (257 bp and 208 bp) of the insulin gene remaining low thereafter (Table 1). In vitro culture of P. coding sequence were amplified from P. obesus and rat islets, respec- tively. B: Quantification of the results at 22.2 mmol/l glucose relative to obesus islets in the presence of a high glucose concentra- those at 3.3 mmol/l glucose. Results are the means ؎ SE of 3 and 7 islet tion for 21 h resulted in 70% depletion of islet insulin preparations from rat and P. obesus, respectively.

DIABETES, VOL. 50, AUGUST 2001 1801 PDX1 DEFICIENCY AND ␤-CELL DYSFUNCTION

FIG. 3. Immunostaining of postnatal P. obesus and rat islets for different ␤-cell markers. Photomicrographs of pancreatic sections from newborn P. obesus double-immunostained with antibodies specific to glucagon, ISL-1, GLUT2, Nkx6.1, and Nkx2.2 and to IPF1/PDX1 (red) and insulin (green). Also shown are sections from SD rat immunostained for IPF1/PDX1 and insulin. Note the lack of staining for IPF1/PDX1 in P. obesus islets. Islet morphology is normal, with positive staining for islet markers. insulin promoter to glucose stimulation or from deleteri- tected by Western blot analysis of nuclear and whole–islet ous effects of the chronic diabetic milieu. To distinguish cell extracts obtained from diabetic or normoglycemic between these alternatives, we studied the insulin gene P. obesus, whereas the expected 45-kDa protein was response after short-term exposure to high glucose in detected in rat islets and various ␤-cell lines (Fig. 4A). In islets of P. obesus and rat. Despite normal basal insulin accordance with these results, no binding to the highly gene expression, no increase in insulin mRNA was ob- conserved IPF1/PDX1 recognition site was observed in served in prediabetic P. obesus islets incubated for 21 h islet nuclear extracts prepared from diabetic and nondia- with 22.2 mmol/l glucose. In contrast, there was a 3.9-fold betic P. obesus (Fig. 4B). In contrast, RIPE3b1-Act and increase in insulin mRNA in SD rat islets cultured at 22.2 IEF1 binding activities were detected in P. obesus islet Ͻ vs. 3.3 mmol/l glucose (P 0.05) (Fig. 2). Thus, insulin extracts, and the binding activity of RIPE3b1-Act was 3–4 P obesus ␤ gene expression in . -cells does not respond to times higher in extracts from diabetic versus nondiabetic glucose stimulation. P. obesus islets (Fig. 4B and C). Taken together, these of normoglycemic and hyperglyce- findings demonstrate that although P. obesus islets show mic P. obesus lack IPF1/PDX1. IPF1/PDX1, RIPE3b1- normal expression of several factors associated with ma- Act, and IEF1 are transcription factors reported to be ␤ implicated in glucose-stimulated insulin gene expression ture -cell function, they completely lack expression of a (4,22,23). To elucidate whether impairment in these tran- functional IPF1/PDX1 protein. scription factors could explain the perturbed insulin ex- To test whether the absence of IPF1/PDX1 in P. obesus ␤ pression in P. obesus, we studied the expression and/or -cells was caused by defects in Ipf1/Pdx1 gene expres- DNA binding activity of these . Immunohisto- sion, RNA was isolated from different batches of P. obesus islets, and RT-PCR was performed using primers corre- chemical analyses using different anti–NH2-terminal and anti–COOH-terminal polyclonal IPF1/PDX1 antibodies sponding to parts of the Ipf1/Pdx1 gene that are highly failed to detect any expression of IPF1/PDX1 in P. obesus conserved between different species, including mouse, rat, islets, whereas rat and mouse ␤-cells exhibited the normal and humans. The homeodomain of Ipf1/Pdx1 was success- positive nuclear staining for IPF1/PDX1 (Fig. 3 and other fully amplified from cDNA derived from several different data not shown). Nevertheless, P. obesus ␤-cells stained species, but not from P. obesus islet cDNA (Fig. 5A). positive for insulin, somatostatin, glucagon, GLUT2, and Furthermore, no PCR product could be obtained from the transcription factors ISL-1, Nkx2.2, and Nkx6.1 (Fig. 3 genomic DNA, even when internal homeodomain primers and other data not shown). Moreover, P. obesus islets that do not span the intron were used (Fig. 5A). Together, exhibited glucose-responsive insulin secretion, albeit with these results imply that Ipf1/Pdx1 gene is absent in P. altered glucose sensitivity (24), and glucokinase activity obesus, that a large deletion encompassing the homeodo- was also present in P. obesus islets, with HE diet further main has occurred, or that the Ipf1/Pdx1 gene is not con- increasing the glucose phosphorylation activity of the served in P. obesus. Southern blot analysis of genomic P. islets (24). obesus DNA using a full-length rat Ipf1/Pdx1 cDNA probe As further confirmation, IPF1/PDX1 could not be de- resulted in the detection of a single ϳ6-kb fragment (Fig.

1802 DIABETES, VOL. 50, AUGUST 2001 G. LEIBOWITZ AND ASSOCIATES

FIG. 4. IPF1/PDX1 expression in P. obesus islets. A: Immunoblotting of P. obesus and SD rat islets and of rat and hamster ␤-cell lines for IPF1/PDX1. Homogenates were prepared and extracted in buffer containing NaCl and Triton X-100, resolved by 12% SDS-PAGE, and immunoblotted using antibodies directed against the COOH-terminal (left panel) or the NH2-terminal (right panel) domain of human IPF1/PDX1. The arrow indicates the migration position of the 45-kDa form of IPF1/PDX1. Each well was loaded with 40 ␮g cell extract protein. B: Mobility shift assay for IPF1/PDX1, RIPE3b1-Act (RIPE), and IEF1 binding in nuclear extracts of P. obesus islets. Arrows indicate the migration position of the DNA-protein complexes. Octa binding was used as an internal control to confirm equal protein loading. Note that nuclear extracts of nondiabetic (ND) and hyperglycemic (DM) P. obesus do not bind the A3 element of the insulin promoter/enhancer. C: Quantitation of RIPE3b1-Act and IEF1 binding. The densitometric readings of the bands in four batches each of islets from normoglycemic (ND) and hyperglycemic P. obesus after 4 days of HE diet (DM) show that RIPE3b1-Act binding is increased in hyperglycemic versus normoglycemic P. obesus, whereas IEF1 binding is unchanged. *P < 0.05 relative to islets from ND animals.

DIABETES, VOL. 50, AUGUST 2001 1803 PDX1 DEFICIENCY AND ␤-CELL DYSFUNCTION

FIG. 5. Ipf1/Pdx1 gene expression in P. obesus. A: RT-PCR for the homeodomain of Ipf1/Pdx1. RNA was extracted from P. obesus rat and human islets and from ␤-cell lines of rat (INS-1), mouse (␤-TC), and hamster (HIT) origin. Islet and ␤-cell line cDNA and P. obesus genomic DNA were subjected to PCR, with 35 cycles of the following (1 min each step): denaturation at 94°C, annealing at 60°C, and extension at 72°C. This was performed with two sets of primers complementary to the 5؅ and 3؅ ends of the Ipf1/Pdx1 homeodomain (HD-1 and HD-2). HD-1 primers cross an intron and amplify a 196-bp fragment from the cDNA, whereas HD-2 do not cross an intron, and they amplify a 94-bp fragment. ␤-actin and Hox3 were used as internal controls for cDNA and genomic DNA PCR amplification, respectively. Note that the amplified Ipf1/Pdx1 fragment was present in islet and ␤-cell cDNA of all species except for P. obesus. B: Southern blot analysis for Ipf1/Pdx1. Genomic DNA was prepared from P. obesus liver (Psam) and from mouse and rat cell lines, digested with EcoR⌱, and hybridized with a full-length rat Ipf1/Pdx1 cDNA probe.

5B). These results may suggest the existence of a highly histochemical analyses using various antibodies raised divergent Ipf1/Pdx1–like gene in P. obesus. against different domains of the protein, nor could it be Ipf1/Pdx1 gene transfer prevents insulin depletion in detected by DNA binding assay. The latter suggests that if P. obesus islets. To ascertain whether the Ipf1/Pdx1 an IPF1/PDX1 variant not recognized by the different deficiency underlies the inability of glucose to regulate antibodies is present in these islets, its function differs insulin gene expression in P. obesus, islets derived from from that of the conserved IPF1/PDX1 found in other prediabetic P. obesus were transduced with an adenoviral species. Because the cDNA sequence of the P. obesus vector expressing Ipf1/Pdx1 or ␤-galactosidase; results IPF1/PDX1 variant has not yet been elucidated, complete from the latter transduction indicated that 50–60% of the assessment of the functional differences between a poten- islet ␤-cells were infected (Fig. 6A). The expression of tial IPF1/PDX1 variant of P. obesus and that of other Ipf1/Pdx1 in P. obesus islets was confirmed by Western blot analysis (Fig. 6B). The insulin content of ␤-galactosi- dase–infected islets was similar to that of uninfected control islets (1.72 Ϯ 0.22 vs. 1.68 Ϯ 0.22 pmol per islet); after 48 h in high glucose, there was a 40–50% decrease in islet insulin content in ␤-galactosidase–infected islets (Fig. 7A). Ipf1/Pdx1 expression in P. obesus islets increased the insulin content of islets cultured at 3.3 mmol/l glucose (2.94 Ϯ 0.26 pmol per islet, P Ͻ 0.01); most importantly, exposure to 22.2 mmol/l glucose caused no reduction in islet insulin content (Fig. 7A). The effect of Ipf1/Pdx1 gene transfer on insulin content was dose-dependent, being most prominent at an MOI of 5 ϫ 106 CFU per islet (data not shown). No increase in insulin mRNA was observed in ␤-galactosidase–infected islets cultured for 48 h at 22.2 mmol/l glucose, whereas the culture of Ipf1/Pdx1–infected islets in 22.2 mmol/l glucose resulted in a nearly threefold increase in islet insulin mRNA compared with islets in 3.3 mmol/l glucose (Fig. 7B and C). These findings provide evidence that the impaired insulin gene expression ob- served in P. obesus ␤-cells reflects the lack of a functional Ipf1/Pdx1 gene.

DISCUSSION

P. obesus seems to be a unique animal in which the FIG. 6. Transduction efficiency of P. obesus islets. Islets were infected Ipf1/Pdx1 gene is not expressed in postnatal pancreatic with adenoviral vectors expressing Ipf1/Pdx1 or ␤-galactosidase (␤- :gal) at MOI 5 ؋ 106 CFU per islet and harvested 48 h postinfection. A islets. Ipf1/Pdx1 is highly conserved in evolution, and the The efficiency of transduction was determined by ␤-galactosidase and homeodomain amino acid sequence of zebrafish IPF1/ insulin staining of AdCMV-␤-gal–infected islets that were plated on PDX1 is 95% identical to that of the mammalian protein extracellular matrix–coated dishes 72 h postinfection and assayed 5–7 days later upon formation of islet monolayer patches. B: Western blot (25). The conserved form of IPF1/PDX1 could not be analysis on whole-cell extracts from isolated P. obesus islets harvested detected in P. obesus islets by Western blot and immuno- 48 h postinfection.

1804 DIABETES, VOL. 50, AUGUST 2001 G. LEIBOWITZ AND ASSOCIATES

species is not possible at this time. The presence of a morphologically normal endocrine and exocrine pancreas and the ␤-cell expression of GLUT2 and Nkx6.1 in P. obe- sus suggest that such a putative candidate gene retains both the developmental functions of the conventional Ipf1/Pdx1 gene and some of its actions in mature ␤-cells. In contrast, it fails to mimic the function of IPF1/PDX1 as a regulator of glucose-dependent insulin gene transcrip- tion. Preliminary studies have shown reduced glucose- regulated proinsulin biosynthesis in P. obesus islets (data not shown) that could be caused by the lack of glucose- regulated insulin gene expression. The failure to detect Ipf1/Pdx1 by PCR on genomic DNA using oligonucleotides corresponding to highly conserved regions of the homeodomain is supportive of the presence of a highly divergent Ipf1/Pdx1 gene with partially re- dundant activities. Furthermore, preliminary analyses of embryos from timed pregnancies failed to detect any conserved IPF1/PDX1 expression in E13.5 P. obesus em- bryos (equivalent to mouse E10.5–11 embryos), although expression of glucagon, ISL-1, and Nkx2.2 was observed (data not shown). Detailed analyses of Ipf1/Pdx1 expres- sion during fetal development is hampered by the low yield of pregnancies in these animals and the subsequent considerable difficulty in retrieving timed embryos. The demonstration that Ipf1/Pdx1 gene transfer to P. obesus islets conferred glucose-responsiveness to the insulin gene confirms the assumption that this factor is essential for the regulation of the insulin gene by glucose (5,6). Consequently, expression of Ipf1/Pdx1 in P. obesus islets increased the basal insulin level of the islets and prevented its rapid depletion on exposure to high glucose, a key pathogenic mechanism in this animal model of diabetes. Interestingly, adult P. obesus on an LE diet are normoglycemic, despite the lack of IPF1/PDX1, and their islets contain normal amounts of insulin. The fact that P. obesus is well adjusted to its natural desert habitat added to the observation that its islet insulin content is normal on the LE diet, indicating that other ␤-cell transcription factors are sufficient to sustain the basal expression of the insulin gene. Thus, IPF1/PDX1 does not seem to be essen- tial for maintaining normoglycemia, provided the animals are fed a low-calorie diet (a situation that puts minimal load on the insulin production capacity). Not all animals develop hyperglycemia on a HE diet, and the prevalence of diabetes varies in different lines of P. obesus, with Ͼ90% in the diabetes-prone line as opposed to 30–40% in the partially diabetes-resistant line (26). In preliminary stud- ies, IPF1/PDX1 was not detected in the partially diabetes- resistant line of P. obesus, indicating that additional factors may be involved in the development of hyperglyce- mia in P. obesus. In addition to new insights into the role of transcrip- tional dysfunction in the development of type 2 diabetes, our findings may have implications for treatment of the disease. In several models of type 2 diabetes, hyperglyce- mia is associated with a loss of IPF1/PDX1 expression as

FIG. 7. Effect of Ipf1/Pdx1 gene transfer to P. obesus islets on insulin mRNA and protein. A: Insulin content in islets infected with adenoviral exposed to 3.3 or 22.2 mmol/l glucose for 48 h. C: Quantitation of the vectors expressing Ipf1/Pdx1 (PDX-1) and ␤-gal at MOI 5 ؋ 106 CFU relative change in insulin mRNA. Results are the means ؎ SE of three per islet and in uninfected islets after 48-h exposure to 3.3 and 22.2 different islet preparations, normalized to ␤-galactosidase–infected -mmol/l glucose. Results are the mean ؎ SE of a representative exper- (␤-gal) islets at 3.3 mmol/l glucose. G-3.3 and G-22.2 denote incuba iment run in quadruplicate. B: Insulin mRNA relative to 18S rRNA in tions at 3.3 and 22.2 mmol/l glucose, respectively. *P < 0.05 and **P < islets infected by AdCMV-Pdx1 (PDX-1) and AdCMV-␤-gal (␤-gal) and 0.01 relative to similarly treated islets at 3.3 mmol/l glucose.

DIABETES, VOL. 50, AUGUST 2001 1805 PDX1 DEFICIENCY AND ␤-CELL DYSFUNCTION well as a loss of its downstream targets, including insulin Weir GC: Reduced insulin, GLUT2, and IDX-1 in ␤-cells after partial and GLUT2 (7–9). Recently, it has been suggested that pancreatectomy. Diabetes 46:258–264, 1997 10. Stoffers DA, Ferrer J, Clarke WL, Habener JF: Early-onset type-II diabetes oxidative stress may play a role in glucose toxicity; indeed, mellitus (MODY4) linked to IPF1. Nat Genet 17:138–139, 1997 antioxidant treatment can prevent decreases in both the 11. Macfarlane WM, Frayling TM, Ellard S, Evans JC, Allen LIS, Bulman MP, binding activity of IPF1/PDX1 and insulin gene expression Ayres S, Shepherd M, Clark P, Millward A, Demaine A, Wilkin T, Docherty (27,28). We therefore postulate that restoring or augment- K, Hattersley AT: Missense mutations in the insulin promoter factor-1 gene ing the expression of Ipf1/Pdx1 in the islets of patients predispose to type 2 diabetes. J Clin Invest 104:R33–R39, 1999 with diabetes might circumvent the deleterious effects of 12. Hani EH, Stoffers DA, Chevre J-C, Durand E, Stanojevic V, Dina C, Habener JF, Froguel P: Defective mutations in the insulin promoter factor-1 (IPF-1) hyperglycemia, hyperlipidemia, and oxidative stress and gene in late-onset type 2 diabetes mellitus. J Clin Invest 104:R41–R48, 1999 prevent the depletion of pancreatic insulin stores. Further- 13. Habener JF, Stoffers DA: A newly discovered role of transcription factors more, most patients undergoing islet transplantation re- involved in pancreas development and the pathogenesis of diabetes main hyperglycemic and require insulin therapy. We have mellitus. Proc Assoc Am Physicians 110:12–21, 1998 shown that exposure of human islets to high glucose 14. Stoffers DA, Zinkin NT, Stanojevic V, Clarke WL, Habener JF: Pancreatic agenesis attributable to a single nucleotide deletion in the human IPF1 decreases Ipf1/Pdx1 and insulin gene expression (20,29). gene coding sequence. Nat Genet 15:106–110, 1997 It has yet to be determined whether augmented expression 15. Cerasi E: Insulin deficiency and insulin resistance in the pathogenesis of of IPF1/PDX1 in transplanted islets improves graft func- NIDDM: is a divorce possible? Diabetologia 38:992–997, 1995 tion in the hyperglycemic environment. 16. Kahn BB: Type 2 diabetes: when insulin secretion fails to compensate for insulin resistance. Cell 92:593–596, 1998 17. Gadot M, Leibowitz G, Shafrir E, Cerasi E, Gross D, Kaiser N: Hyperpro- ACKNOWLEDGMENTS insulinemia and insulin deficiency in the diabetic Psammomys obesus. This work was supported in part by grants from the Yael Endocrinology 135:610–616, 1994 Fund (to G.L.), the Juvenile Diabetes Foundation Interna- 18. Kaiser N, Corcos AP, Sarel I, Cerasi E: Monolayer culture of adult rat tional (I-1998-9 to N.K.), the Israel Science Foundation pancreatic islets on extracellular matrix: modulation of B-cell function by chronic exposure to high glucose. Endocrinology 129:2067–2076, 1991 (249/97 to D.M. and 251/97 to N.K.), and the European 19. Gross DJ, Leibowitz G, Cerasi E, Kaiser N: Increased susceptibility of islets Community (BMH4-CT98-3448 to E.C.). from diabetes-prone Psammomys obesus to the deleterious effects of The authors are grateful to Yaffa Ariav, Ludmilla Eilon, chronic glucose exposure. Endocrinology 137:5610–5615, 1996 and Rachel Oron for dedicated technical assistance and to 20. Marshak S, Leibowitz G, Bertuzzi F, Socci C, Kaiser N, Gross DJ, Cerasi E, Sonya Marshak for advice and help. Melloul D: Impaired ␤-cell functions induced by chronic exposure of cultured human pancreatic islets to high glucose. Diabetes 48:1230–1236, 1999 REFERENCES 21. Seijffers R, Ben David O, Cohen Y, Karasik A, Berezin M, Newgard CB, 1. Melloul D, Cerasi E: Transcriptional regulation of proinsulin synthesis by Ferber S: Increase in PDX-1 levels suppresses insulin gene expression in glucose. In Insulin Secretion and Pancreatic B-Cell Research. Flatt PR, RIN-1046-38 cells. Endocrinology 140:3311–3317, 1999 Lenzen S, Eds. London, Smith-Gordon, 1994, p. 7–13 22. Odagiri H, Wang J, German MS: Function of the human insulin promoter in 2. Ahlgren U, Jonsson J, Jonsson L, Simu K, Edlund H: Beta-cell-specific primary cultured islet cells. J Biol Chem 271:1909–1915, 1996 inactivation of the mouse Ipf1/Pdx1 gene results in loss of the beta-cell 23. Sharma A, Fusco-DeMane D, Henderson E, Efrat S, Stein R: The role of the phenotype and maturity onset diabetes. Genes Dev 15:1763–1768, 1998 insulin control element and RIPE3b1 activators in glucose-stimulated 3. Hart AW, Baeza N, Apelquist Å, Edlund H: Attenuation of FGF-signalling in transcription of the insulin gene. Mol Endocrinol 9:1468–1476, 1995 mouse ␤-cells leads to diabetes. Nature 408:864–868, 2000 24. Nesher R, Gross DJ, Donath MY, Cerasi E, Kaiser N: Interaction between 4. Melloul D, Ben-Neriah Y, Cerasi E: Glucose modulates the binding of an genetic and dietary factors determines ␤-cell function in Psammomys islet-specific factor to a conserved sequence within the rat I and the human obesus, an animal model of type 2 diabetes. Diabetes 48:731–737, 1999 insulin promoters. Proc Natl Acad SciUSA90:3865–3869, 1993 25. Milewski WM, Duguay SJ, Chan SJ, Steiner DF: Conservation of PDX-1 5. Marshak S, Totary H, Cerasi E, Melloul D: Purification of the beta-cell structure, function and expression in zebrafish. Endocrinology 139:1440– glucose-sensitive factor that transactivates the insulin gene differentially in 1449, 1998 normal and transformed islet cells. Proc Natl Acad SciUSA93:15057– 26. Kalman R, Ziv E, Adler JH, Lazarovici G, Bar-On H: The efficiency of the 15062, 1996 sand rat metabolism is responsible for the development of obesity and 6. Macfarlane WM, Shepherd RM, Cosgrove KE, James RF, Dunne MJ, diabetes. J Basic Clin Physiol Pharmacol 4:57–68, 1993 Docherty K: Glucose modulation of insulin mRNA levels is dependent on 27. Kaneto H, Kajimoto Y, Miyagawa J-i, Matsuoka T-a, Fujitani Y, Umayahara transcription factor PDX-1 and occurs independently of changes in intra- Y, Hanafusa T, Matsuzawa Y, Yamasaki Y, Hori M: Beneficial effects of cellular Ca2ϩ. Diabetes 49:418–423, 2000 antioxidants in diabetes: possible protection of pancreatic ␤-cells against 7. Gremlich S, Bonny C, Waeber G, Thorens B: Fatty acids decrease IDX-1 glucose toxicity. Diabetes 48:2398–2406, 1999 expression in rat pancreatic islets and reduce GLUT2, glucokinase, insulin, 28. Tanaka Y, Gleason CE, Tran PO, Harmon JS, Robertson RP: Prevention of and somatostatin levels. J Biol Chem 272:30261–30269, 1997 glucose toxicity in HIT-T15 cells and Zucker diabetic fatty rats by 8. Harmon JS, Gleason CE, Tanaka Y, Oseid EA, Hunter-Berger KK, Robert- antioxidants. Proc Natl Acad SciUSA96:10857–10862, 1999 son RP: In vivo prevention of hyperglycemia also prevents glucotoxic 29. Melloul D, Kaiser N, Cerasi E: Chronic exposure to high glucose impairs effects on PDX-1 and insulin gene expression. Diabetes 48:1995–2000, 1999 multiple ␤-cell functions in cultured human pancreatic islets. Curr Opin- 9. Zangen DH, Bonner-Weir S, Lee CH, Latimer JB, Miller CP, Habener JF, ion Endocrinol Diabetes 7:44–48, 2000

1806 DIABETES, VOL. 50, AUGUST 2001