J Biol Chem, Vol. 275, Issue 5, 3485-3492, February 4, 2000

Pancreatic  Cell-specific Transcription of the pdx-1 Gene
THE ROLE OF CONSERVED UPSTREAM CONTROL REGIONS AND THEIR HEPATIC NUCLEAR FACTOR 3 SITES^*

Kevin Gerrish, Maureen Gannon^§, David Shih^¶, Eva Henderson, Markus Stoffel^¶, Christopher V. E. Wright^§, and Roland Stein^§

From the  Department of Molecular Physiology and Biophysics and ^§ Department of Cell Biology, Vanderbilt Medical Center, Nashville, Tennessee 37232 and ^¶ Laboratory of Metabolic Diseases, Rockefeller University, New York, New York 10021

	ABSTRACT

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

To identify potential transactivators of pdx-1, we sequenced approximately 4.5 kilobases of the 5' promoter region of the human and chicken homologs, assuming that sequences conserved with the mouse gene would contain critical cis-regulatory elements. The sequences associated with hypersensitive site 1 (HSS1) represented the principal area of homology within which three conserved subdomains were apparent: area I (2694 to 2561 base pairs (bp)), area II (2139 to 1958 bp), and area III (1879 to 1799 bp). The identities between the mouse and chicken/human genes are very high, ranging from 78 to 89%, although only areas I and III are present within this region in chicken. Pancreatic  cell-selective expression was shown to be controlled by mouse and human area I or area II, but not area III, from an analysis of pdx-1-driven reporter activity in transfected - and non- cells. Mutational and functional analyses of conserved hepatic nuclear factor 3 (HNF3)-like sites located within area I and area II demonstrated that activation by these regions was mediated by HNF3. To determine if a similar regulatory relationship might exist within the context of the endogenous gene, pdx-1 expression was measured in embryonic stem cells in which one or both alleles of HNF3 were inactivated. pdx-1 mRNA levels induced upon differentiation to embryoid bodies were down-regulated in homozygous null HNF3 cells. Together, these results suggest that the conserved sequences represented by areas I and II define the binding sites for factors such as HNF3, which control islet  cell-selective expression of the pdx-1 gene.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

The PDX-1 homeodomain transcription factor is an essential regulator of pancreatic endocrine cell development and adult islet  cell function. Pancreatic development in homozygous pdx-1^/ mice is blocked at a very early stage (1-3), and endocrine cells in the rostral duodenum (3) and stomach (4) are also reduced or absent. The finding that a human born without a pancreas was homozygous for an inactivating mutation in pdx-1 (5) highlights the critical conserved role that this factor plays in pancreas formation in vertebrates.

In the adult pancreas, PDX-1 is expressed at highest levels within islet  cells (6). PDX-1 can specifically stimulate reporter gene expression from transfected constructs driven by promoter sequences from several genes that are normally transcribed selectively in  cells, including insulin (7-11), glucokinase (12), islet amyloid polypeptide (13-16), and glucose transporter type 2 (GLUT2)¹ (17). The importance of PDX-1 in controlling the expression of genes required for the sensing and regulation of blood glucose levels suggested that an inactivating mutation within the protein would result in dysfunctional  cells. Indeed, mice that are heterozygous mutant carriers of PDX-1 exhibit symptoms of glucose intolerance, suggesting a dosage effect on insulin expression and  cell function (18, 19). More significantly, humans with this condition are susceptible to a form of Type II (non-insulin-dependent) diabetes designated maturity onset diabetes of the young (MODY) (20). In addition, selectively removing PDX-1 from adult islet cells using the Cre-LoxP system results in the development of diabetes in mice, which appears to be at least partially caused by the reduction in insulin and GLUT2 gene expression levels (18).

Although PDX-1 is clearly essential for pancreas formation and islet  cell function, the transcription factors that control its expression are not well characterized. The analysis of pdx-1-driven reporter constructs in transgenic mice has shown that the cis-acting elements mediating pancreas-and duodenum-specific expression are found within the 5'-flanking region of the gene (21, 22). DNaseI and micrococcal nuclease studies performed on the endogenous mouse pdx-1 gene identified three nuclease hypersensitive sites (HSS), which are located approximately between nucleotides 2560 to 1880 (HSS1), 1330 to 800 (HSS2), and 260 to +180 (HSS3) (22). Strikingly, the only sequences that could direct pancreatic  cell-specific expression in transfected cells spanned HSS1 region sequences. Moreover, HSS1 sequences could independently direct  cell-selective transgene expression in vivo (22). Collectively, these results suggested that the HSS1 region was critical for correct development- and differentiation-specific transcription of the pdx-1 gene.

Hepatic nuclear factor 3 (HNF3), a potent transcriptional regulator of the forkhead/winged helix factor family (23-26), is present in islet  cells (22) and implicated in controlling HSS1 selective activation via a binding site located in the mouse pdx-1 gene at 2007 to 1996 base pair(s) (bp) (22). Unfortunately, the role of HNF3 in pdx-1 expression in vivo cannot be resolved from studying HNF3 homozygous null mutant mice, as they die early in embryogenesis before the differentiation of pancreatic endoderm (24, 27). Recent data, however, suggest that potential insight into HNF3 function in the pancreas may be gained by performing experiments with embryonic stem (ES) cells (28). Thus, differentiation of ES cells to embryoid bodies (EBs) induces the expression of genes encoding pancreas-enriched transcription factors, including hepatic nuclear factors HNF3, HNF1, and HNF4 as well their target genes, such as GLUT2 and L-pyruvate kinase (28, 29). Conversely, the expression of HNF1 and HNF4 was compromised in EBs produced from HNF3 null ES cells (28). These results suggested that ES cells might provide a model system allowing insight into the possible roles of HNF3 in the gene interaction hierarchies regulating pdx-1 transcription under different growth and differentiation stimuli.

The three HSS sites detected in pdx-1 presumably reflect an altered chromatin conformation caused by non-histone protein binding (e.g. transcription factors) in expressing versus nonexpressing cells (see Elgin (30), Felsenfeld et al. (31), and Gross and Garrard (32) for a discussion of HSS in gene activation). Since the expression pattern of pdx-1 is very similar in frog (33), chicken (34), rat (21), mouse (6), and human (35), we reasoned that critical cis-acting elements within these HSS regions would be conserved among these species. To begin to address this possibility, approximately 4.5 kilobase(s) (kb) of the promoter region from the human and chicken pdx-1 genes was sequenced and compared with mouse. Interestingly, the HSS1 region contained the only shared areas of significant identity, and these could be resolved into three separate conserved subregions, termed areas I, II, and III. Comparison of reporter constructs driven by mouse and human areas I , II, and III demonstrated that area I and area II can mediate  cell-selective activation. Furthermore, the integrity of conserved HNF3 binding sites present in area I and area II were important for stimulation. We also found that HNF3 inactivation greatly decreased the level of pdx-1 expression that occurs in EBs. Our results strongly suggest that conserved sequences within the HSS1 region define the binding sites for key regulators of pdx-1 transcription.

	EXPERIMENTAL PROCEDURES

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Characterization of 5'-Flanking Regions of Human and Chicken pdx-1-- A human genomic pdx-1 clone (ipf1P1) containing approximately 8 kb of the 5'-flanking region was generously provided by Dr. M. Alan Permutt (Washington University School of Medicine (36)). The chicken pdx-1 locus (cpdx-1 #4) was isolated from a genomic library by probing with a 3' MluI-EcoRV fragment of the rat pdx-1 cDNA (37); the 14-kb insert contained around 4 kb of 5'-flanking sequence and the entire coding region. Approximately 4.5 kb of human and chicken sequence upstream of the transcription start site (+1 bp) was determined, and alignment with the mouse pdx-1 gene was performed using the MacVector DNA analysis program (Oxford Molecular). The GenBank^TM accession numbers for the 5'-flanking sequences of the mouse, human, and chicken pdx-1 genes are AF192495, AF192496, and AF194114, respectively.

Transfection Constructs-- Human and mouse pdx-1 sequences spanning areas I (2694 to 2561 bp), II (2139 to 1958 bp), and III (1879 to 1799 bp) were generated by the polymerase chain reaction (PCR) and cloned directly upstream of the herpes simplex virus thymidine kinase (Tk) promoter region in the chloramphenicol acetyltransferase (CAT) expression vector, pTk(An) (38). These subregion sequences in mouse are located between 2761 to 2457 bp (area I), 2141 to 1923 bp (area II), and 1879 to 1600 bp (area III) and between 2839 to 2521 bp (area I), 2252 to 2023 bp (area II), and 1939 to 1664 bp (area III) in humans. The numbering is relative to the S1 transcription start site (21)). Point mutants within the HNF3 binding elements of the human and mouse area I and area II pTk constructs were generated using the Quik Change mutagenesis kit (Stratagene). The following oligonucleotides were used for mutagenesis: area I, site A1, human 2726 GCAGCTCTTATGGATAAATACCCCAAAAAGGTGTAAACCAATTA 2683, mouse 2649 TCTTATGGATAAATACCCCAAAACGGCTGTAAAC 2615; area I, site A2, human 2723 AACAGCAGCTCTTATGGATACCCCAACAAAAAAGGTGTAAACCA 2679, mouse 2645 CAGCTCTTATGGATACCCCAACAAAAACGGCTGT 2611; area II, site B, human 2121 GCCTGCCACCCCCGGAGTGTGGCATTTGCACTTCTCAACTAATT 2077. The mutated sequences are underlined. Mouse area I oligonucleotides were used to construct the HNF3 binding mutant in the 2917-bp/PstI to 1918-bp/BstEII region of PstBst:pTk and PstBst:pTk M5 (22). The HNF3 mutant in site B of mouse area II was produced from PstBst:pTk M5 by 5' Xmn1 cleavage. Each construct was verified by sequencing.

Cell Transfections-- Monolayer cultures of the pancreatic islet  cell lines, TC3 (22), HIT T-15 2.2.2 (22), and MIN6 (39), as well as the baby hamster kidney (BHK) cells (22) were maintained as described previously. The pdx-1:pTk constructs were transfected by calcium phosphate coprecipitation (HIT T-15 (40), BHK (40)) or LipofectAMINE (TC3, MIN6) procedures (Life Technologies, Inc.). The precipitates (11 µg total) used for the HIT T-15 and BHK transfections contained 1 µg of pdx-1:pTk and pRSVLUC, whereas 1 µg of each plasmid (2 µg total) was used with TC3 and MIN6 cells. The Rous sarcoma virus (RSV) enhancer-driven luciferase (LUC) expression plasmid, pRSVLUC (41), was used as a recovery marker. HIT T-15 and BHK cells were treated with 20% glycerol (2 min) 4 h after the addition of the calcium phosphate DNA precipitate. Extracts were prepared 40 to 48 h after transfection, and LUC and CAT enzymatic assays were performed as described (41, 42). The CAT activity from the reporter constructs were normalized to the LUC activity of the cotransfected internal control plasmid. Each experiment was carried out at least three times.

Electrophoretic Mobility Shift Assays-- Nuclear extracts were prepared from TC3 cells as described previously (43). HNF3 was in vitro transcribed and translated from pGEM1-HNF3 (44) using the TNT coupled reticulocyte lysate system (Promega). Binding reactions (20 µl) were carried out according to Wang et al. (45) in 20 mM HEPES (pH 7.9), 20 mM KCl, 50 mM NaCl, 1 mM dithiothreitol, 1 µg of poly(dI-dC), 10% (v/v) glycerol. Approximately 5 µg of TC3 extract protein was used per gel mobility shift reaction. The double-stranded oligonucleotides used to detect HNF3 binding were end-labeled with polynucleotide kinase and [-³²P]dATP. The probe sequences were as follows: human area I, site A1/A2, 2713 CCGTTTTTGTTTATTTATCCA 2693; human area I, site A1 mutant, 2713 CCGTTTTTGGGGATTTAT 2696; human area I, site A2 mutant, 2709 TTTTGTTGGGGTATCCAT 2692; human area II, site B, 2109 GTGCAAAGTAAACACCC 2093; area II, human site B mutant 2109 GTGCAAATGCCACACCC 2093; mouse area II, site B, 2013 GTGCTAAGCAAACATCCT 1996. The mutated nucleotides are underlined. The rabbit anti-mouse HNF3 antiserum (3 µl) was preincubated with extract protein for 10 min at room temperature before initiation of the DNA binding reaction. The HNF3 antibody was raised against amino acids 1 to 117 of mouse HNF3 and specifically recognized this protein (46). The samples were electrophoresed on 6% nondenaturing polyacrylamide gels at 150 V for 2 h under high ionic strength polyacrylamide gel electrophoresis conditions (10) before drying and autoradiography.

ES cells-- Wild type, heterozygous, and HNF3 null ES cells were maintained in ES cell medium supplemented with 1,000 units of leukemia inhibitory factor on a primary embryonic fibroblast layer as described (47). ES cells were induced to differentiate to EBs in vitro by growing them in bacterial Petri dishes without feeder fibroblasts and leukemia inhibitory factor for 21 days. HNF3-dependent gene expression was measured in EBs produced from three distinct clonal HNF3 null lines (clones B14, 5.1, 5.2) and two different heterozygous lines (clones B13, 4B1). Targeted disruption of HNF3 was performed as described previously (23, 27).

Reverse Transcriptase-PCR-- Total RNA was extracted from ES cells and EBs using TRIzol as described by the manufacturer (Life Technologies, Inc.). Contaminating genomic DNA was removed by treating with 3 µl of RNase-free DNase-I (Roche Molecular Biochemicals)/10 µg of RNA. cDNA was synthesized using moloney leukemia virus reverse transcriptase with dNTPs and random hexamer primers (Stratagene). The cDNAs served as templates for PCR reactions with specific primer pairs to detect hypoxanthine phosphoribosyltransferase, GATA-4, HNF3, and PDX-1 levels. The primer sequences used are: hypoxanthine phosphoribosyltransferase, forward, 5'-AGCGCAAGTTGAATCTGC-3', reverse 5'-AGCGACAATCTACCAGAG-3'; GATA4, forward, 5'-CGCCGCCTGTCCGCTTCC-3', reverse 5'-TTGGGCTTCCGTTTTCTGGTTTGA-3'; HNF3, forward, 5'-ACTGGAGCAGCTACTACG-3', reverse 5'-CCCACTAGGATGACATG-3'; PDX-1, forward, 5'-ACAAGGACCCGTGCGCATTC-3', reverse 5'-CTCGGTTCCATTCGGGAAAG-3'.

	RESULTS

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Conservation of HSS1 Sequences between Mouse, Human, and Chicken pdx-1-- Sequencing of the promoter region of the human and chicken pdx-1 genes and comparison to mouse identified a localized area of substantial identity between all species that corresponds approximately to the previously reported HSS1 region (Fig. 1). In contrast, the level of conservation within the HSS2 region, at approximately 39-48%, is similar to that found for non-HSS sequences. There is also significant conservation of HSS3 sequences in human and mouse, but not chicken, pdx-1 (Fig. 1). An E-box element at 99 to 94 bp, which represents a upstream stimulatory factor regulatory site in the HSS3 region (21), is conserved within the mouse and human pdx-1 genes (data not shown).

View larger version (13K): [in this window] [in a new window]   Fig. 1.   Localized enhanced sequence similarity within HSS1 in the promoter region of mouse, human, and chicken pdx-1 genes. Diagram of the mouse, human, and chicken promoter region. HSS1, 2560 to 1880 bp; HSS2, 1330 to 800 bp; HSS3, -260 to +180 bp. area I, -2694 to -2561 bp; area II, 2139 to 1958 bp; area III, 1879 to 1799 bp. The percent identity of the human and chicken sequences to mouse is indicated with bars below the locus diagram. Mouse to human (the numbering is relative to the S1 transcription start site (21)): 4500/2761, 46%; area I, 89%; 2456/2154, 45%; area II, 78%; 1922/1880, 33%; area III, 84%; 1600/364, 48%; HSS3, 72%. Mouse to chicken (numbering relative to the chicken protein coding ATG codon): 1731/1298, 39%; area I (1297/989), 86%; 988/691, 39%; area III (690/413), 78%; 412/+1, 39%.

The HSS1 region could be subdivided into three distinct domains based on sequence identity, which ranged between mouse and human area I (2694 to 2561 bp), area II (2139 to 1958 bp), and area III (1879 to 1799 bp) from 89, 78, and 84%, respectively, which is higher than the 72% identity shared by the mouse and human promoter-proximal sequences spanning HSS3 (Fig. 2). All three subdomains can be considered to be associated with HSS1 (i.e. 2560 to 1880 bp) based upon the standard deviation (± 150 bp) of the HSS analysis (22). Chicken pdx-1 does not contain a similarly located region homologous to area II, although area I and III sequences are highly conserved between mouse and chicken (Fig. 1). Since Southern blot analysis using mouse and human area II probes has also failed to detect cross-reacting sequences in chicken (data not shown), a chicken equivalent to area II may not exist.

View larger version (49K): [in this window] [in a new window]   Fig. 2.   Sequence identity within area I, II, and III in mouse, human, and chicken pdx-1. The shaded sequences are conserved between all three species. The identity within area I, area II, and area III is greatest between 2694 to 2561 bp, 2139 to 1958 bp, and 1879 to 1799 bp, respectively. The numbering is relative to the mouse pdx-1 gene. The potential (sites A1 and A2) and previously identified (site B (22)) HNF3 binding sites within area I and II are shown. The sequences shown are found within the pdx-1:pTk vectors in Figs. 3 and 5.

Islet  Cell-specific Reporter Gene Expression Is Mediated by Area I or Area II-- We showed previously that mouse pdx-1 sequences spanning the 2917- to 1918-bp region (e.g. PstBst:pTk) directed selective reporter construct expression in transfected pancreatic islet  (i.e. HIT T-15, TC-3) versus non-islet (i.e. BHK, HeLa, H4IIE) cell lines (22). The conserved sequences of areas I and II are entirely contained within this region, but none of area III. To determine if area I, II, or III contributes independently to cell-specific activation, mouse and human pdx-1 sequences spanning each conserved domain were subcloned directly upstream of the Tk minimal promoter in a CAT reporter plasmid. We analyzed their activity in transfected HIT T-15, TC3, MIN6, and BHK cells. The activity of each construct is presented as the ratio of pdx-1:pTk expression in  cells relative to BHK cells.

Area I and II reporter constructs were much more active than area III constructs in islet  cells (Fig. 3). The finding that human and mouse areas I and II gave similar levels of  cell stimulation provides a functional correlation with the sequence conservation of these regions. In contrast, only the human area III construct was active in the transfection assays, with stimulation low compared with area I or area II and only detectable in TC3 and MIN6 cells (Fig. 3). We presume that a nonconserved sequence element(s) is involved in activation of human area III in TC3 and MIN6 cells. Importantly, these results suggested that both area I and area II contribute to the  cell-specific activation pattern observed for the previously tested 2917- to 1918-bp fragment in transfection and transgenic assays (22).

View larger version (11K): [in this window] [in a new window]   Fig. 3.   Area I and area II impart  cell-specific activation. The human and mouse area I, area II, and area III pdx-1:TK constructs were transfected into HIT T-15, TC-3, MIN6, and BHK cells. The CAT activity in each sample was normalized to the cotransfected activity of the pRSVLUC recovery marker. The ratio of the normalized pdx-1:pTk to pTk vector activity is calculated for each cell line. The results are presented as the relative activity of pdx-1:pTk activity ±S.D. in  cells (i.e. HIT T-15 or TC-3) divided by BHK cells. A, HIT T-15 and BHK cells. B, TC-3 and BHK cells. C, MIN6 and BHK cells.

HNF3 Binding to Conserved Sites in Area I and Area II Is Important for Activation in  Cells-- Inspection of the conserved area I, area II, and area III sequences detected two overlapping potential HNF3 binding sites within area I (sites A1 and A2) and one in area II (site B) (Fig. 2). The area II site represents the HNF3 binding element previously determined to contribute in  cell-selective activation of the PstBst:pTk construct (22). The TRANSFAC data base program (48) indicated that few, if any, other potential binding sites for well characterized transcriptional regulators are shared between area I, II, and III sequences.

Gel mobility shift assays were performed on TC3 nuclear extracts to compare the binding properties of the HNF3-like site(s) in the 2713- to 2693-bp region of human area I to the HNF3 site in mouse and human area II. The 2713- to 2693-bp oligonucleotide probe contains the overlapping area I sites (A1 and A2) located at 2707 to 2691 bp and 2704 to 2693 bp (Fig. 4A). A major comigrating protein-DNA complex was detected with the area I and area II oligonucleotide probes (Fig. 4B). The specificity of protein binding to these sites was determined by competition assays using wild type and HNF3 binding-defective oligonucleotides. The two potential HNF3 sites between nucleotides 2713 to 2693 were individually mutated for this analysis (A1m and A2m in Figs. 4, A and B). The wild type sequence competitor but neither binding site mutant effectively competed for protein-DNA complex formation (Fig. 4B). The competition results indicated that HNF3 can bind to the 2713- to 2693-bp region of area I but to only one of the two potential sites. This proposal is also supported by the comigration of the single, common TC3 nuclear complex of area I and area II with in vitro translated HNF3 (Fig. 4C). To test directly for the interaction of HNF3 with the oligonucleotide probes, we analyzed whether an HNF3-specific antibody affected the protein-DNA complex formed with area I or area II probes in TC3 extracts. The quantitative elimination of area I and II complex formation after preincubation of extracts with HNF3 antiserum (Fig. 4D) demonstrates that HNF3 can bind and potentially activate area I and area II in  cells.

View larger version (12K): [in this window] [in a new window]   Fig. 4.   HNF3 binds to the 2713- to 2693-bp region of area I. A, the sequence of the human 2706/2691- and 2105/2093-bp regions are indicated. There are two overlapping HNF3 consensus binding sites (VAWTRTTKRYTY, where V = A, C, or G; W = A or T; K = G or T; Y = pyrimidine, C, or T; and R = purine, G, or A in Fig. 3C (55)) within nucleotides 2713 to 2693 (sites A1 and A2). The 2105 to 2093-bp element (site B) corresponds to the HNF3 binding site in the mouse pdx-1 gene at 2007 to 1995 bp (22). The mutated (m) nucleotides within the HNF3 consensus binding site are in lowercase. The band denoted by the asterisk is not seen in all TC3 nuclear extract preparations. B, binding to the 2713 to 2693 bp (labeled A1 and A2) and 2109 to 2093 bp (labeled B) probes in TC3 extracts is specifically competed by a 200-fold excess of unlabeled wild type (WT) but not mutant competitor. The HNF3 binding is indicated by the arrow. C, the in vitro translated (IVT) HNF3 binding complex comigrates with the one formed in TC3 nuclear extracts. D, the HNF3 antibody was preincubated with the TC3 extract before addition of the A1 and A2 and B probes.

To determine if HNF3 regulates area I- and area II-driven expression, the mutations that eliminated HNF3 binding to the A1/A2 and B elements were incorporated into the human and mouse pdx-1:pTk constructs, and their activities were measured in the HIT T-15  cell line. Area I-driven activity was decreased to a similar low level in either single or double HNF3 binding site mutants (Fig. 5A). In contrast, preventing HNF3 binding in area II did not significantly reduce its activity (Fig. 5B). These pdx-1:pTk constructs also showed the same regulatory properties in TC3 cells(data not shown). These data imply that HNF3 has a more significant role in area I-mediated activation than area II, at least assayed as separated control regions. In addition, these results also support the proposal that the 2713- to 2693-bp region of area I contains only one functional HNF3 binding site, which is contained within the A1 to A2 sequences.

View larger version (14K): [in this window] [in a new window]   Fig. 5.   HNF3 binding is involved in area I:pTk activity. The conserved area I and area II sequences in each pdx-1:pTk construct are boxed. The HNF3 binding mutants (m) are labeled (i.e. A1m, A2m, or Bm), and the boxed regions are shaded; the nucleotide alterations in areas I and II are within the A/T-rich sequences of the HNF3 element that specifically eliminated binding in vitro in Fig. 4. HIT T-15 cells were transfected with wild type and mutant area I:pTk (A) and area II:pTk (B) constructs. The relative activity of pdx-1:pTk to pTk is presented ±S.D.

The lack of an effect on  cell expression of mutations in the area II HNF3 site (site B; Fig. 5B) appeared to contradict Wu et al. (22), where it was reported that mutation of this site reduced the area I and II activity when they were present together in the construct PstBst:pTk. However, we considered that the discrepancy may arise from the different contexts of area II and how assaying it alone or linked to area I affects its potential for regulation by HNF3. As a consequence, we tested how activation of the larger PstBst:pTk construct was affected by mutation of the area I HNF3 binding site. Mutation of either the area I or area II HNF3 sites led to a 3-fold decrease in HIT T-15  cell activity (Fig. 6). Moreover, preventing HNF3 binding in both area I and area II reduced PstBst:pTk activity to the same extent as the single site mutants (Fig. 6). These results demonstrate that HNF3 binding in area I and area II is necessary for full 2917-bp/PstI to 1918-bp/BstEII region activation.

View larger version (12K): [in this window] [in a new window]   Fig. 6.   HNF3 binding within areas I and II is necessary for PstBst:pTk activation. The boxes represent the area I and area II sequences within the mouse PstBst region. The HNF3 binding site mutants in area I (A1m) and area II (Bm) are labeled, and the boxed regions containing the mutant site are shaded. HIT T-15 cells were transfected with wild type and mutant PstBst:pTk constructs. The relative activity of the PstBst:pTk construct to pTk is presented ±S.D.

HNF3 Is Required for pdx-1 Expression in Embryoid Bodies-- The early embryonic lethal phenotype of mouse embryos homozygous null for HNF3 prohibits analysis of the role of this factor in endogenous pdx-1 transcription (24, 27). However, recent studies suggested that ES cells might be an experimental system that could provide insight into the function of HNF3 in pdx-1 activation. When ES cells are grown in suspension in the absence of leukemia inhibitory factor, they differentiate to form visceral endoderm that expresses transcription factors (e.g. HNF3, HNF1, HNF4) and other gene products (e.g. GLUT2, L-pyruvate kinase) that are enriched in hepatocytes and islet  cells (28, 29). Interestingly, many of these enriched products are down-regulated in EBs produced from HNF3^/ ES cells (28). Since, our results strongly suggested that HNF3 directly activates pdx-1 transcription in  cells, we tested if its expression was induced upon differentiation of ES cells to EBs. Steady-state mRNA levels of HNF3 and GATA4, a marker of visceral endoderm as well as definitive endoderm and mesoderm, were compared with pdx-1 in 21-day-old EBs by reverse transcriptase-PCR. Each sample contained similar amounts of mRNA as shown by the amplification of hypoxanthine phosphoribosyltransferase mRNA (Fig. 7). pdx-1 mRNA levels were induced in differentiated EBs and, as expected, were HNF3 and GATA4 (Fig. 7) (28).

View larger version (51K): [in this window] [in a new window]   Fig. 7.   pdx-1 expression in EBs is dependent upon HNF3. HNF3 wild type (+/+), heterozygous (+/; clones B13, 4B1), and homozygous null (/; clones B14, 5.1, 5.2) ES-cells and differentiated EBs were assayed for PDX-1, HNF3, hypoxanthine phosphoribosyltransferase (HPRT), and GATA4 mRNA expression by reverse transcriptase-PCR. The hypoxanthine phosphoribosyltransferase levels show that each sample contained comparable amounts of cDNA; no product was amplified in the absence of reverse transcriptase (hypoxanthine phosphoribosyltransferase (HPRT-RT lane).

We next analyzed the possible significance of HNF3 in endogenous pdx-1 gene transcription by determining if pdx-1 induction was altered in EBs produced from HNF3 heterozygous and null ES cells. Strikingly, pdx-1 expression was down-regulated in HNF3^/ EBs in comparison to heterozygous or wild type cells, whereas GATA4 mRNA levels were unaffected (Fig. 7). Similar reductions of pdx-1 expression were found in three independently derived HNF3^/ cell lines. These data provide additional support for HNF3 as a direct mediator of pdx-1 transcription.

	DISCUSSION

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

The present study provides further insight into the mechanisms controlling pdx-1 transcription. Three highly conserved subdomains, areas I , II, and III, were detected by sequence comparisons between the 5'-flanking regions of the mouse and human pdx-1 genes, whereas chicken pdx-1 contains only areas I and III. These sequences are located within the vicinity of HSS1, one of three HSS detected previously in our analysis of pdx-1 chromatin structure in  cells. Although there is no large scale sequence conservation between the subregions of HSS1, area I and area II, but not area III, were approximately equally efficient in independently driving pancreatic  cell-selective reporter gene expression in transfected cell lines. These results suggest that the conserved sequences within area I and area II contain binding sites for a different sets of factors that contribute to pdx-1 transcriptional activation in  cells in mammals. Among these, we have provided support for a central role for HNF3 with two principal observations. First,  cell activation mediated by area I and/or area II was compromised when their HNF3 binding sites were mutated, and second, endogenous pdx-1 expression was substantially reduced when HNF3^/ ES cells were differentiated to EBs. Together, these observations strongly suggest that the programs controlling pdx-1 transcription during pancreatic differentiation and in the adult islet are principally defined through interactions with conserved sequences in the HSS1 region and that HNF3 is a component of these regulatory networks.

pdx-1 is transcribed in pancreatic progenitors of the embryo (2, 3) and in the adult in the endocrine lineages of the mature islet (predominantly in the  cells (6)), exocrine pancreas (22), rostral duodenum (3), and antral stomach (4). Our studies have demonstrated that transgene reporter constructs that include 4.5 kb of sequence upstream of the mouse pdx-1 transcription start site are sufficient for appropriate developmental and adult specific expression in the pancreas and duodenum (22). Furthermore, mouse sequences from 2917 to 1918 bp can independently direct pancreatic  cell-specific transgene expression in vivo (22). This region encompasses areas I and II, which are each capable of directing  cell-specific reporter expression (Fig. 3), and none of the area III sequences. Consequently, these results strongly imply that distinct, and in some cases separable, regulatory sequences within the 5'-flanking region of the pdx-1 gene control islet versus duodenal or exocrine expression. Although specific non-islet control sequences have not yet been isolated, the high degree of conservation within area III between all three vertebrate species studied here may suggest that this subregion of HSS1 is associated with such a function. It is also possible that these tissue-specific regulatory elements occur in less extensively conserved regions, including those corresponding to HSS2 (1330 to 800 bp) and HSS3 (260 to +180 bp). However, the results obtained from transfection (22)- and transgenic-based assays² have demonstrated that the HSS3 region cannot independently direct reporter gene expression to either islet  cells or other areas of the pdx-1 expression domain.

We propose that the HSS1 in the mammalian pdx-1 gene is controlled by the combined tissue-specific transcription factor activities associated with the conserved sequences in areas I , II, and III. Although our sequence comparisons were only conducted on the mouse and human genes (Figs. 1 and 2), a recent Southern blot analysis indicates that all three subregions are located in a comparable region of the rat gene (data not shown). In contrast, an area II-equivalent is not found in a similar location in chicken pdx-1 (Fig. 1). How might the absence of area II affect chicken pdx-1 expression? The general similarity in pdx-1 expression patterns between chicken and mammals might mean that area I and III, but not area II, are primarily responsible for the regionalized expression of pdx-1. On the other hand, it is possible that area II imparts transcriptional properties unique to the mammalian gene, perhaps leading to specific differences between birds and mammals in the timing and/or level of pdx-1 expression in different tissues. In mouse, for example, pdx-1 expression precedes islet-1 during pancreatic development (1), but the order is reversed in chicken (34). In addition, two kinds of islets have been uniquely described in the avian pancreas, containing different proportions of glucagon and insulin cells (49) whose cellular composition may be regulated by differential expression of pdx-1.

Area I and area II both contain a single effective HNF3 binding site, but only area I appeared mutationally sensitive when the subregions were tested independently in transfected  cells (Fig. 5). In contrast, HNF3 binding to area I or area II was necessary for the full activity of a pdx-1-driven reporter construct spanning both subregions (see PstBst:pTk A1m and PstBst:pTk Bm in Fig. 6). Further support for a role of this  cell-enriched factor in pdx-1 expression was provided by the observation that endogenous expression is reduced in HNF3^/ EBs (Fig. 7). While our data and those of others (50) fit a model for HNF3 as a direct activator of pdx-1 transcription, we remain cautious in our interpretation of the general relevance of the EB data. For example, studies of gene regulation in the presence and absence of HNF3 illustrate a circumstance where quite profound differences in putative target gene activation were found between differentiated EBs in vitro (28) and endodermally derived tissues in the animal (51). More definitive experiments need to be performed to establish the role of HNF3 in pdx-1 transcription in vivo. Unfortunately, HNF3 homozygous null mutant mice die prior to the differentiation of pancreatic endoderm (23, 27), preventing their use in evaluating HNF3 function in pdx-1 expression. Interestingly, since pdx-1^/ mice are apancreatic (2, 3), the specific function of PDX-1 in the islet was also unclear until  cell-specific PDX-1 mutant mice were generated using the Cre-LoxP system to produce cell-type gene inactivation in  cells by expression of insulin enhancer/promoter-driven Cre recombinase. Continued pdx-1 expression was then shown to be required for expression of various  cell-enriched target genes, including insulin, GLUT2, and IAPP (18). It remains to be seen if tissue-specific gene inactivation strategies applied to HNF3 would lead to compromised  cell function because of reduced expression of pdx-1 target genes.

Our data suggest that pdx-1 transcription in  cells is mediated by functional interactions between HNF3 and other transcription factors binding to evolutionarily conserved sequences of area I and II. The lack of sequence similarity between these regions indicates that a unique and highly complex set of regulators act in conjunction with HNF3 to control transcriptional stimulation. Strikingly, area I also contains potential binding sites for HNF1, HNF1, and HNF4, each of which are transcription factors mutated in families with a form of diabetes termed maturity-onset diabetes of the young (52-54). This raises the possibility that identifying the regulators of HSS1 activity may also provide insight into why  cell function is affected under certain diabetic situations.

	ACKNOWLEDGEMENTS

HNF3 antisera was generously provided by Dr. Brigid Hogan. The sequencing of the 5'-flanking region of chicken and human pdx-1 genes was performed, in part, by the DNA Sequencing Core Facility in the Vanderbilt University Cancer Center.

	FOOTNOTES

^* This work was supported by National Institutes of Health Grants NIH RO1 DK50203 (to R. S.) and NIH R01 HD 28062 and NIH DK 42502 (to C. V. E. W.), Juvenile Diabetes Foundation International postdoctoral fellowship Grant 397019 (to M. G.), and in part by the Vanderbilt University Diabetes Research and Training Center Molecular Biology Core Laboratory (Public Health Service NIH Grant P60 DK20593).The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

The nucleotide sequence(s) reported in this paper has been submitted to the GenBank^TM/EMBL Data Bank with accession number(s) AF192495, AF192496, and AF194114.

To whom correspondence should be addressed. Tel.: 615-322-7026; Fax: 615-322-7236; E-mail: Roland.Stein@mcmail.vanderbilt.edu.

² M. Gannon and C. V. E. Wright, unpublished results.

	ABBREVIATIONS

The abbreviations used are: GLUT2, glucose transporter type 2; HSS, hypersensitive site; HNF, hepatic nuclear factor; bp, base pair(s); ES, embryonic stem; EB, embryoid bodies; kb, kilobase(s); Tk, thymidine kinase; CAT, chloramphenicol acetyltransferase; BHK, baby hamster kidney; RSV, Rous sarcoma virus; PCR, polymerase chain reaction.

	REFERENCES

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES