Pdx1 and BETA2/NeuroD1 Participate in a Transcriptional Complex That Mediates Short-range DNA Looping at the Insulin Gene

Abstract

The activity of the insulin gene, Ins, in islet β cells is thought to arise in part from the synergistic action of the transcription factors Pdx1 and BETA2/NeuroD1. We asked how the binding of these factors to A and E elements many tens or hundreds of base pairs upstream of the start site could influence activity of transcriptional machinery. We therefore tested the hypothesis that the complex of Pdx1 and BETA2/NeuroD1 maintains a DNA conformation such that distal regions of the gene are brought into proximity of the promoter and coding region. We show by coimmunoprecipitation that Pdx1 and BETA2/NeuroD1 exist within a complex and that the two physically interact with one another in this complex as assessed by fluorescence resonance energy transfer. Consistent with this interaction, we found that the two factors simultaneously occupy the same fragment of the Ins gene in β cell lines using the chromatin immunoprecipitation/re-chromatin immunoprecipitation assay. Using a modification of the chromosome conformation capture assay in vitro and in β cells, we observed that Pdx1 and BETA2/NeuroD1 mediate looping of a segment of Ins that brings EcoRI sites located at –623 and +761 bp (relative to the transcriptional start site) in proximity to one another. This looping appears to be dependent in vitro upon an intact A3 binding element, but not upon the E2 element. Based on our findings, we propose a model whereby Pdx1 and BETA2/NeuroD1 physically interact to form a nucleoprotein complex on the Ins gene that mediates formation of a short DNA loop. Our results suggest that such short loop conformations may be a general mechanism to permit interactions between transcription factors and basal transcriptional machinery.

Transcription factors function as crucial “switches” in the embryonic development of the mammalian pancreas, as they control the expression of genes in a defined temporal and spatial pattern. Of particular interest recently has been the role of transcription factors in the development of the β cells of the pancreatic islets of Langerhans. Islet β cells are solely responsible for the secretion of physiologically relevant amounts of insulin into the circulation, and their dysfunction is associated with the development of not only Type 1 but also Type 2 diabetes mellitus (1). The ParaHox factor Pdx1 and the basic helix-loop-helix factor BETA2/NeuroD1 (henceforth referred to as “NeuroD1”) are necessary for the formation of the pancreas and β cells, respectively. Pdx1 is expressed broadly in the early pancreas, and pdx1-null mice (and humans with homozygous defects) completely lack a pancreas (2, 3). Pdx1 later becomes restricted to β cells, where it activates an array of β cell-specific genes and is necessary for normal glucose homeostasis (for a review, see Ref. 4). NeuroD1 is expressed in developing endocrine cells of the pancreas, and its deletion in mice results in the severe reduction of endocrine cell formation and the eventual development of diabetes (5, 6). In the mature pancreas, NeuroD1 expression is limited to β cells, where it heterodimerizes with the ubiquitous basic helix-loop-helix protein E47 to activate genes necessary for β cell function (7–9).

A common downstream target of both Pdx1 and NeuroD1 in the mature β cell is the gene encoding preproinsulin (referred to here as the “Ins” gene). Multiple discrete DNA elements (A-, E-, and C-boxes) within the 5′ regulatory region (the “enhancer”) of the Ins gene serve as binding sites for several β cell-specific and ubiquitous transcription factors, and a similar genetic organization has been described for a host of other β cell genes, including Iapp, pdx1, Glut2, and Gck (10). Based on DNA binding studies in vitro and reporter gene analysis in transfected mammalian cells, Pdx1 is believed to bind to the A-boxes and NeuroD1 to the E-boxes (11, 12). Importantly, in the rodent and human Ins genes, tandemly occurring E/A-boxes (referred to as the “E2/A3” and “E1/A1” elements) are situated within a crucial 350-bp regulatory region upstream of the transcriptional start site (13). In reporter gene assays, these tandem elements synergistically activate gene transcription response to Pdx1 and NeuroD1 overexpression (12, 14). The mechanisms underlying synergy between these factors remain poorly defined but may involve a physical interaction between the two that allosterically enhances their binding to DNA and/or enhances their recruitment of cofactors to the gene (12, 14).

Another potential mechanism not implicated previously in activation of the Ins gene is DNA looping. A looped DNA structure may allow factors bound at more distant regions of the gene enhancer to gain access to elements of the basal transcriptional machinery, thereby allowing for recruitment or activation of these components (15). Whether such looping occurs in the Ins enhancer and whether Pdx1 and NeuroD1 might contribute to the formation of such a loop have never been investigated. In this report, we present evidence that Pdx1 and NeuroD1 physically interact within the living nucleus and form a transcriptional complex on the endogenous Ins gene. Studies in vitro suggest that the complex involving Pdx1 and NeuroD1 (with its heterodimeric partner E47) leads to a short-range DNA loop on the mouse Ins gene that brings more distal elements of the enhancer in proximity to the promoter region. We propose that this loop may allow for interactions between components of this complex and the basal transcriptional machinery and thereby contribute to synergistic activation of Ins gene transcription.

EXPERIMENTAL PROCEDURES

Cells, Antibodies, and Vectors—The mouse insulinoma cell line βTC3 and the mouse pancreatic ductal cell line mPAC L20 were maintained in Dulbecco's modified Eagle's medium as described previously (16). Mouse monoclonal anti-Pdx1 antiserum was from R&D Systems. Mouse monoclonal anti-NeuroD1 antiserum was a gift from Dr. U. Jhala (University of California at San Diego). Goat anti-NeuroD1 antiserum (Santa Cruz Biotechnology catalog no. sc-1086) and rabbit polyclonal antiserum against acetylated histone H3 were from Millipore.

Expression constructs encoding the monomeric variants of green, yellow, and cyan fluorescent proteins were generated through oligonucleotide-directed mutagenesis (QuikChange®, Stratagene) of the plasmids encoding the respective dimeric variants in the vectors pEGFP, pEYFP, and pECFP (Clontech). Fluorescent fusion proteins used in this study were constructed by subcloning the cDNAs encoding Pdx1, NeuroD1, and Nkx6.1 in-frame with the coding sequences of the respective monomeric fluorescent proteins using SacI and SalI sites. The vector containing the mouse I Ins gene for chromosome conformation capture (3C)3 assays (pCRIns3C) was prepared by PCR-amplifying a 1384-bp EcoRI fragment of the mouse I Ins gene from NIH3T3 genomic DNA and cloning it into vector pCR2.1 using T/A overhangs (Invitrogen). Mutagenesis of pCRIns3C was performed using oligonucleotide-directed mutagenesis. The following oligonucleotides were used for the mutagenesis (top strands shown): E2 mutation (mutation underlined), 5′-GGGAACTGGTTCATCAGGGCGGCCGCTCCCTTATTAAGAC-3′; and A3 mutation (mutation underlined), 5′-GGCCATCTGGTCCCTTATCGCGGCCGCGATAACCCTAAGACTAAG-3′. All vectors were verified by restriction enzyme digestion and automated nucleotide sequencing.

Immunoblot Analysis—Whole cell extracts were prepared by lysing ∼1 × 10⁶ cells in a buffer containing 1% sodium dodecyl sulfate, and 5 μg of extract were resolved by electrophoresis on a 12% SDS-polyacrylamide gel followed by immunoblotting with anti-Pdx1 or anti-NeuroD1 antiserum. Immunoblots were visualized using the ECL-Plus® system (Amersham Biosciences) and quantitated by scanning densitometry using ImageQuant® software (Amersham Biosciences).

Transient Transfection of Cell Lines—For luciferase assays, a total of 5 μg of plasmid DNA consisting of 4 μg of a luciferase reporter plasmid (Ins minienhancer pFoxLuc5FF1) and 0.2 μg of pBAT12-Pdx1, pBAT12-NeuroD1, and/or pBAT14-E47 were mixed with Reagent L (Amaxa, Inc.) and transfected into 1 × 10⁶ mPAC L20 cells using an Amaxa Nucleofector (program T-20) according to the manufacturer's protocol. Cells were harvested 48 h after transfection, and luciferase activities were measured using a commercially available luciferase assay substrate (Promega) and a Sirius luminometer (Berthold Detection Systems). To measure activation of the endogenous Ins promoter in mPAC L20 cells, only reporter DNA was omitted in transfections, and cells were harvested after 96 h for isolation of total RNA and for subsequent real-time reverse transcription-PCR of the Ins transcript as detailed previously (17).

For small interfering RNA (siRNA) knockdown experiments, 2 × 10⁶ βTC3 cells were transfected with 3.2 μg of double-stranded RNAs against Pdx1 (siRNA 47, 5′-GGGAACUUAACCUAGGCGUUU-3′, and siRNA 48, 5′-GGUUAAGAAUAACGAAAGGUU-3′) or NeuroD1 (siRNA 78, 5′-CGAUUAGAGGCACGUCAGUUU-3′, and siRNA 28, 5′-CGAAUCCACUGUGCGUACAUU-3′) using an Amaxa Nucleofector, Reagent V, and program D23 according to the manufacturer's protocol. After 96 h, cells were processed for isolation of total RNA or whole cell extracts as described previously (17). Sample statistics were calculated by one-way analysis of variance followed by the Tukey post test using GraphPad Prism 5.00 software for the Mac.

Quantitative Chromatin Immunoprecipitation (ChIP) and Re-ChIP—ChIP assays, including quantitation of coimmunoprecipitated DNA fragments by real-time PCR using the threshold cycle methodology, were performed as described previously (16). All primer sequences were detailed previously (18). Re-ChIP experiments were performed as described (19). Briefly, following the initial ChIP with the first antibody, protein-DNA complexes were eluted from agarose beads by incubation with 50 μl of 10 mm dithiothreitol at 37 °C for 30 min, and then the eluate was diluted to a final volume of 1 ml of ChIP buffer (50 mm Tris-Cl, pH 8.1, containing 1% Triton X-100, 0.1% deoxycholate, 150 mm NaCl, and 5 mm EDTA). The ChIP protocol was then repeated with the second antibody prior to analysis of recovered DNA fragments by real-time PCR. Sample statistics were calculated using one-way analysis of variance followed by the Tukey post test.

Coimmunoprecipitation Assays—Coimmunoprecipitation of endogenous Pdx1 and NeuroD1 from whole cell lysates of 10⁷ βTC3 cells was performed essentially as described previously using anti-Pdx1 antiserum (20) followed by electrophoresis on a 10% polyacrylamide gel and immunoblotting using monoclonal anti-NeuroD1 antibody.

Fluorescence Resonance Energy Transfer (FRET) Experiments—mPAC L20 cells were transiently transfected with plasmids encoding fluorescent fusion proteins and maintained on glass coverslips in phenol red-free culture medium for 48 h prior to fluorescence microscopy. Images for FRET were acquired on an inverted Olympus IX-70 wide-field microscope equipped with a60× oil objective lens. The Linux-based ISee acquisition software (ISee Imaging Systems) was then used to obtain images and determine gray-level intensity profiles. The filter combinations for sequential filter-based FRET were as follows: 440/20 nm excitation and 480/40 nm emission for ECFP and 545/35 nm emission for EYFP and the 525DRLP dichroic mirror (Chroma Technology Corp.). Data were collected as 12-bit monochromatic images using a cooled charge-couple device camera (CH260, Photometrics, Ltd.). Special care was taken to avoid detector saturation, and exposure times were maintained among all samples for effective post-acquisition analysis of FRET efficiency. Precision FRET data analysis was used to remove spectral bleed-through and variation in the fluorophore expression level (or concentration) for the donor and acceptor molecules (21). The acquired images were processed using NIH ImageJ software, and the precision FRET algorithm was used to calculate the energy transfer efficiency and the distance between donor and acceptor molecules.

Modified 3C Assay—3C assays were performed as described previously (22) but with some modifications. For in vitro 3C assays, a reaction was prepared using 2 μl of in vitro transcribed/translated protein expressed from vectors pET-Pdx1, pET-NeuroD1, and pBAT13-E47 and 40 ng of linearized plasmid pCRIns3C. Reactions were incubated for 45 min at 30 °C in the presence of HeLa cell nuclear extract (to provide basal transcriptional proteins). The protein-DNA complexes were then fixed for 10 min at room temperature with 1% formaldehyde followed by addition of glycine to a final concentration of 0.125 m. The protein-DNA complexes were ethanol-precipitated overnight at –20 °C and resuspended in 30 μl of 10 mm Tris (pH 7.5) with 0.1 mm EDTA. The DNA was digested with EcoRI for 1 h at 37 °C, and the enzyme was heat-inactivated at 65 °C for 20 min. The reaction was diluted 2-fold in ligation buffer (30 mm Tris-HCl, pH 7.8, 10 mm MgCl₂, 10 mm dithiothreitol, 1 mm ATP, and 1% Triton X-100), and 20 ng of the DNA were then subjected to a ligation reaction in a 40-μl volume using 2000 units of T4 DNA ligase for 5 h at 16°C. As a carrier, 40 μg of herring sperm DNA were added, and samples were brought to a volume of 100 μl and treated with proteinase K to digest protein and reverse cross-links. The DNA was purified by phenol/chloroform extraction and ethanol precipitation. For quantitation of ligation products, the DNA was subjected to SYBR Green-based real-time PCR analysis using primers to detect a novel 250-bp product that would occur only in the presence of ligase (forward and reverse primers, 5′-TCTTTTTCTCTGGCATTTATTGTC-3′ and 5′-TCATTGGTCAACTGGGCTG-3′, respectively). As an internal control for input and recovery of DNA, PCR amplification was also performed using primers corresponding to a fragment of the mouse I Ins gene that was unaffected by digestion or ligation (forward and reverse primers, 5′-TCAGCCAAAGATGAAGAAGGTCTC-3′ and 5′-TCAAACACTTGCCTGGTGC-3′, respectively). PCR cycling parameters were as follows: 40 cycles at 95 °C for 30 s, 55 °C for 30 s, and 72 °C for 30 s.

3C assays with 10⁷ βTC3 cells were performed in a similar fashion with a few exceptions. Cross-linking was performed in 1% formaldehyde on adherent cells in situ, and cells were lysed as described previously (23). EcoRI digestion was performed overnight, and a 1:4 dilution of digested DNA was used in the ligation reaction. Subsequent PCR analysis of ligated products was performed using the same internal control primers and primers to detect a novel 197-bp product (forward and reverse primers, 5′-TCTGGCATTTATTGTCATGTTAGCA-2′ and 5′-CCCAACAACCCAAGATATGTACCT-3′, respectively). PCR cycling parameters were as follows: 40 cycles at 95 °C for 30 s, 64 °C for 30 s, and 72 °C for 30 s. In all cases, sample statistics were calculated using one-way analysis of variance followed by the Tukey post test.

RESULTS

Pdx1 and NeuroD1 Synergistically Transactivate Insulin Gene Transcription—Prior studies have demonstrated that Pdx1 and NeuroD1 (with its heterodimer partner E47) synergistically activate a “minienhancer” of the rat I Ins gene that consists of five tandem repeats of the E2/A3 elements (known as the 5-Far-Flat reporter) (12, 14). For our studies, we employed cells derived from mouse pancreatic ductal epithelium (mPAC L20) that contain low levels of Pdx1 and no discernible NeuroD1 but are capable of transdifferentiating into β-like cells upon overexpression of NeuroD1 (24). Consistent with prior findings in non-pancreatic cell types, Fig. 1A demonstrates that whereas transient transfection of Pdx1 and NeuroD1 individually activates the 5-Far-Flat reporter ∼16-fold each in mPAC L20 cells, the combination of Pdx1 with NeuroD1 causes a striking 56-fold activation.

FIGURE 1.

Pdx1, NeuroD1, and E47 synergistically activate the Ins gene in mPAC L20 cells. A, a reporter plasmid containing the luciferase gene under control of the minimal prolactin promoter and five copies of the rat Ins I E2/A3 minienhancer (5-Far-Flat) was cotransfected into mPAC L20 cells with cytomegalovirus promoter-driven expression plasmids encoding the indicated proteins. -Fold induction of luciferase activity was calculated as the activity of cells transfected with empty expression vector (EV) alone defined as 1. Data represent the mean ± S.E. for 10 independent transfections. B, cytomegalovirus promoter-driven expression plasmids encoding Pdx1, NeuroD1, and E47 were transfected into mPAC L20 cells. Real-time reverse transcription-PCR was performed to detect mouse I Ins transcript. -Fold induction of transcript was calculated as the activity of cells transfected with empty expression vector alone defined as 1. Data represent the mean ± S.E. of six independent transfections. * indicates that the value is statistically different (p < 0.05) compared with the expression vector; # indicates that the value is statistically different (p < 0.05) from all doubly transfected values.

Given the limitations imposed by the chromatin structure of endogenous genes, we next asked whether the synergistic activation of Ins reporter genes, as presented here, might be applicable to the endogenous Ins gene. As shown in Fig. 1B, transfection of Pdx1 or NeuroD1 caused no significant activation of the endogenous gene in mPAC cells as determined by real-time reverse transcription-PCR. However, cotransfection of both factors led to a 4.5-fold activation of the gene. Notably, the absolute level of activation of the Ins gene in these cells still remains only a small fraction of that observed in intact islets or β cell lines (∼10^–5; data not shown), implying that limitations at the level of chromatin or of expression of other transcription factors/cofactors prevent high-level expression of the gene in these non-β cells. Overexpression of the NeuroD1 heterodimerization partner E47 did not substantially enhance gene activation in these experiments, presumably because sufficient endogenous E47 is present in this cell line.

Pdx1 and NeuroD1 Are Required to Fully Activate the Endogenous Insulin Gene in β Cells—To demonstrate that full activity of the endogenous Ins gene in β cells and islets requires actions of Pdx1 and NeuroD1, we next performed siRNA studies in βTC3 cells (a mouse β cell-derived cell line) to diminish Pdx1 and NeuroD1 protein levels. Whereas two different siRNAs (47 and 48) against Pdx1 led to 50–60% reductions in Pdx1 protein with no discernible effect on NeuroD1 protein levels, two siRNAs (78 and 28) against NeuroD1 caused 40–50% reductions in both Pdx1 and NeuroD1 protein levels (Fig. 2, A and B); this latter finding is consistent with prior studies that suggest that maintenance of Pdx1 expression in mature β cells may be dependent upon NeuroD1 (6, 25). Fig. 2C demonstrates that siRNAs 47 and 48 against Pdx1 result in small but significant 20–30% reductions in Ins mRNA levels; siRNAs 78 and 28 against NeuroD1 result in even greater 90 and 70% reductions, respectively, in Ins mRNA levels. The greater reductions in Ins mRNA levels with siRNAs against NeuroD1 likely result from the combined knockdown of both NeuroD1 and Pdx1 in these cells. Taken together, the data in Fig. 2 suggest that NeuroD1 and Pdx1 are required for full activation of the Ins gene.

FIGURE 2.

The Ins gene activity is dependent upon Pdx1 and NeuroD1. βTC3 β cells were transfected with negative control siRNA (si-Control) and two different siRNA sequences against Pdx1 (si-Pdx1) or NeuroD1 (si-NeuroD1) and harvested after 96 h. A, representative immunoblots for Pdx1 and NeuroD1 and histone H3 as a loading control following transfection with siRNAs. B, quantitation of Pdx1 and NeuroD1 protein levels following transfection with siRNAs. Data represent the mean ± S.E. of at least four independent siRNA transfections. C, real-time reverse transcription-PCR analysis of Ins mRNA after transfection of βTC3 cells with the indicated siRNAs. Data are expressed as -fold changes relative to cells treated with negative control siRNA. Data represent the mean ± S.E. of six independent siRNA transfections. * indicates that the value is statistically different (p < 0.05) compared with control siRNA-treated cells.

Pdx1 and NeuroD1 Physically Interact in the Nucleus—Evidence for the physical interaction between Pdx1 and NeuroD1 comes primarily from functional reporter assays (as in Fig. 1 above) and from glutathione S-transferase fusion pulldown studies in vitro (12, 14). To determine whether the two proteins interact endogenously in a cellular context, we performed coimmunoprecipitation studies using extracts from βTC3 cells. As shown in Fig. 3A, immunoprecipitation of Pdx1 protein from these extracts using Pdx1 antiserum resulted in coimmunoprecipitation of NeuroD1, a result that was not observed when normal rabbit serum was used instead. These findings suggest that the two proteins are components of an intracellular complex.

FIGURE 3.

Pdx1 and NeuroD1 physically interact with each other. A, extracts from βTC3 cells were immunoprecipitated (IP) using the antiserum against Pdx1 and then subjected to immunoblot (IB) analysis using monoclonal anti-NeuroD1 antibody. B, mPAC L20 cells were cotransfected with the 5-Far-Flat luciferase reporter plasmid and cytomegalovirus promoter-driven expression plasmids encoding the indicated fusion proteins. -Fold induction of luciferase activity was calculated as the activity of cells transfected with empty expression vector alone defined as 1. Data represent the mean ± S.E. of five independent transfections. C and D, mPAC L20 cells were transfected with a 2:3 ratio of donor (ECFP-Pdx1 or ECFP-Nkx6.1) to acceptor (EYFP-NeuroD1) expression plasmid as indicated. Images are representative examples of doubly transfected cells. Dex, donor excitation; Dem, donor emission; Aem, acceptor emission; UFRET, uncorrected FRET image; PFRET, precision (corrected) FRET image. The histograms shown below each image display corresponding pixel count as a function of pixel intensity. FRET percent efficiency (E%) is derived from the mean ± S.E. of at least 10 cells.

To observe the interactions of Pdx1 and NeuroD within the living nucleus, we utilized fluorescence microscopy with fluorescent protein fusions of Pdx1 and NeuroD1. To ensure that fluorescent protein fusions of Pdx1 and NeuroD1 remain functional, we first tested the ability of enhanced EGFP-Pdx1 and EGFP-NeuroD1 fusion proteins to activate the 5-Far-Flat minienhancer. As shown in Fig. 3B, both EGFP-Pdx1 and EGFP-NeuroD1 transactivate the minienhancer 3- and 7-fold, respectively, in mPAC L20 cells, and the combination of the two proteins synergistically activates the minienhancer 18-fold. We next used fluorescent protein fusions in FRET studies. FRET measures the transfer of excitation energy from a donor fluorophore to an acceptor fluorophore over a distance of <7 nm (26). For these studies, proteins were fused to either ECFP (donor fluorophore) or EYFP (acceptor fluorophore) and transiently transfected into mPAC L20 cells. Each fluorophore-specific image was collected at similar gray-level intensities, and identical conditions were then used to acquire images from doubly-transfected cells using the FRET filter set, this allowed the measurement of the donor and acceptor spectral bleed-through background signals in the FRET channel. In a cell coexpressing ECFP-Pdx1 and EYFP-NeuroD1 (Fig. 3C), the mean gray-level intensity in the corrected FRET image (precision FRET) was 638 units after adjustment for bleed-through with control cells individually expressing ECFP-Pdx1 or EYFP-NeuroD1. Based on an analysis of 10 cells, this persistent intensity in the corrected FRET channel translated to a FRET efficiency of 24.8%. These results suggest the existence of a physical interaction (<7 nm) between Pdx1 and NeuroD1 in the nucleus. To rule out the possibility that the observed FRET is nonspecific, next we repeated these experiments using another β cell-specific homeodomain factor, Nkx6.1. As shown in Fig. 3D, in cells coexpressing ECFP-Nkx6.1 and EYFP-NeuroD1, the FRET efficiency was 0.8% (based on a population analysis of 10 cells). These latter results imply that NeuroD1 likely does not interact with homeodomain proteins in general.

Pdx1 and NeuroD1 Occupy the Same Fragment of the Endogenous Insulin Gene Enhancer—To demonstrate that Pdx1 and NeuroD1 simultaneously occupy the same fragment of the Ins gene enhancer, we next performed ChIP and re-ChIP using extract from βTC3 cells. As shown in Fig. 4 (A and B), ChIP analysis using antiserum to Pdx1 or NeuroD1 reveals that Pdx1 and NeuroD1 individually occupy the same fragment of the proximal Ins enhancer (encompassing a genomic region within 500 bp of the transcriptional start site) but do not occupy the inactive myoD1 gene enhancer. Following ChIP using anti-serum to Pdx1, we then performed re-ChIP using antiserum against NeuroD1. This re-ChIP analysis (Fig. 4C) reveals that each factor simultaneously occupies the proximal Ins enhancer and therefore both factors are likely components of the same transcriptional complex on the gene. When the re-ChIP was repeated using the reverse antiserum combination (NeuroD1 first, then Pdx1), there was a trend in the data suggesting simultaneous occupancy (Fig. 4D). As expected, neither factor simultaneously occupies the inactive myoD1 gene in β cells (Fig. 4, C and D).

FIGURE 4.

Pdx1 and NeuroD1 occupy the same fragment of the mouse Ins gene. βTC3 cells were subjected to ChIP or re-ChIP assay. Real-time PCR was used to quantitate recovery of the proximal Ins and myoD1 promoters, and results are expressed as the percent recovery of each gene fragment relative to input DNA. A, ChIP using normal rabbit serum (NRS) or Pdx1 antibody (Ab). B, ChIP using normal rabbit serum or NeuroD1 antibody. C, re-ChIP assay using normal rabbit serum or Pdx1 antibody followed by NeuroD1 antibody. D, re-ChIP assay using normal rabbit serum or NeuroD antibody followed by Pdx1 antibody. Data represent the mean ± S.E. of four independent experiments. * indicates that the value is significantly different (p < 0.05) compared with normal rabbit serum.

Pdx1 and NeuroD1 Mediate DNA Looping at the Insulin Gene—A looped DNA structure may allow factors bound at more distant regions of a gene enhancer to gain physical access to elements of the basal transcriptional machinery, thereby allowing for recruitment or activation of these components (15). To investigate whether the interaction between Pdx1 and NeuroD1 could mediate the formation of a DNA loop at the Ins promoter, we first developed and tested a novel in vitro modification of the 3C assay. The 3C assay has been used primarily to demonstrate DNA looping across expansive regions of DNA thousands of base pairs in length (22). The assay is based on the principle that if looping of DNA occurs (as a result of protein-protein interactions) such that distant restriction sites are brought into close proximity, then restriction of these ends followed by religation would generate a unique DNA product in which the frequency of formation would be significantly greater than if the looping did not occur. Product abundance is then detected by quantitative real-time PCR using primers that cross the ligation site (see schematic in Fig. 5A).

FIGURE 5.

Pdx1, NeuroD1, and E47 induce looping of the Ins gene in vitro. A, schematic presentation of the 3C assay using the mouse I Ins gene as an example. Ovals represent interacting proteins; numbers indicate positions in base pairs relative to the transcriptional start site (+1); arrows indicate positions of PCR primers; the double arrow indicates the position of the PCR amplicon; and RI represents the EcoRI site. B, 2% agarose gel following in vitro 3C reactions containing the indicated proteins. C, real-time PCR data from in vitro 3C reactions containing the indicated proteins and the mouse I Ins gene as a template. The quantity of looped product formed in the absence of added protein was defined as 1. * indicates that the value is significantly different (p < 0.05) compared with the “no protein” sample. Data represent the mean ± S.E. of at least four independent experiments.

Within the mouse I Ins gene, EcoRI restriction sites are located at–623 and +761 bp (relative to the transcriptional start site), encompassing a region of ∼1400 bp. We hypothesized that if occupancy of Pdx1 and NeuroD1 (with its heterodimeric partner E47) at the E2/A3 and E1/A1 elements occurred such that a protein complex formed, this complex may lead to enhanced looping of the DNA, allowing the two EcoRI sites to come into proximity to one another (see Fig. 5A). We first performed the assay in vitro using rabbit reticulocyte lysate-transcribed/translated proteins, HeLa nuclear extract (to provide basal transcriptional machinery), and a fragment of the mouse I Ins gene. Successful production of full-length Pdx1, NeuroD1, and E47 occurred using appropriately programmed rabbit reticulocyte lysate in vitro, as demonstrated by [³⁵S]Met incorporation (data not shown). By conventional PCR analysis, Fig. 5B shows that in the absence of ligase, no “looped product” was detected. However, in the presence of ligase, this product is detectable, and its abundance is substantially enhanced upon the addition of rabbit reticulocyte lysate programmed for Pdx1, NeuroD1, and E47. In all cases, a control PCR product (representing an unaffected portion of the Ins gene) was equivalently detectable (Fig. 5B). We next applied real-time PCR analysis to these in vitro studies. As shown in Fig. 5C, whereas Pdx1, NeuroD1, or E47 individually did not significantly enhance looping beyond that observed with unprogrammed rabbit reticulocyte lysate, a trend to a greater (5–10-fold) increase in looping occurred upon the addition of pairs of these proteins. Importantly, looping was statistically significantly enhanced to a more than additive 25-fold when all three proteins were added. These results suggest that cooperative looping of DNA at the Ins gene can occur in the presence of Pdx1 and NeuroD1 (with its heterodimerization partner E47).

DNA Looping Is Dependent upon an Intact A3 Element—To determine whether the cooperative looping observed in Fig. 5C is dependent upon the binding of Pdx1 and NeuroD1/E47 to respective A and E elements, we next mutated these elements and determined their effects on DNA looping. As shown in Fig. 6A, mutation of the E2 element did not affect the ability of the three proteins to cause synergistic looping of DNA. By contrast, Fig. 6B shows that mutation of the A3 element completely abrogated loop formation. These results are consistent with studies emphasizing the particular importance of the A3 element in Ins gene transactivation (27).

FIGURE 6.

DNA looping in vitro is dependent upon an intact A3 element. In vitro 3C reactions were performed as in Fig. 5 but with the use of mutated mouse I Ins genes as templates as indicated in the schematics at the top of each panel. A, E2 element mutation as a template. * indicates that the value is significantly different (p < 0.05) compared with the no protein sample. B, A3 element mutation as a template. The quantity of looped product formed in the absence of added protein was defined as 1. Data represent the mean ± S.E. of at least four independent experiments. EV, expression vector.

Finally, to determine whether looping of the endogenous Ins gene in β cells is dependent upon Pdx1 and NeuroD1, we performed siRNA knockdown of Pdx1 and NeuroD1 in βTC3 cells. As shown in Fig. 7, formation of the looped product in the presence of ligase is diminished 3–5-fold following depletion of either Pdx1 or NeuroD1, consistent with the requirement of either of these factors for the formation of looped DNA on the endogenous Ins gene. The extent of Pdx1 and NeuroD1 knockdown in the experiments in Fig. 7 (∼40–60% compared with control siRNA) is included in the quantification shown in Fig. 2B.

FIGURE 7.

Depletion of Pdx1 and NeuroD1 in βTC3 cells diminishes looping at the mouse I Ins gene. βTC3 cells were transfected with the indicated siRNAs, and then cells were harvested for the 3C assay 96 h later. The levels of protein knockdown from these experiments are averaged into the data shown in Fig. 2B. Real-time PCR was used to quantitate the presence of the looped PCR product. The quantity of looped product formed in the presence of control siRNA (siControl) was defined as 1. siPdx, Pdx1 siRNA; siNeuroD1, NeuroD1 siRNA. * indicates that the value is significantly different (p < 0.05) compared with the no protein sample. Data represent the mean ± S.E. of at least four independent experiments.

DISCUSSION

To date, the prevailing model of Ins gene transcription has focused on the concept of combinatorial factor synergy; that is, that a group of transcription factors within the β cell (no one factor that, by itself, is entirely exclusive to the β cell) occupies the Ins gene enhancer and collectively synergizes to draw basal transcriptional machinery to the promoter (28). Although the collective occurrence of transcription factors may be necessary for gene activation, to date no studies have definitively shown that these factors occur as “complexes” that reside on the Ins gene.

To address the role of transcription factor complexes in Ins gene activation, we employed mPAC L20 cells. mPAC L20 cells are derived from pancreatic ductal cells (29), and based on gene expression profiling these cells have been suggested to represent a potential β cell “precursor” (24). Our reporter gene results in transfected mPAC L20 cells suggest that the combination of factors Pdx1 and NeuroD1 can activate the Ins gene in a synergistic fashion when neither factor alone is capable. Importantly, our findings also extend to the endogenous Ins gene in mPAC L20 cells, where expression of both Pdx1 and NeuroD1 leads to measurable activity of the gene. However, we note that the absolute level of endogenous gene activation that we observed (based on threshold cycle numbers) is still several orders of magnitude lower than that observed in insulinoma cells or primary islets. This finding emphasizes that full activation requires additional transcription factors (10) and the appropriate chromatin structural accessibility to these factors, particularly in this cell type (30). In this regard, our studies suggest that activation of the Ins gene during cellular differentiation may occur in a gradual fashion as the developing cell acquires necessary transcription factors and appropriately remodels its chromatin.

At least two mechanisms have been proposed for the synergy between Pdx1 and NeuroD1. First, from interaction studies in vitro, it has been proposed that Pdx1 and NeuroD1 physically interact with one another. Based on electrophoretic mobility shift assays, this interaction has been shown to increase each factor's affinity for its respective DNA binding site (12). Our coimmunoprecipitation studies presented here demonstrate for the first time that NeuroD1 and Pdx1 exist as a complex in β cells, and colocalization and FRET studies suggest that this complex likely involves a physical interaction between the two proteins. Second, both Pdx1 and NeuroD1 are known to directly interact with and recruit the transcriptional co-activator p300 to the Ins gene (9, 17, 31, 32). p300 may serve as both a chromatin-modifying factor (through acetylation of histone proteins) and a “bridge” to link the transcriptional complex to components of the basal transcriptional machinery (33). In this regard, we demonstrate here using the re-ChIP assay that both Pdx1 and NeuroD1 simultaneously bind to the Ins enhancer; thus, the co-occupancy of the Ins enhancer by both factors would be predicted to augment p300 recruitment and hence gene transcription.

Another previously unexplored mechanism that may contribute to synergistic activation by Pdx1 and NeuroD1 is DNA looping. Here, the term “loop” refers to a conformation of DNA such that distant cis elements of a DNA strand are physically in proximity to one another. DNA looping between distant DNA elements may facilitate to bring a locally high concentration of transcription factors, cofactors, and chromatin-modifying factors near the transcriptional start site of genes and thereby contribute to transcriptional activation. This mechanism has been used to explain why enhancer elements located hundreds to thousands of base pairs away from the transcriptional start site can contribute to gene activation (15). Using the 3C assay, looping of this nature has been demonstrated to provide proximity between the β-globin locus control region and the β-major globin promoter, which are located ∼50 kilobase pairs apart (23, 34). Interestingly, formation of this long-range loop at the β-globin locus is at least partially mediated by the transcription factor GATA-1 and its cofactor FOG-1, suggesting that these factors serve as anchors in stabilizing the loop (34).

Our studies of Ins gene regulation provide a new perspective on looping and suggest that looping can occur over a much shorter span of hundreds of base pairs in relatively nucleosome-free segments of DNA. The occurrence of this short-range DNA loop is consistent with the observation that only ∼300 bp of the gene upstream of the promoter are sufficient to confer high-level, cell-specific activity of the promoter in both cell-based systems and transgenic animals (13). We show here using a novel application of the 3C technique in vitro that two EcoRI sites located at –623 and +761 bp (relative to the transcriptional start site) are brought into close proximity of one another such that ligation of these EcoRI sites is statistically enhanced by the presence of Pdx1, NeuroD1, and E47. Although each factor alone contributes little to DNA looping in this system, the combination of factors leads to a synergistic increase in looping. Interestingly, mutation of the A3 Pdx1 binding site abolishes the synergistic looping, whereas mutation of the E2 site does not. This finding emphasizes both the importance of the A3 site in Ins gene activation (27, 28) and the potential redundancy of E-box elements within the enhancer. Alternatively, these data could suggest that DNA binding by basic helix-loop-helix factors is not necessary for looping to occur, only their interaction with Pdx1. Consistent with these observations in vitro, we also found that loss of Pdx1 (using siRNAs against either Pdx1 or NeuroD1) within β cells leads to a decrease in both Ins transcription and looping at the endogenous gene.

Taken together, our data provide a framework for modeling the molecular events that lead to Ins gene activation in islet β cells. Central to this model is the formation of a DNA loop that brings into proximity a distal region at about –600 bp near the coding region of the gene. We recognize that our data do not strictly establish a cause- and-effect relationship between DNA looping and transcription at the Ins gene, only a correlation. With this limitation in mind, why might Pdx1-mediated looping be essential to gene activation? Our prior studies have shown that depletion of Pdx1 in βTC3 cells results in the loss of Ser² phosphorylation of the C-terminal domain of RNA polymerase II at the Ins promoter (18). This modification is necessary for the conversion of the initiation isoform of RNA polymerase II to the elongation isoform (35). It is therefore tempting to speculate that looping mediated by Pdx1 might allow a transcriptional complex to form in the proximity of the promoter such that cofactors that harbor kinase activity permit the activation of RNA polymerase II. We therefore propose that the transcriptional complex mediating Ins gene activation should be viewed in the context of the loop model depicted in Fig. 8. Currently, our laboratory is investigating Pdx1-interacting proteins that harbor enzymatic activities that alter both RNA polymerase II activity and local chromatin structure.

FIGURE 8.

Model of transcriptional complex formation at the Ins gene. The figure shows schematic representations of Ins with the positions of the A and E elements and major transcription factors that are known to regulate its transcription. The upper schematic shows the “classic” model of the complexes that form on the gene (adapted from Refs. 10 and 28). The lower schematic shows the proposed loop model, in which factors bound at more distal elements are brought into proximity of the promoter region with the potential to influence RNA polymerase II (RNA pol II) activity. The figure is not meant to be exhaustive or explicit, as many known factors have been omitted for clarity, and the interactions shown may be either physical or functional in nature.