The Krüppel-Like Protein Gli-Similar 3 (Glis3) Functions as a Key Regulator of Insulin Transcription

Abstract

Transcriptional regulation of insulin in pancreatic β-cells is mediated primarily through enhancer elements located within the 5′ upstream regulatory region of the preproinsulin gene. Recently, the Krüppel-like transcription factor, Gli-similar 3 (Glis3), was shown to bind the insulin (INS) promoter and positively influence insulin transcription. In this report, we examined in detail the synergistic activation of insulin transcription by Glis3 with coregulators, CREB-binding protein (CBP)/p300, pancreatic and duodenal homeobox 1 (Pdx1), neuronal differentiation 1 (NeuroD1), and v-maf musculoaponeurotic fibrosarcoma oncogene homolog A (MafA). Our data show that Glis3 expression, the binding of Glis3 to GlisBS, and its recruitment of CBP are required for optimal activation of the insulin promoter in pancreatic β-cells not only by Glis3, but also by Pdx1, MafA, and NeuroD1. Mutations in the GlisBS or small interfering RNA−directed knockdown of GLIS3 diminished insulin promoter activation by Pdx1, NeuroD1, and MafA, and neither Pdx1 nor MafA was able to stably associate with the insulin promoter when the GlisBS were mutated. In addition, a GlisBS mutation in the INS promoter implicated in the development of neonatal diabetes similarly abated activation by Pdx1, NeuroD1, and MafA that could be reversed by increased expression of exogenous Glis3. We therefore propose that recruitment of CBP/p300 by Glis3 provides a scaffold for the formation of a larger transcriptional regulatory complex that stabilizes the binding of Pdx1, NeuroD1, and MafA complexes to their respective binding sites within the insulin promoter. Taken together, these results indicate that Glis3 plays a pivotal role in the transcriptional regulation of insulin and may serve as an important therapeutic target for the treatment of diabetes.

Gli-similar 3 (Glis3) is a Krüppel-like zinc finger transcription factor that plays a critical role in the generation of pancreatic β-cells (1, 2). In humans, GLIS3 deficiency has been linked to the development of a rare syndrome characterized by neonatal diabetes and congenital hypothyroidism (3, 4). In addition, genome-wide association studies have identified GLIS3 as a risk locus for both type 1 and 2 diabetes (3, 5–9). In mice, ubiquitous knockout of Glis3 gives rise to pups with neonatal diabetes characterized by hyperglycemia and hypoinsulinemia that survive only several days after birth (10–12). The diabetic phenotype presented by Glis3 knockout mice appears to be related to their paucity of insulin-producing β-cells in the pancreas and has indicated that Glis3 is probably required for the commitment of pancreatic progenitor cells to a β-cell lineage. Furthermore, Glis3 has also been reported to be expressed in mature pancreatic β-cells and to positively regulate insulin transcription (11, 13, 14). Together, these studies indicate that Glis3 plays a critical role in the regulation of β-cell development and endocrine function, including insulin expression in mature β-cells.

Insulin, produced and secreted by pancreatic β-cells, plays a key role in the regulation of blood glucose levels. Preproinsulin gene expression (hereafter referred to as insulin) as well as the secretion of the processed hormone are under complex controls. The transcriptional regulation of insulin gene expression is mediated by several transcription factors that recognize specific cis-acting enhancer elements located within the proximal insulin promoter (15–17). Three key transcription factors that bind within this region are pancreatic and duodenal homeobox 1 (Pdx1; also known as IPF1), neuronal differentiation 1 (NeuroD1; also known as β2), and v-maf musculoaponeurotic fibrosarcoma oncogene homolog A (MafA; also known as RIPE3b1) which bind to their cognate enhancers termed A-boxes, E-boxes, and C-boxes, respectively. These factors act together to synergistically activate insulin through direct interactions and through the recruitment of the ubiquitous coactivator, CREB-binding protein (CBP)/p300 (18–21). Pdx1, NeuroD1, and MafA have important functions in pancreatic development as well as in the maintenance of mature pancreatic endocrine cells (15, 16, 22, 23).

Recently, Glis3 was shown to positively regulate the transcription of the insulin gene by binding 2 conserved Glis enhancer elements (GlisBS) located near the A-, E-, and C-boxes within the proximal insulin promoter (11, 13). In this report, we characterize in greater detail the role of Glis3 in the transcriptional regulation of the insulin gene. We demonstrate that Glis3 acts synergistically with Pdx1, NeuroD1, and MafA to activate the insulin promoter. We further show that both the Glis3 N terminus and C-terminal transactivation domain associate with CBP/p300 and that the observed synergism between the different transcription factors relies in large part on the recruitment of CBP/p300 by Glis3 to the insulin promoter. Our data indicate that Glis3 functions as a key regulator of insulin transcription and that optimal transactivation of the Ins2 promoter by Pdx1, NeuroD1, and MafA is contingent on the binding of Glis3 at the GlisBS. Chromatin immunoprecipitation (ChIP) analyses suggest that neither Pdx1 nor MafA stably associates with the insulin promoter in the absence of functional GlisBS. Finally, we show that a single nucleotide mutation within the GlisBS of the human INS promoter that is responsible for the development of neonatal diabetes in several patients compromised the ability of Pdx1, NeuroD1, and MafA to activate the promoter in the absence of exogenously expressed Glis3. Based on these findings, we propose a model whereby recruitment of CBP/p300 by Glis3 provides a scaffold for the formation of a transcriptional regulatory complex that stabilizes binding by Pdx1, NeuroD1, and MafA to their respective binding sites.

Materials and Methods

Cells and growth conditions

Rat insulinoma INS1 832/13 cells, a generous gift from Dr H. Hohmeier (Duke University), were maintained in RPMI 1640 medium supplemented with 10% fetal calf serum, 10 mM HEPES, 2 mM glutamine, 1 mM sodium pyruvate, 100 U/mL penicillin, 100 μg/mL streptomycin, and 50 μM β-mercaptoethanol. HEK293T and the mouse pancreatic β-cell line βTC-6 were purchased from American Type Culture Collection and cultured in DMEM containing 10% fetal bovine serum.

Generation of reporter and expression plasmids

The generation of p-3xFLAG-CMV10-Glis3, p-3xFLAG-CMV10-Glis3-ΔC748, p-3xFLAG-CMV10-Glis3-ZFmut, and the LUC reporter plasmids p-INS-700-Luc and p-mIP-696-Luc, under control of the human and mouse insulin promoter, respectively, was described previously (11, 24, 25). The plasmids p-3xFLAG-CMV10-Glis3-ΔN496, p-3xFLAG-CMV10-Glis3-ΔN653, and p-3xFLAG-CMV10-Glis3-ΔN748 were generated by amplification of the indicated fragment by PCR and insertion into the EcoRI and BamHI sites of the p3xFLAG-CMV10 vector (Sigma-Aldrich). The plasmids pCMV6-CBP-HA, pCMV6-CBP-ΔC750-HA, pCMV6-CBP-ΔN959-HA, pCMV6-CBP-HAT, pCMV6-CBP-CH3, pCMV6-CBP-Q, pCMV6-Pdx1-HA, and pCMV6-NeuroD1-HA were generated by PCR amplification of the respective fragment and insertion into the HindIII and NheI sites of pCMV6-HA (Origene). p-3xFLAG-CMV10-Pdx1, p-3xFLAG-CMV10-NeuroD1, p-3xFLAG-Pdx1-ΔN141, p-3xFLAG-CMV10-NeuroD1-ΔC156, p-3xFLAG-CMV10-MafA, and p-3xFLAG-CMV10-MafA-ΔN173 were generated by amplification of the indicated inserts and subcloning into the XbaI and BamHI sites of p-CMV10–3xFLAG (Sigma-Aldrich). p-Glis3-(496–748)-VP16 was created by amplifying the region of Glis3 coding for amino acids (aa) 496–748 and inserting the fragment into the EcoRI and BamHI sites of the VP16 vector (Clontech). pM-Glis3-ΔN653 and pM-Glis3-ΔN748 were created by PCR amplification of the respective fragments and subcloning into the pM expression vector (Clontech). The construct p-E1A-Myc was generated by amplifying the 12s adenoviral E1A protein cDNA by PCR and directionally inserting it into the EcoRI and XhoI sites of pCMV-Myc (Clontech). The mIP-696-Luc GlisBS mutants were described previously (11). The mIP-696-Luc A-, E-, and C-box mutants were generated by site-directed mutagenesis using the QuikChange site-directed mutagenesis kit (Stratagene), following the manufacturer's protocol. The primer sets used are as follows with the mutated bases underlined: A-box(−78), 5′-CCTGGAGCCCTTACTGGGTCAAACAGC; A-box(−211), 5′-GTTAAGACTCTACTTACCCT-AGGAC; E-box(−105), 5′-CCTCTGGCCATCTTCTGACCTACCC; and C-box(−130), 5′-GGAAACTACAGCTTCAATCCCTCTGGCCATCTGC. The construct mIP-696-UAS-Luc was generated by site-directed mutagenesis using the following primers with the mutated bases underlined: GlisBS(−99)UAS1, 5′-CCATCTGCTGACCCGGACCACCTGGAG-CCC; GlisBS(−99)UAS2, 5′-GCTGACCCGGAGTACCTGGAGCCCTTAATGG; GlisBS(−99)UAS3, 5′-GCTGACCCGGAG-TACTGTCAGCCCTTAATGG; GlisBS(−99)UAS4, 5′-CCCGGAGTACTGTCCTCC-GTTAATGGG-TCAAACAGC; GlisBS(−263)UAS1, 5′-CCCACTCCGGAGTAATGTCCCCTGC; and GlisBS (−263)UAS2, 5′-CCACTC-CGGAGTACTGTCCTCCGCTGTGAACT. p-INS-700(c93g)-Luc was generated by site-directed mutagenesis using the following primer with the mutated bases underlined: GlisBS1(c93g), 5′-GCCATCTGCCGACCCCCCGACCCCAGGCCCTAATGGGC.

Transfection and reporter assays

Cells were plated in 12-well dishes at 1 × 10⁵ cells/well and 24 hours later were transfected with pCMV-β-galactosidase, and the indicated plasmids were diluted in Opti-MEM (Invitrogen) using either Lipofectamine 2000 (Invitrogen) for HEK293T cells or Lipofectamine LTX with PLUS reagent for INS1 cells, following the manufacturer's protocol. After 24 hours, cells were harvested, and luciferase and β-galactosidase activities were measured using a luciferase assay kit (Promega) and a luminometric β-galactosidase detection kit (Clontech), respectively, following the manufacturer's protocols. Each data point was assayed in triplicate, and each experiment was performed at least twice. Relative luciferase activity was calculated. All values underwent ANOVA and Tukey-Kramer comparison tests using InStat software (GraphPad Software Inc), and data are presented as means ± SE.

Coimmunoprecipitation analysis

HEK293T cells were transiently transfected with the specified plasmids using Lipofectamine 2000 reagent (Invitrogen), following the manufacturer's protocol. At 48 hours after transfection, cells were harvested by scraping in radioimmunoprecipitation assay buffer (25 mM Tris, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 20 mm sodium molybdate, and 0.5% Nonidet P-40) containing protease inhibitor cocktails I and II (Sigma-Aldrich). Cell lysates were centrifuged at 16,000 × g for 10 minutes at 4°C. A portion of the supernatants was incubated at room temperature for 10 minutes with Dynabeads (Invitrogen) conjugated to high-affinity anti-hemagglutinin (HA) antibody (Roche) or anti-M2 FLAG antibody (Sigma-Aldrich). Magnetic beads were washed 3 times with 200 μl of ice-cold PBS (137 mM NaCl, 10 mM phosphate, and 2.7 mM KCl, pH 7.4). Bound protein complexes and input fractions were examined by Western blot analysis using mouse anti-FLAG or rat anti-HA antibodies.

ChIP assays

HEK293T cells grown in 100-mm dishes were transiently transfected with p-mIP-696-Luc or the indicated mutants along with p-3xFLAG-CMV10-Glis3, -Pdx1, or -MafA as specified using Lipofectamine 2000 reagent (Invitrogen), following the manufacturer's protocol. At 48 hours later, ChIP assays were performed using a ChIP assay kit (Millipore), following the manufacturer's protocol with minor revisions. In brief, the cells were cross-linked with 1% formaldehyde for 10 minutes at room temperature, and the reaction was stopped by the addition of 1× glycine (125 mM). After 3 washes with ice-cold PBS, the cells were harvested by scraping, and the nuclei were isolated after incubation in hypotonic buffer (10 mM Tris HCl, pH 7.4, 1.5 mM MgCl₂, 10 mM KCl, and 0.5 mM dithiothreitol) for 20 minutes on ice and vigorous vortexing. The chromatin was then sonicated and incubated for 1 hour with anti-M2-FLAG agarose beads (Sigma-Aldrich) at 4°C. The immobilized complexes were eluted with excess 3xFLAG peptide and cross-linking was reversed by incubation with 20 mM NaCl overnight at 65°C. Proteinase K and RNase A digestion for 2 hours at 42°C was performed before DNA purification. Chromatin immunoprecipitated DNA was quantified by quantitative PCR in triplicate experiments using primers amplifying the Ins2 promoter, 5′-TGCTCAGCCAAGGACAAAGA and 5′-CCAAACACTTCCCTGGTGCT, or hGAPDH, 5′-TCCCTTGGGTATATGGTACCTTG and CCACTTGATTTTGGAGGGATCTC. Data are presented as the percentage of chromatin immunoprecipitated DNA relative to input fractions ± SE.

Results

GlisBS play a significant role in the regulation of basal Ins2 promoter activity

Previously, we and others reported that exogenous Glis3 was capable of activating rodent insulin 2 (mIns2 and rIns2) genes as well as the human insulin (INS) gene by binding 2 conserved GlisBS within their respective proximal promoter regions (11, 13, 25). To investigate the role of the GlisBS in more detail, the regulation of a mouse Ins2-driven luciferase reporter by endogenous proteins in the β-like cell line INS1 832/13 was assessed. Site-directed mutagenesis was used to render both GlisBS (located at nucleotides [nt] −99 and −263 relative to the Ins2 transcriptional start site) nonfunctional. We examined the effect of these GlisBS mutations on Ins2 luciferase reporter (p-mIP-696-Luc) activation by themselves or in combination with mutations in the A-, E-, and C-boxes, which bind the known Ins2 transcriptional regulators, Pdx1, NeuroD1, and MafA, respectively (Figure 1A). Figure 1B shows that the endogenous proteins present in INS1 832/13 cells were capable of robustly activating the Ins2 promoter relative to cells transfected with empty, pGL4.10 vector. Mutation of the 2 A-boxes (located at nt −78 and −211) resulted in a >80% reduction in basal activity, whereas loss of the E-box reduced activity by approximately 60%. Mutation of the C-box (located between nt −119 and −130) reduced the Ins2 reporter activity in INS1 832/13 cells by approximately 20%, although the effect was not statistically significant. Loss of the A, E, and C elements simultaneously led to a statistically different decrease (>90%) in activation. Mutation of the 2 GlisBS decreased transactivation of the Ins2 promoter by 90%, and activity was reduced further when the GlisBS were mutated in combination with the A-, E-, and C-boxes to a level similar to that of pGL4.10 empty vector.

Figure 1.

Glis3 and GlisBS are critical for optimal Ins2 promoter activation. A, Schematic representation of the mIns2 promoter region used in p-mIP-696-Luc reporter constructs. The different enhancer elements are indicated. B and C, INS1 832/13 cells were transfected with pGL4.10, p-mIP-696-Luc, or the specified mutants and pCMV10–3xFLAG empty vector or p-CMV10–3xFLAG-Glis3 as indicated. After 48 hours, cells were harvested and assayed for luciferase and β-galactosidase activities, and the fold change was plotted. Each bar represents the mean ± SE. Schematic diagrams to the left of each bar visually display mutations of enhancer elements. a, statistically different from p-mIP-696-Luc activation (P < .01); ns, not statistically significant. D, INS1 832/13 cells were transfected with pGL4.10 or p-mIP-696-Luc and the indicated FLAG-tagged expression constructs. After 48 hours, reporter activity was analyzed as described for B and C. Each bar represents the mean ± SE. a, statistically different from the pGL4.10 control (P < .01); b, statistically different from pGL4.10 and a samples (P < .01); c, statistically different from pGL4.10, a, and b (P < .01); d, statistically different from pGL4.10, a, b, and c (P < .01).

To determine whether optimal activation of Ins2 by Glis3 was dependent on the presence of functional A-, E-, or C-boxes, INS1 cells were cotransfected with mIP-696-Luc or mutant mIP-696-Luc reporter plasmids with or without a FLAG-Glis3 expression vector. As shown in Figure 1C, mutation of the A-, E-, or C-boxes had little effect on the ability of Glis3 to activate the Ins2 promoter; however, when the mutations in the A-, E-, and C-boxes were combined, Glis3-mediated transactivation of the Ins2 reporter was reduced by 75%. These results suggested that the GlisBS and A, E, and C sites are mutually required for optimal Ins2 activation.

To obtain further insights into the cooperation between Glis3 and Pdx1, NeuroD1, and MafA, we analyzed the effects of exogenous expression of these proteins on Ins2 activation in INS1 cells. Figure 1D shows that Pdx1 or NeuroD1 enhanced luciferase reporter activity <2-fold, whereas MafA increased activity about 5-fold, and joint expression of Pdx1, NeuroD1, and MafA activated the reporter approximately 8-fold. Strikingly, expression of Glis3 alone stimulated reporter activity >20-fold over basal levels. Coexpression of Pdx1, NeuroD1, or MafA with Glis3 enhanced the activation of the reporter roughly 2-fold, whereas all 4 transcription factors did not cause a further enhancement in luciferase activity, possibly because maximal levels of activation might have been reached. These data collectively demonstrate that Glis3 is a powerful regulator of insulin transcription and suggest that Glis3 acts cooperatively with Pdx1, NeuroD1, and MafA to induce optimal activation of the Ins2 promoter in β-cells.

Glis3 zinc finger domain (ZFD) and transactivation domain (TAD) are required for its synergism with Pdx1 and NeuroD1

Insulin transcription in β-cells is under complex regulatory control by endogenous transcription factors (15, 16). This makes assessment of the contribution of each individual factor to Ins2 activation difficult. Therefore, to further investigate the cooperation between Glis3, Pdx1, NeuroD1, and MafA, we analyzed the transactivation of Ins2 in HEK293T cells that do not express insulin or the β-cell−specific regulatory factors, Pdx1, NeuroD1, and MafA. As shown in Figure 2A, exogenous expression of Glis3 in HEK293T cells increased Ins2 promoter activation about 10-fold, whereas transiently expressed Pdx1 had little effect on Ins2 promoter activity; however, coexpression of Glis3 and Pdx1 resulted in a synergistic activation of the Ins2 reporter that was 6-fold higher than that by Glis3 alone. Coexpression of Glis3 with NeuroD1, which by itself only slightly activated the Ins2 promoter, also synergistically increased the activation of the Ins2 promoter to a level similar to that observed for Glis3 and Pdx1. The synergistic activation by Glis3 with either Pdx1 or NeuroD1 required the DNA binding and TAD of Glis3 as evidenced by the inability of the Glis3 C-terminal deletion mutant, Glis3-ΔC748, which lacks the TAD (24), and the zinc finger mutant, Glis3-ZFmut, which is unable to bind GlisBS (24), to activate the Ins2 reporter synergistically with Pdx1 or NeuroD1 (Figure 2A). Expression of exogenous MafA alone enhanced reporter activity about 5-fold, but when coexpressed with Glis3, Ins2 promoter activity was increased approximately 30-fold. The synergy between Glis3 and MafA was nearly lost when the Glis3 ZFD was mutated, whereas the activation was significantly (80%) less when MafA was coexpressed with Glis3-ΔC748. Deletion of the Glis3 N terminus up to aa 496 (Glis3-ΔN496) caused about a 30% reduction in Glis3-mediated activation of the reporter, whereas cotransfection with Pdx1 or NeuroD1 resulted in a 6-fold increase similar to the synergistic effect obtained with full-length Glis3. However, the synergy between Glis3-ΔN496 and MafA was roughly half of that observed between MafA and full-length Glis3. Synergistic activation by Glis3 with each of the 3 proteins was lost when they were coexpressed with a Glis3 mutant lacking both the N terminus and C-terminal TAD (Glis3 496–748). These results suggest that the Glis3 N terminus contributes to the observed synergism between Glis3 and MafA and that maximal synergy with MafA requires both the Glis3 N and C termini.

Figure 2.

The Glis3 ZFD and TAD are required for its synergism with Pdx1, NeuroD1, and MafA. A and B, HEK293T cells were transfected with p-mIP-696-Luc or p-mIP-696-UAS-Luc and pCMV10–3xFLAG empty vector, pCMV10–3xFLAG-Glis3, or the Glis3 mutant indicated by the schematic diagrams (left of the y-axes). Each sample was cotransfected with pCMV10–3xFLAG empty vector (white bars), pCMV10–3xFLAG-Pdx1 (pale gray bars), pCMV10–3xFLAG-NeuroD1 (dark gray bars), or pCMV10–3xFLAG-MafA (hatched bars). After 48 hours, cells were harvested and assayed for luciferase and β-galactosidase activities and normalized relative Luc activity (nRLU) was plotted. Each bar represents the mean ± SE. The white X indicates mutation of ZF3. VP16, herpes simplex virion protein 16 transactivation domain; Gal4, Gal4 DNA-binding domain. Lower inset: schematic cartoon depicting the replacement of GlisBS in p-mIP-696-Luc used in experiments indicated by * with UAS (Gal4 enhancer) in p-mIP-696-UAS-Luc used in experiments indicated by #. a, statistically different from equivalent samples expressing pCMV10–3xFLAG empty vector (P < .01).

To determine whether the TAD of Glis3 was specifically required for this synergy, we investigated whether synergistic activation of the Ins2 promoter with Pdx1, NeuroD1, and MafA was lost if we swapped the Glis3 TAD with the activation domain of herpes simplex virion protein 16 (VP16). To this end, the ability of a Glis3 chimera, Glis3 (496–748)-VP16, in which the Glis3 ZFD was fused to the TAD VP16, to enhance the activation of the mIP-696-Luc reporter by Pdx1, NeuroD1, and MafA was examined. As shown in Figure 2B, Glis3 (496–748)-VP16 activated the luciferase reporter. This activation was considerably increased by expression of exogenous Pdx1, NeuroD1, or MafA, although not as efficiently as by Glis3-ΔN496 containing the native Glis3 TAD. These results indicated that the Glis3 TAD is not specifically required for synergistic activation with these β-cell transcription factors. We likewise were curious about whether synergistic activation of Ins2 with Pdx1, NeuroD1, and MafA would be maintained if we swapped out the Glis3 ZFD with the DNA-binding domain (DBD) of Gal4 while replacing the GlisBS within the Ins2 reporter with Gal4 upstream activation sequence (UAS) binding sites. Gal4(DBD)-Glis3(ΔN653), consisting of a fusion between the Gal4(DBD) and the Glis3 C terminus, was able to activate the reporter plasmid mIP-696-UAS-Luc, in which both GlisBS1 and -2 were replaced by Gal4 enhancer elements (UAS) (Figure 2B; lower inset). Moreover, replacement of the Glis3 ZFD with the Gal4(DBD) did not greatly affect the ability of Pdx1, NeuroD1, or MafA to synergistically activate mIP-696-UAS-Luc, suggesting that the Glis3 ZFD is not specifically required for the observed synergism. Finally, a Glis3 chimera lacking a TAD in which the N terminus (aa 1–480) was fused to the Gal4(DBD) failed by itself to activate mIP-696-UAS-Luc, modestly enhanced the activation together with Pdx1 and NeuroD1, and exhibited significant synergism with MafA. Collectively, these results indicate that Glis3 synergism with Pdx1, NeuroD1, and MafA requires its C-terminal transactivation as well as its DBD.

Synergistic activation of insulin transcription by Glis3, Pdx1, NeuroD1, and MafA requires DNA binding of each factor

To determine the role of each individual enhancer element in the observed synergism between Glis3 and Pdx1, NeuroD1, and MafA, the effect of A-, E-, or C-box mutations on the activation of mIP-696-Luc was examined. Consistent with previous reports using other cell lines (19, 21, 26–29), coexpression of Pdx1 and NeuroD1 in HEK293T cells synergistically activated the insulin promoter, and coexpression with MafA increased activation further by 3-fold (Figure 3A). As demonstrated above, coexpression of Pdx1 or NeuroD1 enhanced Glis3-mediated Ins2 activation significantly (10-fold). This synergy was almost totally ablated when, respectively, the A-boxes or the E-box, were mutated. These data indicated that binding of Pdx1 or NeuroD1 to the insulin promoter is required for their synergistic activation with Glis3. When both Pdx1 and NeuroD1 were coexpressed with Glis3, the Ins2 promoter activity was stimulated by approximately 500%. Mutation of either the A-boxes or the E-box resulted in a 6- or 4-fold decrease in activity, respectively, indicating that binding of Pdx1 and NeuroD1 to their response elements contributed to this synergism to a similar degree. Coexpression of MafA with Glis3 enhanced Glis3-mediated activation of the reporter approximately 20-fold, and, as observed for Pdx1 and NeuroD1, this synergism required MafA binding at the C-box. Activation of mIP-696-Luc was predictably increased further when Pdx1, NeuroD1, and MafA were simultaneously coexpressed with Glis3. Mutation of each of the A-, E-, or C-boxes reduced the level of activity by approximately 50%, again suggesting an approximately equal contribution of each element in the synergism of Pdx1, NeuroD1, and MafA with Glis3. Interestingly, activation resulting from coexpression of Pdx1, NeuroD1, MafA, and Glis3 was reduced 95% when the 2 GlisBS were mutated, suggesting that Glis3 binding to the promoter contributed disproportionately more to the observed synergism than binding of the other 3 transcription factors.

Figure 3.

Cooperative Ins2 activation by Glis3, Pdx1, NeuroD1, and MafA requires binding of each protein to its respective DNA binding site. A, HEK293T cells were transfected with p-mIP-696-Luc (black bars) or the indicated mutant along with pCMV10–3xFLAG empty vector, pCMV10–3xFLAG-Glis3, -Pdx1, -NeuroD1, or -MafA alone or in combination, as specified. After 48 hours, cells were harvested and assayed for luciferase and β-galactosidase activities and normalized relative Luc activity (nRLU) was plotted. Each bar represents the mean ± SE. B, Glis3 synergistically activates the human INS promoter (p-INS-700-Luc) with Pdx1, NeuroD1, and MafA. HEK293T cells were transfected with p-INS-700-Luc and pCMV10–3xFLAG or the specified combination of FLAG-tagged expression plasmids. After 48 hours, cells were harvested and assayed for luciferase and β-galactosidase activities and nRLU was plotted. Each bar represents the mean ± SE. a, statistically different from pCMV10–3xFLAG empty vector (P < .01); b, statistically different from pCMV10–3xFLAG empty vector and a (P < .01); c, statistically different from pCMV10–3xFLAG, a, and b (P < .01); d, statistically different from pCMV10–3xFLAG, a, b, and c (P < .01); e, statistically different from pCMV10–3xFLAG, a, b, c, and d (P < .01); f, statistically different from pCMV10–3xFLAG, a, b, c, d and e, (P < .01).

A similar synergistic cooperation between Glis3, Pdx1, NeuroD1, and MafA could be observed when a luciferase reporter driven by the human INS promoter (p-INS-700-Luc) was transfected into HEK293T cells, suggesting that the data obtained with the mouse Ins2 promoter are probably pertinent to transcriptional regulation of the human insulin promoter as well (Figure 3B). Taken together, these data support our conclusion that the 4 proteins act together to optimally activate the Ins2 or INS promoter and that binding of each transcription factor to their respective response elements was required for optimal synergistic activation and further show that the HEK293T cell line can be valuable for studying transactivation of Ins2 by Glis3 and its cooperation with Pdx1, NeuroD1, and MafA.

The C terminus of CBP interacts with Glis3

The cooperation observed between Glis3, Pdx1, NeuroD1, and MafA probably involves interactions between coactivator complexes recruited by these transcription factors to the Ins2 promoter. Both Pdx1 and NeuroD1 have been reported to recruit, via their TADs, the ubiquitous coactivators, CBP and p300 (18, 19, 30–32). Because Glis3 binds the proximal Ins2 promoter near the Pdx1, NeuroD1, and MafA enhancer sites, we were interested in studying the possibility that the observed synergistic activation involved recruitment of CBP by Glis3 to the Ins2 promoter. Coimmunoprecipitation assays were used to identify whether or not CBP interacted with Glis3 in HEK293T cells. Cells were cotransfected with HA-tagged CBP or pCMV6-HA empty vector and FLAG-tagged Glis3, Pdx1, NeuroD1, or MafA, and protein complexes were subsequently immunoprecipitated with an anti-HA antibody. A representative Western blot (Figure 4A) demonstrated that CBP was associated with Glis3 as well as Pdx1, NeuroD1, and MafA, which served as positive controls. Glis3 did not coimmunoprecipitate with the N-terminal half of CBP (CBP-ΔC750-HA) but was pulled down by the C-terminal half of CBP (CBP-ΔN959-HA) containing the histone acetyltransferase (HAT), C/H3, and Q-rich domains (see Supplemental Figure 1A published on The Endocrine Society's Journals Online web site at http://mend.endojournals.org). These data suggest that Glis3 interacts with the C terminus of CBP. The interaction between FLAG-Glis3 and CBP-ΔN959-HA was further validated when the inverse immunoprecipitation experiment was performed using an anti-M2 FLAG antibody (Supplemental Figure 1B).

Figure 4.

Glis3 interacts with the C terminus of CBP through its transactivation domain. A, HEK293T cells were transfected with pCMV6-HA empty vector or pCMV6-CBP-HA and cotransfected with pCMV10–3xFLAG-Glis3, -NeuroD1, -MafA, or -Pdx1 as indicated. After 48 hours, cells were harvested and coimmunoprecipitations were performed as described under Materials and Methods using a rat anti-HA antibody. Immunoprecipitated (IP) proteins were examined by Western blot (WB) analysis using either horseradish peroxidase (HRP)-conjugated mouse anti-FLAG or rat anti-HA and an HRP-conjugated goat anti-rat antibody. B and C, HEK293T cells were transfected with the indicated Glis3 mutants and cotransfected with either pCMV6-HA empty vector or pCMV6-CBP-ΔN959-HA. After 48 hours, cells were harvested, and coimmunoprecipitations were conducted as described in A. D, HEK293T cells were transfected with Glis3-ΔC480 or Glis3-ΔN653 and cotransfected with either pCMV6-HA empty vector or the indicated CBP mutants. After 48 hours, cells were harvested, and coimmunoprecipitations were conducted as described in A. E, Schematic representation of Glis3 and CBP. Regions that interact with Glis3 N or C terminus or CBP are indicated by brackets. Numbers represent amino acid positions. N, N-terminal conserved region; ZF, zinc finger domain; TAD, transactivation domain; C/H1, 3, cysteine/histidine regions 1 and 3; KIX, CREB interacting domain; HAT, histone acetyltransferase domain; Q, glutamine-rich region.

To ascertain which region(s) of Glis3 were required for its association with CBP, the ability of 3 Glis3 mutants to interact with CBP-ΔN959-HA was examined. This analysis showed that both Glis3-ΔC480 containing the Glis3 N terminus and Glis3-ΔN653 containing the C terminus of Glis3 coimmunoprecipitated with CBP-ΔN959-HA, whereas the Glis3 ZFDs (aa 496–653) failed to do so (Supplemental Figure 1C). To narrow down the regions required for the interaction, further truncations at the Glis3 C terminus were made, and the results in Figure 4B show that the Glis3 TAD (aa 748–935) was sufficient for interaction with CBP. Analysis of additional N-terminal truncations demonstrated that the region up to aa 389 was not capable of associating with CBP, in contrast with Glis3-ΔC480, which pulled down efficiently with CBP (Figure 4C). These data indicated that the interaction of the C-terminal domains of CBP with Glis3 is mediated through the proline-rich region of the Glis3 N terminus located between aa 389 and 480 and the activation domain of Glis3 (aa 749–935) (Figure 4E).

To identify the specific domains of CBP that mediate its association with the N- and C-terminal regions of Glis3, coimmunoprecipitation assays were conducted with CBP-ΔN959-HA or HA-tagged CBP mutants expressing the HAT domain (aa 1194–1671), the C/H3 domain (aa 1626–1851), or the Q-rich region (aa 1841–2441) and either FLAG-tagged Glis3-ΔC480 or Glis3-ΔN653. These experiments indicated that Glis3-ΔC480 interacted principally with the C/H3 domain of CBP, whereas Glis3-ΔN653 preferentially associated with the Q-rich region of CBP (Figure 4D) as depicted schematically in Figure 4E. Both fragments also showed a weak interaction with the HAT domain of CBP.

CBP mediates Glis3 synergy with Pdx1, NeuroD1, and MafA

Because CBP interacted with Glis3, Pdx1, NeuroD1, and MafA, we were interested in determining whether CBP played a role in mediating the synergistic activation of the Ins2 promoter by these transcription factors. We therefore examined the effect of exogenous CBP overexpression on the activation of mIP-696-Luc by the 4 transcription factors in HEK293T cells. Ectopic expression of CBP greatly increased the activation of mIP-696-Luc by Glis3 (Figure 5A). Likewise, coexpression of CBP enhanced the activation of the Ins2 reporter by Pdx1 and NeuroD1 and to a lesser extent that by MafA and dramatically induced the activation in cells coexpressing Pdx1, NeuroD1, and MafA. However, CBP caused only a small increase in Ins2 promoter activation in cells in which Glis3 was coexpressed with MafA and had a minimal effect in cells coexpressing Glis3, Pdx1, and NeuroD1 or all 4 factors. The observation that CBP only slightly increased Ins2 promoter activation in cells expressing Glis3 together with the other transcription factors might be due to the likelihood that near-maximal levels of activation were being reached.

Figure 5.

CBP mediates synergistic activation of Ins2 by Glis3 with Pdx1, NeuroD1, and MafA. HEK293T cells were transfected with p-mIP-696-Luc and pCMV10–3xFLAG-Glis3, -Pdx1, -NeuroD1, or -MafA as indicated. Cells were cotransfected with pCMV6-HA empty vector (white bars), pCMV6-CBP-HA (gray bars), or p-CMV-Myc-E1A (black bars). After 48 hours, cells were harvested and assayed for luciferase and β-galactosidase activities and the normalized relative Luc activity (nRLU) was plotted. Each bar represents mean ± SE. a, statistically different from pCMV6-HA empty vector (P < .01). B, HEK293T cells were transfected with p-mIP-696-Luc and pCMV10–3xFLAG-Glis3. Cells were cotransfected with the indicated full-length expression constructs (wild-type [WT]; black bars) or TAD-deficient mutants (TAD mut; white bars) of Pdx-1, NeuroD1, and MafA. After 48 hours, luciferase assays were conducted as described in A. a, statistically different from pCMV10–3xFLAG empty vector (P < .01); b, statistically different from pCMV10–3xFLAG empty vector and a (P < .01).

To examine the role of endogenous CBP/p300, the effect of the 12S adenoviral E1A oncoprotein, a potent inhibitor of CBP/p300 function (33–36), on the activation of the Ins2 promoter was analyzed. As shown in Figure 5A, E1A overexpression in HEK293T cells reduced Glis3-mediated transactivation of Ins2 by about 60%. The presence of E1A nearly eliminated the activation of the insulin promoter by Pdx1 and NeuroD1 or MafA, alone or in combination, as has been demonstrated previously (19, 21, 31). Moreover, E1A dramatically decreased the synergistic stimulation of Ins2 promoter activity when Glis3 was coexpressed with Pdx1 and NeuroD1, with MafA, or with all 3 factors. Similarly, in INS1 cells E1A inhibited the activation of the Ins2 promoter by endogenous transcription factors and in cells overexpressing Glis3 by 90% (Supplemental Figure 2A). Furthermore, the ability of Pdx1, NeuroD1, or MafA to synergistically activate the Ins2 reporter with Glis3 was lost when their CBP/p300 interacting TADs (18, 20, 31) were removed (Figure 5C). Our observations suggest that the synergistic activation of Ins2 by Glis3, Pdx1, NeuroD1, and MafA relies in large part on the recruitment of CBP/p300. CBP/p300 has been reported to function in part as a bridge or scaffold mediating the assembly of multiprotein transcriptional complexes (37) and helps to promote the cooperation between different transcription factors.

Glis3 is essential for Pdx1/NeuroD1-mediated Ins2 activation in HEK293T cells

The data presented thus far indicate that Glis3 cooperatively activates the Ins2 promoter together with Pdx1, NeuroD1, and MafA through mutual interactions with CBP/p300. The data further suggest that the GlisBS contributed disproportionately more to synergistic activation than did the A-, E-, or C-boxes (Figure 3A). To examine the role of the GlisBS in the synergistic Ins2 activation by Pdx1, NeuroD1, and MafA in more detail, the effect of mutations in the 2 GlisBS on the activation of the mIP-696-Luc reporter was assessed in HEK293T cells. Intriguingly, mutation of both GlisBS elements greatly reduced (85%) the activation of the Ins2 reporter in cells coexpressing Pdx1, NeuroD1, and MafA (Figure 6A). Mutation of either GlisBS1 or GlisBS2 reduced activity by approximately 50%, suggesting that the 2 GlisBS are almost equally important to Pdx1-, NeuroD1-, and MafA-mediated transactivation. To determine whether this decrease in activity was due to the loss of GlisBS binding by endogenous GLIS3, the effect of GLIS3 knockdown on Pdx1-, NeuroD1-, and MafA-directed Ins2 activation was examined. Transfection of GLIS3 small interfering RNA (siRNA) reduced endogenous GLIS3 transcript levels by approximately 85% and greatly diminished the ability of Pdx1, NeuroD1, and MafA to activate mIP-696-Luc compared with that of cells transfected with scrambled control siRNA (Figure 6B). These findings indicate that Glis3 binding at the GlisBS is essential for the transcriptional activation of Ins2 by Pdx1, NeuroD1, and MafA.

Figure 6.

GlisBS and endogenous Glis3 are necessary for optimal Ins2 transactivation by Pdx1, NeuroD1, and MafA. A, HEK293T cells were transfected with p-mIP-696-Luc or the indicated GlisBS mutant. Cells were cotransfected with pCMV10–3xFLAG empty vector (white bars) or pCMV10–3xFLAG-Pdx1, -NeuroD1, and -MafA (black bars). After 48 hours, cells were harvested and assayed for luciferase and β-galactosidase activities and the normalized relative Luc activity (nRLU) was plotted. Each bar represents the mean ± SE. a, statistically different from p-mIP-696-Luc (P < .01); b, statistically different from p-mIP-696-Luc and p-mIP-696-GlisBS1 or GlisBS2 (P < .01). B, HEK293T cells were transfected with p-mIP-696-Luc and the indicated FLAG expression constructs. Cells were cotransfected with scrambled control siRNA (Scr; white bars) or GLIS3 siRNA (black bars). After 48 hours, luciferase assays were conducted as described in A. a, statistically different from scrambled control siRNA (P < .01).

GlisBS are required for stable binding of Pdx1 and MafA to the Ins2 promoter

To determine whether the effect of the GlisBS mutations on the ability of Pdx1, NeuroD1, and MafA to activate the Ins2 promoter in HEK293T was related to a reduction in the recruitment of these factors to their respective enhancer elements, their interaction with the Ins2 proximal promoter was investigated by ChIP-PCR analysis. ChIP-PCR was performed in HEK293T cells expressing either p-mIP-696-Luc or p-mIP-696-GlisBS-Luc as well as FLAG-tagged Pdx1, MafA, or Glis3. This analysis showed that Glis3 was associated with the proximal Ins2 promoter region and that the interaction was significantly diminished when the 2 GlisBS were mutated (Figure 7). Likewise, Pdx-1 and MafA were recruited to the proximal Ins2 promoter, and their association was greatly decreased when, respectively, the A- and C-boxes were mutated. Most interestingly, mutation of the 2 GlisBS significantly reduced the association of Pdx-1 and MafA with the proximal Ins2 promoter region. These observations suggest that Glis3 binding at the GlisBS promotes or stabilizes the recruitment of Pdx1 and MafA protein complexes to the Ins2 promoter.

Figure 7.

GlisBS are required for stable binding of Pdx1 and MafA to the Ins2 promoter. HEK293T cells were transfected with p-mIP-696-Luc or the indicated mutant. Cells were cotransfected with pCMV10–3xFLAG empty vector or pCMV10–3xFLAG-Pdx1 and -NeuroD1, or pCMV10–3xFLAG-MafA as indicated. After 48 hours, protein-DNA complexes were immunoprecipitated using anti-M2 FLAG-conjugated agarose beads as described in Materials and Methods. Immunoprecipitated DNA was quantified by quantitative PCR using primers amplifying a region of the mIns2 promoter (black bars) or hGAPDH (gray bars). Each bar represents chromatin immunoprecipitated DNA as a percentage of the input (1:100 dilution) ± SE. a, statistically different from p-mIP-696-Luc with FLAG empty vector (P < .01); b, statistically different from p-mIP-696-Luc with FLAG empty vector and p-mIP-696 Luc with FLAG Pdx1, MafA, or Glis3 (P < .01).

Pdx1, NeuroD1, and MafA cannot optimally activate the human mutant INS-700(c93g) promoter

Recently, a recessive c to g mutation (c93g), which was linked to the development of neonatal diabetes in 7 families, was identified at position −93 relative to the transcriptional start site of the INS gene (38). This mutation occurs within the proximal promoter region of the INS gene and is located within a putative GlisBS (11, 13). To examine the effect of this mutation on the activation of the INS promoter by endogenous transcription factors, we compared the transcriptional activation of the wild-type INS-700 and the mutant INS-700(c93g) promoter in INS1 cells. As expected, the c93g mutation within the GlisBS reduced the activation of the promoter by endogenous proteins by approximately 70%, whereas increased expression of exogenous FLAG-Glis3 restored promoter activity (Figure 8A and Supplemental Figure 3A). Similarly, activation of the INS promoter was reduced by >85% in βTC-6 cells and restored by exogenous Glis3 expression (Supplemental Figure 3, B and C). Moreover, the c93g mutation significantly reduced the activation of the INS promoter by exogenously expressed Pdx1, NeuroD1, and MafA in HEK293T cells and coexpression of exogenous Glis3 increased activation synergistically in a concentration-dependent manner (Figure 8B). These observations are consistent with the hypothesis that overexpression of Glis3 might be able to overcome a reduced affinity for the mutated binding site and restore INS promoter activation.

Figure 8.

A SNP in the GlisBS within the INS proximal promoter reduces INS transactivation by Pdx1, NeuroD1, and MafA. A, INS1 832/13 cells were transfected with pGL4.10 empty vector, p-INS-700-Luc or p-INS-700(c93g)-Luc and pCMV10–3xFLAG empty vector or increasing amounts of p-CMV10–3xFLAG-Glis3 (5, 20, or 50 ng of plasmid/well). After 48 hours, cells were harvested and assayed for luciferase and β-galactosidase activities, and the normalized relative Luc activity (nRLU) was plotted. Each bar represents the mean ± SE. a, statistically different from hINS-700-Luc and pCMV10–3xFLAG empty vector (P < .01). B, HEK293T cells were transfected with p-INS-700-Luc or p-INS-700-(c93g)-Luc and pCMV10–3xFLAG empty vector or pCMV10–3xFLAG-Pdx1, -NeuroD1, and –MafA in combination with or without increasing concentrations of p-CMV10–3xFLAG-Glis3 as indicated. After 48 hours, cells were harvested and assayed for luciferase and β-galactosidase activities, and the fold increase relative to p-INS-700-Luc and empty vector was plotted. Each bar represents the mean ± SE. Dashed lines indicate fold activation generated by p-INS-700-Luc and empty vector or p-INS-700-Luc and pCMV10–3xFLAG-Pdx1, -NeuroD1, and –MafA in combination. a, statistically different from p-INS-700-Luc and pCMV10–3xFLAG empty vector (P < .01); b, statistically different from p-INS-700-Luc and pCMV10–3xFLAG-Pdx1, -NeuroD1, and –MafA (P < .01).

Discussion

Previous studies identified Glis3 as a critical factor in the development of the endocrine pancreas and demonstrated that Glis3 regulates Ins2 transcription through its interaction with 2 GlisBS located at −99 and −263, relative to the Ins2 transcriptional start site (11, 13). In this report, we examined in detail the activation of insulin transcription by Glis3 and its interaction with 3 additional transcription factors integral to β-cell development and maintenance (Pdx1, NeuroD1, and MafA). Our study shows that Glis3 synergistically activated the mouse and human insulin promoters when coexpressed together with Pdx1, NeuroD1, and MafA individually or in combination (Figures 1D and 3, A and B). The cooperation between Glis3 and Pdx1, NeuroD1, and MafA required binding of Glis3 to the GlisBS as evidenced by the lack of synergism by a ZFD mutant of Glis3 that is unable to bind the GlisBS (Figure 2A). Taken together, these observations indicated that Glis3 plays a key role in the optimal transcriptional activation of Ins2 by acting cooperatively with Pdx1, NeuroD1, and MafA at the Ins2 promoter.

The importance of Glis3 and GlisBS in Ins2 activation was further demonstrated by data showing that mutation of the 2 GlisBS greatly reduced the activation of the Ins2 promoter by exogenously expressed Glis3 (Figure 1C) and almost totally abolished the activation by endogenous transcription factors in INS1 cells (Figure 1B). Moreover, the GlisBS mutations dramatically reduced the activation of Ins2 by exogenous Pdx1, NeuroD1, and MafA in HEK293T cells (Figure 6A), whereas knockdown of endogenous GLIS3 expression greatly inhibited the activation of the Ins2 promoter by Pdx1, NeuroD1, and MafA (Figure 6B), suggesting that optimal activation of Ins2 by these factors appears to depend on the interaction of Glis3 with GlisBS. The significance of Glis3 and the GlisBS in the transcriptional activation of Ins2 by the 4 transcription factors was further supported by data showing that although loss of either the A-, E-, or C-box or GlisBS diminished the activity of the Ins2 promoter, the loss of the GlisBS had by far the greatest effect on promoter activation (Figure 3A). These observations suggest that the 2 GlisBS located within the Ins2 promoter are not only critical for the transactivation of insulin by Glis3 but are also integral for the transcriptional activation of insulin by Pdx1, NeuroD1, and MafA.

The transactivation of target genes by transcription factors is mediated in part via the recruitment of intermediary factors that communicate with the basic transcriptional machinery. This includes recruitment of coactivators, such as CBP and p300, which through their intrinsic HAT activity and their ability to act as scaffolds for the assembly of multiprotein transcriptional complexes function as key regulators of RNA polymerase II−mediated transcription (37, 39). In this study, we demonstrate that Glis3 is able to interact with CBP as has been reported previously for Pdx1, NeuroD1, and c-Maf (18, 19, 30–32). A proline-rich region at the Glis3 N terminus as well as its C-terminal TAD mediate the interaction with CBP, whereas the C/H3 domain and Q-rich region within the C terminus of CBP are involved in the interaction with Glis3. Exogenous expression of CBP enhanced Glis3-induced transcriptional activation of the Ins2 promoter as well as that by Pdx1, NeuroD1, and MafA. Furthermore, replacement of the Glis3 TAD with the VP16-AD, which can recruit CBP/p300 (37, 40), also resulted in synergistic activation of the insulin promoter with Pdx1, NeuroD1, or MafA (Figure 2B). The adenoviral oncoprotein, E1A, which inhibits CBP activity (41), greatly reduced the transcriptional activation of the Ins2 promoter by Glis3 as well as the synergistic activation by Glis3 and Pdx1, NeuroD1, and MafA, further suggesting that recruitment of CBP is integral to these processes. Synergism between Glis3 and Pdx1, NeuroD1, and MafA is dependent on the C-terminal transactivation domain of Glis3, which was shown to associate with the Q-rich region of CBP (Figure 4D). Interestingly, the Q-rich region of CBP/p300 has previously been demonstrated to be crucial for the activation of insulin transcription in β-cells (20). Together, these observations suggest that CBP is part of a coactivator complex recruited by Glis3 and that both the optimal activation of Ins2 and the synergism between Glis3 and the other transcription factors are dependent on interactions with CBP, as well as CBP activity (Figure 9).

Figure 9.

Model of the proposed role of Glis3 and CBP in the formation of the transcriptional activation complex regulating insulin gene transcription. Top, In the absence of Glis3 or functional GlisBS, Pdx1, NeuroD1, and MafA only transiently associate with the Ins2 promoter. Bottom, Glis3 binds to the GlisBS in the proximal insulin promoter and recruits the coactivator, CBP/p300, whereas Pdx1, NeuroD1, and MafA complexes bind to adjacent A-, E-, and C-boxes, respectively. Binding of Glis3 to GlisBS and recruitment of CBP/p300 might provide a scaffold for the formation of a larger transcriptional regulatory complex, possibly involving additional CBP/p300 proteins, which modifies the chromatin and stabilizes the binding of Pdx1, NeuroD1, and MafA complexes to their respective binding sites. The activation complex activates insulin transcriptional through interaction with the general transcriptional machinery (GTF).

The association of CBP with all 4 factors and the fact that their binding sites are in close proximity to each other at the proximal INS and Ins2 promoter, suggested that Glis3 is a part of a larger transcriptional regulatory complex. This hypothesis is supported by studies showing that Pdx1, NeuroD1, and MafA are able to interact directly with each other (21, 42). Moreover, Glis3 has been reported to interact with Pdx1, NeuroD1, and MafA in INS1 cells (13); however, whether those associations are direct or mediated through common association with coactivators such as CBP/p300 has not yet been elucidated. Our data showing that the GlisBS are critical for activation of Ins2 by Pdx1, NeuroD1, and MafA suggest a key role for Glis3 and the GlisBS in the regulation of insulin by such a multiprotein complex. This conclusion is further supported by ChIP analysis, which demonstrated that mutations in the GlisBS diminished the association of Pdx1 and MafA to their respective enhancer elements (Figure 7). Taken together, these data suggest that binding of Glis3 to the GlisBS might stabilize Pdx1 and MafA activator complexes at the Ins2 promoter. A schematic model of the proposed Glis3-CBP/p300 transcriptional complex is depicted in Figure 9. In this model, Glis3 recruits ≥1 CBP/p300 proteins and consequently promotes the assembly and stabilization of a multiprotein transcription factor complex containing MafA, NeuroD1, and Pdx1.

There is evidence that the mechanisms implicated in the pathogenesis of various diseases, including type 2 diabetes, coalesce around a limited set of core pathways and networks. A recent, large-scale association analysis revealed that type 2 diabetes susceptibility loci are enriched for transcription factors interacting with CBP, suggesting that the CBP-linked transcription factors play an important role in type 2 diabetes susceptibility (43). The interaction between Glis3 and CBP reported in this study, the ability of Glis3 to transcriptionally regulate Ins2, and data from genome-wide association studies identifying GLIS3 as a risk locus for both type 1 and 2 diabetes (3, 6–9, 13, 44) are consistent with this concept and suggest that Glis3 is part of such a network. Although the mechanisms by which noncoding single nucleotide polymorphisms (SNPs) within GLIS3 lead to an increased risk for type 2 diabetes are not yet understood, these SNPs might affect the transcriptional activation and level of expression of GLIS3, possibly involving regulation by noncoding RNAs. Subsequent reduction in GLIS3 protein might then affect the recruitment of CBP and the stability of transcriptional enhancer complexes at the Ins2 promoter. Recently, recessive homozygous mutations resulting in single nucleotide changes in the human insulin promoter within the region that corresponds to GlisBS1 were found to be associated with the development of neonatal diabetes in 7 families (38). We showed that one such mutation in which a c to g mutation at position −93 relative to the transcriptional start site within the human insulin promoter reduced INS promoter activity in INS1 and βTC6 cells by 70% and 85%, respectively (Figure 8A and Supplemental Figure 3), consistent with a previous report using INS1 or βTC3 cells (45) and similar to the GlisBS mutations within the Ins2 promoter (Figure 1B). The reduced activation of the INS(c93g) promoter might be due to decreased affinity of Glis3 for the mutated binding site. In fact, in vitro studies have indicated that mutation of the corresponding cytosine within the consensus GlisBS significantly diminished its affinity for Glis3 (24). Increased expression of Glis3 restored the activation of INS(c93g) to levels similar to that of the wild-type promoter and enhanced the greatly reduced activation of the INS(c93g) by exogenous Pdx1, NeuroD1, and MafA (Figure 8, A and B). It is possible that high levels of Glis3 expression compensate for the reduced affinity of Glis3 for the c93g mutant GlisBS and restore INS activation. These findings are consistent with our conclusion that the stability of the transactivation complex containing CBP/p300, Pdx1, MafA, and NeuroD1 at the insulin promoter might be compromised in the absence of functional GlisBS.

In summary, the data presented in this report show that Glis3 plays a critical role in the transcriptional regulation of insulin in pancreatic β-cells by recruiting CBP/p300, which may serve as a scaffold for the formation of a larger transcriptional regulatory complex containing Pdx1, NeuroD1, and MafA at the insulin promoter. Our data further suggest that the binding of Glis3 to the GlisBS and its recruitment of CBP/p300 are necessary to stabilize the interaction between Pdx1, NeuroD1, and MafA and their respective binding sites within the insulin promoter. This conclusion is consistent with findings showing that sustained expression of Glis3 is required for pancreatic β-cell generation in mice and humans, as well as for the maintenance of adult β-cell functions, including insulin expression (43). Future studies are needed to confirm the mechanisms underlying the transcriptional regulation of insulin by Glis3 in a more physiological context. Such studies are, however, challenging at the current time because Glis3 knockout mice fail to develop mature β-cells, and a suitable Glis3 antibody is not currently available. A better understanding of the role Glis3 plays in regulating insulin gene expression provides new insights not only into the control of βcell physiology but also into mechanisms by which Glis3 dysfunction results in the development of diabetes (3, 6–9, 13, 44) and may therefore lead to novel therapeutic strategies in the management of diabetes.