Transcription Factors Recognizing Overlapping C1-A2 Binding Sites Positively Regulate Insulin Gene Expression*

Transcription factors binding the insulin enhancer region, RIPE3b, mediate β-cell type-specific and glucose-responsive expression of the insulin gene. Earlier studies demonstrate that activator present in the β-cell-specific RIPE3b1-binding complex is critical for these actions. The DNA binding activity of the RIPE3b1 activator is induced in response to glucose stimulation and is inhibited under glucotoxic conditions. The C1 element within the RIPE3b region has been implicated as the binding site for RIPE3b1 activator. The RIPE3b region also contains an additional element, A2, which shares homology with the A elements in the insulin enhancer. Transcription factors (PDX-1 and HNF-1α) binding to A elements are critical regulators of insulin gene expression and/or pancreatic development. Hence, to understand the roles of C1 and A2 elements in regulating insulin gene expression, we have systematically mutated the RIPE3b region and analyzed the effect of these mutations on gene expression. Our results demonstrate that both C1 and A2 elements together constitute the binding site for the RIPE3b1 activator. In addition to C1-A2 (RIPE3b) binding complexes, three binding complexes that specifically recognize A2 elements are found in nuclear extracts from insulinoma cell lines; the A2.2 complex is detected only in insulin-producing cell lines. Furthermore, two base pairs in the A2 element were critical for binding of both RIPE3b1 and A2.2 activators. Transient transfection results indicate that both C1-A2 and A2-specific binding activators cooperatively activate insulin gene expression. In addition, RIPE3b1- and A2-specific activators respond differently to glucose, suggesting that their overlapping binding specificity and functional cooperation may play an important role in regulating insulin gene expression.

In adult mammals, the insulin gene is expressed only in the pancreatic ␤-cells in the islets of Langerhans. In studies using transgenic animals and transient transfection analysis, the proximal 5Ј-flanking region of the insulin promoter was shown sufficient for directing ␤-cell-specific expression of the insulin gene (1)(2)(3)(4)(5)(6)(7). Further, mutational analysis of the promoter proximal region identified several cis-acting enhancer elements that are important for insulin expression. The insulin enhancer elements, with the exception of E-box elements, have been classified based on the nucleotide sequence of the element (8). Most of these enhancer elements are well conserved in various species, suggesting the presence of a common regulatory mechanism(s) controlling insulin expression.
Three conserved insulin enhancer elements, A3 (Ϫ201 to Ϫ196 bp) 1 (9 -13), RIPE3b/C1-A2 (Ϫ126 to Ϫ101 bp) (14), and E1 (Ϫ100 to Ϫ91 bp) (4,15,16), play an important role in regulating cell-specific expression of the insulin gene. Factors binding these sites have a very limited cellular distribution. Transcription factors that bind and activate expression from two of the three conserved insulin enhancer elements (A3 and E1) have been cloned. PDX-1, a member of the homeodomain family of transcription factor expressed in cells of the pancreas and duodenum, binds the A3 element (17)(18)(19)(20)(21)(22). Heterodimers of ubiquitously distributed basic helix-loop-helix family members (E2A and HEB (23,24)) and the cell type enriched basic helixloop-helix member (BETA2 (25,26)) bind the E1 element. BE-TA2 is expressed in all pancreatic endocrine cell types, in some intestinal endocrine cells, and in the brain (25)(26)(27). Limited but not identical cellular distribution of transcription factors PDX-1 and BETA2 suggests that cell-specific insulin gene expression may result because of the presence of a unique combination of enhancer element binding factors in the pancreatic ␤-cells. Interestingly, one of the RIPE3b element-binding complexes is expressed only in insulin-producing cells (14,28) and may play a critical role in regulating the cell type specificity of insulin gene expression.
The RIPE3b binding activity has been extensively characterized in nuclear extracts from both insulin-producing and noninsulin-producing cell lines (14,(27)(28)(29)(30)(31). Two specific RIPE3bbinding complexes have been identified: 1) a cell-specific complex, RIPE3b1, which is detected only in pancreatic ␤-cell lines, and 2) the RIPE3b2-binding complex that is detected in nuclear extracts from all cell lines examined to date. The importance of the factor(s) binding to the RIPE3b element in mediating cell type-specific insulin gene expression is further emphasized from transgenic mice studies (5). It was demonstrated that an enhancer construct containing both RIPE3b and E1 elements, RIPE3, could correctly regulate temporal and spatial expression of the transgene in vivo (5), whereas a construct containing the E1 element alone was unable to induce expression in transgenic animals (32). In all RIPE3 transgenic lines, the transgene (growth hormone) expression was detected in ␤-cells, whereas in other lines, expression of growth hormone was also noted in ␣-cells (5).
In addition to regulating cell type-specific expression, factors binding to the enhancer elements, A3, E1, and RIPE3b, also regulate glucose-mediated alterations in insulin gene expression (30,31,(33)(34)(35)(36). Our earlier studies demonstrated that the binding activity of the cell-specific RIPE3b1 activator play a key role in regulating this glucose responsiveness (30,31).
Although the binding activity of the RIPE3b1 activator is induced in response to acute change in glucose concentration, under glucotoxic conditions, as observed in HIT T-15 cells cultured chronically in the presence of high concentrations of glucose, the RIPE3b1 and PDX-1 binding activities were inhibited (37)(38)(39). PDX-1 protein and mRNA levels were also inhibited in rat islets, following development of diabetes after partial pancreatectomy (40). These observations suggest that in vitro and in vivo glucotoxic conditions regulate the levels and activity of insulin gene transcription factors. Interestingly, the binding activity of the RIPE3b1 factor, but not of PDX-1, is inhibited under glucotoxic conditions in the insulinoma cell line ␤TC-6 (41). This suggests that the inhibition of the RIPE3b1 activator binding may be the primary or initial defect under glucotoxic conditions. The cell type-specific and glucose-responsive transcription factors of the insulin gene also play an important role in pancreatic development and differentiation of ␤-cells. Mice homozygous for the null mutation in BETA2/NeuroD have a striking reduction in the number of ␤-cells and fail to develop mature islets, develop severe diabetes, and die perinatally (26). The pdx-1 knockout mice have a more profound phenotype, being apancreatic; these animals also develop extreme hyperglycemia and die perinatally (20,21,42). During embryonic development, PDX-1 is also expressed in exocrine and ductal epithelial cells, whereas in the adult pancreas expression is predominantly restricted to ␤-cells (20 -22, 43). However, expression of PDX-1 is induced in the ductal epithelium during pancreatic regeneration in adult animals (44).
Similar to the PDX-1 and BETA2 factors, the RIPE3b1 activator, is an important regulator of cell type-specific and glucose-responsive expression of the insulin gene and may play an important role in pancreatic development and/or ␤-cell differentiation. The C1 element (Fig. 1) in the RIPE3b region has been implicated as the binding site for the RIPE3b1 activator (14,28,31). Factors binding the C1 element were shown to functionally co-operate with E1 binding factors (BETA2-E12/ E47) in regulating insulin gene expression (14,25,31,45). In addition to C1, the RIPE3b region also contains an A element, A2 ( Fig. 1 and Refs. 8, 46, and 47). The homeodomain family of transcription factors (PDX-1, HNF1 ␣, and Isl-1) that are important for pancreatic development bind insulin A elements (48 -51). Furthermore, factors binding to A elements are implicated in synergistically interacting with other insulin gene transcription factors and regulating gene expression (52)(53)(54).
Hence, in this study, we have further characterized the RIPE3b region to better understand the role of C1 and A2 elements in regulating insulin gene expression. We demonstrate that the presence of both C1 and A2 elements are essential for binding of the RIPE3b1 activator. We also identified factors that require the A2 element but not the C1 element for binding. Interestingly, one of the A2 element-binding complexes, A2.2, is selectively expressed in insulin-producing cell lines. We further demonstrate that two base pairs in the A2 element are critical for the formation of both ␤-cell-specific DNA-binding complexes, RIPE3b1 and A2.2. Results from transient transfection analysis indicate that C1-A2 (RIPE3b1) and A2-specific factors positively regulate insulin gene expression. A mutation that prevents binding of both activators shows greater inhibition of insulin gene expression than a mutation that prevents binding of either of these factors alone. Based on these results we suggest that the presence of two ␤-cell-specific factors with overlapping DNA-binding sites may provide an important means of regulating insulin gene expression in response to various metabolic/environmental signals.

EXPERIMENTAL PROCEDURES
Tissue Culture-The HIT T-15 cell line was obtained from American Type Culture Collection (Manassas, VA), and all experiments were conducted with cells between passage numbers 68 and 80. ␤TC-3 and ␤TC⅐TET cell lines have been described earlier (55,56) and were provided by Dr. Shimon Efrat. Non-insulin-producing cell lines used in this study are: HeLa cells (human cervical carcinoma), baby hamster kidney cells, and IM Duct (immorto Ϫmouse ductal cell line, provided by Dr. Susan Bonner-Weir, Boston, MA). Cells were maintained in Dulbecco's modified Eagle's medium supplemented with Ham's F-12 (22 mM glucose), 10% fetal bovine serum, penicillin, and streptomycin. For the preparation of medium with different glucose concentration, glucosefree Dulbecco's modified Eagle's medium and dialyzed fetal bovine serum (Life Technolgies Inc.) were used. Glucose levels in the medium were adjusted to the desired concentration by addition of filter sterilized glucose.
Extract Preparation-HIT T-15 nuclear extracts, prepared using a large scale nuclear extract preparation protocol (57), were generously provided by Dr. Larry Moss (Tufts University, Boston, MA). Other insulin-producing and non-insulin-producing cell lines were cultured for 4 -6 days, and nuclear extracts were prepared as described (58). To prepare extracts from HIT T-15 cells grown in the presence of different concentrations of glucose, cells were cultured for 4 days in Dulbecco's modified Eagle's medium/Ham's F-12 medium. Medium was then removed, cells were washed twice with phosphate-buffered saline, and medium containing the desired concentration of glucose was added. Cells were harvested after 48 h, and nuclear extracts were prepared (58).
Electrophoretic Mobility Shift Assays-Oligonucleotides used as probes and competitors (Table I) were synthesized either by Sigma-Genosys (The Woodlands, TX) or by the DNA core facility at the Joslin Diabetes Center. In addition to the oligonucleotides listed in Table I, a series of two base pairs substitution (A 7 C and G 7 T) mutations in the RIPE3b (Ϫ126 to Ϫ101 bp) oligonucleotide were synthesized and named according to the two mutated nucleotides (e.g. Ϫ126⅐125m oligonucleotide is mutated at positions Ϫ126 and Ϫ125 bp in the RIPE3b oligonucleotide). Double-stranded oligonucleotides were radiolabeled with [␣-32 P]dCTP and the Klenow fragment of DNA polymerase I and used as probes. Binding reaction conditions for the RIPE3b, Ϫ139 to Ϫ113 and thyroid transcription factor-1 probes were identical (10 mM Tris, pH 7.8, 150 mM NaCl, 2 mM EDTA, 8 mM dithiothreitol, 0.4 mg of poly(dI-dC), 5% glycerol, 50 -100 fmol of probe, and 3-5 g of nuclear extracts). Binding reaction conditions for A3 (FLAT-E) and major late transcription factor probes have been described (31,37). Competition experiments were performed by simultaneous addition of radiolabeled probe and excess unlabeled competitors to the binding reaction.
The anti-IDX-1 (PDX-1) antibody was generously provided by Dr. Joel Habener (Massachusetts General Hospital, Boston, MA), whereas monoclonal anti-Nkx2.2 antibody was purchased from the Developmental Studies Hybridoma Bank at the University of Iowa (Iowa City, IA). For antibody supershift experiments, antibodies were preincubated with the nuclear extract for 20 min at room temperature, followed by another 20-min incubation in the presence of the radiolabeled probe. Binding reactions were then loaded onto a 6% nondenaturing polyacrylamide gel and were run in Tris-glycine-EDTA buffer (31). After completion of the run, gels were dried and scanned on a Molecular Dynam- ics PhosphorImager (Sunnyvale, CA), and bands were quantitated using Imagequant software.
DNA Constructs-The insulin reporter construct-238 WT LUC has been described earlier (31). The QuickChange site-directed mutagenesis kit (Stratagene, La Jolla, CA) was used to construct additional plasmids with specific mutations in the insulin enhancer. Doublestranded oligonucleotides containing desired mutations (underlined nucleotides) were synthesized to construct Ϫ125⅐124m LUC (Ϫ138 CA GGC AAG TGT TTT TAA ACT GCA GCT TCA GCC CCT CTG G, Ϫ100 bp), Ϫ122⅐121m LUC (Ϫ135 GC AAG TGT TTG GAC CCT GCA GCT TCA GCC CCT CTG G, Ϫ100 bp), and Ϫ110⅐109m LUC (Ϫ135 GC AAG TGT TTG GAA ACT GCA GCT TCC TCC CCT CTG GCC, Ϫ98 bp). The Ϫ238 WT LUC plasmid was used as the template with these oligonucleotides to construct reporter plasmids with mutations at the desired positions in the insulin enhancer. Plasmids were sequenced at the DNA core facility at Joslin Diabetes Center to confirm the construction of each mutant plasmid. To avoid the effect of nonspecific mutations outside the insulin enhancer region on the reporter gene expression, the plasmids were digested with restriction enzymes to release the mutated insulin enhancer fragments. These fragments were then used to replace the wild type insulin enhancer from the Ϫ238 WT LUC construct, resulting in the generation of reporter plasmids that differed only at the specific mutations in the insulin enhancer.
Transient Transfections and Luciferase Assays-Approximately 18 h before transfection, 5 ϫ 10 5 HIT T-15 or ␤TC-3 cells were plated onto 60-mm 2 plates. Reporter plasmids (2.5 or 2.0 g for HIT T-15 or ␤TC-3 cell line, respectively) were co-transfected with 0.1 g of a pRL-Null (Promega, Madison, WI) plasmid, as an internal control. Although the pRL-Null plasmid lacks any of the eukaryotic enhancer elements, significant levels of constitutive Renilla luciferase expression have been observed from this construct (59). We also observed a linear relationship between the reporter gene expression and the amount of the pRL-Null plasmid used (data not shown). Plasmid DNA were transfected using LipofectAMINE (HIT T-15 cells, DNA:LipofectAMINE ratio of 1:4) or LipofectAMINE PLUS (␤TC-3 cells, DNA:LipofectAMINE PLUS ratio of 1:5) as descried earlier (31). To determine reporter gene expression a commercial Dual Luciferase Kit (Promega) was used. Whole cell extracts were prepared 48 h after transfection, using passive lysis buffer. Extracts were used to determine firefly (reporter plasmids) and Renilla (internal control) luciferase expression using the Moonlight 3010 Luminometer (Analytical Luminescence Laboratory, Sparks, MD). The ratio of the firefly:Renilla luciferase activity represents the normalized reporter gene expression. Transfection experiments were repeated several times, with at least two or three different plasmid preparations.

C1 and A2 Elements Together Constitute the Binding Site for the RIPE3b1 and RIPE3b2 Factors-Two sequencespecific
RIPE3b (Ϫ126 to Ϫ101 bp) binding complexes have been identified. The RIPE3b1 complex is detected in insulin-producing cells, whereas the RIPE3b2 complex is detected in nuclear extracts from all cell lines examined to date (14,(27)(28)(29)(30)(31). Mutating two nucleotides (Ϫ112 and Ϫ111, a TC to CG substitution) in the RIPE3b probe prevented the formation of both the RIPE3b1 and the RIPE3b2 complexes (14,28,31). Methylation interference analysis indicated that the RIPE3b1 complex exerted a strong interference at Gs Ϫ107, Ϫ108, and Ϫ111 and weak interference at G Ϫ114 on the bottom strand, suggesting that the binding site may span from Ϫ114 to Ϫ107 bp (14). Based on these observations and other results, it was suggested that the RIPE3b region may contain two elements, C1 (Ϫ116 to Ϫ107) (for the binding of RIPE3b1 and RIPE3b2 factors) and A2 element (Ϫ126 to Ϫ113) (8,46,47). The C1 and A2 elements share a 4-bp segment (Ϫ116 to Ϫ113 bp) that contains the G Ϫ114 (bottom strand) implicated in the binding of the RIPE3b1 factor. Hence, to define the binding site for the RIPE3b1 activator as well as to characterize the role of the A2 element in this binding, we made a series of 2-bp substitution mutations in the RIPE3b oligonucleotide and studied the effect of these mutations on the RIPE3b1 binding.
Electrophoretic mobility shift assays (EMSA) were performed using radiolabeled RIPE3b (Ϫ126 to Ϫ101) probe and HIT T-15 nuclear extracts, in the presence or absence of 50-fold excess of unlabeled competitors ( Fig. 2A). As expected, a 50-fold excess of the previously described unlabeled mutant RIPE3b oligonucleotide (having TC to CG substitution at Ϫ112 and Ϫ111 bp) did not compete for binding to either RIPE3b1 or RIPE3b2 factors, whereas the unlabeled wild type RIPE3b oligonucleotide successfully competed for binding of these factors. Competition between the RIPE3b probe and a series of RIPE3b oligonucleotides with mutations in two consecutive nucleotides showed interesting binding profiles for the RIPE3b1 and RIPE3b2 factors ( Fig. 2A). Mutations in the nucleotides from Ϫ112 to Ϫ107 bp (C1 region) inhibited the ability of excess unlabeled competitors to compete as effectively as the wild type oligonucleotide, confirming the importance of this region for binding of these factors. Interestingly, the oligonucleotides with mutations in the A2 element (Ϫ124 to Ϫ117 bp) were also ineffective competitors implying a role of A2 element in the binding of the RIPE3b1 and RIPE3b2 factors. I This list shows the nucleotide sequences corresponding to the top strand of wild type and mutant oligonucleotides used in EMSA. Underlined nucleotides represents sites of mutated base pairs in the wild type oligonucleotides. A series of two base pair substitution (A N C and G N T) mutations in the RIPE3b oligonucleotide that were named according to two mutated nucleotides is not included in the table. The position of A3 and MLTF (31,37) and TTF-1 (60) oligonucleotides in the corresponding genes has been described. Oligonucleotides Ϫ150 to Ϫ125 (A2) and Ϫ148 to Ϫ122 (A2) represent rat insulin I and human insulin sequence homologous to rat insulin II Ϫ139 to Ϫ113 sequence.

TAG GTG TAG GCC ACG TGA CCG GGT GTT C
Substitution mutations in the four nucleotides Ϫ116 to Ϫ113 bp (residing between these two regions) had no effect on their ability to compete as effectively as the wild type oligonucleotide. These results indicate that binding of the RIPE3b1 and the RIPE3b2 complexes require sequences present in both the C1 and A2 elements. Our results also suggest that nucleotides at Ϫ110 to Ϫ109 bp and Ϫ122 to Ϫ121 bp are most critical for the binding of these factors (Fig. 2B). Because two distinct elements, C1 and A2, in the RIPE3b region are required for binding of the RIPE3b1 and the RIPE3b2 factors, we next asked whether the exact spacing between these elements was essential for the binding activity. Competitor oligonucleotides were synthesized with an insertion of two nucleotides or a deletion of two or five nucleotides between the C1 and A2 elements. As shown in Fig. 2C, altering the distance between the two elements prevented these oligonucleotides from competing for binding to these factors. Results presented in Fig. 2 suggest that the RIPE3b1 and RIPE3b2 factors bind a large (Ϫ107 to Ϫ124 bp) DNA-binding element of the insulin enhancer region. The RIPE3b binding activity not only requires the nucleotide sequence present in both the C1 and A2 elements but also requires the exact distance between these two elements. Hence, we suggest that RIPE3b1 and RIPE3b2 factors bind a composite C1-A2 element.
A ␤-Cell-specific Factor Selectively Binds the A2 Element-Because the RIPE3b1 factor binds the C1-A2 element, we were interested in identifying factors that selectively bind either the C1 or A2 elements. Factors that selectively bind one of these elements could have a potential regulatory role in modulating the DNA binding and transcriptional activation mediated by the RIPE3b1 activator. We consistently observed only two specific RIPE3b (Ϫ126 to Ϫ101 bp) binding complexes when nuclear extracts from insulin-producing cell lines were used. When the C1 element-specific probe (Ϫ116 to Ϫ101 bp) was used in EMSA with HIT T-15 nuclear extracts, only nonspecific DNA-binding complexes were detected (data not shown). These observations indicate that insulin-producing cells may not contain a C1 element-specific binding factor.
To identify factors that selectively bind the A2 element, we designed a new probe, Ϫ139 to Ϫ113 bp, that included the A2 element and additional sequences upstream of the RIPE3b region (Fig. 3A). In EMSA using radiolabeled Ϫ139 to Ϫ113 bp probe and HIT T-15 nuclear extracts, we detected three binding complexes with gel mobilities distinct from both the RIPE3b1 and RIPE3b2 complexes (Fig. 3B). Competition analysis indicated that these complexes bind specifically to the Ϫ139 to Ϫ113 bp region of the insulin enhancer. Formation of A2 complexes was not inhibited in the presence of excess unlabeled RIPE3b, C1, A2 (Ϫ126 to Ϫ113) or sequences upstream of the RIPE3b region (Ϫ139 to Ϫ127) (Fig. 3). These observations suggest that the formation of A2 complexes requires the rat insulin II enhancer region from Ϫ139 to Ϫ127 bp, in addition to sequence spanning A2 element (Ϫ126 to Ϫ113). Furthermore, these results demonstrate that the A2-specific complexes selectively bind to the Ϫ139 to Ϫ113 region of the insulin enhancer, independent of the presence of the C1 element. Results presented in Fig. 3 also demonstrate that formation of the RIPE3b1 and RIPE3b2 complexes is not inhibited in the presence of excess unlabeled C1, A2, or Ϫ139 to Ϫ113 regions of the insulin enhancer. Taken together, these observations demonstrate the presence of factors with overlapping C1-A2 and A2-specific DNA binding specificity in nuclear extracts from insulin-producing cells.
The region corresponding to the Ϫ139 to Ϫ113 bp of the rat insulin II gene is highly conserved in the rat insulin I and the human insulin genes (Fig. 4A). Competition analysis was performed to confirm that the A2 factors that bind the rat insulin II Ϫ139 to Ϫ113 probe would successfully bind the corresponding region from rat insulin I and the human insulin gene (Fig.  4B). EMSA were performed with the rat insulin II Ϫ139 to Ϫ113 probe and HIT T-15 nuclear extracts in the absence or presence of increasing concentrations of unlabeled excess Ϫ139 to Ϫ113 bp oligonucleotide or the corresponding region from the rat insulin I and human insulin gene and the RIPE3b oligonucleotide. Formation of A2-specific complexes was inhibited by excess unlabeled Ϫ139 to Ϫ113 oligonucleotide and its homologous regions from rat I and human insulin gene but not by the RIPE3b oligonucleotide. These observations suggest that the DNA-binding site recognized by A2-specific factors is highly conserved.
Because tissue-or cell type-specific transcription factors are major regulators of critical genes in a given cell or tissue, we next determined whether A2-specific factors were selectively expressed in insulin-producing cells. Nuclear extracts from insulin-producing and non-insulin-producing cell lines were analyzed for their ability to form A2-specific complexes. Nuclear extracts from HIT T-15, ␤TC-3, and ␤TC⅐TET (insulinproducing cell lines) and baby hamster kidney, HeLa, and a pancreatic ductal cell line IM Duct (non-insulin-producing cells) were incubated with radiolabeled Ϫ139 to Ϫ113 probe from rat insulin II gene and analyzed by EMSA. As shown in Fig. 5, the A2.1 binding activity was detected in all cell lines, whereas the A2.2-binding complex was formed only with the nuclear extracts from insulin-producing cell lines. These results demonstrate that the A2.2 factor is selectively expressed in the insulin-producing cells. In addition, the A2.2 factor binds a conserved region in the insulin enhancer that overlaps with the binding site for RIPE3b1. The presence of two distinct ␤-cell-specific DNA-binding factors with overlapping DNAbinding sites suggest an important role for these factors in regulating insulin expression.

A2-specific Factors Are Distinct from PDX-1 and Nkx2.2-
The A2 element has been classified based on its homology to other A elements. The homeodomain family of transcription factors, such as PDX-1, bind these A elements (48 -51). In addition to homology with the A elements, the Ϫ139 to Ϫ113 bp region of rat insulin II also contains a consensus-binding element (CAAGTG) for the Nkx2 family of transcription factors (60 -62). Nkx2.2, a member of Nkx2 family, is important for differentiation of pancreatic ␤-cells (63). Hence, we performed competition analysis and antibody supershift assay to determine whether the PDX-1 and Nkx2.2 are present in the A2specific complexes.
Nuclear extracts from HIT T-15 cells were incubated with radiolabeled A3 or Ϫ139 to Ϫ113 probes, in the absence or presence of excess cold A3 and Ϫ139 to Ϫ113 bp oligonucleotide. In addition, anti-PDX-1 antibody was added to binding reactions containing the A3 and Ϫ139 to Ϫ113 probes, and the reactions were analyzed by EMSA. As shown in Fig. 6A, formation of the A2-binding complex is not inhibited by the presence of excess unlabeled A3 oligonucleotide. In addition, the anti-PDX-1 antibody neither prevented the formation of the A2-specific complexes nor altered their mobility, suggesting that the A2-specific complex does not contain PDX-1. These observations are consistent with the converse experiment, where presence of the excess unlabeled Ϫ139 to Ϫ113 oligonucleotide did not prevent formation of PDX-1 or other A3-specific complexes (Fig. 6A). We used a similar strategy to identify the

FIG. 4. The A2 element is conserved within human (hIns), rat I (rIns I), and rat II (rIns II) insulin genes. A, DNA sequence from
Ϫ139 to Ϫ113 bp region of rat insulin II enhancer and corresponding sequence from rat insulin I and human insulin genes are shown. The asterisk denotes nucleotides that are identical within these sequences, and ϩ or Ϫ symbol represent conserved or distinct nucleotide substitutions, respectively. B, the ability of the A2 elements from rat insulin I and human insulin gene to compete with rat insulin II A2 element binding activity in HIT T-15 nuclear extract was analyzed. HIT T-15 extract was incubated with rat insulin II Ϫ139 to Ϫ113 bp probe in the absence or presence of excess cold Ϫ139 to Ϫ113 bp oligonucleotide from rat II or corresponding region from rat I or human insulin genes or rat II RIPE3b oligonucleotide as competitors, and reactions were analyzed by EMSA. Positions of A2-specific complexes A2.1, A2.2, and A2.3 are indicated.
FIG. 5. The A2.2-binding complex is detected only in insulinproducing cell lines. Nuclear extracts from insulin-producing (HIT T-15, ␤TC-3, and ␤TC⅐TET) and non-insulin-producing (baby hamster kidney, HeLa, and IM Duct) cell lines were incubated with radiolabeled rat insulin II Ϫ139 to Ϫ113 probe. Binding reactions were analyzed by EMSA and positions of A2-specific complexes are indicated. A2.2 complex (asterisk) was formed only in the presence of nuclear extract from insulin-producing cell lines has been indicated.
presence of Nkx2.2 in A2-specific complexes. Because the Nkx2.2 binding site is not known, we used the binding element from the rat thyroglobulin gene for the thyroid transcription factor-1, a Nk2 family member (60). As shown in Fig. 6B, Nkx2.2 is not part of the A2-specific complexes. Additionally, oligonucleotides with mutation in the Nkx consensus sequence at positions Ϫ133 to Ϫ132 (Ϫ133⅐132m) competed as effectively as the wild type Ϫ139 to Ϫ113 oligonucleotide, suggesting this sequence was not required for binding of A2-specific factors. These results support the conclusion that the A2 binding factors are distinct from other homeodomain proteins and the Nkx2 family of transcription factors. These important observations suggest that the ␤-cell-specific A2.2 activator may represent a novel factor.
The Binding Activity of A2-specific Factors Is Not Regulated by Glucose-DNA binding activity of the RIPE3b1 activator is induced in response to an increase in glucose concentrations (28,30,31). This increased binding of the RIPE3b1 activator to the insulin enhancer may be responsible for the observed induction of insulin gene expression. To test whether glucose can stimulate binding activity of A2-specific factors, nuclear extracts were prepared from HIT T-15 cells grown in the presence of different concentrations of glucose. Equal concentrations of nuclear extracts from HIT T-15 cells grown in 0.2, 1.0, 5.0, and 22.0 mM glucose were incubated with RIPE3b and Ϫ139 to Ϫ113 probes, and binding reactions were analyzed by EMSA (Fig. 7). As a control, equal concentration of HIT T15 nuclear extracts were also analyzed for binding to Adenoviral major late transcription factor (data not shown). As we showed earlier (31), increasing concentrations of glucose significantly induced the binding of the RIPE3b1 activator ( Fig. 7) but had no effect on the binding activity of ubiquitously expressed major late transcription factor (data not shown). However, when the same experiment was performed with Ϫ139 to Ϫ113 probe under identical conditions, there was no difference in the binding activity of any A2-specific activators in response to alteration in glucose concentrations. These results demonstrate that glucose differentially regulates binding activity of the RIPE3band A2-specific activators.
Nucleotides at Positions Ϫ122 and Ϫ121 bp Are Essential for the Binding of Both RIPE3b (C1-A2) and A2-specific Activators-An oligonucleotide probe spanning the binding region of both RIPE3b-and A2-specific factors, Ϫ139 to Ϫ101, can bind all five (RIPE3b1, REIP3b2, A2.1, A2.2, and A2.3) DNA-binding complexes (data not shown). To analyze the role of overlapping binding regions in the binding of these factors, various mutations in the larger Ϫ139 to Ϫ101 oligonucleotide (LM) were constructed. To determine the role of nucleotides in the overlapping region in the formation of C1-A2 and A2-specific DNA-binding complexes, competition analysis was performed.
HIT T-15 nuclear extract was incubated with the RIPE3b (Ϫ126 to Ϫ101) or Ϫ139 to Ϫ113 probes, in the absence or presence of excess wild type (Ϫ139 to Ϫ101, RIPE3b, Ϫ116 to Ϫ101 and Ϫ139 to Ϫ113) or mutant (Ϫ110⅐109LM, Ϫ114⅐113LM, Ϫ122⅐121LM, Ϫ125⅐124LM, and Ϫ127⅐126LM) oligonucleotides. These binding reactions were analyzed by EMSA, and results are presented in Fig. 8. Presence of excess cold wild type oligonucleotides in the binding reaction had the expected effect (see Figs. 2 and 3) on binding of factors to either the RIPE3b or Ϫ139 to Ϫ113 probe. Mutating nucleotides Ϫ110 to Ϫ109 bp and Ϫ122 to Ϫ121 bp in the larger Ϫ139 to Ϫ101 oligonucleotide prevented these oligonucleotides from competing with the RIPE3b probe for binding of RIPE3b1 and RIPE3b2 factors, whereas mutations at Ϫ114 to Ϫ113 bp, Ϫ125 to Ϫ124 bp, and Ϫ127 to Ϫ126 bp positions had no effect on binding of these factors. These observations confirm that binding of RIPE3b1 and RIPE3b2 factors depends on nucleotides at positions Ϫ110 to Ϫ109 bp and Ϫ122 to Ϫ121 bp in the insulin enhancer, whereas the other six nucleotides can be substituted without any significant effect on the binding activity. The results from competition analysis for binding of A2-specific complexes to Ϫ139 to Ϫ113 probe were extremely interesting. The formation of the A2-specific complexes, like the C1-A2 complexes, depends on nucleotides at positions Ϫ122 to Ϫ121 bp in the insulin enhancer but not on nucleotides at positions Ϫ110 to Ϫ109 bp. Furthermore, nucleotides at positions Ϫ125 to Ϫ124 bp were only essential for formation of the A2-specific complexes and not for C1-A2 specific complexes. We also used these oligonucleotides as probes and confirmed the requirement of these nucleotides for binding of the RIPE3b-and A2specific complexes (data not shown). These observations suggest that the rat insulin enhancer region from Ϫ139 to Ϫ101 bp represents the binding site for two distinct ␤-cell-specific factors, and nucleotides at positions Ϫ122 and Ϫ121 bp are critical for the binding of each factor. Furthermore, these observations suggest that mutations at positions Ϫ110 to Ϫ109 bp or Ϫ125 to Ϫ124 bp can selectively prevent binding of C1-A2 or A2specific factors, respectively.
Both C1-A2 and A2-specific Activators Positively Regulate Insulin Gene Expression-To determine the role of C1-A2 and A2-specific factors in regulating insulin gene expression, mutations were made in the rat insulin II enhancer (Ϫ238 to ϩ2 bp) construct driving expression of the LUC reporter gene. Wild type (Ϫ238 WT LUC) and the mutant insulin enhancer luciferase reporter constructs (Ϫ110⅐109m LUC, Ϫ122⅐121m LUC, and Ϫ125⅐124m LUC), were transfected into two insulin-producing cell lines, HIT T-15 and ␤TC-3. 48 h after transfection, cells were harvested, and whole cell extracts was prepared as described under "Experimental Procedures." Luciferase activity from insulin reporter constructs was normalized for transfection efficiency, and the results are presented relative to the activity of Ϫ238 WT LUC construct (Fig. 9).
Earlier studies showed that an insulin enhancer construct with mutations at Ϫ112 to Ϫ111 bp prevents formation of the RIPE3b1 and RIPE3b2 complexes and inhibits insulin gene expression (14,28,31). Similarly we observed that mutations at Ϫ110 to Ϫ109 bp that prevents formation of C1-A2 specific binding complexes, significantly inhibit (76 and 87% in HIT T-15 and ␤TC-3 cell lines, respectively) insulin gene expression. Mutations at positiona Ϫ125 to Ϫ124 that selectively prevent binding of A2-specific factors also inhibited insulin gene expression by 43 and 37% in HIT T-15 and ␤TC-3 cell lines, respectively. Interestingly, mutations that prevent binding of both C1-A2 and A2-specific binding factors (Ϫ122-121 bp) showed significantly more inhibition (92 and 96% in HIT T-15 and ␤TC-3 cell lines, respectively) of insulin gene expression than mutations that prevented binding of either of these factors. This inhibition of insulin gene expression was consistently more (about 1.7-2.0-fold) than that that can be accounted by simple additive contributions of these factors. Hence, we suggest that C1-A2 (RIPE3b) and A2-specific binding factors may functionally cooperate with each other and positively regulate insulin gene expression. Mutations that prevents binding of these factors to the insulin enhancer can drastically inhibit insulin gene expression even when the binding site for other key transcription factors (PDX-1 and BETA2) is unaffected, emphasizing the importance of these factors on insulin gene expression. We are currently investigating the mechanism by which C1-A2 and A2-specific factors contact overlapping DNA-binding sites and cooperatively activate insulin gene expression. DISCUSSION The present study demonstrates that ␤-cell-specific RIPE3b1 and A2.2 factors recognize overlapping DNA-binding sites within the insulin enhancer. These factors can each positively regulate insulin gene expression as well as cooperatively activate insulin gene expression. Because the binding activity of RIPE3b1 activator is regulated by glucose ( Fig. 7 and Refs. 28, 30, and 31) and phosphorylation (28), overlapping binding specificity and functional cooperation between A2.2 and RIPE3b1 activator may play major roles in regulating insulin expression in response to various metabolic stimulus. These observations suggest that like RIPE3b1, the A2.2 activator is an important regulator of insulin gene expression. Initial characterization indicates that A2-specific binding factors are distinct from PDX-1 (Fig. 6A). Excess unlabeled A3 oligonucleotide did not prevent formation of A2-specific complexes (Fig. 6A), suggesting that A2 binding factors might belong to a different family of transcription factors than the PDX-1 and other A element binding homeodomain factors. Similarly, anti-Nkx2.2 antibody did not recognize any of the A2-specific complexes, and mutating the Nk2 consensus sequence did not affect the formation of A2 complexes. These results indicate that A2-specific activators are novel and distinct from the Nk2 and other homeodomain family of transcription factors.
The importance of the RIPE3b region in mediating ␤-cellspecific and glucose-responsive insulin gene expression is well established (5, 14, 27-31, 37, 41). Although two insulin enhancer elements, C1 and A2, were described in the RIPE3b region, we were unable to detect any C1 selective binding FIG. 8. Nucleotides at ؊122 and ؊121 bp in the rat insulin II enhancer are critical for RIPE3b (C1-A2) and ؊139 to ؊113 (A2-specific) binding activity. The RIPE3b and Ϫ139 to Ϫ113 bp binding activity in the HIT T-15 nuclear extract was analyzed by EMSA. Cold competitors (50ϫ) used were either wild type rat insulin II oligonucleotides (Ϫ139 to Ϫ101 bp, RIPE3b, Ϫ116 to Ϫ101 bp, and Ϫ139 to Ϫ113 bp) or were mutant Ϫ139 to Ϫ101 bp oligonucleotide with the mutations in the indicated base pairs. Positions of specific DNA-binding complexes are indicated. An asterisk denotes mutation critical for the formation of C1-A2 and A2-specific binding complexes, whereas a pound sign denotes mutations that selectively prevents formation of either C1-A2 or A2-specific complexes. activity in insulin-producing cells. Our results demonstrate that C1 binding factors (RIPE3b1 and RIPE3b2) require sequences present in the A2 element. Similarly, A2-specific factors require additional nucleotides upstream of the RIPE3b region for their binding (Figs. 2 and 3). These observations suggest a need to redefine the enhancer elements in the RIPE3b region. Because we were unable to identify any C1specific binding factor, we suggest that designation of the C1 region as a DNA-binding element may not be accurate. We suggest that the binding element for RIPE3b1 and RIPE3b2 activator may accurately be described as C1-A2 or the CA element. Similarly, additional nucleotides must be included within the A2 element to describe it as a binding site for A2-specific factors. These criteria would help interpret earlier studies that describe mutations and deletions in C1 and A2 elements and their effect on insulin gene expression (47,64). For example, Tomonari and co-workers (47) reported the role of the A2 (GG1) element in regulating human insulin gene expression. Their results obtained from an A2 deletion construct would not only inhibit activation from the A2-specific factor but also from the RIPE3b1 activator. Similarly, other mutations in the A2 region described in the their study would also inhibit activation mediated by both RIPE3b1 and A2.2 factors. In addition, functional cooperation between the C1-A2 and A2specific factors should be considered prior to interpreting results from earlier studies. We therefore suggest that a clear designation of the C1-A2 and A2 elements would be helpful in analyzing the role of this region in insulin gene expression.
Mutational analysis of the RIPE3b region has provided important information regarding the RIPE3b1 and RIPE3b2 activators. Binding of these factors requires at least a 16-bp-long region (Ϫ123 to Ϫ107 bp; Figs. 2 and 8) of the rat insulin II enhancer. Nucleotides critical for binding of these factors are separated into two distinct regions, Ϫ112 to Ϫ107 bp and Ϫ123 to Ϫ117 bp. Although the nucleotides between these regions can be mutated without any effect on binding activity and insulin gene expression ( Fig. 2A, data not shown, and Ref. 28), attempts to alter spacing between these regions by either insertion or deletion mutations prevented DNA binding (Fig. 2C). These observations suggest that RIPE3b1 and RIPE3b2 factors not only recognize specific nucleotide sequence but also the recognize specific structure/organization of these nucleotides. The two critical nucleotide regions are spaced apart by about one helical turn of DNA, suggesting that RIPE3b1 and RIPE3b2 activators may recognize only one surface of the DNA helix. This is consistent with the results from methylation interference analysis of RIPE3b1 and RIPE3b2 complexes by Shieh and colleagues (14,29), showing a strong interference at Gs Ϫ107, Ϫ108, and Ϫ111 and a weak interference at Ϫ114 bp. Furthermore, results from the methylation interference experiment demonstrated interference at Gs only on the bottom strand and not on the top strand. Our mutational analysis (Fig.  2) and transient transfection data (Fig. 9) are consistent with the importance of nucleotides Ϫ107, Ϫ108, and Ϫ111 bp in insulin gene expression. The role of weak interference at Ϫ114 bp is unclear, because Ϫ113 and Ϫ114 bp are not critical for binding nor gene expression (data not shown). Also, Zhao and co-workers (28) demonstrated that mutating Ϫ114 bp had no effect on insulin gene expression, suggesting that weak interference at Ϫ114 bp is not critical for binding of the RIPE3b1 factor and insulin gene expression. Our results demonstrate the requirement of a large binding region for the RIPE3b1 and the RIPE3b2 factors (Ϫ123 to Ϫ107 bp; Figs. 2 and 8); however, it is unclear why no interference was observed at Gs on the bottom strand at positions Ϫ117 and Ϫ120 bp.
RIPE3b1 and RIPE3b2 factors have identical DNA binding properties, suggesting that they may belong to the same family of transcription factors. These observations are in contrast with the results described by Shieh and colleagues (14,29) in that RIPE3b1 and RIPE3b2 complexes have distinct methylation interference profiles and recognize distinct DNA-binding sites. Rip-1, one of the components of the RIPE3b2 complex, was cloned and identified as the hamster homologue of the mouse and human gene Smbp-2, a protein that binds to the immunoglobulin m chain switch region and was also cloned as glial factor-1 (29). Glial factor-binding element (GFE) selectively competes for the formation of the RIPE3b2 complex but not for the RIPE3b1 complex. Although the nucleotide sequence in the GFE element shares some homology with the C1 element, no homology in the A2 region can be observed (29). Similarly, GFE lacks significant homology with that for Smbp-2 (Sm), and Shieh and colleagues (29) suggested that the Rip-1 may bind both sites by recognizing a structure rather than the DNA sequence. However, based on sequence requirements for the binding of RIPE3b1 and RIPE3b2 factors (Fig. 2), it is unclear how the GFE element can selectively inhibit the formation of RIPE3b2 complex. One possible explanation could be that the binding specificity of Rip-1 might be distinct from that of the multi-protein RIPE3b2 complex. Consequently, excess unlabeled GFE oligonucleotide will selectively remove Rip1 from the DNA binding reaction, thereby preventing formation of RIPE3b2 complex.
The salient finding of the present study is that two base pairs are critical for DNA binding of two ␤-cell-specific transcriptional activators. Several possibilities could explain how two factors can recognize the same base pairs and not hinder the action of the other factor. It is possible that the different factors recognize the binding site on different alleles within a cell or in different cells in a population. Transient transfection analysis of a cell population does not discriminate between these possibilities and the binding of both factors to the same allele. Nonetheless, because C1-A2 and A2-specific activators cooperatively activate insulin gene expression, these factors may bind the same element on an allele.
Other examples of transcription factors binding to an overlapping binding site that result in either activation or inhibition of gene expression exist. Overlapping binding specificity of SP1 and C/EBPs in the C/EBP␣ promoter has been implicated in regulating adipocyte differentiation (65). During normal growth conditions, the SP1 factor binds to the overlapping C/EBP binding site and prevents binding of other factors. Differentiation signals decrease levels of the SP1 factor, which facilitates access of C/EBPs to the binding site and results in the activation of the C/EBP␣ gene expression (65). Similarly, there are several examples where both factors recognizing overlapping binding sites positively activate gene expression (66 -69). Overlapping binding of transcription factors NF-B and STAT6 can positively and negatively regulate eotaxin and E-selectin gene expression, respectively (69,70). In the Eselectin promoter, the STAT6 binding site shares 5 bp with the NF-B binding site, whereas in the eotaxin gene only 4 bp are shared. Matsukura and colleagues (69) suggested that this difference in DNA structure may permit simultaneous binding of these factors to overlapping binding sites and activate eotaxin gene expression. Similarly, it is possible that the RIPE3b1 activator may bind only on one surface of the C1-A2 region (as discussed above), which permits simultaneous occupation of overlapping regions by the A2.2-specific activator, providing a possible mechanism by which both these factors can positively regulate insulin gene expression. Transcription factors recognizing overlapping binding sites can integrate responses from multiple positive and negative signals by inhibiting or cooperatively activating gene expression (65)(66)(67)(68)(69)(70). Key insulin gene transcription factors must respond to multiple signals required to mediate ␤-cell-specific and glucose-responsive insulin gene expression as well as regulate pancreatic development and differentiation of ␤-cells. Because insulin gene expression is regulated in response to metabolic and environmental signals, key insulin gene transcription factors may integrate these responses by functionally cooperating with each other and/or binding overlapping sites. We suggest that the RIPE3b1 and A2.2 activators may represent one such regulatory unit. Because phosphorylation and alterations in glucose concentration regulate the activity of the RIPE3b1 factor, we suggest that RIPE3b1 and A2.2 activators may play a major role in regulating insulin gene expression in response to metabolic signals. In addition, overlapping binding specificity of these factors may also have a role in regulating pancreatic development and/or differentiation of ␤-cells. Availability of cloned RIPE3b1 and A2.2 factors would permit better understanding of the mechanisms by which these factors bind and regulate insulin gene expression; we are currently pursuing these objectives.