Neurogenin3 and Hepatic Nuclear Factor 1 Cooperate in Activating Pancreatic Expression of Pax4*

During fetal development, paired/homeodomain transcription factor Pax4 controls the formation of the insulin-producing β cells and the somatostatin-producing δ cells in the islets of Langerhans in the pancreas. Targeting of Pax4 expression to the islet lineage in the fetal pancreas depends on a short sequence located ∼2 kb upstream of the transcription initiation site of the PAX4 gene. This short sequence contains binding sites for homeodomain transcription factors PDX1 and hepatic nuclear factor (HNF)1, nuclear receptor HNF4α, and basic helix-loop-helix factor Neurogenin3. In the current study we demonstrate that the HNF1α and Neurogenin3 binding sites are critical for activity of the region through synergy between the two proteins. Synergy involves a physical interaction between the factors and requires the activation domains of both factors. Furthermore, exogenous expression of Neurogenin3 is sufficient to induce expression of the endogenous pax4 gene in the mouse pancreatic ductal cell line mPAC, which already expresses HNF1α, whereas expression of both Neurogenin3 and HNF1α are necessary to activate the pax4 gene in the fibroblast cell line NIH3T3. These data demonstrate how Neurogenin3 and HNF1α activate the pax4 gene during the cascade of gene expression events that control pancreatic endocrine cell development.

Basic helix-loop-helix factor Neurogenin3 has been shown to control the differentiation of all pancreatic endocrine cell types. Neurogenin3 is expressed early in endocrine progenitor cells in the fetal pancreas but not in mature endocrine cells (11)(12)(13). Targeted disruption of the neurogenin3 gene in mice results in the total agenesis of endocrine cells (10), and transgenic overexpression of Neurogenin3 using the pdx1 promoter drives the precocious differentiation of a large population of endocrine cells that are largely ␣ cells (11,14). The preponderance of ␣ cells suggests that additional factors are important for diverting endocrine precursor cells from the ␣ cell fate to the alternate endocrine lineages.
One such factor may be the paired homeodomain transcription factor, Pax4. Mice homozygous for a targeted disruption of the pax4 gene develop ␣ cells but few ␤ or ␦ cells, demonstrating its role in endocrine cell type differentiation (7). Like Neu-rogenin3, Pax4 expression in the pancreas crests during the period of peak ␤-cell differentiation in the fetus and is absent from mature islets, demonstrating that it functions only during the development of the endocrine cells (15).
Investigation of the mechanisms that control Pax4 expression has provided insight into the factors that lie upstream of Pax4 in the pathway for differentiation of insulin-producing ␤ cells. A previous study of the human PAX4 gene promoter identified a 50-bp region (that is highly conserved in the mouse promoter) (16) located between Ϫ1910 and Ϫ1960 bp relative to the transcription start site that is critical for promoter activity (16 -18). Removal of this region ablated promoter activity in transgenic mice, and two copies of the region activated the heterologous HSV-TK promoter 12-fold in ␤ cells (18). The 50-bp sequence contains several transcription factor binding motifs, including an E box capable of binding bHLH family members such as Neurogenin3 along with sites capable of binding factors including hepatic nuclear factor (HNF) 1 1␣, HNF4␣, and Pdx1.
In the current study we investigated the role of the human PAX4 promoter binding factors in controlling PAX4 gene expression. Mutation of the binding sites in the promoter demonstrated that the most influential promoter elements are the HNF1␣ binding site and the E box. In co-transfection experiments, a combination of HNF1␣ and Neurogenin3 strongly * This work was supported by National Institutes of Health Grants DK553401 and DK21344. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. activates the promoter, but neither factor can activate alone. The activation domains of both factors are necessary for this synergy, and the two proteins physically interact. Neurogenin3 alone is sufficient to induce the endogenous mouse pax4 gene expression in a mouse pancreatic ductal cell line, which may be attributed to endogenous expression of HNF1␣ in these cells. Remarkably, the combination of Neurogenin3 and HNF1␣ can stimulate transcription of the endogenous pax4 gene in the non-pancreatic mouse fibroblast cell line NIH3T3.

MATERIALS AND METHODS
Cell Culture and Transfection-␤TC-3 and ␣ TC1.6 cells were grown in Dulbecco's modified Eagle's medium supplemented with 2.5% fetal bovine serum and 15% horse serum. NIH3T3 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. Cells were split into 6-well plates 24 h prior to transfection with 1 ϫ 10 6 cells/well for ␣ TC1.6 and ␤TC3 cells and 5 ϫ 10 4 cells/well for NIH3T3 cells. 2 g of reporter construct was used per well; 50 ng of any co-transfected transcription factor cDNA was used per well. The cationic lipid Transfast (Promega) was used for all transfections according to the manufacturer's instructions. Cells were harvested 48 h after transfection, and luciferase assays were performed with the Promega luciferase assay system according to the manufacturer's instructions. Luciferase activity was corrected for protein concentration.
Construction and Purification of Recombinant Adenoviruses-Mouse neurogenin3 cDNA was cloned into the pShuttle vector and a recombinant adenovirus was constructed using the Adeno-X TM expression system (Clontech). A control adenovirus containing the bacterial ␤-galac-tosidase gene was a generous gift from Dr. C. Newgard (Duke University, Durham, NC). Adenoviruses were produced, purified, and desalted through a Sephadex G-25 column as described previously (19,20).
Viral Infection-mPAC (clones L20 and L4S2) and NIH3T3 cells were seeded in 6-well plates at a density of 5 ϫ 10 5 (mPAC) or 5 ϫ 10 4 (NIH3T3) cells/well and treated the next day at a multiplicity of infection of 10 viral particles/cell with AdCMV-Bgal or AdCMV-Neurogenin3 at 37°C for 2 h. Cells were then washed with phosphate-buffered saline and cultured in fresh medium for 48 h prior to harvesting. Adenovirusassisted transfection of NIH3T3 cells was carried out as previously described (21). Briefly, plasmid DNA (4 g) was mixed with 4 g of 25-kDa polyethylenimine (PEI; Aldrich) in Opti-MEM 1 serum-free medium (Invitrogen) and then combined with the AdCMV-Bgal or Ad-CMV-Neurogenin3 adenovirus. The DNA/PEI/adenovirus mixture was then added to the cells and incubated at 37°C for 2 h prior to washing with phosphate-buffered saline and culturing in fresh medium.
RT-PCR-48 h after virus infection, cells were collected and total RNA was isolated using the RNeasy kit (Qiagen). RNA samples were treated with DNase to remove contamination of genomic DNA. Firststrand cDNA was prepared using 3 g of total RNA, using the Superscript II RT kit and random hexamer primers (Invitrogen) in a total volume of 20 l according to the manufacturer's instructions. 1 l of the cDNA product was used as a template for PCR reactions under the following conditions: an initial incubation at 95°C for 3 min, followed by 25 cycles (Neurogenin3/␤-actin) or 35 cycles (HNF1␣/Pax4) of 95°C for 30 s, 55°C for 30 s, and 72°C for 30 s. Primer sets were as follows: mouse Pax 4 forward GTGTTGGCTCCAGTTCTTCC and reverse AAC-CAAACCCTCACCGTGTC; neurogenin3 forward CGCACCATGGCGC-FIG. 1. Mutation of the PAX4 minienhancer. a, the sequence of the minienhancer from the PAX4 gene promoter is shown with the binding sites for HNF4, HNF1␣, and Neurogenin3 (Ngn3) underlined. The bases altered in the mutant minienhancers are indicated above the sequence. b, wild type and mutant PAX4 minienhancers were linked to the HSV-TK promoter driving the luciferase reporter gene as shown. These constructs were transfected into ␤TC3 cells. Luciferase activities of all samples were determined 48 h after transfection and are expressed relative to the activity of the promoterless backbone vector (pFOX-luc1). Results are expressed as the mean Ϯ S.E. of data from experiments performed in triplicate on at least three separate occasions.

FIG. 2. Cooperative activation by
Neurogenin3 and HNF1␣. The wild type PAX4 minienhancer luciferase reporter construct was transfected into NIH3T3 cells. Plasmids containing the indicated transcription factor cDNAs ligated downstream of the CMV promoter were co-transfected with the reporter construct. Luciferase activities of all samples were determined 48 h after transfection and are expressed relative to the activity in cells transfected with the reporter construct and the empty CMV expression plasmid (pBAT12). Results are expressed as the mean Ϯ S.E. of data from experiments performed in triplicate on at least three separate occasions.
TaqMan Assay-TaqMan assay was used for quantification of pax4 mRNA levels in infected NIH3T3 cells. First-strand cDNA was generated as described above. PCR amplification of Pax4 and ␤-actin (as endogenous control) was carried out in separate tubes using 1 l of cDNA in a PCR mix containing AmpliTaq Gold DNA polymerase (0.75 unit) and buffer (PE Biosystems, TaqMan PCR core reagent kit N808 -0228), AmpErase uracyl N-glycosilase (0.5 unit), dUTP (400 uM), dNTP (dATP, dCTP, dGTP, 200 M each), and 5.5 mM MgCl 2 . The following primers and TaqMan probes were used: Pax4 forward CATCCCAGGC-CTATCTCCAAC, reverse CCACATATGAGGAGGAAGCCAC, TaqMan probe TACTGGGACTGCCAATCCCTCCTTCC; ␤-actin forward CCGG-GACCTGACAGACTACCT, reverse GATTTCCCTCTCAGCTGTGGTG, TaqMan probe TCCTGACCGAGCGTGGCTACAGC. TaqMan probes were used at a concentration of 500 nM and PCR primers were used at a concentration of 100 nM in a 50-ul reaction volume. PCR reactions were run in an ABI PRISM 7700 sequence detection system under the following conditions: 50°C 2 min, 95°C 10 min, and then 40 cycles 95°C 15 s, 60°C 1 min. Data were analyzed using the comparative C T method (⌬ ⌬ C T ) as described in the manufacturer's manual (PE Applied Biosystems). Efficiencies of Pax4 and ␤-actin amplifications were found to be equal in preliminary validation experiments.
In Vitro Protein-Protein Interaction Assay-Glutathione S-transferase (GST) fusion proteins were produced in Escherichia coli BL21 competent cells via the pPIG plasmid system (22). In vitro translated and [ 35 S]methionine-labeled proteins were prepared using the TNTcoupled reticulocyte lysate system (Promega) according to the manufacturer's instructions. 15 l of labeled protein were mixed with 1 g of GST fusion protein bound to 20 l of glutathione-agarose beads in a total volume of 600 l of interaction buffer (40 mM HEPES (pH7.5), 50 mM KCL, 5mM MgCl 2 , 0.2 mM EDTA, 1 mM dithiothreitol, 0.5% (v/v) Nonidet P-40). Samples were then incubated for 1 h at 4°C with gentle rocking, and the beads were then washed three times with interaction buffer. The bound proteins were eluted with 25 l of Laemmli buffer. 15-l aliquots of the eluted proteins were fractionated on SDS-polyacrylamide gels, dried by heat and vacuumed and autoradiographed.
Plasmid Construction-Deletion mutants of Neurogenin3 and HNF1␣ were generated by PCR and cloned into BglII and BamHI sites of the eukaryotic expression vector pBat12 (15) or PCDNA3 (Invitrogen), both of which express the inserted cDNA under the control of the CMV promoter.

RESULTS
Mutational Analysis of the PAX4 Promoter-A previous study demonstrated the importance of a 50-bp region within the pax4 promoter for transcriptional activity of the entire promoter, both in transgenic animals and in pancreatic cell lines (18). Furthermore, two tandem copies of the region increased activity of a heterologous promoter 12-fold in the pancreatic ␤-cell line ␤TC3 but not in the NIH3T3 fibroblast line, thus exhibiting the same cell specificity as the native promoter. We set out to determine which transcription factor binding motifs within this region were most critical for promoter activity by introducing mutations into the promoter sequences previously shown to bind HNF4␣, HNF1␣, and pancreatic bHLH factors (E-box sequence). The resulting constructs were transfected into ␤TC3 cells, and their relative transcriptional activities were determined (Fig. 1). Mutations in either the E box or the HNF1␣ binding motif ablated the activity of the 50-bp minienhancer when linked to the heterologous HSV-TK promoter, whereas a mutation in the HNF4␣ site resulted in a less dramatic reduction in promoter activity of ϳ50% (Fig. 1). All the constructs tested showed no significant activity above background in either the pancreatic ␣-cell line ␣TC1.6 or NIH3T3 cells (data not shown).
Cooperativity between HNF1␣ and Neurogenin3-To compare the relative importance of the pancreatic transcription factors that bind to the critical elements within the minien-hancer, NIH3T3 cells were co-transfected with the minienhancer HSV-TK reporter construct and CMV promoter-driven expression plasmids containing cDNAs encoding the transcription factors (Fig. 2). Because the previous study demonstrated that no individual transcription factor was sufficient to activate the heterologous promoter in NIH3T3 cells (18), we transfected combinations of the pancreatic transcription factors. HNF1␣ and the bHLH dimer composed of Neurogenin3 and its ubiquitous partner E47 cooperated in activating the minienhancer; the combination of the two activated the minienhancer to a level more than 20-fold over basal activity. No further increase was produced by the inclusion of other pancreatic transcription factors, including HNF4␣ and PDX1, neither of which activated the heterologous promoter in any combination with other factors (data not shown).
HNF1␤, often a heterodimeric partner for HNF1␣, was also able to synergize with Neurogenin3/E47, although to a lesser extent than HNF1␣ (Fig. 2a). The inclusion of both HNF1␣ and HNF1␤ gave no greater synergy with Neurogenin3/E47 than HNF1␣ alone. Similarly, substituting the bHLH proteins Neu-roD1 or MyoD for Neurogenin3 resulted in decreased levels of synergy, whereas the bHLH protein Meso1 produced no synergy at all (Fig. 2b).
Decreased Synergy in Pancreatic Cell Lines-Synergy was tested further in pancreatic cell lines ␣ TC1.6 and ␤TC3. In ␣ TC1.6 cells, both Neurogenin3/E47 and HNF1␣ activated the PAX4 minienhancer ϳ1.5-fold; the combination of both factors only activated to a level marginally greater than the sum of the activation produced by each factor alone (Fig. 3a).
In ␤TC3 cells, expression of Neurogenin3/E47 alone was sufficient to induce activation of the minienhancer (ϳ7-fold), but the addition of HNF1␣ produced no further activation (Fig.  3a). We hypothesized that the marked effect of Neurogenin3/ E47 alone might be due to cooperation with endogenous HNF1␣, which is present in ␣ TC1.6 and ␤TC3 cells but not in 3T3 cells (Fig. 3b). To test this hypothesis, we co-transfected a plasmid expressing a dominant negative form of HNF1␣ (T547E548fsdelTG/ter) (23) along with the Neurogenin3/E47 expression plasmids into ␤TC3 cells. Addition of the plasmid expressing the dominant negative HNF1␣ mutant eliminated the stimulatory effect of Neurogenin3/E47.
Synergy Requires Activation Domains of Both Factors-To map regions of HNF1␣ and Neurogenin3 required for their synergistic interaction, cDNAs encoding truncated variants of both proteins were engineered and co-transfected as shown into NIH3T3 cells along with the minienhancer reporter construct (Fig. 4a). The carboxyl-terminal proximal region of Neuroge-nin3, which contains a transcriptional activation domain, is required for synergy, whereas the amino-terminal region is not.
HNF1␣ contains three separate activation domains, all of which are located on the carboxyl-terminal side of the basic helix-loop-helix motif (24). Deletion of activation domain II (construct ⌬ 288 -317) appears to have had no effect on synergy (Fig. 4b), but the removal of activation domains I (construct 1-485) or III (construct ⌬ 288 -496) diminished the ability of HNF1␣ to synergize with wild type Neurogenin3/E47. In addition, the HNF1 dimerization domain, amino acids 1-33, which is required for forming HNF1-HNF1 dimers, is also important for producing full synergistic activation.
To determine whether the synergy between Neurogenin3/ E47 and HNF1␣ is associated with a physical interaction between the two factors, we tested the ability of the two proteins to interact in vitro. A chimeric protein consisting of the bHLH domain and carboxyl terminus of Neurogenin3 fused to GST was mixed with 35 S-labeled HNF1␣. The GST-Neurogenin3 fusion protein was able to bring down the labeled HNF1␣ protein, whereas the GST protein alone could not, demonstrating that the two transcription factors can physically interact in vitro.
Synergistic Activation of the Endogenous pax4 Gene-The human PAX4 minienhancer experiments demonstrate that Neurogenin3/E47 and HNF1␣ can synergize in an artificial system, but we next asked whether such an interaction could activate the endogenous mouse pax4 gene. Mouse pancreatic ductal cell lines (mPAC L4S2 and mPAC L20) and NIH3T3 cells were infected with an adenovirus expressing Neurogenin3 under the control of the CMV promoter. In both mPAC cell lines, which already express high levels of HNF1␣, Neuroge-nin3 expression alone was sufficient to induce transcription of the endogenous pax4 gene as gauged by RT-PCR (Fig. 5a). In NIH3T3 cells, however, HNF1␣ was needed in addition to Neurogenin3 to significantly induce expression of the native pax4 gene (Fig. 5b).

DISCUSSION
The current study demonstrates that the function of the critical Ϫ1910-bp element of the PAX4 gene promoter depends on two sites, a bHLH binding site (E box) and an HNF1␣ binding site. The pancreatic transcription factors Neurogenin3 (along with its heterodimeric partner E47) and HNF1 bind to these sites, physically interact, and synergistically activate both the promoter and the endogenous pax4 gene. As has been observed for many other transcription factor interactions (25)(26)(27)(28)(29), we demonstrated a physical interaction between the two proteins and showed that the activation domains of both proteins were required for synergy. Therefore, cooperativity between HNF1␣ and Neurogenin3 may result from the cooperative recruitment of transcriptional co-activators to the transcription complex on the pax4 gene.
Neurogenin3 is not the only pancreatic bHLH factor that can dimerize with E47 and bind the PAX4 promoter E box, 2 and it seems possible that others, especially neuroD1, could substitute for Neurogenin3 on the PAX4 promoter in vivo. Two observations suggest that Neurogenin3 is the most important bHLH activator of the PAX4 promoter. First, Neurogenin3 and Pax4 have almost identical expression patterns during pancreatic development, with maximum expression between embryonic days 13 and 15 and no expression in mature islets (15). Second, in conjunction with heterodimeric partner E47, Neu-rogenin3 exhibits a higher affinity for the PAX4 promoter E box than for E boxes from other pancreatic promoters. 3 Finally, the current study demonstrates that the Neurogenin3/E47 heterodimer is a better activator of the PAX4 minienhancer than any of the other bHLH dimers tested.
Similarly, HNF1␤ could possibly substitute for HNF1␣. HNF1␣ and HNF1␤ are related proteins sharing 93% sequence identity in their DNA binding domains, 75% identity in their dimerization domains, and 47% identity in their carboxyl-terminal activation domains (30). HNF1␣ and HNF1␤ can form homodimers or heterodimers with each other, producing complexes with very similar or identical DNA binding characteristics. Both factors are found in the developing pancreas when Pax4 is present, and both factors can synergize with Neuroge-nin3 and activate the Pax4 promoter. Both factors are expressed broadly in the early pancreatic bud, although HNF1␤ may appear earlier (31,32). Either or both may play a role in Pax4 expression.
The role of the HNF4␣ binding site adjacent to the HNF1 binding site in the PAX4 promoter remains uncertain. The wild type minienhancer is maximally stimulated by the combination of Neurogenin3/E47 and HNF1␣, with no further activation produced from the addition of HNF4␣ 1 or HNF4␣ 2 (data not shown). Mutation of the HNF4␣ site reduced the activity of the 2 S. Smith, H. Watada, and M. German, unpublished data. 3 S. Smith, H. Watada, and M. German, submitted for publication.
FIG. 5. Synergistic activation of the endogenous pax4 gene. a, two different clonal lines (L4S2 and L20) of the mouse pancreatic ductal mPAC cells were infected with adenovirus expressing either ␤-galactosidase or Neurogenin3 from the CMV promoter as indicated. Subsequent to transfection, RNA was purified, and the relative levels of transcripts for Neurogenin3, Pax4, HNF1␣, and ␤-actin were determined by RT-PCR. RNA from untransfected ␤TC-3 cells was purified and used as a positive control for RT-PCR reactions. b, the fibroblast cell line NIH3T3 was infected with adenovirus expressing either ␤-galactosidase or Neurogenin3, either alone or in combination with plasmids expressing the other factors shown. Subsequent to transfections, RNA was purified, and TaqMan quantitative RT-PCR was used to determine the abundance of pax4 mRNA relative to the abundance of ␤-actin mRNA in the infected cells. Results are expressed as the mean Ϯ S.E. of three determinations, with the background signal from cells infected with ␤-galactosidase virus set at 1. minienhancer, but the same mutated minienhancer could still be potently activated by a combination of Neurogenin3 and HNF1␣ (data not shown). Furthermore, Neurogenin3 and HNF1␣ alone were sufficient to induce expression of the endogenous pax4 gene in NIH3T3 cells (which have no endogenous HNF4␣), although the level of expression is far below the level seen in ␤TC3 cells. It is possible that the absence of an appropriate ligand or other interacting proteins normally present in vivo could reduce the activity of HNF4␣ in these assays. We cannot rule out a role for HNF4␣ in pax4 gene expression.
The pancreatic transcription factor PDX1 can also bind to the HNF1 site in the PAX4 promoter. We found no circumstances, however, in which PDX1 could activate the PAX4 promoter by itself or in combination with other factors. In view of the fact that PDX1 is not co-expressed with Neurogenin3 in islet progenitor cells (11), some of which also co-express Pax4 (18), it seems unlikely that PDX1 plays a direct role in Pax4 expression during normal development.
These data strongly support a model in which HNF1 proteins and Neurogenin3 cooperate in activating Pax4 expression and ␤-cell differentiation in the developing pancreas and place these factors upstream of Pax4 on an increasingly detailed map of the hierarchy of factors that control islet cell determination and differentiation. Ultimately, our understanding of this pathway may lead to the development of methods for driving progenitor cells to differentiate into pancreatic islet cells for patients with diabetes. The ability of this cascade to function in cells as disparate as pancreatic duct cells and NIH3T3 fibroblasts provides encouraging evidence that this approach to islet cell production may eventually succeed.