Differential regulation of islet-specific glucose-6-phosphatase catalytic subunit-related protein gene transcription by Pax-6 and Pdx-1.

Islet-specific glucose-6-phosphatase catalytic subunit-related protein (IGRP) is selectively expressed in islet beta cells and is a major autoantigen in a mouse model of type I diabetes. The analysis of IGRP-chloramphenicol acetyltransferase (CAT) fusion gene expression through transient transfection of islet-derived betaTC-3 cells revealed that a promoter region, located between -273 and -254, is essential for high IGRP-CAT fusion gene expression. The sequence of this promoter region does not match that for any known islet-enriched transcription factor. However, data derived from gel retardation assays, a modified ligation-mediated polymerase chain reaction in situ footprinting technique and a SDS-polyacrylamide separation/renaturation procedure led to the hypothesis that this protein might be Pax-6, a conclusion that was confirmed by gel supershift assays. Additional experiments revealed a second non-consensus Pax-6 binding site in the -306/-274 IGRP promoter region. Pax-6 binding to these elements is unusual in that it appears to require both its homeo and paired domains. Interestingly, loss of Pax-6 binding to the -273/ -246 element is compensated by Pax-6 binding to the -306/-274 element and vice versa. Gel retardation assays revealed that another islet-enriched transcription factor, namely Pdx-1, binds four non-consensus elements in the IGRP promoter. However, mutation of these elements has little effect on IGRP fusion gene expression. Although chromatin immunoprecipitation assays show that both Pax-6 and Pdx-1 bind to the IGRP promoter within intact cells, in contrast to the critical role of these factors in beta cell-specific insulin gene expression, IGRP gene transcription appears to require Pax-6 but not Pdx-1.

Islet-specific glucose-6-phosphatase catalytic subunitrelated protein (IGRP) is selectively expressed in islet ␤ cells and is a major autoantigen in a mouse model of type I diabetes. The analysis of IGRP-chloramphenicol acetyltransferase (CAT) fusion gene expression through transient transfection of islet-derived ␤TC-3 cells revealed that a promoter region, located between ؊273 and ؊254, is essential for high IGRP-CAT fusion gene expression. The sequence of this promoter region does not match that for any known islet-enriched transcription factor. However, data derived from gel retardation assays, a modified ligation-mediated polymerase chain reaction in situ footprinting technique and a SDS-polyacrylamide separation/renaturation procedure led to the hypothesis that this protein might be Pax-6, a conclusion that was confirmed by gel supershift assays. Additional experiments revealed a second non-consensus Pax-6 binding site in the ؊306/؊274 IGRP promoter region. Pax-6 binding to these elements is unusual in that it appears to require both its homeo and paired domains. Interestingly, loss of Pax-6 binding to the ؊273/ ؊246 element is compensated by Pax-6 binding to the ؊306/؊274 element and vice versa. Gel retardation assays revealed that another islet-enriched transcription factor, namely Pdx-1, binds four non-consensus elements in the IGRP promoter. However, mutation of these elements has little effect on IGRP fusion gene expression. Although chromatin immunoprecipitation assays show that both Pax-6 and Pdx-1 bind to the IGRP promoter within intact cells, in contrast to the critical role of these factors in ␤ cell-specific insulin gene expression, IGRP gene transcription appears to require Pax-6 but not Pdx-1.
Islet-specific glucose-6-phosphatase catalytic-subunit-related protein (IGRP) 1 is a putative enzyme specifically ex-pressed in the insulin producing ␤ cells of the Islets of Langerhans (1)(2)(3)(4). As its name implies, this protein is a homologue of the catalytic subunit of glucose-6-phosphatase (G6Pase), an enzyme that catalyzes the conversion of glucose 6-phosphate to glucose and orthophosphate (5)(6)(7)(8). This activity is restricted to a limited number of tissues, principally the liver and kidney (5)(6)(7)(8). In the liver, G6Pase is positioned at the junction of the metabolic pathways that mediate glycogenolysis and gluconeogenesis and thereby catalyzes the final step in hepatic glucose production. The role of G6Pase has also been studied in the context of other tissues, notably the pancreatic ␤ cell (5)(6)(7)(8).
Interest in this cell type stems from the hypothesis that elevated G6Pase activity in certain settings, such as type II diabetes, could lead to the uncoupling of glucose-stimulated insulin secretion by reversing a required step in this process, the phosphorylation of glucose. Reports have varied, however, as to the level of and even existence of G6Pase activity/expression in islets (5)(6)(7)(8). Interestingly, in studies that have been able to demonstrate islet G6Pase activity, different kinetic characteristics from those of the hepatic enzyme were observed, suggesting the presence of a unique islet G6Pase (1). The cloning of IGRP served to identify a candidate for this hypothetical enzyme (1,2). Although initial studies failed to demonstrate any glucose 6-phosphate hydrolysis by IGRP (1), an independent study recently suggested that IGRP may possess low G6Pase activity (9).
Although the role of IGRP in ␤ cell function remains unclear, a recent report identified IGRP as an autoantigen in a mouse model of Type I diabetes (10). This form of diabetes is caused by an autoimmune reaction that leads to the destruction of the pancreatic ␤ cell (11)(12)(13). Current evidence suggests that autoreactive CD8-positive T cells play a key role in disease pathogenesis (14). In the NOD (non-obese diabetic) mouse, which spontaneously develops a disease similar to type I diabetes in humans, a substantial portion (40%) of the CD8 cells infiltrating the islet are IGRP-reactive (10). These results suggest that IGRP is a major autoantigen in this murine model. Whether these observations directly translate to humans is unknown, however, numerous similarities between disease etiology in NOD mice and Type I diabetics have been noted (15,16).
Our work has focused on identifying the transcription factors that control IGRP gene expression (2,3,17,18). Originally, this work was undertaken in an attempt to identify novel, isletenriched transcription factors important for pancreatic development and/or function. We believe this approach is reasonable given that similar work, focused on other islet-specific genes, has led to the identification of such proteins (19 -21). With data now implicating IGRP as an autoantigen in Type I diabetes, the possibility exists that this work could also be useful for the design of strategies aimed at preventing immune recognition through suppression of ␤ cell antigens. The viability of this approach has already been investigated using antisense RNA directed against glutamic acid decarboxylase (GAD), another protein identified as a type I diabetes autoantigen. In transgenic NOD mice expressing this antisense RNA, it was noted that suppression of GAD protein levels was associated with lower incidence of diabetes (22). Although controversial (23), the results of this study suggest that the manipulation of other autoantigens, such as IGRP, should be investigated as a potential means to slow or prevent ␤ cell destruction and type I diabetes.
Previous studies investigating the molecular mechanisms that determine the islet-specific expression of the IGRP gene established the boundaries of a functional promoter (2). This sequence, which consists of 306 bp upstream of the transcription start site, is sufficient to drive expression of reporter genes specifically in ␤ cell-derived cell lines in vitro and pancreatic islets in vivo (24). Further, the results of a 5Ј deletion analysis, coupled with in situ footprinting, indicated that multiple cisacting elements within this 306-bp region contribute to IGRP promoter activity (17). A recent publication examining the importance of two conserved E-Box motifs extended these findings by demonstrating that the transcription factors NeuroD/ BETA2 and Upstream Stimulatory Factor contribute to IGRP promoter activity (18). In this report we show that IGRP promoter activity in insulinoma cell lines is highly dependent on the binding of the transcription factor Pax-6 to two non-consensus binding sites. Interestingly, despite interacting directly with the IGRP promoter in situ, our data do not support a similar critical role for the transcription factor Pdx-1.

Fusion Gene Plasmid Construction
The construction of a rat insulin II-chloramphenicol acetyltransferase (CAT) fusion gene, containing promoter sequence from Ϫ238 to ϩ2 has been previously described (26) as has a site-directed mutation of the A3 Pdx-1 binding site generated in the context of the Ϫ238 to ϩ2 promoter region (27). The construction of mouse IGRP-CAT fusion genes, containing promoter sequence from Ϫ306 to ϩ3, Ϫ273 to ϩ3, and Ϫ254 to ϩ3, has been previously described (2,17).
The Pax-6 binding site located between Ϫ273 and Ϫ246 in the IGRP promoter was mutated by site-directed mutagenesis within the context of the Ϫ273 to ϩ3 promoter fragment. Two constructs, designated Ϫ273 MUT 1 and Ϫ273 MUT 2 ( Fig. 1), were generated using PCR and the following oligonucleotides as the 5Ј primers: 5Ј-GCGGATCCAAGCT-(Ϫ273)TTAAGGGAAATcgcagtTATGAAAAATTGCAAG-3Ј (MUT 1) and 5Ј-GCGGATCCAAGCT(Ϫ273)TTAAGcactgaGAACTGTATGAAAAAT-TG-3Ј (MUT 2). BamHI cloning sites are underlined, and the mutated sequences are in lowercase letters. The 3Ј PCR primer (5Ј-CCGCTC-GAGATCCAGATCCTC-3Ј; XhoI cloning site underlined) and the 5Ј PCR primers were designed to ensure that the 3Ј and 5Ј junctions of the IGRP promoter were the same as those in the wild-type Ϫ273 IGRP-CAT fusion gene construct.
A three-step PCR strategy (28) was used to create a site-directed mutation of the Ϫ273/Ϫ246 Pax-6 motif within the context of the Ϫ306 to ϩ3 IGRP promoter fragment. The resulting construct, designated Ϫ306 Pax MUT A, was generated using two complementary PCR primers that were designed to mutate nucleotides within the Ϫ273/Ϫ246 Pax-6 motif. The sequence of the sense strand oligonucleotide was as follows (mutated nucleotides in lowercase): 5Ј-(Ϫ277)GCCCTTAAGG-GAAATcgcagtTATGAAAAATTGC-3Ј. With the Ϫ306 IGRP-CAT plasmid as the template, this sense strand oligonucleotide was used in conjunction with a 3Ј PCR primer to generate the 3Ј half of the IGRP promoter, whereas the complementary antisense strand oligonucleotide was used in conjunction with a 5Ј PCR primer to generate the 5Ј half of the IGRP promoter. The 3Ј primer was the same as described above. The 5Ј PCR primer (5Ј-CGGGATCCAAGCT(Ϫ306)CTAGCCAAGC-3Ј; BamHI cloning site underlined) was designed to conserve the junction between the IGRP promoter and CAT vector to be the same as that in the wild-type Ϫ306 IGRP-CAT fusion gene plasmid. The PCR products from each reaction pair were then combined and used themselves as both primer and template in a second PCR reaction step to a generate small amount of the full-length, mutated IGRP promoter fragment. Finally, the 5Ј and 3Ј PCR primers were then used to amplify this fragment.
Similarly, the Pax-6 binding site located between Ϫ306 and Ϫ274 in the IGRP promoter was targeted by site-directed mutagenesis within the context of the Ϫ306 to ϩ3 promoter fragment. The construct, designated Ϫ306 Pax MUT B (Fig. 10), was generated, with the wild-type Ϫ306 IGRP-CAT plasmid as the template, using PCR with the same 3Ј PCR primer as described above and the following oligonucleotide as the 5Ј primer: 5Ј-CGGGATCCAAGCT(Ϫ306)CTAGCCAAGCGTAGAtcggG-CTAATGTCTGCC. A BamHI cloning site is underlined, and the mutated sequence is in lowercase letters. The 5Ј junction of the resulting IGRP promoter fragment was the same as that in the wild-type Ϫ306 IGRP-CAT fusion gene construct. The mutation selectively targets the Pax-6 paired domain binding site. In addition, the Ϫ306 Pax MUT A construct described above served as the template for a PCR reaction with the Pax MUT B primer that led to the creation of a construct designated Ϫ306 Pax MUT AB, which contains mutations in both Pax-6 paired domain binding sites in the IGRP promoter.
Using a combination of unique restriction enzyme sites in the IGRP promoter and the three-step PCR strategy described above, the mutations shown in Table I were introduced in combination into Pdx-1 binding sites 1, 2, 3, and 4 to generate the plasmid IGRP Ϫ306 Quad MUT (Fig. 9). Promoter fragments generated by PCR were completely sequenced to ensure the absence of polymerase errors. Plasmid constructs were purified by centrifugation through cesium chloride gradients (29).

CAT and Luciferase Assays
Transfected ␤TC-3 cells were harvested by trypsin digestion and then solubilized in passive lysis buffer (Promega). After two cycles of freeze/thawing, firefly luciferase activity was assayed as described previously (30). The remaining ␤TC-3 lysate was heated for 10 min at 65°C, and cellular debris was removed by centrifugation. CAT assays were then performed on the supernatant as previously described (2). To correct for variations in transfection efficiency, the results are expressed as the ratio of CAT:firefly luciferase activity. In addition, three independent preparations of each IGRP-CAT plasmid construct were analyzed to obtain the data shown in each figure.

Gel Retardation Assay
Labeled Probes-Sense and antisense oligonucleotides representing various IGRP promoter sequences, as indicated in the figure legends, were synthesized with BamHI-compatible ends and subsequently gelpurified, annealed, and labeled with [␣-32 P]dATP using the Klenow fragment of Escherichia coli DNA polymerase I to a specific activity of ϳ2.5 Ci/pmol (29).
High Salt Nuclear Extract Preparation-High salt ␤TC-3 nuclear extract was prepared as described previously (31,32), except that nuclei were lysed by resuspension in a buffer containing 800 mM NaCl, 20 mM Hepes, pH 7.9, 0.75 mM spermidine, 0.15 mM spermine, 0.2 mM EDTA, 2 mM EGTA, 2 mM dithiothreitol (DTT), 25% glycerol, and 1 mM phenylmethylsulfonyl fluoride. After incubation for 30 min at 4°C to ensure complete lysis, samples were centrifuged at 100,000 rpm in a Beckman TLA 100.3 rotor for 40 min at 4°C. The supernatant was dialyzed against buffer containing 100 mM KCl, 20 mM Hepes, pH 7.9, 0.2 mM EDTA, 0.2 mM EGTA, 2 mM DTT, 20% glycerol, and 1 mM phenylmethylsulfonyl fluoride. The protein concentration of the nuclear extracts was determined using the Bio-Rad assay and was typically ϳ1 g l Ϫ1 .
Low Salt Nuclear Extract Preparation-Low salt ␤TC-3, ␣TC-6, Y1, HeLa, and H4IIE nuclear extracts were prepared as previously described (31,32), except that the nuclear pellet was extracted with 20 mM HEPES, pH 7.8, 0.75 mM spermidine, 0.15 mM spermine, 0.2 mM EDTA, 2 mM EGTA, 2 mM DTT, 25% glycerol containing 200 mM NaCl, instead of 0.4 M ammonium sulfate, and the supernatant was used directly in gel retardation assays. The protein concentration of the nuclear extracts was determined using the Bio-Rad assay and was typically ϳ1 g l Ϫ1 .
Binding Assays-ϳ14 fmol of radiolabeled probe (ϳ50,000 cpm) was incubated with 1.5-3 g of the indicated nuclear extract in a final 20-l reaction volume. For experiments using high salt nuclear extract, these reactions contained 20 mM Hepes, pH 7.9, 0.1 mM EDTA, 0.1 mM EGTA, 10% glycerol (v/v), 1 mM DDT, 1 g of poly(dI-dC)⅐poly(dI-dC), and 100 mM KCl. In experiments using low salt nuclear extract the final concentrations of EDTA, EGTA, and glycerol were 0.1 mM, 1 mM, and 12.5%, respectively. In addition, the binding reaction contained 0.375 mM spermidine, 0.075 mM spermine, and 100 mM NaCl instead of 100 mM KCl. After incubation at room temperature for 20 min, samples were loaded onto a 6% polyacrylamide gel containing 1ϫ TGE (25 mM Tris base, 190 mM glycine, 1 mM EDTA) and 2.5% (v/v) glycerol. Samples were electrophoresed for 1.5 h at 150 V in 1ϫ TGE buffer before the gel was dried and exposed to Kodak XB film with intensifying screens.
Competition Experiments and Gel Supershifts-For competition experiments, unlabeled competitor DNA was mixed with the radiolabeled oligomer at the indicated molar excess prior to addition of nuclear extract. For supershift experiments, specific antisera (1 l) were preincubated with ␤TC-3 nuclear extract for 10 min on ice prior to the addition of the labeled IGRP WT oligonucleotide probe. All subsequent steps were carried out as described above.

Fractionation of Cell Proteins by SDS-PAGE, Elution, and Renaturation
Nuclear extract proteins were boiled for 5 min in SDS-PAGE sample buffer (62.5 mM Tris, pH 6.8, 2% SDS, 10% glycerol, 5 mM EDTA, 100 mM DTT, 0.007% bromphenol blue) and fractionated on 1.5-mm thick 15% SDS-polyacrylamide gels using standard methods (29). For molecular weight standardization, samples containing high molecular weight markers (Rainbow Molecular Weight markers, Amersham Biosciences) were run in lanes adjacent to those containing experimental samples. Proteins were electrophoretically transferred from SDS-PAGE gel lanes to polyvinylidene fluoride (PVDF) membranes (Immobilon-P, Millipore) using the Bio-Rad Trans-Blot SD blotting apparatus according to the manufacturer's instructions. Proteins were eluted from ϳ2-mm slices of PVDF membrane and renatured overnight in 100 l of buffer containing 25 mM Hepes, pH 7.9, 100 mM KCl, 0.1 mM EDTA, 1 mM DTT, 10% glycerol, 1% Triton X-100 (Surfact-Amps X-100, Pierce), and 0.1 mg/ml bovine serum albumin. Elutions were carried out in microcentrifuge tubes at 4°C with vigorous shaking. Samples were then flash-frozen on dry ice and later thawed for analysis in gel retardation experiments.

Genomic DNA Isolation, in Situ and in Vitro Methylation, and Piperidine Cleavage
To analyze protein binding to the IGRP promoter, promoter methylation by DMS in vitro and in situ was compared, with differences interpreted as being indicative of trans-acting factor binding in situ. Methylation of genomic DNA and strand cleavage by piperidine and heating was based on the Maxam and Gilbert method of DNA sequencing (33). The relative abundance of piperidine-cleaved DNA fragments, and hence the level of DNA methylation, was then assayed using the ligation-mediated polymerase chain reaction (LMPCR) footprinting technique (34).
Isolation and purification of ␤TC-3 genomic DNA and its subsequent in vitro methylation was performed exactly as previously described (17). Similarly, methylation of ␤TC-3 genomic DNA in situ and its subsequent isolation and purification were also performed exactly as previously described (17). DNA was isolated from single 55-cm 2 cell culture dishes with a typical yield of ϳ200 g per dish. After purification, DNA that had been methylated in situ or in vitro was treated using two separate methods for base displacement and strand scission as described by Maxam and Gilbert (33). In method A, DNA was resuspended in 180 l of water and incubated with 20 l of piperidine at 90°C for 30 min, as previously reported (17). This method yields DNA products predominately cleaved at methylated guanine (G) residues with only very weak cleavage of methylated adenine (A) residues (33). Alternatively, in method B, DNA was resuspended in 180 l of phosphatebuffered water (10 mM sodium phosphate, pH 7) and heated at 90°C for 15 min. 20 l of piperidine was then added, and the incubation was continued at 90°C for 30 min. This method yields DNA products cleaved at both methylated guanine (G) and methylated adenine (A) residues, although cleavage at methylated guanine residues still predominates (33). In both methods, samples were then lyophilized to remove the piperidine. The DNA was resuspended in 100 l of water and lyophilized another two times. The DNA was then resuspended in 100 l of water, centrifuged briefly to pellet any debris, and ethanolprecipitated. Finally, the DNA was resuspended in 10 mM Tris-HCl (pH 8.0) and 1 mM EDTA to a final concentration of 1 g/l.

Ligation-mediated PCR
The ligation-mediated PCR (LMPCR) technique was used to analyze protein DNA interactions over the sense strand of the Ϫ273 to Ϫ246 region of the IGRP promoter in intact ␤TC-3 cells by comparing the level of G and A residue methylation over this region in situ and in vitro (34). LMPCR was performed exactly as previously described (17). Briefly, three nested oligonucleotides (Primer Set A (17)) were used for each step of the LMPCR method, namely the first strand synthesis reaction, the amplification reaction, and the labeling reaction (17). The reaction products were separated by electrophoresis on 6% polyacrylamide/urea/TBE gels (29) and visualized by autoradiography.

RESULTS
The Ϫ273/Ϫ254 IGRP Promoter Region Is Required for Maximal IGRP Fusion Gene Expression-In a previous study it was shown that the progressive 5Ј deletion of IGRP promoter sequence, starting at Ϫ306, leads to a stepwise drop in promoter activity as measured by the analysis of IGRP-CAT fusion gene expression (17). The most dramatic reduction in activity occurred when the sequence located between Ϫ273 and Ϫ254 was removed, an observation that pointed to this element as being particularly important for maintaining transcription of the IGRP gene (17). This region is highly conserved between the mouse, rat, and human IGRP promoters (3). Promoter-scanning software (35) identified a number of binding sites for previously characterized transcription factors in the Ϫ273/ Ϫ254 promoter region, including binding sites for STAT (signal transducer and activator of transcription) and NFAT (nuclear factor activated in T cells). These candidates, however, were eliminated based on gel retardation competition studies using consensus binding sites for these factors (data not shown).
To define the boundaries of the putative novel element present in the Ϫ273/Ϫ254 promoter region, the functional effect of introducing two block mutations in the Ϫ273/Ϫ254 region of the IGRP promoter on fusion gene expression was investigated (Fig. 1). These mutations, designated Ϫ273 MUT 1 and Ϫ273 MUT 2, were generated in the context of the Ϫ273 to ϩ3 IGRP promoter region, and the level of reporter gene expression directed by fusion genes containing these mutations was then analyzed by transient transfection of ␤TC-3 cells (Fig. 1). For comparison, the CAT expression directed by the wild-type Ϫ273 IGRP-CAT and Ϫ254 IGRP-CAT fusion genes was also determined. The results of this analysis reveal that both block mutations lead to a large reduction in IGRP promoter activity (Fig. 1). Importantly, the level of basal CAT expression directed by these mutants is equivalent to that seen when the Ϫ273/ Ϫ254 region was completely deleted, as is the case with the Ϫ254 5Ј deletion end-point construct (Fig. 1). This result suggests that sequences encompassed by both block mutations are required for the activity conferred by the Ϫ273/Ϫ254 promoter region. Almost identical results were obtained when these fusion genes were analyzed by transient transfection of hamster insulinoma tumor (HIT) cells (data not shown).
A modified LMPCR in Situ Footprinting Method Reveals Broad Protein Binding over the Endogenous Ϫ273/Ϫ246 IGRP Promoter Region in ␤TC-3 Cells-To further define the boundaries of the putative novel element present in the Ϫ273/Ϫ254 promoter sequence, protein binding to this IGRP promoter region was analyzed using in situ footprinting. IGRP promoter methylation by dimethyl sulfate (DMS) in vitro and in ␤TC-3 cells in situ was compared, with differences interpreted as being indicative of trans-acting factor binding in situ. DMS methylates DNA on adenine (A) and guanine (G) residues and freely permeates cell membranes (36). The binding of a transacting factor to a gene promoter can result in either an increase or decrease in the level of methylation (34,36). A decrease in the level of methylation is interpreted to occur as a result of steric hindrance, whereas an increase is interpreted to occur if the trans-acting factor causes a change in DNA conformation leading to greater reactivity of a base with DMS (36,37).
In vitro and in situ methylated genomic DNA was cleaved either using piperidine under conditions that yielded DNA products predominately cleaved at guanine (G) residues with only weak cleavage of adenine (A) residues (33), as previously described (17). Alternatively, methylated genomic DNA was cleaved using piperidine under conditions that yielded DNA products cleaved at both G and A residues (33). The relative abundance of these piperidine cleaved DNA fragments, and hence the level of DNA methylation, was then assayed using the LMPCR footprinting technique (17,34). Differences in the frequency of methylation at a given G residue in situ result in the over or under representation of that piperidine-cleaved fragment in the subsequent PCR amplification in comparison to in vitro methylated DNA (Fig. 2) (17,34). Fig. 2 shows the in situ footprinting analysis of trans-acting factor binding to the sense strand of the Ϫ277/Ϫ245 IGRP promoter region. The positions of specific G and A residues in the IGRP promoter sequence are indicated, as are increases or decreases in the level of IGRP promoter methylation comparing in vitro (lane 1) and in situ (lanes 2-4) methylated ␤TC-3 genomic DNA. Only changes that were consistent between duplicate experiments are indicated. When methylated genomic DNA was cleaved using piperidine under conditions that yielded DNA products predominately cleaved at G residues, with only weak cleavage of A residues, hypomethylation of the guanine residues at Ϫ269, Ϫ268, and Ϫ257 and hypermethylation of the guanine residues at Ϫ267 and Ϫ262 was detected ( Fig. 2; compare lanes 1 and 4), as previously described (17). In contrast, when methylated genomic DNA was cleaved using piperidine under conditions that yielded DNA products cleaved at both G and A residues, an additional hypermethylation of the adenine residues at Ϫ255 was detected ( Fig. 2; compare lanes 1 and 2). None of these differences in methylation were influenced by the presence or absence of serum in the culture medium ( Fig. 2; compare lanes 2 and 3). The results of this in situ footprinting analysis suggest either that multiple factors bind the Ϫ273/Ϫ254 IGRP promoter re-FIG. 1. Two mutations within the ؊273 to ؊254 IGRP promoter region reduce basal fusion gene expression. ␤TC-3 cells were transiently co-transfected, as described under "Experimental Procedures," using a LipofectAMINE solution containing various IGRP-CAT fusion genes (2 g) and an expression vector encoding firefly luciferase (0.5 g). The IGRP-CAT fusion genes incorporated either the wild-type promoter sequence, located between Ϫ273 and ϩ3 (Ϫ273 WT) or between Ϫ254 and ϩ3 (Ϫ254 WT), or contained site-directed mutations of two promoter regions that are conserved between the mouse, rat, and human IGRP promoters (3). These mutations, designated Ϫ273 MUT 1 and Ϫ273 MUT 2, were generated within the context of the Ϫ273 to ϩ3 promoter fragment. Following transfection, ␤TC-3 cells were incubated for 18 -20 h in serum-containing medium. Cells were then harvested, and CAT and firefly luciferase activity were assayed as previously described (2,30). Results are presented as the ratio of CAT:luciferase activity expressed as a percentage relative to the value obtained with the Ϫ273 WT fusion gene and represent the mean of three experiments Ϯ S.E., each using an independent preparation of all fusion gene plasmids.  3 and 4). Methylated DNA was cleaved using a modified (lanes 1-3) or standard (lane 4) protocol and an LMPCR reaction was then performed as described under "Experimental Procedures." The autoradiograph shown is representative of two experiments. Only increases (q) or decreases (E) in band intensity, comparing in situ to in vitro methylated DNA, that were consistent between both gels are indicated. gion or that a single factor makes broad contacts across this entire promoter region.
Specific Protein Binding to the Ϫ273/Ϫ254 IGRP Promoter Region in Vitro-The fusion gene analysis (Fig. 1) indicated that one or more transcription factors required for expression of the IGRP gene interacts with the Ϫ273/Ϫ254 region of the IGRP promoter and that two block mutations in this region (Fig. 1) disrupt binding of this factor or factors. To identify the factor(s) binding this element, a long oligonucleotide that includes sequence 3Ј of the Ϫ254 5Ј deletion end point was used as the labeled probe in a gel retardation assay. When this labeled oligonucleotide, representing the wild-type (WT) IGRP promoter sequence from Ϫ273 to Ϫ246, was incubated with nuclear extract prepared from ␤TC-3 cells, four protein-DNA complexes were detected (Fig. 3A). Competition experiments, in which a 100-fold molar excess of unlabeled DNA was included with the labeled probe, were used to correlate protein binding with basal IGRP gene expression. The wild type (WT) Ϫ273/Ϫ246 oligonucleotide competed effectively for the formation of three of these protein-DNA complexes, designated 1, 2, and 3 (Fig. 3A), indicating that complex 4 must represent a nonspecific protein-DNA interaction. By contrast, oligonucleotides, designated Ϫ273/Ϫ246 MUT 1 and Ϫ273/Ϫ246 MUT 2, that contain mutations identical to those described in the Ϫ273 MUT 1 and MUT 2 constructs, respectively (Fig. 1), failed to compete with the labeled probe for formation of two complexes, designated 3 and 4 (Fig. 3A). This indicates that complex 3 ( Fig.  3A; see arrow) represents a specific protein-DNA interaction and that its formation correlates with basal gene expression conferred by the Ϫ273/Ϫ254 IGRP promoter region. In contrast, complexes 1 and 2 must either represent nonspecific protein-DNA interactions or protein binding to sequences outside the regions spanning the MUT 1 and MUT 2 mutations. A comparison of protein binding to the Ϫ273/Ϫ246 WT oligonucleotide using nuclear extracts isolated from other cell lines suggests that the factor in complex 3 is restricted in its expression (Fig.  3B). Thus, of the cell lines analyzed, complex 3 was only detected using ␤TC-3 and ␣TC-6 nuclear extracts (Fig. 3B).
To further characterize the protein(s) present in complex 3, nuclear extracts were subjected to a fractionation/renaturation protocol, a technique that allows for the determination of the molecular weight of transcription factors that retain their DNA binding activity in this assay. ␤TC-3 extract was fractionated by SDS-polyacrylamide gel electrophoresis, and proteins were transferred to a PVDF membrane. Eluates from slices of this membrane were then subjected to a gel retardation assay using the Ϫ273/Ϫ246 WT oligonucleotide probe. Fig. 3C shows the results of the fractionation/renaturation procedure and includes a lane with crude nuclear extract (NE) so that the migration of fractionated binding activities can be compared with the complex of interest (see arrow). This gel retardation assay indicates that the eluate from slice D contains a binding activity that co-migrates with the complex of interest observed in crude extracts (Fig. 3C). From comparison with molecular weight standards it was concluded that proteins eluting from this slice were ϳ45 kDa in size. The reconstitution of this binding activity from a single membrane slice also suggested that the complex may contain a single protein species. Transcription factors that bind to DNA as heterodimers, in contrast, may be separated from their partners during fractionation depending on their respective molecular weights.
Pax-6 Stimulates IGRP Fusion Gene Expression through the Ϫ273/Ϫ246 Promoter Region-Although computer analyses do not identify the Ϫ273/Ϫ246 promoter region as a Pax-6 binding site, the ability to recognize a broad DNA element (consistent with the in situ footprinting analysis (Fig. 2)), the expression in both ␤TC-3 and ␣TC-6 nuclear extracts (Fig. 3B), and the molecular mass of ϳ45 kDa (Fig. 3C) are all characteristics of Pax-6 (38, 39). Therefore, a Pax-6 antiserum was tested for its FIG. 3. A 45-kDa protein specifically binds the ؊273/؊246 IGRP promoter sequence. A, a labeled oligonucleotide representing the wild-type (WT) Ϫ273/Ϫ246 IGRP promoter sequence was incubated in the absence (Ϫ) or presence of a 100-fold molar excess of the unlabeled competitor (comp.) DNAs shown, representing the WT or mutated (MUT 1 and MUT 2) Ϫ273/Ϫ246 IGRP sequence (Fig. 1). Low salt ␤TC-3 nuclear extract was then added, and protein binding analyzed using the gel retardation assay as described under "Experimental Procedures." In the representative autoradiograph shown only the retarded complexes are visible and not the free probe, which was present in excess. The formation of a single complex, complex 3, indicated by the arrow, is selectively reduced by competition with the WT and not the MUT oligonucleotides. B, the labeled oligonucleotide representing the wild-type Ϫ273/Ϫ246 IGRP promoter sequence was incubated with either low salt ␤TC-3, ␣TC-6, Y1, HeLa, or H4IIE nuclear extract and protein binding was then analyzed using the gel retardation assay as described under "Experimental Procedures." The representative autoradiograph only shows the retarded complexes and not the free probe, which was present in excess. C, a fractionation and renaturation technique was used to analyze protein binding to the Ϫ273/Ϫ246 IGRP oligonucleotide as described under "Experimental Procedures." Crude low salt ␤TC-3 nuclear extract (NE) and PVDF membrane slice eluates (lanes A-G) were screened for binding activity using the gel retardation assay in conjunction with the Ϫ273/Ϫ246 labeled probe. MW, approximate molecular mass (kilodaltons). A representative autoradiograph is shown. In panels B and C the specific protein-DNA complex, whose formation is selectively reduced by competition with the WT and not the MUT oligonucleotides, is indicated by the arrow. ability to alter formation of the specific Ϫ273/Ϫ246 complex in gel retardation assays. Fig. 4 demonstrates that preincubating ␤TC-3 nuclear extracts with antibodies raised against this protein results in the disappearance of the specific complex and the appearance of a lower mobility, or supershifted complex. As a control, antibodies raised to various other transcription factors known to be enriched in pancreatic islets were tested for their ability to alter the abundance or migration of the Ϫ273/ Ϫ246 complex in gel retardation assays, but they were without effect (data not shown). This result suggests that the Ϫ273/ Ϫ246 complex contained the transcription factor Pax-6. Pax-6 is able to bind DNA as a monomer (38), an observation that is consistent with the ability to retain DNA binding activity after fractionation/renaturation (Fig. 3C).
To confirm this conclusion, competition experiments were performed using Pax-6 binding sites from promoters of other genes expressed in pancreatic islets (Fig. 5A). Pax-6 has been shown to interact with the C2 element of the rat insulin I promoter, and various lines of evidence indicate that this factor is required for the expression of the insulin gene (39). In addition to the regulation of ␤ cell-specific genes such as insulin, Pax-6 is also expressed in the ␣ cells of the pancreatic islet and has been shown to activate the glucagon promoter through two elements designated G1 and G3 (39,40). Consistent with the results of the supershift analysis (Fig. 4), when unlabeled oligonucleotides representing these three Pax-6 sites were included in gel retardation reactions with the Ϫ273/Ϫ246-labeled probe, they competed effectively for the formation of the Pax-6 protein-DNA complex in a concentration-dependent manner. Representative autoradiographs are shown in Fig. 5B, whereas Fig. 5C shows quantified data from multiple experiments. These comparisons indicate that the C2, G1, and G3 elements all compete more effectively for Pax-6 binding than the IGRP element (Fig. 5B) and that Pax-6 binds to the G1 element with ϳ10-fold higher affinity than the IGRP element (Fig. 5C).
The results of these competition studies demonstrate that the affinity of Pax-6 for the IGRP Ϫ273/Ϫ246 element is relatively weak compared with other characterized binding sites. However, given that the promoter-scanning software did not identify this site even at low stringency, the question arises as to how this protein can bind to the IGRP element at all, albeit with low affinity. Pax-6 has two separate DNA binding domains, a paired-type domain and a homeodomain (41). Most studies, including the determination of the consensus Pax-6 binding site, have focused on the paired domain (38). The Ϫ273/Ϫ246 IGRP sequence bears a very weak resemblance to a consensus Pax-6 paired domain binding site (Fig. 5A), and thus is not recognized by the promoter-scanning software. It has been observed, however, that the homeodomain can bind to sites opposed to paired domain binding sites and can increase the affinity of Pax-6 for DNA in a cooperative manner (41). Interestingly, there is an A/T-rich sequence in the Ϫ273/Ϫ246 region of the IGRP promoter, which is similar to the consensus TAAT recognized by many homeodomains (Fig. 5A). It is possible that, although the Ϫ273/Ϫ246 IGRP promoter region contains neither a strong Pax-6 paired domain binding site nor a strong homeodomain binding site, such that neither of these sequences by themselves would be predicted to bind the two respective Pax-6 DNA binding domains, these weak binding sites could cooperate to mediate an interaction. Consistent with this idea, a truncated Ϫ273/Ϫ250 IGRP oligonucleotide probe, in which the 3Ј A/T-rich sequence has been removed, no longer binds Pax-6 (data not shown). Collectively, the results of these experiments suggest that, although Pax-6 binds to the Ϫ273/ Ϫ246 region relatively weakly, this interaction is critical for maintaining maximum IGRP promoter activity.
Pdx-1 Binds to Four Sites in the IGRP Promoter-The studies described above were driven by the initial observation that deletion of the Ϫ273/Ϫ254 IGRP promoter region resulted in a large drop in IGRP promoter activity. As an alternative approach aimed at identifying transcription factors relevant to IGRP expression, candidates were sought that had been well studied in the context of ␤ cell gene expression. The homeodomain transcription factor Pdx-1 fulfills this criterion based on the numerous reports investigating its role in the expression of ␤ cell-specific or enriched proteins such as insulin, GLUT2, IAPP, and even Pdx-1 itself (42). The function of this transcription factor has been most closely examined in the context of studies investigating the cell type-specific activity of the insulin promoter. Although not sufficient by itself to mediate transcriptional activation, Pdx-1 is critical for insulin promoter activity and most likely functions in a higher order complex containing other islet-enriched factors (43).
To determine if Pdx-1 interacts with the IGRP promoter, potential binding sites were first identified. Based on previously studied Pdx-1 binding sites, it appears that this transcription factor recognizes a core TAAT motif, which as mentioned above, is recognized by many different homeodomain proteins (44). A search of the IGRP promoter identified three such motifs, all of which are conserved across human, mouse, and rat species. The flanking sequences surrounding these TAAT motifs do not match those of consensus Pdx-1 binding sites, so promoter-scanning software (35) did not predict that these motifs would bind Pdx-1. Nevertheless, the potential of these sites to interact with Pdx-1 was examined using oligonucleotide probes representing all three sites in gel retardation experiments. These probes were designated Site 1, Site 3, and Site 4 and are depicted in Table I along with another probe labeled Site 2, which is discussed below. When the labeled Site 1 oligonucleotide was incubated with nuclear extract prepared from ␤TC-3 cells a single, major protein-DNA complex was formed (Fig. 6A). As described above, the specificity of this protein-DNA interaction was investigated by including various cold competitors in the gel retardation assay. Fig. 6A shows that the wild-type (WT) unlabeled Site 1 oligonucleotide competed for formation of the major protein-DNA complex, whereas an unrelated (UNR) oligonucleotide was completely ineffective, demonstrating the specificity of this interaction. Further, an oligonucleotide containing a block mutation in the Site 1 TAAT motif (Table I) 4. Pax-6 binds the ؊273/؊246 IGRP promoter sequence. High salt ␤TC-3 cell nuclear extract was incubated in the absence (Ϫ) or presence of the indicated antisera for 10 min on ice prior to the addition of the labeled oligonucleotide representing the wild-type (WT) Ϫ273/ Ϫ246 IGRP promoter sequence, and incubation was continued for an additional 20 min at room temperature. Protein binding was then analyzed as described under "Experimental Procedures." In the representative autoradiograph shown only the retarded complexes are visible and not the free probe, which was present in excess. The specific protein-DNA complex, whose formation is selectively reduced by competition with the WT and not the MUT Ϫ273/Ϫ246 IGRP oligonucleotides, is indicated by the arrow.
FIG. 5. Relative affinity of Pax-6 binding sites in the insulin, glucagon, and IGRP promoters. A, Pax-6 binding sites in the insulin, glucagon, and IGRP promoters and the consensus Pax-6 binding site are shown. This consensus was taken from Ref. 38. Increases (q) or decreases (E) in DMS methylation between in situ versus in vitro methylated mouse ␤TC-3 cell DNA over the Ϫ273 to Ϫ244 IGRP promoter region are shown; this information was derived from Fig. 3. B, prior to the addition of high salt ␤TC-3 cell nuclear extract, the labeled oligonucleotide representing the wild-type Ϫ273/Ϫ246 IGRP promoter sequence was incubated in the absence (Ϫ) or presence of the indicated molar excess of the unlabeled competitor (comp.) DNAs shown, representing the wild-type Ϫ270/Ϫ246 IGRP sequence, the insulin C2 element, and the glucagon G1 and G3 elements. The IGRP Ϫ273/Ϫ246 and Ϫ270/Ϫ246 oligonucleotides bind Pax-6 with the same affinities (data not shown). Protein binding was then analyzed as described under "Experimental Procedures." In the representative autoradiographs shown only the retarded complex is visible and not the free probe, which was present in excess. C, the protein binding shown in panel C was quantified by using a Packard Instant Imager to count 32 P associated with the retarded complex. The data represent the mean Ϯ S.E. of three experiments.
gesting that the formation of the specific complex was partially dependent on the TAAT motif.
In contrast to this result with the Site 1 probe, when the labeled Site 3 and Site 4 oligonucleotides were incubated with ␤TC-3 nuclear extract several protein-DNA complexes were formed (Fig. 6, B and C). The wild type Site 3 and Site 4 oligonucleotides competed effectively for the formation of all of these protein-DNA complexes, whereas oligonucleotides containing mutations in the Site 3 and Site 4 TAAT motifs (Table  I) failed to compete with the labeled probe for formation of two complexes, designated X and Y (Fig. 6, B and C). This indicates these complexes represent specific protein-DNA interactions and that the other bands detected in the assay must either represent nonspecific protein-DNA interactions or protein binding to sequences outside the regions spanning the TAAT mutations.
To assess the presence of Pdx-1 in these specific protein-DNA complexes, ␤TC-3 nuclear extract was preincubated with antisera raised to the amino terminus of this transcription factor (25). As can be seen in Fig. 7 (lane 3), this had no appreciable effect on formation of the complex formed with the Site 1 probe. In contrast, preincubation with this antiserum resulted in the disappearance of the specific complex Y and the appearance of a lower mobility, or supershifted complex, with both the Site 3 and Site 4 probes.
Also included in these supershift experiments was an oligonucleotide probe designated Site 2, which does not contain a TAAT motif (Table I). This probe encompasses a region in the IGRP promoter from Ϫ197 to Ϫ173 that, when deleted, results in a loss in promoter activity (17). Originally, this probe was used in gel retardation assays to identify the transcription factors responsible for this activity. These experiments resulted in the identification of two binding activities that mapped to an A/T-rich region of the Site 2 probe and comigrated with complexes X and Y formed using Site 3 and 4 probes (data not shown). Interestingly, despite lacking the core TAAT motif, Pdx-1 was observed to interact with this probe as evidenced by the Pdx-1 antibody-induced supershift of one of the specific complexes, complex Y, that forms with the Site 2 probe (Fig. 7, lane 3, Site 2 panel).
The identity of the specific complex that forms with the Site 1 probe and complex X that forms with the Site 2, 3, and 4 probes was next investigated. Based on the results of the competition analysis described above, it was apparent that the formation of these complexes was dependent on a TAAT motif (Fig. 6), or in the case of Site 2, an A/T-rich sequence (data not shown). These observations suggested the potential binding of another homeodomain protein. The molecular weight of the potential homeodomain protein present in the specific complex that forms with the Site 1 probe was estimated by utilizing the fractionation renaturation protocol described earlier. After fractionation of the ␤TC-3 nuclear extract and renaturation, binding activity was detected using the eluate from a particular slice that co-migrates with the specific complex that forms with the Site 1 probe (data not shown). From comparison with molecular mass standards, it was apparent that proteins eluting from this slice were ϳ45 kDa in size (data not shown). Because Pax-6 is also present in this slice (Fig. 3C) and because Pax-6   (Table I), or an unrelated (UNR) DNA sequence. High salt ␤TC-3 nuclear extract was then added, and protein binding was analyzed using the gel retardation assay as described under "Experimental Procedures." In the representative autoradiograph shown only the retarded complexes are visible and not the free probe, which was present in excess. The formation of a single complex, indicated by the arrow, is selectively reduced by competition with the WT and not the UNR Site 1 oligonucleotides, whereas the formation of two complexes, designated X and Y, are selectively reduced by competition with the WT and not the MUT Site 3 and Site 4 oligonucleotides. contains a homeodomain, antibodies to this factor were tested for their ability to alter the formation of the specific complex that forms with the Site 1 probe. Preincubation of ␤TC-3 nuclear extracts with Pax-6 antiserum reduced the abundance of this complex and resulted in the appearance of a slower migrating complex (Fig. 7, lane 4, Site 1 panel), demonstrating the presence of Pax-6. Further, these antibodies also disrupted and/or supershifted complex X, which formed when the Site 2, 3, and 4 probes were used (Fig. 7, lane 4). These experiments thus identify the unknown homeodomain protein binding to Sites 1 through 4 as Pax-6.
In the course of these binding analyses it was found that the ability to detect Pax-6 using some of the indicated probes depended on the salt concentration used for nuclear extraction. Thus, as an example, under high salt (800 mM NaCl) extraction conditions, both Pdx-1 and Pax-6 are observed to interact with the Site 2, 3, and 4 probes (Fig. 7, lanes 1-4). In contrast, the Pax-6 interaction is almost undetectable using proteins extracted from nuclei under low salt (200 mM NaCl) conditions (Fig. 7, lanes 5-8). Similarly, Pax-6 binding to the Ϫ270/Ϫ246 IGRP promoter sequence is reduced using ␤TC-3 nuclear ex-tract prepared from nuclei under low salt nuclear conditions (Fig. 7). These observations most likely reflect differences in the extraction efficiency of Pax-6. We were able to take advantage of this to demonstrate Pdx-1 binding to the Site 1 probe by using ␤TC-3 nuclear extract prepared from nuclei under low salt nuclear conditions (Fig. 7, lane 7, Site 1 panel). The inability to detect Pdx-1 binding to this probe using high salt extracts may be explained by a combination of both larger amounts of Pax-6 protein in these extracts and different affinities of Pdx-1 and Pax-6 for the Site 1 probe. In summary, these gel retardation experiments demonstrate that both Pdx-1 and Pax-6 can bind to four and five different sites, respectively, in the IGRP promoter. Importantly, we do not detect Pdx-1 binding to the Ϫ270/Ϫ246 IGRP promoter sequence under any condition tested (Fig. 7, Ϫ270/Ϫ246 panel).
Pax-6 and Pdx-1 Bind to the IGRP Promoter in Situ-Although the gel retardation experiments were able detect binding of Pax-6 and Pdx-1 to oligonucleotides representing IGRP promoter sequences in vitro, they do not address whether these interactions occur inside intact cells. This caveat was addressed by using chromatin immunoprecipitation (ChIP) as- FIG. 7. Pdx-1 and Pax-6 bind the Site 1, Site 2, Site 3, and Site 4 IGRP promoter regions. High or low salt ␤TC-3 cell nuclear extract was incubated in the absence (Ϫ) or presence of the indicated antisera for 10 min on ice prior to the addition of a labeled oligonucleotide probe representing the indicated IGRP promoter regions, and the incubation was continued for an additional 20 min at room temperature. Protein binding was then analyzed as described under "Experimental Procedures." In the representative autoradiograph shown only the retarded complexes are visible and not the free probe, which was present in excess. says (45). ␤TC-3 cells were first subjected to formaldehydemediated cross-linking to preserve DNA-protein interactions. Once fragmented by sonication, the chromatin isolated from these cells was subjected to immunoprecipitation using antisera to Pax-6 and Pdx-1, or IgG as a control. To detect the IGRP promoter in the resulting immunoprecipitates, purified genomic DNA isolated from the pellets was used as a template for PCR using primers representing the IGRP promoter sequence between Ϫ414 and Ϫ119. Fig. 8 shows the results of this PCR analysis and demonstrates that the IGRP promoter is enriched in the Pdx-1 and Pax-6 immunoprecipitates compared with the IgG control. To test the specificity of the antibodyprotein interactions, these immunoprecipitates were also analyzed for the presence of the PEPCK promoter, a target not predicted to associate with Pdx-1 or Pax-6, using PCR primers that amplify the PEPCK promoter sequence between Ϫ434 and Ϫ96. As expected little (Pdx-1) or no (Pax-6) enrichment of PEPCK promoter was detected in the Pdx-1 or Pax-6 immunoprecipitates compared with the IgG control (Fig. 8). The low level of enrichment of the PEPCK promoter in the Pdx-1 immunopellet compared with the IgG control probably reflects a nonspecific interaction of the Pdx-1 anti-serum with chromatin, because this level of enrichment is observed when other negative control targets are examined (data not shown). The results of these ChIP assays demonstrate that both Pdx-1 and Pax-6 interact specifically with the IGRP promoter inside intact cells.

Pdx-1 Binding to the IGRP Promoter Has Little Effect on IGRP Fusion
Gene Expression-To address the functional relevance of Pdx-1 and/or Pax-6 binding to Sites 1 through 4, a fusion gene, designated IGRP Ϫ306 Quad MUT, containing mutations in all four sites was constructed. These mutations, which disrupt binding of both proteins to these elements in vitro ( Fig. 6 and data not shown for Site 2), were introduced into an IGRP-CAT fusion gene by site-directed mutagenesis in the context of the Ϫ306 to ϩ3 IGRP promoter region. Although the mutation introduced into Site 1 only partially reduces Pax-6 binding (Fig. 6), it abolishes Pdx-1 binding (data not shown). The level of reporter gene expression directed by this fusion gene containing the quadruple mutation was analyzed by transient transfection of ␤TC-3 cells and compared with that directed by the wild-type Ϫ306 IGRP-CAT fusion gene (Fig. 9). Fig. 9 shows that, even with all four sites disrupted, there is no significant reduction in reporter gene expression, suggesting that Pdx-1/Pax-6 binding to these sites is not required for high IGRP fusion gene expression. One trivial explanation for these results could be that Pdx-1 is down-regulated in this particular cell line and thus the effects of disrupting these binding sites on IGRP fusion gene expression are blunted. To control for such a phenomenon we examined the effects of mutating the A3 element in the rat insulin II promoter. This element has previously been shown to interact with Pdx-1, and mutations in this site lead to a large (greater than 70%) reduction in promoter activity (27). Consistent with these previous observations, an insulin-CAT fusion gene containing a mutation of the A3 element that abolishes Pdx-1 binding directed greatly reduced reporter gene expression compared with the activity of the wild-type rat insulin II promoter (ϳ70% reduction in activity) (Fig. 9).
Loss of Pax-6 Binding to the Ϫ273/Ϫ246 Element Is Compensated by Pax-6 Binding to Site 1 and Vice Versa-The experiment described above shows that disruption of Pdx-1 and/or FIG. 8. The IGRP promoter binds Pdx-1 and Pax-6 in situ. Pdx-1 and Pax-6 binding to the IGRP promoter were analyzed in situ using the chromatin immunoprecipitation (ChIP) assay. Chromatin from formaldehyde-treated ␤TC-3 cells was immunoprecipitated using anti-Pdx-1 or anti-Pax-6 antibodies or, as a control, using IgG. The presence of the IGRP promoter and the PEPCK promoter in the chromatin preparation prior to immunoprecipitation (1:100 input) and in the immunoprecipitates was then assayed using PCR as described under "Experimental Procedures." Panel A shows a representative autoradiograph, whereas panel B shows the mean data Ϯ S.E. of three experiments. A Packard Instant Imager was used to count 32 P associated with the amplified promoter fragments. The IGRP and PEPCK promoter were amplified with similar efficiencies in the chromatin preparation (data not shown).
FIG. 9. Mutation of the four TAAT-like motifs in the IGRP promoter causes a minimal reduction in basal fusion gene expression. ␤TC-3 cells were transiently co-transfected, as described under "Experimental Procedures," using a LipofectAMINE solution containing various IGRP-CAT or Insulin-CAT fusion genes (2 g) and an expression vector encoding firefly luciferase (0.5 g). The IGRP-CAT fusion genes incorporated either the wild-type promoter sequence, located between Ϫ306 and ϩ3 (Ϫ306 WT), or contained a quadruple mutation of all four TAAT-like motifs (IGRP Ϫ306 Quad MUT) within the context of the Ϫ306 to ϩ3 promoter fragment. The Insulin-CAT fusion genes incorporated either the wild-type promoter sequence, located between Ϫ238 and ϩ2 (Insulin Ϫ238 WT), or contained a sitedirected mutation of the insulin A3 element (Insulin Ϫ238 A3 MUT), generated within the context of the Ϫ238 to ϩ2 promoter fragment. Following transfection, ␤TC-3 cells were incubated for 18 -20 h in serum-containing medium. Cells were then harvested, and CAT and firefly luciferase activity were assayed as previously described (2,30). Results are presented as the ratio of CAT:luciferase activity, expressed as a percentage relative to the value obtained with the wild-type IGRP and Insulin fusion genes, and represent the mean of three experiments Ϯ S.E., each using an independent preparation of all fusion gene plasmids.
Pax-6 binding to Sites 1 through 4 has a minimal effect on IGRP promoter activity. In contrast, it was demonstrated that disruption of Pax-6 binding to the Ϫ273/Ϫ246 element leads to a large reduction in IGRP promoter activity (Fig. 1). Unlike the mutations in Sites 1 through 4, which were introduced into the full-length promoter extending from Ϫ306 to ϩ3, the mutations in the Ϫ276/Ϫ246 element were originally introduced into an IGRP-CAT fusion gene, designated Ϫ273 IGRP-CAT, that is truncated at the 5Ј end. This truncated fusion gene therefore lacks the sequence that spans the Ϫ306 to Ϫ274 region, where Site 1 is located. This point became relevant when the effect of disrupting Pax-6 binding to the Ϫ273/Ϫ246 element in the context of the full-length promoter was assessed. Thus, a fusion gene, designated Ϫ306 Pax MUT A, was generated that contains the same mutation as present in the Ϫ273 MUT 1 fusion gene (Fig. 1), except that the mutation was introduced in the context of the full-length Ϫ306 IGRP-CAT fusion gene. As shown in Fig. 10, following transient transfection of ␤TC-3 cells, the Ϫ306 Pax MUT A fusion gene directed a similar level of CAT expression as directed by the wild-type Ϫ306 IGRP-CAT fusion gene. This result suggested that Pax-6 binding to Site 1 may compensate for the loss of Pax-6 binding to the Ϫ273/Ϫ246 element. The observation that deletion (17) or mutation ( Fig. 9) of the Ϫ306 to Ϫ274 IGRP promoter region has little effect on fusion gene expression suggests that the reverse also applies, namely that Pax-6 binding to the Ϫ273/Ϫ246 element may compensate for the loss of Pax-6 binding to Site 1.
Site 1 differs from Sites 2-4 in that sequence analysis suggests that it contains a very weak putative Pax-6 paired domain binding motif (Fig. 10A), thus explaining why the mutation introduced into the TAAT motif in Site 1 only partially reduces Pax-6 binding, whereas mutations introduced into the TAAT motif in Sites 2-4 completely abolish Pax-6 binding ( Fig.  6 and data not shown for Site 2). In addition, when using low FIG. 10. Loss of Pax-6 binding to the ؊273/؊246 element is compensated by Pax-6 binding to Site 1 and vice versa. A, increases (q) or decreases (E) in DMS methylation between in situ versus in vitro methylated mouse ␤TC-3 cell DNA over the Ϫ306 to Ϫ273 IGRP promoter region are shown. This information was derived from Ref. 17. B, ␤TC-3 cells were transiently co-transfected, as described under "Experimental Procedures," using a LipofectAMINE solution containing various IGRP-CAT fusion genes (2 g) and an expression vector encoding firefly luciferase (0.5 g). The IGRP-CAT fusion genes incorporated either the wild-type promoter sequence, located between Ϫ306 and ϩ3 (Ϫ306 WT), or contained mutations in the Pax-6 paired domain binding sites in the Ϫ273/Ϫ254 element (Ϫ306 Pax MUT A), Site 1 (Ϫ306 Pax MUT B), or both (Ϫ306 Pax MUT AB) all generated within the context of the Ϫ306 to ϩ3 promoter fragment. Following transfection, ␤TC-3 cells were incubated for 18 -20 h in serum-containing medium. Cells were then harvested, and CAT and firefly luciferase activity were assayed as previously described (2,30). Results are presented as the ratio of CAT:luciferase activity, expressed as a percentage relative to the value obtained with the wild-type Ϫ306 IGRP-CAT fusion gene, and represent the mean of three experiments Ϯ S.E., each using an independent preparation of all fusion gene plasmids. salt ␤TC-3 nuclear extract Pax-6 binding to the Site 2, 3, and 4 elements is no longer detected (Fig. 7). This suggests that Pax-6 interacts with higher affinity to the Site 1 than Site 2, 3, and 4 elements. As with the Ϫ273/Ϫ246 Pax-6 binding site, promoterscanning software (35) does not predict Pax-6 binding to Site 1.
To address the role of Pax-6 binding to Site 1 further, a mutation that targets the putative Pax-6 paired domain binding motif in Site 1 (Fig. 10A) was introduced in the context of the full-length, wild-type Ϫ306 IGRP-CAT and Ϫ306 Pax MUT A fusion genes, resulting in the generation of constructs designated Ϫ306 Pax MUT B and Ϫ306 Pax MUT AB, respectively (Fig. 10B). This paired domain binding motif mutation partially disrupts Pax-6 binding to Site 1 leaving Pdx-1 binding intact (data not shown). Consistent with the hypothesis that both Pax-6 binding elements in the IGRP promoter compensate for each other, this Site 1 mutation has no effect by itself as assessed by comparing the level of CAT expression directed by the wild-type Ϫ306 IGRP-CAT and Ϫ306 Pax MUT B fusion genes (Fig. 10). However, when coupled with the mutation of the Ϫ273/Ϫ254 Pax-6 binding motif, fusion gene expression was reduced by ϳ50% (Fig. 10). Thus, loss of Pax-6 binding at either site individually is compensated for by Pax-6 interaction at the reciprocal site. This result explains why disruption of Pax-6 binding to the Ϫ273/Ϫ246 element has a significant impact on promoter activity in the context of a truncated IGRP promoter lacking Site 1, but not in the context of the full-length IGRP promoter. Site-directed mutagenesis of both Pax-6 paired domain binding motifs, as present in the Ϫ306 Pax MUT AB construct, reduces fusion gene expression to a lesser degree than deletion of both elements, as occurs in the Ϫ254 5Ј deletion construct (compare Figs. 1 and 10). We hypothesize that this difference is explained by residual Pax-6 homeodomain binding to the TAAT-like motif in Site 1, consistent with the observation that mutation of the Pax-6 paired domain binding motif in Site 1 only partially disrupts Pax-6 binding (data not shown).

DISCUSSION
The results of this study provide insight into the molecular mechanisms controlling IGRP gene transcription. A role for Pax-6 was demonstrated based on data derived from in vitro and in situ binding studies, coupled with functional studies utilizing promoter fusion genes. Although Pax-6 is capable of interacting with at least five separate elements in the IGRP promoter in vitro (Fig. 7), only two of these, located in the Ϫ306/Ϫ273 (Site 1) and Ϫ273/Ϫ246 regions, appear critical for promoter activity (Figs. 1, 9, and 10). In fact, it was the marked loss in promoter activity observed when the Ϫ273/Ϫ246 IGRP promoter region was removed in the original 5Ј deletion analysis (17) that directed our attention to this element and eventually led to the identification of Pax-6 as the factor binding this region. However, this reduction in promoter activity was not detected when this element is mutated using site-directed mutagenesis, because Pax-6 binding to Site 1 can compensate for the loss of Pax-6 binding to the Ϫ273/Ϫ246 element (Fig.  10B). A key result that led to the identification of Pax-6 as a candidate for the factor binding the Ϫ273/Ϫ246 element was the use of a modified in situ footprinting method, the results of which suggested that a single factor may make unusually broad contacts across this promoter region (Fig. 2). We have previously demonstrated the utility of this methodology by showing that multiple trans-acting factor binding sites identified in the IGRP promoter in ␤TC-3 cells by in situ footprinting correlate with regions of the IGRP promoter identified as being important for basal IGRP-CAT fusion gene expression in ␤TC-3 cells (17).
As an alternative method for identifying the factors control-ling IGRP expression, we also took advantage of the accumulating literature regarding the transcription factors controlling the expression of other ␤ cell-specific or -enriched proteins, such as insulin, to identify candidate proteins. Our efforts here were focused on the transcription factor Pdx-1, which was shown to interact with the IGRP promoter at four separate elements in vitro together with Pax-6 ( Fig. 7). Both of these proteins bind to TAAT or TAAT-like motifs within these elements ( Fig. 6 and data not shown), presumably through their respective homeodomains. It is interesting that this dependence of binding on the presence of the TAAT sequence is not absolute, as reflected in the observation that Pdx-1 and Pax-6 were able to interact with the element designated Site 2, which contains the sequence TAAA (Table I). This observation suggests that the homeodomains in these proteins exhibit rather loose specificity in vitro. Because similar binding characteristics have been observed for other homeoproteins (44,46,47), the question arises: how do these proteins achieve target specificity in vivo? One route through which this could occur is by cooperative binding with other transcription factors. With regard to Pdx-1, such a model is consistent with work showing that this factor can interact with the transcription factor Neu-roD/BETA2 to support cooperative binding to the A3 and E2 elements, respectively, in the rat insulin I promoter (48). Based on data derived from site-directed mutagenesis ( Fig.  9), it appears that Pdx-1 does not contribute significantly to IGRP promoter activity. Thus, it would seem that speculation about potential Pdx-1 interacting proteins with regard to IGRP promoter regulation is unwarranted. It is possible, however, that an interaction between Pdx-1 and another transcription factor could allow this protein to interact with the promoter in the absence of a binding site, thus masking the effects of site-directed mutagenesis. Consistent with this idea is the fact that Pdx-1 is associated with the IGRP promoter in situ, which suggests it plays a role in regulating IGRP expression (Fig. 8). Alternatively, the simpler explanation, that Pdx-1 is not critical for maximum IGRP expression, is consistent with studies on glucokinase gene expression. Thus, although Pdx-1 is able to interact with the glucokinase promoter in vitro (49), no alteration in glucokinase protein is observed in mice heterozygous for a Pdx-1 null allele or in animals whose ␤ cells completely lack Pdx-1 (50,51). Measurements of IGRP expression levels in the absence of Pdx-1 may definitively answer this question.
In addition to its role in regulating basal gene transcription, Pdx-1 also appears to contribute to the modulation of gene expression by glucose in ␤ cells. Transcription of the insulin gene, for example, is induced when ␤ cells are exposed to elevated glucose concentrations, and studies aimed at identifying the transcription factors responsible for this effect have shown that disruption of Pdx-1 binding to the insulin promoter impairs the glucose response (52). Consistent with the idea that Pdx-1 is a glucose-responsive factor, it has been demonstrated that this transcription factor translocates from the cytoplasm, or nuclear periphery, to the nucleus in response to elevated glucose levels (53,54). In light of these data, it seems reasonable to hypothesize that, although we are not able to ascribe a major role for Pdx-1 in the maintenance of basal IGRP promoter activity, this transcription factor may play a role in the regulation of the IGRP gene by metabolites such as glucose.
Similarly, in addition to its role in regulating basal gene transcription, Pax-6 also appears to contribute to the regulation of glucagon expression by insulin in ␣ cells. Pax-6 binds the G3 element in the glucagon promoter (55), an element that is critical for the regulation of glucagon gene transcription by insulin (56). However, although this element is sufficient to confer insulin-regulated reporter gene expression in a heterol-ogous context (56,57), it is apparent that other promoter elements contribute to the regulation of glucagon gene transcription by insulin (57). Thus, although Pax-6 binds the promoters of several genes whose expressions are islet-enriched (58), it is possible that Pax-6 may only mediate an insulin response in specific promoter contexts. Future studies investigating the possible effects of glucose and insulin on IGRP promoter activity will address these issues.
Because IGRP is only one of a small group of proteins known to be highly enriched in pancreatic ␤ cells, it offers a relatively unique opportunity to study the molecular basis for selective gene expression within this cell type. Our future goals include identifying the additional transcription factors required for the expression of this gene with an emphasis on novel transcription factors, not previously implicated in ␤ cell-specific gene expression. The identification of such factors is important because of their potential relevance to both pancreatic/␤ cell development and efforts to produce ␤ cells from stem cells in situ (59). In addition, knowledge of these transcription factors may be useful to the design of future therapies that aim to suppress IGRP gene expression in the event that IGRP, as in the NOD mouse, is a major autoantigen in human type I diabetes.