Loss of HNF-1α Function in Mice Leads to Abnormal Expression of Genes Involved in Pancreatic Islet Development and Metabolism

The construction of expression vector pCMV-pHNF-4α was described previously (15). The vector pcDNA3-pHNF-1α was generated by cloning a 2.8-kb NotI/BamHI fragment from pBJ5-pHNF-1α (a gift from G. Crabtree, Stanford University, Stanford, CA) containing the entire rat HNF-1α coding sequence into pcDNA3. The pcDNA3-pSHP-1 expression vector was constructed by polymerase chain reaction (PCR) amplification of an 850-bp mouse Shp fragment that contained the entire open reading frame using mouse liver cDNA as a template. This fragment was cloned into the EcoRI site of pcDNA3.

The reporter plasmid pGL2-pSHP-1 was constructed by PCR amplification of the human SHP-1 promoter (Genbank accession no. AF044316), using oligonucleotides SHP-Pm-1 (5′-TCCTAGACTGGACAGTGGGCA-3′) and SHP-Pm-2 (5′-CTTCCAGCTCTCTG GCTCTGT-3′) and subsequent cloning of the 582-bp fragment into plasmid pPCR2.1. A KpnI/XhoI fragment from this plasmid was then subcloned into the pGL2, inserting the SHP promoter upstream of the luciferase reporter gene. Vector pGL2-pSHPmut was generated by PCR using oligonucleotides SHPmut-F (5′-TCCTAGACTGGACAGTAA AAAAAGCCT-3′) and SHP-Pm-2, as well as pGL2-pSHP as a template. This fragment, containing a mutated HNF-4 binding site, was cloned into pPCR2.1 before insertion into pGL2. The sequences of all PCR-generated clones were confirmed by dideoxynucleotide sequencing.

Pancreatic islets were isolated from 4- to 6-week-old Hnf-1α^−/− mice and their heterozygous (Hnf-1α^+/−) and wild-type (Hnf-1α^+/+) littermates. We used collagenase digestion and differential centrifugation through Ficoll gradients, with a modification of procedures previously described (32). Islets were cultured overnight in RPMI 1640 medium supplemented with 10% fetal calf serum and 11.1 mmol/l glucose. Whole pancreata were removed from newborn mice. Total RNA was then extracted using TRIzol reagent (Gibco-BRL) and following the manufacturer’s instructions. Contaminating genomic DNA was removed using 1 μl RNase free DNase-I (Boehringer) per 5 μg RNA.

Conditions used for reverse transcriptase (RT)-PCR followed the method of Wilson and Melton (33), with minimal modifications. cDNAs were synthesized using MMLV-RT (Stratagene) with dNTPs and random hexamer primers (Stratagene). These cDNAs provided templates for PCRs using specific primers at annealing temperatures of 60–65°C in the presence of dNTPs and Taq polymerase. Typically, between 20 and 28 cycles were used for amplification in the linear range. The sequences of the primers that were used for the amplification of cDNAs are available upon request from M.S.

Glut2 immunofluorescence was performed on 5-μm sections from 6-week-old Hnf-1α^+/+, Hnf-1α^+/−, and Hnf-1α^−/− mice that were prepared as previously described (27). Sections were blocked for 30 min with normal rabbit serum at a concentration of 1:100 in phosphate-buffered saline (PBS). Sections were stained with rabbit anti-mouse Glut2 antibody (Chemicon International, Temecula, CA) in PBS at a dilution of 1:200. The secondary antibody, a TRITC-labeled guinea pig anti-rabbit IgG, was applied at a concentration of 1:100 in PBS. Sections were washed twice in PBS between each step and after the application of the secondary antibody and were mounted in PBS:glycerol, 1:9.

Pyruvate kinase enzyme activity was measured in extracts of islets isolated from four mice. After overnight culture in RPMI 1640 containing 11.6 mmol/l glucose, islets were hand picked, washed twice in PBS, and transferred to a homogenization buffer consisting of 200 mmol/l mannitol, 70 mmol/l sucrose, 5 mmol/l potassium HEPES buffer (pH 7.5), and 1 mmol/l dithiothreitol. The islets were washed twice in buffer containing 1 mmol/l EGTA and then twice in buffer without EGTA. The islets were then sonicated (Ultrasonics Sonicator Cell Disrupter, Model W 185 F; Heat Systems, Plainview, NY) and centrifuged at 105,000 × g for 60 min at 4°C. The protein content of the islet extracts was measured (Biorad, #500-0006), and the activity was measured in equivalent amounts of protein from the different mice. Pyruvate kinase enzyme activity was measured in a reaction buffer consisting of 80 mmol/l triethanolamine chloride buffer (pH 7.5), 5 mmol/l MgSO₄, 100 mmol/l KCl, 2.5 mmol/l ADP, 0.14 mmol/l NADH, 1 mmol/l dithiothreitol, and 7 mmol/l phosphoenolpyruvate and 2.5 units of lactic dehydrogenase (all reagents were purchased from Sigma, St. Louis, MO). Each reaction was carried out in a final volume of 1 ml and was started by the addition of the islet extract. NADH levels were measured as fluorescence emitted at 448 nm in response to excitation at 340 nm for 10 min (31). The rates of consumption of NADH were similar in Hnf-1α^+/+ (n = 4), Hnf-1α^+/− (n = 5), and Hnf-1α^−/− (n = 4) mice.

In vitro translated proteins or nuclear cell extracts were incubated with ³²P-labeled ds-stranded oligonucleotide probes containing the wild-type or mutant HNF-4 binding sites in the SHP promoter (sequences: 5′-TGGACAGTGGGCAAAGTCC-TCCC-3′ and 5′-TGGACAGTAAAAAAAGTCCT CCC-3′, respectively). The 15-μl reaction mixture contained 20 mmol/l Hepes buffer (pH 7.9), 40 mmol/l KCl, 1 mmol/l MgCl₂, 0.1 mmol/l EGTA, 0.5 mmol/l DTT, 4% Ficoll, and 2 μg of poly(dIdC) at 25°C for 20 min. Supershift analysis was performed by incubating the antibody with the protein or nuclear extracts for 5 min on ice before adding the probe. The reaction mixture was loaded on a 6% nondenaturing polyacrylamide gel containing 0.25× TBE buffer (0.023 mol/l Tris-borate and 0.5 mmol/l EDTA) and run at 4°C. Nuclear extracts were prepared as described (15).

Cos7 and HepG2 cells were grown on minimum essential medium supplemented with 10% fetal calf serum and Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 15% fetal calf serum, respectively. HIT-T15 cells were grown in DMEM supplemented with 5% fetal calf serum and 10% horse serum. A modified calcium phosphate precipitation procedure was used for transient transfections (15). In short, 0.5 ml of the precipitate, containing 1 μg CMV-LacZ, 0.5 μg luciferase reporter construct, and the indicated amount of expression vectors and carrier DNA (up to 10 μg), were added per 60-mm dish. Luciferase was normalized for transfection efficiency by the corresponding β-galactosidase activity (34).

Results are given as mean ± SE. Statistical analyses were performed by using a Student’s t test, and the null hypothesis was rejected at the 0.05 level.

Insulin secretory responses to glucose are markedly reduced in Hnf-1α^−/− mice compared with littermate controls. Methylpyruvate, an insulin secretagogue that freely permeates the mitochondrial membrane, can suppress ATP-dependent K⁺-channel activity in Hnf-1α^−/− pancreatic islets, suggesting that the defect in glucose metabolism is upstream of mitochondrial glucose metabolism. To test which genes involved in glucose metabolism were affected by the disruption of Hnf-1α, we measured the mRNA levels of glucose transporters, amino acid transporters, and all glycolytic enzymes. We also measured the proximal rate-limiting enzymes of oxidative phosphorylation, including components of the pyruvate dehydrogenase complex (pyruvate decarboxylase and dihydrolipoyl dehydrogenase), aconitase, isocitrate dehydrogenase, and α-ketoglutarate dehydrogenase complex (2-oxoglutarate dehydrogenase, dihydrolipoyl succinyl-transferase, and dihydrolipoyl dehydrogenase) as well as Ucp2. Furthermore, we determined the mRNA levels of the sulfonylurea receptor-1 (Sur-1), ATP-dependent K⁺-channel (Kir6.2), prohormone convertase-2 and -3 (Pc2, Pc3), pterin-4α-carbinolamine dehydratase (Pcd, DCoh), insulin, glucagon, somatostatin, pancreatic polypeptide (YY), and pancreatic peptide (PP).

Hypoxanthine phosphoribosyltransferase expression levels were determined to show that each sample contained equivalent amounts of mRNA. No products were amplified in the absence of reverse transcriptase, indicating the absence of genomic DNA in the mRNA samples. The cDNA of each sample was derived from one animal, and five to six animals of each genotype (Hnf-1α^+/+, Hnf-1α^+/−, and Hnf-1α^−/−) were studied. RNAs for genes encoding L-Pk, Gck, Glut2, neutral and neutral and basic amino acid transporter (Nbat), insulin (Ins-1), glucagon, and Dcoh were significantly reduced (Fig. 1A). The absence of Glut2 protein in mice that lack Hnf-1α was also confirmed by immunohistochemistry of pancreatic sections (Fig. 2). The expression of other glucose (Glut1) and amino acid (neutral amino acid transporter) transporters remained unchanged (data not shown). The expression of the remaining glycolytic enzymes, including phosphoglucose isomerase, phosphofructokinase, aldolase B, triosephosphate isomerase, glucosephosphate dehydrogenase, phosphoglycerate kinase, phosphoglycerate mutase, 2,3-bisphosphoglycerate mutase, and α-, β-, and γ-enolases were also similar in islets of Hnf-1α^+/+ and Hnf-1α^−/− animals (Fig. 1B, data not shown). Expression of mitochondrial genes involved in the TCA cycle, such as Ogdh, as well as Kir6.2, Sur-1, Pc2, and Pc3, were also unchanged (Fig. 1B, data not shown). The expression of different islet cell type markers, such as Iapp, somatostatin, and YY genes, did not differ, suggesting that the various islet cell types contributed equally in wild-type and Hnf-1α^−/− animals (data not shown). As expected, glucagon expression was slightly reduced in hyperglycemic Hnf-1α^−/− mice (Fig. 1A).

Because hyperglycemia and hypoinsulinemia in Hnf-1α^−/− animals can lead to secondary changes in gene expression (35), the expression of the above-mentioned genes were also studied in pancreata of newborn animals that had not been exposed to chronic hyperglycemia. We found that the expression of Glut2, L-Pk, Dcoh, and Ins-1 was already decreased shortly after birth (Fig. 1A), suggesting that Hnf-1α deficiency leads to the reduced expression of these genes. In contrast, glucagon and Gck mRNA levels were not altered in newborn pancreata (Fig. 1A), demonstrating that the decreased expression in Hnf-1α^−/− adult islets is likely due to hyperglycemia or hypoinsulinemia in these mice. Finally, we found that the expression of L-Pk was markedly reduced in islets of adult and newborn Hnf-1α^−/− animals. This transcriptional defect was specific for L-Pk, and expression levels of the muscle isoform (M₂Pk) were unchanged in Hnf-1α^−/− mice. To estimate the effect of the reduced L-Pk expression on overall pyruvate kinase activity, we measured total pyruvate kinase activity in islets of Hnf-1α^+/+ and Hnf-1α^−/− mice (data not shown). No differences were detected in total pyruvate kinase activity, supporting previous findings that M₂Pk is the major pyruvate kinase isoform in pancreatic islets (36,37).

The effect of Hnf-1α deficiency on the expression of islet-enriched transcription factors was determined in islets of 4-week-old Hnf-1α^−/−, Hnf-1α^+/−, and Hnf-1α^+/+ littermates by RT-PCR. Steady-state mRNA levels of Hnf-4α, an upstream regulator of Hnf-1α, were reduced in islets of Hnf-1α^−/− mice, suggesting that Hnf-1α is critical for the normal expression of Hnf-4α in pancreatic islets (Fig. 1A). We also found that expression levels of several other islet-enriched transcription factors, including Pdx-1, Neuro-D1, and Shp-1, were reduced ∼2- to 11-fold in islets of Hnf-1α^−/− mice (Fig. 1A). In contrast, the expression levels of Hnf-1β, -3α, -3β, -3γ, -4γ, -6, Nkx-2.2, Pax-4, Pax-6, Isl-1, Cdx2/3, and Lim-1 did not differ between Hnf-1α^−/− and Hnf-1α^+/+ mice (Fig. 1A, data not shown). Steady-state mRNA levels were also measured in pancreata of newborn mice. The expression levels of Pdx-1 and Neuro-D1 were already reduced in newborn mice (Fig. 1A). This indicates that Hnf-1α deficiency rather than nonspecific effects as a result of hyperglycemia is responsible for the abnormal expression of Pdx-1 and Neuro-D1.

To test whether the changes in gene expression profiles of pancreatic islets are also present in the liver of Hnf-1α^−/− animals, we measured mRNA concentrations of all the above-mentioned genes in Hnf-1α null, heterozygous, and wild-type mice. In contrast to pancreatic islets, the expression of Glut1, Glut2, and L-Pk was increased in Hnf-1α^−/− mice (Fig. 3). The expression of the remaining glycolytic and mitochondrial genes was similar in mice of different Hnf-1α genotypes (data not shown). Consistent with previous observations (21,22), transcript levels of Hnf-1β were increased, whereas steady-state mRNA of DCoH and Igf-1 were reduced in Hnf-1α null liver (Fig. 3). The expression of Hnf-3α, -3β, -3γ, and -4α were unchanged, whereas Hnf-6 was reduced eightfold in Hnf-1α^−/− mice compared with wild-type littermates (Fig. 3). We also tested 100 additional genes, including genes of apolipoprotein, amino acid, and lipid metabolism. We found that in the liver of Hnf-1α^−/− mice, mRNA levels of Nbat were reduced 2.2-fold compared with wild-type littermates (Fig. 3). In summary, these data demonstrate that differences exist in the gene expression profiles of islets versus liver in Hnf-1α^−/− mice. Therefore, in the liver, other mechanisms that can activate Hnf-1α target genes, such as Glut2 and L-Pk, and compensate for the loss of Hnf-1α function must be operative.

Shp-1 is an orphan nuclear receptor that heterodimerizes with various nuclear hormone receptors, including Hnf-4α (38,39). Our in vivo expression data indicate that Shp-1 expression was reduced in adult islets of Hnf-1α null mice but not in newborn islets (Fig. 1). We then analyzed the promoter regions of human and murine Shp-1 but found no Hnf-1 binding sites (40). In contrast, a putative HNF-4α consensus binding sequence was identified at position −551 to −570 bp in the human SHP-1 promoter. This binding site was also conserved at a similar location in the murine Shp-1 promoter (Genbank accession nos. for murine and human sequences are AF044315 and AF044316, respectively). We therefore hypothesized that the downregulation of Shp-1 may be due to the reduced Hnf-4α message in Hnf-1α^−/− mice.

To further study the transcriptional regulation of Shp-1, we cloned a 582 bp of the 5′ human SHP-1 promoter containing the putative HNF-4 binding site and linked it to a luciferase reporter gene (pGL2-pSHP-1). Cotransfection of this reporter plasmid with an HNF-4α expression vector (pcDNA3-pHNF-4α) showed an approximate two- to fivefold dose-dependent activation of the SHP-1 promoter in three different cell lines, Cos7, HepG2, and HIT-T15 (Figs. 4A–C). In contrast, HNF-1α expression did not result in the transcriptional activation of this reporter construct (Figs. 4A–C). To confirm that HNF-4α can directly activate the SHP-1 promoter, we mutated one hexamer half of the HNF-4 binding site in the reporter plasmid pGL2-pSHP-1. Transcriptional activation of HNF-4α on the resulting SHP-1 reporter plasmid (pGL2-pSHPmut) was abolished (Fig. 4D–F), indicating that the HNF-4 binding site in the SHP-1 promoter is functionally important for its expression.

Electrophoretic mobility shift assays were performed to investigate whether HNF-4α can bind to the putative HNF-4 binding site in the SHP-1 promoter. A major DNA/protein complex was detected with a ³²P-labeled oligonucleotide that contained the HNF-4 binding site (Fig. 5, lane 2). This binding activity could be competed by an unlabeled excess of cold HNF-4 binding oligonucleotide (Fig. 5, lane 3). Furthermore, a supershift of the complex was observed after preincubation of extract with a monospecific anti–HNF-4α antiserum (α-445) but not with anti–STAT-1 antibodies (Fig. 5, lanes 4 and 5), demonstrating that HNF-4α can bind to the HNF-4 site in the SHP-1 promoter. As expected, HNF-4α was not able to bind to the mutated HNF-4 site (Fig. 5, lane 7).

To test whether SHP-1 can inhibit its own transcriptional activation, we cotransfected SHP-1 expression vector pcDNA3-pSHP-1 with constant quantities of pCMV-pHNF-4α and pGL2-pSHP-1 into Cos7, HIT-T15, and HepG2 cells. Luciferase activity decreased in a dose-dependent manner when increasing amounts of pcDNA3-pSHP-1 vector were coexpressed with HNF-4α (Figs. 4G–I). These data suggest that SHP-1 can inhibit its own transcription through a negative transcriptional feedback loop that is mediated by HNF-4α (Fig. 6).

Previous studies demonstrated that the insulin secretory responses to glucose are impaired in Hnf-1α^−/− mice (27). Physiological studies have shown that K_ATP in β-cells of mutant mice are insensitive to suppression by glucose or glyceraldehyde but are normally sensitive to ATP. Flux of glucose through glycolysis in islets of mutant mice is reduced, but mitochondrial substrates, such as methylpyruvate, can inhibit K_ATP, elevate [Ca²⁺], and restore the insulin secretion defect (28). These results indicate that insulin secretion is impaired by a defect proximal to mitochondrial metabolism in the β-cell glycolytic signaling pathway. However, a recent in vitro study showed that expressing a dominant negative form of Hnf-1α, HNF1α-P291fsinsC, in insulinoma INS-1 cells affects mitochondrial function by reducing the expression of Ogdh and increasing the expression of Ucp2 (31). The net result of these expressing alterations is to impair mitochondrial function such as oxidation and ATP generation and to dissipate the mitochondrial membrane potential, leading to decreased glucose-stimulated insulin secretion (GSIS). These inconsistent findings have not been resolved. In addition, the molecular basis by which Hnf-1α deficiency causes impaired GSIS are poorly understood.

In this study, we investigated the transcriptional defect resulting from Hnf-1α deficiency. We studied steady-state mRNA levels in adult and newborn islets of genes that are critical for β-cell function. Gene transcript levels were also measured in the pancreas of newborn animals to control for secondary effects that hyperglycemia may have on gene expression profiles in HNF-1α^−/− mice. Newborn Hnf-1α^−/− mice are normoglycemic (87 vs. 90 mg/dl in HNF-1α^+/+ and HNF-1α^−/− newborn mice, respectively), and blood glucose levels of HNF-1α^+/− females are indistinguishable from HNF-1α^+/+ females during the last week of pregnancy (118 vs. 121 mg/dl in HNF-1α^+/+ and HNF-1α^+/− pregnant mice, respectively). This indicates that newborn islets of HNF-1α^−/− mice were not exposed to chronic hyperglycemia. We found reduced expression of three genes involved in the glycolytic signaling pathway, namely Glut2, Gck, and L-Pk in Hnf-1α^−/− mice (Fig. 1A). In addition, we found diminished expression of Nbat, an amino acid transporter that transports both neutral and basic amino acids into the cell (Fig. 1A). Reduced expression of Glut2, Gck, and Nbat are likely to contribute to the defect in GSIS because mutant mice that lack Glut2 or Gck are hyperglycemic and hypoinsulinemic as a result of impaired GSIS (41,42). The downregulation of Nbat may reduce the transport of basic and neutral amino acids into the β-cell. Furthermore, the reduced expression of Nbat offers a molecular explanation of the decreased cellular ATP production in response to leucine in cells that express a dominant negative form of the HNF-1α protein (31). In contrast, it is unlikely that decreased L-Pk expression levels contribute significantly to the β-cell defect in HNF-1α^−/− mice because we did not observe a significant reduction in the overall enzyme activity of islet pyruvate kinase in Hnf-1α^−/− mice. Therefore, our findings agree with previous studies that indicated that M₂Pk is the major pyruvate kinase isoform in pancreatic islets (36,37). Our data suggest, then, that the lack of Glut2, Gck, and Nbat, which was not compensated for by increased expression levels of other glucose and amino acid transporters, may contribute to the diminished glucose or amino acid–stimulated insulin secretion rates in HNF-1α^−/− animals.

To study whether Hnf-1α deficiency also impairs the expression of genes that are critical for mitochondrial function, we measured mRNA levels of numerous mitochondrial genes, including Ogdh and Ucp2. Ogdh constitutes the rate-limiting enzyme in the mitochondrial TCA cycle, and Ucp2 uncouples respiration from oxidative phosphorylation and inhibits the efficiency of ATP synthesis. In contrast to a previous report (31), the expression of Ogdh and Ucp2 were not changed in the absence of Hnf-1α (Fig. 1B). We also did not find the expression of other mitochondrial enzymes involved in the TCA cycle to be affected by Hnf-1α deficiency (Fig. 1B). Several reasons may account for these discrepancies. First, the previous study investigated the consequences of Hnf-1α deficiency by using a deoxycycline-inducible dominant negative Hnf-1α. Deoxycycline has been shown to be cytotoxic and thus may alter gene expression in treated cells (43,44). Second, because Hnf-1α can either homodimerize or heterodimerize with Hnf-1β, the dominant negative Hnf-1α used in the previous study may inhibit the function of both Hnf-1α and Hnf-1β (45). As a result, reduced Ogdh and increased Ucp2 expression may be due to impaired Hnf-1β function and not solely to Hnf-1α. Finally, the previous study was done in a transformed insulinoma cell line (INS-1), and there are differences between INS-1 cell line (46) and islets (used in the present study) that may explain the differences in Hnf-1α–regulated genes in the mitochondria.

In addition to the transcriptional defect in genes regulating β-cell glycolytic signaling, we found the expression levels of several other transcription factors to be significantly reduced in mice with Hnf-1α deficiency. The steady-state mRNAs of Hnf-4α, Pdx-1, and Neuro-D1/Beta-2, as well as the dimerization cofactor Dcoh, were reduced shortly after birth in islets of Hnf-1α^−/− mice, suggesting that these changes are caused by HNF-1α deficiency and are likely to contribute to the molecular defects in Hnf-1α^−/− islets. Decreased expression of several of these genes may be mediated directly by Hnf-1α. For instance, a functional HNF-1 binding site was identified recently in the HNF-4α promoter P2 that is located ∼46 kb upstream of the hepatic (P1) promoter. Transcription from the P2 promoter is responsible for the expression of the most abundant HNF-4α isoform in pancreatic β-cells, and HNF-1α is a potent activator of HNF-4α transcription from the P2 promoter in vitro (47). Our finding suggests that in pancreatic islets, Hnf-1α and Hnf-4α form an autoregulatory loop and that a haploinsufficiency in either of these transcription factors causes an expression disequilibrium (Fig. 6). This observation may also explain a possible molecular mechanism for the similar pancreatic β-cell defects in MODY1 and MODY3 patients (25,48,49). It is interesting that the expression of the insulin promoter factor Ipf-1/Pdx-1 was reduced in a dose-dependent manner in Hnf-1α^+/+, Hnf-1α^+/−, and Hnf-1α^−/− mice. Pdx-1 is an essential transcription factor for pancreas development and β-cell function, including insulin gene transcription (50,51). Expression levels of Pdx-1 in Hnf-1α^−/− islets were reduced 30–50% (Fig. 1A). This reduction was sufficient for a significant downregulation of Pdx-1 target genes, such as Gck and insulin (51,52,53,54). Functional Hnf-1α as well as Hnf-4α binding sites have been reported in evolutionarily conserved enhancer elements of the Pdx-1 promoter (11,12). Therefore, decreased Pdx-1 expression in Hnf-1α^−/− islets may also be the result of reduced HNF-1α and HNF-4α gene expression. We also found that the mRNA levels of Neuro-D1, a helix-loop-helix transcription factor that can activate insulin gene transcription in vitro (55), were reduced 80% in Hnf-1α^−/− mice. Despite the reduced expression of Pdx-1 and Neuro-D1, mRNA levels of Ins-2 were indistinguishable in islets of HNF-1α^+/+ and HNF-1α^−/− mice (Fig. 1A). This finding is in agreement with other reports (27,29) that did not find significantly reduced insulin mRNA levels in HNF-1α^−/− mice and may be explained by compensation through other transcriptional activators of the insulin gene, such as Nkx2.2, Cdx2/3, Isl-1, and Pax4. In contrast, we found that Ins-1 expression was reduced ∼50% in islets of in HNF-1α^−/− mice (Fig. 1A). This reduction is most likely a direct effect of Hnf-1α deficiency, mediated by an Hnf-1 binding element in the Ins-1 promoter that is lacking in the murine Ins-2 promoter (56).

It is worth noting that expression levels of other islet transcription factors that are important for β-cell metabolism, differentiation, and insulin gene transcription, including Hnf-1β, -3α, -3β, -3γ, -4γ, -6, Cdx-2, Nkx2.2, Isl-1, Pax4, and Pax6, were unchanged in Hnf-1α^−/− mice. These findings support our hypothesis that loss of Hnf-1α function causes a pleiotropic but distinct defect in the β-cell transcriptional network essential for its normal function. It is interesting that the transcriptional control of this network and the regulation of downstream target genes are distinct in pancreatic islets and in the liver. For instance, the expression of Glut2, Gck, and L-Pk was normal or increased in the liver but significantly decreased in islets of Hnf-1α^−/− mice compared with wild-type littermates. These differences may be explained by compensatory upregulation of transcriptional regulators (e.g., Hnf-1β) in the liver, reduced expression of liver-enriched transcription factors absent in adult pancreatic islets (e.g., Hnf-6), or differential regulation of transcription factors and downstream targets through tissue-specific promoters (e.g., Hnf-4α and Gck).

Finally, in vitro studies have suggested that SHP-1 may heterodimerize with HNF-4α and inhibit its transcriptional activity (39). Mutations in SHP-1 have also been identified in humans and may be associated with mild obesity in Japanese study subjects (57). Shp-1 expression was markedly reduced in islets of adult Hnf-1α^−/− mice, suggesting that Shp-1 may be regulated by Hnf-1α, Hnf-4α, Pdx-1, or Neuro-D. To better understand the molecular basis of Shp-1 gene regulation, we studied the Shp-1 promoter. The 5′ regulatory sequence of Shp-1 lacked binding sites for Hnf-1α, Pdx-1, and Neuro-D but contained a conserved and functional Hnf-4 site. The biochemical characterization of this site and in vitro transactivation studies indicate that Hnf-4α is a transcriptional activator of the Shp-1 and that the reduced Shp-1 mRNA levels in pancreatic islets of HNF-1α^−/− mice likely are due to the reduced expression of HNF-4α in Hnf-1α–deficient islets. In addition, our data indicate that Shp-1 may regulate its own expression by inhibiting Hnf-4α function, thus forming a negative feedback loop, and indirectly (through Hnf-4α) regulate the activity of Hnf-1α, Pdx-1, and Neuro-D (Fig. 6). Additional studies, including targeted mutagenesis of the Shp-1 gene in mice, are needed to shed light on the biological role of Shp-1 in pancreatic islet function.

This work was supported in part by the American Diabetes Association, National Institutes of Health Grant R01-DK-55033-01 (M.S.), Human Frontier Science Program Organization research grant (M.P. and M.S.), and National Institutes of Health Medical Scientists Training Program Grant GM07739 (D.Q.S.). M.S. is an Irma Hirschl Scholar, Pew Scholar, and Robert and Harriet Heilbrunn Professor. K.N.M. was supported by the GATEWAYS to the Laboratory Program.