The PDX1 Homeodomain Transcription Factor Negatively Regulates the Pancreatic Ductal Cell-specific Keratin 19 Promoter*

Keratin 19 is a member of the cytokeratin family that is critical for maintenance of cellular architecture and organization, especially of epithelia. The pancreas has three distinct cell types, ductal, acinar, and islet, each with different functions. Embryologically, the pancreatic and duodenal homeobox 1 (PDX1) homeodomain protein is critical for the initiation of all pancreatic lineages; however, the later differentiation of the endocrine pancreas is uniquely dependent upon high PDX1 expression, whereas PDX1 is down-regulated in the ductal and acinar cell lineages. We find that this down-regulation may be required for normal ductal expression of cytokeratin K19. The K19 promoter-reporter gene assay demonstrates that ectopic PDX1 inhibits K19 reporter gene activity in primary pancreatic ductal cells. This is reinforced by our findings that retrovirally mediated stable transduction of PDX1 in primary pancreatic ductal cells suppresses K19 expression, and short interfering RNA to PDX1 in Min6 insulinoma cells results in the induction of normally undetectable K19. Complementary functional and biochemical approaches led to the unexpected finding that a multimeric complex of PDX1 and two members of the TALE homeodomain factor family, MEIS1a and PBX1b, regulates K19 gene transcription through a specific cis-regulatory element (–341 to –325) upstream of the K19 transcription start site. These data suggest a unifying mechanism whereby PDX1, myeloid ecotropic viral insertion site (MEIS), and pre-B-cell leukemia transcription factor 1 (PBX) may regulate ductal and acinar lineage specification during pancreatic development. Specifically, concomitant PDX1 suppression and MEIS isoform expression result in proper ductal and acinar lineage specification. Furthermore, PDX1 may inhibit the ductal differentiation program in the pancreatic endocrine compartment, particularly beta cells.

Cytokeratins are members of the intermediate filament family that are critical to the maintenance of cell and tissue integrity (1). In addition, they influence membrane and subcellular localization of proteins. The cytokeratin family consists of at least 20 members that are categorized as acidic type I, comprising keratins 9 -20, or basic type II, comprising keratins 1-8. Typically, cytokeratins form heterodimers between one type I member and one type II member. Keratin 19 (K19) 2 is expressed in epithelia and substitutes for keratin 18 in heterodimerization with keratin 8. Among the pancreatic cell types, K19 is specifically expressed in pancreatic ducts in vivo and in primary pancreatic ductal cells in vitro that our laboratory has successfully isolated and characterized (2). We have previously demonstrated that K19 expression is modulated by the KLF4 and Sp1 zinc-finger transcription factors, contributing to its tissue specificity in the pancreas (3). This activity is mediated by a short cis-regulatory region containing an overlapping binding site for KLF4 and Sp1 within the K19 promoter. KLF4 has a higher binding affinity and is the predominant binding factor in pancreatic ductal cells with low Sp1 protein levels (3). PDX1 (pancreatic and duodenal homeobox 1) is a Hox type homeodomain transcription factor that is critical for the transcriptional regulation of beta cell development (4 -9). PDX1 expression and function are noteworthy in the emergence of pancreatic buds from the endoderm and the maintenance of putative pancreatic progenitor cells during development, and it is, thereafter, highly expressed in the endocrine beta cell lineage. Within the pancreas PDX1 expression is minimally detected in ductal and acinar cells. Mice deficient for Pdx1 exhibit pancreatic agenesis (10 -12). Humans with germ line PDX1 mutations develop early (designated maturity onset dia-betes of the young, or MODY4) and late-onset forms of type 2 diabetes (13)(14)(15)(16)(17)(18)(19). Use of a tet-regulatory system for modulation of PDX1 expression in utero has shown that, in addition to its necessity for early pancreatic development, PDX1 is also required later for the formation of acinar cell compartments. In the absence of PDX1, acini do not form, and in addition the precursor epithelium develops a truncated ductal tree consisting of immature duct-like cells. Thus, the temporal and spatial regulation of PDX1 expression appears to be critical for cell fate determination during development and has implications for cell autonomous and non-autonomous behavior during adult differentiation and regeneration (20 -24).
It is precisely the potential interplay between PDX1 and ductal cell morphogenesis that motivated us to investigate the impact of PDX1 upon ductal epithelial cell-specific gene expression and, in this particular context, K19 gene transcription. We report herein that the transcription factor PDX1 negatively regulates K19 expression in pancreatic cells. K19 expression is reduced in pancreatic ductal cells that stably express PDX1 mediated by retroviral transduction, and conversely, Min6 insulinoma cells that normally express high levels of PDX1 exhibit increased K19 expression after short interfering RNA (siRNA) knockdown of PDX1. We identified a cis-regulatory region of 16 nucleotides in the K19 promoter that is negatively regulated by PDX1 and is distinct from the region regulated by KLF4 and Sp1. Surprisingly, consensus binding sites for the MEIS (myeloid ecotropic viral insertion site) and PBX (pre-B-cell leukemia transcription factor) three-amino acid loop extension (TALE) homeodomain transcription factors were also found in this region. This 16-bp regulatory region is modulated by MEIS1a acting as a positive regulator, whereas PBX1b functions as a repressor of K19 reporter gene activity. Complex formation of PDX1 with MEIS1a and PBX1b leads to PDX1mediated repression of K19. We conclude that low level PDX1 expression, as observed in ductal cells, permits higher levels of K19 expression as compared with islet cells that express high levels of PDX1 and undetectable levels of K19. This dynamic interplay between PDX1 and K19 has important implications for cell fate decisions during development and regeneration.

EXPERIMENTAL PROCEDURES
Cell Lines-The mouse primary pancreatic ductal cell line, WT-PDC, was isolated and characterized as previously described and maintained in a serum-free Dulbecco's modified Eagle's medium/ F-12 medium (2). The mouse insulinoma ␤-cell line, Min6, was maintained in Dulbecco's modified Eagle's medium (high glucose and supplemented with pyroxidine hydrochloride) with 10% heatinactivated fetal calf serum. HeLa and PANC-1 cells were grown in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum. Cell lines were kept as subconfluent monolayers and were maintained in a 5% CO 2 humidified incubator at 37°C.
Transient Transfections and Luciferase Activity Assays-Cells were plated 24 h before transfection in 24-well plates and transiently transfected with 0.15 g of pGL3-Basic or pGL3-K19 plasmid, 1 ng of pRL-CMV plasmid expressing the renilla luciferase reporter gene with or without 50 ng to 0.3 g of pCMX-PDX1, 0.3 g of pCS2-MEIS1a, 0.3 g of pCS2-PBX1a, 0.3 g of pcDNA1.1-PBX1b expression plasmids using Lipofectamine 2000 (Invitrogen) as directed by the manufacturer. The total amount of DNA in each transfection was adjusted with the corresponding empty expression vector. pGL3-Basic was used as a standard control. Twenty-four hours after transfection, the cells were harvested, assayed for firefly luciferase activity and normalized to renilla luciferase activity using the Dual-luciferase Reporter assay system (Promega). Luciferase activities were detected by the Orion microplate luminometer (Berthold detection system).
In Vitro Transcription-Translation-MEIS1a, MEIS1b, and PBX1a were in vitro transcribed and translated with the SP6 TNT rabbit reticulocyte lysate coupled transcription-translation system (Promega) according to the manufacturer's protocol. MEIS2b, PBX1b, and PDX1 were translated using the T7 polymerase from a similar coupled reaction kit. Controls of the translation efficiency were performed using [ 35 S]methionine (Redivue, Amersham Biosciences) in the reaction mixes, and samples were resolved by SDS-PAGE (4 -12%).
EMSAs-For EMSAs, nuclear extracts were prepared from WT-PDC cells and Min6 cells as described previously (28). Synthetic oligonucleotides and their respective complementary oligonucleotides for K19-352 probe (Ϫ352 to Ϫ323, 5Ј-GGTG-TGATTTCTAAGGGTGTCAAATTCCTGG-3Ј), TSE2-PDX1 (5Ј-GATCTCAGTAATTAATCATGCA-3Ј) (29), mutant K19-352⌬ (5Ј-GAGGGGTGTGATTTCT-⌬-GGAGGTTTTAAAG-GGCC-3Ј), mutant K19-352m (5ЈGGTGTGATTTCTAAGatca-TCAAATTCCTGGAGG-3Ј), mutant K19-352m2 (5Ј-CTAAG-GGTGTCAAtcagCTGGAGGTTTTAAAGG-3Ј), and mutant 352m4 (5Ј-GTGATTTCTAAGGGTGatcaATTCCTGGAGG-3Ј) were diluted to a final concentration of 5 M (lowercase letters represent mutated nucleotides, and bold letters represent region of interest in the K19 promoter). The sense oligonucleotides were end-labeled using T4 polynucleotide kinase (New England Biolabs, Beverly, MA) in the presence of 50 Ci of [␥-32 P]ATP and then purified using Microspin G-25 columns (Amersham Biosciences) following the manufacturer's instructions. The complementary antisense oligonucleotides were then added to the radiolabeled oligonucleotides and annealed by heating at 100°C for 5 min followed by slow cooling to room temperature. EMSA experiments were carried out by mixing the following components to the reaction mixture: 5 g of nuclear extract, ϳ0.5 pmol of radiolabeled probe, 25 mM HEPES, pH 7.9, 150 mM KCl, 10% glycerol, 5 mM dithiothreitol, and 0.5 g of poly(dI-dC). The various components were incubated at room temperature for 30 min. Nuclear extracts from PDC and Min6 were preincubated in the presence or absence of 100-fold excess of competitor DNAs at room temperature for 20 min before the addition to the reaction mixture. For supershift analysis, nuclear extracts were preincubated with 3 l of goat polyclonal antibody to PDX1, mouse anti-PBX, mouse anti-MEIS, or purified immunoglobulin G at room temperature for 20 min before the addition of the labeled probes. Free and bound DNA were separated on a 4% non-denaturing polyacrylamide gel, which was run at a constant voltage of 120 V in Tris-glycine buffer. After drying the gel, the results were visualized by phosphorimaging or exposed to BioMax MR film (Eastman Kodak Co.).
Western Blot Analysis-For immunoblot analysis, cells were washed with phosphate-buffered saline and lysed with radioimmune precipitation assay buffer (150 mM NaCl, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 50 mM Tris-HCl, pH 7.5, and a mixture of protease inhibitors (Complete mini, Roche Applied Science)). Protein concentration was determined with the Bradford reagent (Bio-Rad). A total of 10 g of total protein was resolved by SDS-PAGE (4 -12%) and transferred to polyvinylidene difluoride membrane (Immobilon, Millipore Corp., Bedford, MA). To check for equal protein transfer, the membranes were stained briefly with Ponceau S solution (Sigma). Blocking was performed in 5% milk, 10 mM Tris-HCl, pH 7.4, 150 mM NaCl, and 0.1% Tween 20 for 1 h before incubation with primary antibodies. Horseradish peroxidase-conjugated secondary antibodies were used according to the manufacturer's protocol. Immunoreactivity was visualized using the ECLϩ system (Amersham Biosciences) and exposed to BioMax MR film.
DNA Affinity Precipitation Assay-5Ј-Biotinylated oligonucleotides and their respective complementary oligonucleotides were synthesized, gel-purified by Integrated DNA Technologies (Coralville, IA), and annealed in Tris-EDTA buffer with 150 mM NaCl by heating at 100°C for 5 min followed by a slow cooling to room temperature. The positive control oligonucleotide (CMV2) was 5Ј-biotin-TAATCAATTACGGGGTCATTA-3Ј. The negative control oligonucleotide (scramble) was 5Ј-biotin-GCCGCC-GCCGCCGCCGCCGC-3Ј. K19 oligonucleotides were 5Ј-biotin K19-352 (as described above). The biotinylated probes (2 g) were incubated on ice with 20 g of Min6 or 60 g of WT-PDC nuclear extract in 400 l of binding buffer (20 mM HEPES, pH 7.9, 10% glycerol, 0.2 mM EDTA, 1.5 mM MgCl 2 , 50 mM KCl, 1 mM dithiothreitol, 0.25% Triton X-100). Increasing amounts of excess competitor (0 -50-fold excess) were included in the binding reaction for competition assays. The negative control was done by omitting the biotinylated probes. After 30 min of incubation, 20 l of streptavidin-agarose beads prewashed 3 times with the binding buffer were added to each reaction mixture, and the reaction was conducted for an additional hour on ice with gentle shaking. The streptavidin-agarose beads were washed 4 times with 1 ml of binding buffer before adding 30 l of protein sample buffer (with 1% 2-␤ mercaptoethanol). All the samples were denatured by heating at 85°C for 5 min and resolved by SDS-PAGE (4 -12%).
Quantitative Chromatin Immunoprecipitation (ChIP)-ChIP assays were performed as described previously (30,31) with a few modifications. Briefly, for each antiserum, one confluent 10-cm plate of pancreatic ductal cells (ϳ1 ϫ 10 7 cells) was cross-linked with 1% formaldehyde in phosphate-buffered saline for 10 min at room temperature and quenched with glycine to a final concentration of 0.125 M. Chromatin was sonicated to create ϳ500-bp fragments in size and precleared with normal goat or mouse IgG (Santa Cruz) overnight at 4°C. After removal of an aliquot for analysis as input, precleared chromatin was divided equally for immunoprecipitation with either goat polyclonal anti-PDX1 antibody, mouse monoclonal anti-MEIS, or normal IgG for 3 h at 4°C. Data were analyzed quantitatively in duplicate by real-time PCR.
GST Pulldown Assays-The GST pull-down assays were carried out using the ProFound TM pulldown GST protein-protein interaction kit (Pierce) according to the manufacturer's protocol. Briefly, GST fusion proteins or GST alone were produced in Escherichia coli BL21-Gold(DE3)pLysS cells induced with 1 mM isopropyl-␤-D-thiogalactopyranoside for 3 h at 37°C. The purified GST proteins were then incubated with immobilized glutathione for 1 h at 4°C. After 3 washes (ProFound lysis buffer: Tris-buffered saline, 1:1), the immobilized baits were incubated overnight at 4°C with 5 l of in vitro translated 35 S-labeled PDX1, MEIS1a, or PBX1b diluted in washing buffer. The beads were then washed 4 times, the captured proteins were eluted in 100 l of sample buffer and heated at 85°C for 5 min, and 20 l of proteins were fractionated by SDS-PAGE (4 -12%). The gels were dried under vacuum at 80°C and autoradiographed.
Viral Infection and RNA Isolation-Min6 cells were infected with recombinant adenovirus containing a siRNA sequence designed to target either PDX1 or luciferase as described (32). Briefly, 5 ϫ 10 5 cells were infected at a multiplicity of infection of 2500 plaque-forming units/cell for 6.5 h. Cells were harvested 72 h post-infection and processed for RNA isolation using Trizol (Invitrogen). RNA samples were treated with DNase I (Ambion), analyzed for integrity with an Agilent 2100 Bioanalyzer, and reverse-transcribed with Superscript II (Invitrogen) using oligo(dT) for priming.
Full-length mouse PDX1 cDNA was obtained by digesting pCMX-PDX1 expression vector with the restriction enzymes SalI and BamHI. The extremities of the insert were then filled blunt with T4 DNA polymerase, and the PDX1 cDNA was inserted into the SnaBI-digested pBABE/puro retroviral vector (33). Retrovirus preparation and WT-PDC infection and selection were conducted as previously described (34).
Reverse Transcription-PCR and Real-time PCR Analysis-Total RNA was isolated from monolayer cultures using Trizol reagent, and cDNA was synthesized by oligo(dT) priming from 1 g of total RNA using a Superscript first-strand synthesis system (Invitrogen) according to the manufacturer's directions. The semiquantitative analysis of transcripts encoding various isoforms of MEIS and PBX was carried out with a mixture of cDNA derived from WT-PDC or Min6, 0.2 M each of the sense and antisense primers, 0.2 M dNTP, and 2.5 units of Pfu Turbo DNA polymerase (Stratagene) in a final reaction volume of 50 l. The specific primers to detect MEIS1, -2, and -3 and PBX1a, -1b, -2, and -3 were described elsewhere (35). The PCR program was 90°C for 2 min and 30 cycles of 94°C for 30 s, 60°C for 30 s, and 68°C for 1 min with a final extension at 68°C for 7 min. Data shown were obtained with 30 PCR cycles. Analysis of digested and undigested PCR products was done by electrophoresis on a 1% agarose gel.
Densitometry Measurements and Statistical Analysis-Results are expressed as mean Ϯ S.E. Densitometry measurements were performed using Scion Image Beta 4.02 software (Frederick, MD) and calibrated with the ␤-actin signal. Analysis of variance with a Tukey post hoc test was used for statistical analysis. A p Ͻ 0.05 was considered statistically significant.

Expression of PDX1 in Primary Pancreatic Ductal Cells-To
determine the level of expression of PDX1 in our primary pancreatic cell lines, designated as wild-type pancreatic ductal cells (WT-PDC), we performed Western blots on whole cell lysates. The results show that PDX1 is expressed at low levels in the pancreatic ductal cells, and its expression is significantly higher in the insulinoma cell line Min6 (PDX1 is 22 Ϯ 3.5-fold higher in Min6 cells compared with WT-PDC, p Ͻ 0.01). PDX1 expression was undetectable in PANC-1 and HeLa cells (Fig. 1).
PDX1 Negatively Regulates K19 Expression-The K19-1970 plasmid containing a luciferase reporter gene under the control of the 5Ј regulatory region of the mouse K19 gene was constructed as previously described (3). A set of 5Ј deletion constructs were generated using a PCR-based technique, and the reporter plasmids were transiently transfected into WT-PDC ( Fig. 2A). PDX1 had no effect on the basal luciferase activity of the empty pGL3-Basic vector (Fig. 2B). Data from the luciferase reporter assays revealed that co-expression of PDX1 in WT-PDC negatively regulated the K19 promoter. PDX1 partially suppressed K19 expression by about 50% when using the K19 promoter constructs from Ϫ1970/ϩ46 to Ϫ341/ϩ46. However, PDX1 had no effect on K19 expression when the K19 promoter was reduced to Ϫ325/ϩ46 and Ϫ288/ϩ46. This result suggests that the negative regulation of K19 by PDX1 is alleviated when the region between Ϫ341 and Ϫ325 of the K19 promoter is deleted. Interestingly, this region does not contain an obvious AT-rich consensus DNA binding site for PDX1.
PDX1 Occupies the K19 Promoter in Vivo-To provide in vivo evidence for PDX1 binding to the K19 promoter, we carried out ChIP in WT-PDC and analyzed the results using quantitative real-time PCR. The compiled data from 5 independent ChIP experiments demonstrate that PDX1 specifically occupies the K19 promoter in WT-PDC, resulting in an ϳ2.5-fold higher enrichment than observed with the control IgG immunopre-cipitation (Fig. 3A). In these assays the lack of PDX1 occupancy at the albumin promoter serves as a negative control. These data suggest that PDX1 associates with the K19 promoter despite the absence of a TAA(T/T)TAT consensus sequence, perhaps via interaction with another/other DNA binding factor(s).
To further confirm the interaction of PDX1 with the K19 promoter, the DNA affinity precipitation assay was carried out using both Min6 and WT-PDC nuclear extracts and a doublestranded K19 oligonucleotide. As a positive control for PDX1 binding, we used a well established element taken from the human cytomegalovirus immediate early (CMV IE) promoter (36). Our results showed that the K19-352 and the CMV oligonucleotides were able to pull down PDX1 protein from both  . PDX1 occupancy of the K19 promoter in PDC cells and DNA affinity precipitation assay. A, after formaldehyde cross-linking and sonication, chromatin from pancreatic ductal cells was immunoprecipitated with either anti-PDX1 antiserum or normal IgG. The data were analyzed quantitatively using real-time PCR with primers designed to amplify the putative PDX1 binding region of the K19 promoter (Ϫ337 to Ϫ253) or a distal TAATcontaining element of the albumin promoter. For each primer pair, ChIP signals were compared with a dilution series of input chromatin to account for differing amplification efficiencies and are expressed here as such. These data represent the means Ϯ S.E. of five independent experiments. The asterisk signifies a p value of 0.01. B and C, 20 g of nuclear extracts (NE) from Min6 (B) or WT-PDC (C) were incubated with 2 g of biotinylated K19-352, CMV2, or scramble oligonucleotides (containing GCC repeats) for 30 min. The DNAprotein complexes were then precipitated by streptavidin-agarose beads followed by extensive washes, and the reaction mixtures were resolved by SDS-PAGE. PDX1 expression was detected by Western blot. D, increasing amounts (2-50-fold excess) of unlabeled K19-352 oligonucleotides were used as specific competitors.
Min6 and WT-PDC nuclear extracts, whereas a scrambled oligonucleotide with GC repeats did not (Fig. 3, B and C). The binding specificity was further confirmed by competition assays with cold K19 oligonucleotide incubated with WT-PDC nuclear extract (Fig. 3D) and Min6 nuclear extracts (data not shown). The cold K19 oligonucleotide competed for binding in a dose-dependent manner that almost completely abrogated binding at 10-fold or greater concentration of the biotinylated oligonucleotide.
PDX1 Negatively Regulates K19 Expression in Min6 and WT-PDC Cell Lines-We infected WT-PDC with a retrovirus expressing PDX1 or the control empty pBABE vector. We confirmed higher expression of PDX1 by Western blot and showed by real-time PCR that induction of PDX1 results in a 54% decrease of K19 expression (Fig. 4A). Conversely, to comple-ment the PDX1 overexpression studies, we employed a siRNAmediated approach to knockdown endogenous PDX1 in Min6 cells, which normally express high levels of PDX1 and undetectable levels of K19. Min6 cells were infected with an adenovirus encoding siRNA designed to target either PDX1 (AdsiPDX1) or luciferase (AdsiLUC), and we confirmed specific reduction of PDX1 at both the protein and mRNA levels (Figs. 4, B and C, respectively) (32). Furthermore, we observed a 2.5-fold increase in K19 transcript levels as assessed by quantitative real-time PCR in the AdsiPDX1-infected cells relative to both untreated cells and those infected with AdsiLUC (Fig. 4C). These data support a role for PDX1 in the negative transcriptional regulation of K19 and suggest that the high PDX1 levels in beta cells might contribute to the repression of K19 expression in these cells.
Analysis of the Upstream K19 Promoter Region by EMSA-The cis-regulatory region of 16 nucleotides between Ϫ341 and Ϫ325 delineated by the promoter-luciferase reporter gene assays to be regulated by PDX1 was then analyzed by EMSAs. Nuclear extracts from WT-PDC were used. The specificity of the retarded band observed was verified by competition with 100-fold excess of unlabeled K19-352 probe, whereas no competition was apparent with the nonspecific oligonucleotide (Fig. 5A). Supershift assays were performed by preincubating the WT-PDC nuclear extracts with increasing amounts of anti-PDX1 antibody before the addition of the radiolabeled K19-352 probe. These results suggest that PDX1 may interact with the K19-352 probe (Fig. 5B).

Mutagenesis of the Region Regulated by PDX1 in the K19 Promoter Reveals Interaction with MEIS and PBX-Systematic
analysis of the 5Ј upstream region of K19 that is regulated by PDX1 revealed the presence of consensus binding sites for MEIS (TGTCA) and PBX (TGATT), two members of the TALE family of homeodomain proteins (37). Previous work has demonstrated that PDX1 is able to form a trimeric complex with PBX1b and MEIS2b to activate the elastase ELA1 mini-enhancer in HeLa cells (27,38).
We used reverse transcription-PCR to determine MEIS and PBX expression levels in WT-PDC and Min6 cells. Total RNA was purified, and the cDNA was synthesized by oligo(dT) priming. Using specific primers that were described previously for MEIS1, -2, and -3 and PBX1a, -1b, -2, and -3 (35), the PCR results demonstrate that MEIS 2 and 3 are present in both WT-PDC and Min6 cells, whereas MEIS1 is detected only in WT-PDC. Both cell lines contain several isoforms of PBX, namely PBX1, -2, and -3 (Fig. 6A). To further quantify expression levels, real-time PCR was performed for PBX1 and MEIS2 (Fig. 6B). These results reveal significantly greater levels of both PBX1 and MEIS2 in WT-PDC compared with Min6 cells. The differences in MEIS and PBX isoform expression may contribute to the differential regulation of K19 expression in distinct pancreatic cell types.
Deletion of the 16-bp region abrogated PDX1-mediated repression of K19 and demonstrated a decrease in promoter activity that suggests a loss of positive regulatory elements (Fig.  7B). Mutations 352m and 352m2 had little effect upon PDX1mediated repression (Fig. 7B and data not shown), whereas mutation 352m4, which mutates part of the consensus binding site for MEIS, relieved the PDX1 repression of K19 transcription (Fig. 7B). We then co-expressed the MEIS1a and MEIS1b expression vectors along with the K19 reporter-luciferase gene in WT-PDC. The results of these transfection assays demonstrate that MEIS1a up-regulates K19 expression, whereas MEIS2b does not (Fig. 7C). PBX1a co-expressed with the K19 reporter-luciferase gene does not appear to directly regulate K19 transcription (Fig. 7D). Noticeably, expression of PBX1b in WT-PDC appears to negatively regulate K19 in a manner comparable with PDX1 (Fig. 7E). Moreover, co-expression of MEIS1a and/or PBX1a in WT-PDC that express high levels of ectopic PDX1 is not able to rescue the negative regulation of K19 by PDX1 ( Fig. 7E and data not shown). Similar results are observed for PBX1b expressed in WT-PDC, with MEIS1a unable to rescue the down-regulation of K19 by PBX1b (Fig.  7E). These results indicate that a high level of PDX1 or PBX1b in WT-PDC is sufficient to down-regulate K19 expression even in the presence of the transactivator MEIS1a or MEIS1b.
MEIS1a Interacts with the K19 Promoter-EMSA was performed to determine whether the MEIS and PBX homeodomain proteins are involved in the direct binding to the 16-bp region of the K19 promoter. Mouse monoclonal anti-MEIS and anti-PBX antibodies were added to the reaction mixtures containing WT-PDC nuclear extracts before the addition of the radiolabeled probe. The antibody against PDX1 specifically eliminates the retarded-mobility complex as demonstrated previously. The antibody against MEIS supershifted the same complex, although the level of intensity was different from the PDX1 EMSA. Of note, the MEIS supershift revealed two bands or a doublet, perhaps consistent with two MEIS isoforms, MEIS1a and MEIS1b. However, the antibody against PBX did FIGURE 5. EMSA with PDX1 and K19 promoter radiolabeled cis element probes. A, the radiolabeled K19-352 probe located between Ϫ352 and Ϫ325 of the K19 promoter was incubated in the presence of 5 g of WT-PDC nuclear extract for 30 min at room temperature in presence or absence of a 100ϫ excess-fold nonspecific (NS, TFIID double-strand oligonucleotide obtained from Santa Cruz) or specific (S, cold K19-352 oligonucleotide) competitor. The arrow indicates bound probe, and the star indicates free probe. B, radiolabeled K19-352 probe was incubated in the presence of 5 g of WT-PDC nuclear extract (NE) for 20 min at room temperature. For supershift assay the WT-PDC nuclear extracts were preincubated as indicated with increasing amounts (0.5-3 l) of control goat IgG or goat anti-PDX1 for 20 min on ice before incubation with radiolabeled probe. PDX1 antibody was able to shift the retarded band as indicated by an arrow.

FIGURE 6. MEIS and PBX gene expression in WT-PDC and Min6 cells.
A, MEIS1, -2, and -3 and PBX1, -2, and -3 gene expression in WT-PDC and Min6 cells. Reverse transcription-PCR was performed using reverse transcription products and primers specific for each of the MEIS and PBX isoforms. For PBX1, the lower band corresponds to PBX1b, whereas the upper band corresponds to PBX1a. B, relative PBX1 and MEIS2 transcript levels in WT-PDC (designated as PDC) and Min6 cells by real-time PCR. PBX1 and MEIS2 mRNA levels were assessed using quantitative real-time PCR from either PDC or Min6 lysates (n ϭ 3 of each done independently). PBX1 and MEIS2 primers were designed and optimized for linear amplification of cDNA to ensure accurate quantification, and the Ct values were normalized to hypoxanthineguanine phosphoribosyltransferase (HPRT). Error bars represent the means Ϯ S.E. *, p Ͻ 0.01; **, is p Ͻ 0.001. DECEMBER 15, 2006 • VOLUME 281 • NUMBER 50 not appear to interfere with the same complex formation (Fig.  8A), although we cannot rule out the lack of efficiency of this antibody in EMSA assays. These results suggest that in WT-PDC, MEIS binds to the K19-352 probe.

PDX1 and Keratin 19 Gene Regulation
To confirm in vivo occupancy of MEIS protein to the K19 promoter, the ChIP assay was performed in WT-PDC using the mouse monoclonal anti-MEIS1/2/3 antibody for immunoprecipitation. Real-time PCR results indicate a 5-fold enrichment using mouse anti-MEIS as compared with control IgG (Fig. 8B). The albumin promoter served as a negative control, showing no MEIS occupancy. These results were further confirmed by the DNA affinity precipitation assay (Fig. 8C). The same membranes used previously to demonstrate the interaction of PDX1 with the K19 promoter were probed with anti-MEIS antibody. The results demonstrated that MEIS proteins indeed interact with the Ϫ352/Ϫ323 fragment of the K19 promoter and that this interaction is specific since cold K19 oligonucleotide abrogated the signal in a dose-dependent manner. In addition, we showed an interaction of PBX1 to the same region of K19 promoter, most likely to the consensus binding site located at Ϫ345/Ϫ350 of the K19 promoter (Fig. 8C).
MEIS1a Interacts with PDX1 Independently of PBX1-Next, we performed GST pulldown assays to determine whether PDX1 was involved in a trimeric complex with MEIS1a and PBX1b. GST-PDX1 fusion protein was incubated with in vitro translated 35 S-labeled MEIS1a, and the result was visualized by autoradiography. The results show that GST-PDX1 interacts with MEIS1a (Fig. 9). Interestingly, GST-PDX1 also interacts with in vitro translated MEIS1b (data not shown). Of note, the addition of PBX1b in the pulldown reactions does not appear to increase MEIS1a interaction with PDX1, suggesting that PBX1b has little function in stabilizing the PDX1/MEIS1a complex. Furthermore, GST-PDX1 (144 -283), which contains the homeodomain and C terminus domain of PDX1, and GST-PDX1 (206 -283), which contains the C terminus domain of PDX1, are still able to interact with in vitro translated MEIS1a (data not shown and Fig. 9), thereby indicating that the C terminus of PDX1 mediates the interaction with MEIS1a and that this interaction is PBX1-independent. The N terminus domain of PDX1 has been shown to interact with PBX1, whereas our data suggest a new interaction between the C terminus domains of PDX1 with MEIS1a.

DISCUSSION
The objective of this study was to understand the molecular basis underlying the regulation of pancreatic ductal epithelial cell-specific K19 promoter gene activity by the PDX1 homeodomain protein. We found that PDX1 negatively regulates K19 expression in WT-PDC and Min6 cells as demonstrated by a reduction in K19 expression in WT-PDC expressing ectopic PDX1, whereas the introduction of a siRNA to PDX1 in Min6 leads to increased K19 expression. More specifically, PDX1 regulates a 16-bp region located between Ϫ341 and Ϫ325 upstream of the transcription start site of the K19 gene. This region does not contain any consensus DNA binding sites for PDX1. Instead, the consensus binding sites for two homeodomain transcription factors MEIS1 (located at position Ϫ332 to Ϫ337) and PBX1 (located at position Ϫ345 to Ϫ350) are present.
MEIS1 and PBX1 are members of the TALE superclass of atypical homeodomain-containing proteins (39). Additionally, MEIS and PBX have several isoforms. In our primary pancreatic ductal cells, MEIS1, -2, and -3 are present, whereas in Min6 cells, MEIS2 and -3 are present, but MEIS1 is absent. Similar analysis reveals that PBX1b, -,2 and -3 are present in primary pancreatic ductal cells, but PBX1a is absent. However, in Min6 cells, PBX1a, -1b, -2, and -3 are present. HOX proteins, such as PDX1, can form complexes with either PBX or MEIS isoforms to result in a heterotrimeric complex. To that end it has been demonstrated that PDX1 forms a multimeric complex with PBX1b and MEIS2b to regulate the 10-bp B element of the transcriptional enhancer of the pancreatic elastase I gene promoter (27). The complex binds to overlapping half-sites for PDX1 and PBX in the promoter. Whereas in pancreatic acinar cells the B element requires other elements of the ELA1 enhancer for promoter activity, in beta cells the B element can activate a promoter in the absence of other enhancer elements.
Our results indicate that a multimeric complex of PDX1, MEIS1a, and PBX1b modulates K19 transcriptional activity and that the in vitro complex between PDX1 and MEIS1a is most critical. MEIS1a alone can induce K19 gene transcription through direct binding with its cognate DNA consensus binding site (also evident with MEIS1b). Mutation of the MEIS consensus binding site (located Ϫ332 to Ϫ337) abolished MEIS1a positive regulation of K19 gene transcription and, interestingly, abolished also PDX1-mediated repression. Noticeably, the recruitment of PDX1 or PBX1b negates the MEIS1a-mediated effect, thereby leading to negative regulation of K19 gene transcription. Our data suggest that PDX1 regulates K19 gene transcription indirectly by interaction of its C terminus domain with MEIS1a and, thus, potentially modulating MEIS1a binding to DNA. However, we have no direct evidence for DNA binding of PBX1b to the K19 promoter, suggesting that PDX1 alone or possibly through the known PDX1/PBX1b complex interacts with MEIS1a, and this heterocomplex of PDX1/MEIS1a and/or PDX1/PBX1b/MEIS1a modulates K19 gene transcription in a manner that may suppress the effects of direct DNA binding by MEIS1a. Thus, the nature of the association between PDX1 and the TALE proteins, MEIS, and PBX controls gene transcription in pancreatic acinar cells (27,37) and ductal cells. Furthermore, we would propose that the particular Arrows indicate bands shifted and supershifted. B, after formaldehyde cross-linking and sonication, chromatin from WT-PDC was immunoprecipitated with either anti-MEIS antiserum or normal IgG. The data were analyzed quantitatively using real-time PCR with primers designed to amplify the putative MEIS binding region of the K19 promoter (Ϫ337 to Ϫ253) or a distal TAAT-containing element of the albumin promoter. For each primer pair ChIP signals were compared with a dilution series of input chromatin to account for differing amplification efficiencies and are expressed here as such. These data represent the means Ϯ S.E. of two independent experiments done in duplicates. The asterisk signifies a p valueϽ0.02. C, 20 g of nuclear extracts from WT-PDC were incubated with 2 g of biotinylated K19-352 oligonucleotide, CMV2 oligonucleotide, scramble oligonucleotide, or increasing amounts (2-50-fold excess) of unlabeled K19-352 oligonucleotide (for competition assay) for 30 min. The DNA-protein complexes were then precipitated by streptavidin-agarose beads and washed, and the reaction mixtures were resolved by SDS-PAGE. MEIS and PBX interactions were detected by Western blot. FIGURE 9. The C terminus domain of PDX1 interacts with MEIS1a. GST pulldown assays are shown. A purified fusion protein consisting of GST fused to the intact PDX1 or the C terminus domain of PDX1 (206 -283) was incubated with in vitro transcribed-translated 35 S-labeled MEIS1a alone or in combination with 35 S-labeled PBX1b. GST incubated with 35 S-labeled MEIS1a served as negative control. After washing, proteins bound to glutathione beads were resolved by SDS-PAGE, dried and exposed to film. Western blot staining for PDX1 was used as a control for equal protein loading. DECEMBER 15, 2006 • VOLUME 281 • NUMBER 50 MEIS isoform is critical in the regulation of gene transcription in acinar (MEIS2b) versus ductal cells (MEIS1a/1b), whereas PBX1b expression is likely constant. We would also suggest that whereas in acinar cells PDX1/MEIS2b/PBX1b (domain B) and p48/PTF1 (domain A) are necessary for the acinar specific elastase transcriptional activity, PDX1/ MEIS1a/PBX1b (domain B "equivalent") and KLF4 (domain A equivalent) are necessary for the ductal K19 transcriptional activity (3,27,37). This does not preclude the possibility of recruitment of coactivators (e.g. p300, CBP (cAMPresponse element-binding protein (CREB)-binding protein)) and/or co-repressors by PDX1 (30, 40 -46). Our results are further highlighted by the functional characterization of these transcription factors in primary pancreatic ductal cells that provide biological relevance. The potential role of PDX1 in ductal cells either as a marker of putative progenitor cells or during states of pancreatic regeneration has been reported previously (47)(48)(49). Additionally, phosphorylated PDX1 has been noted in ductal cells and in islet cells; however, how this form of PDX1 modulates PDX1-mediated gene expression is not known (50).

PDX1 and Keratin 19 Gene Regulation
A model that emerges from our studies would complement developmental studies that indicate PDX1 is necessary for early pancreatic development and proper specification of the endocrine lineage (20). PDX1 may be necessary for proper formation of the acinar cellular compartment. In the absence of PDX1, achieved by a regulatable system, acini do not form in the appropriate fashion, and yet, immature ductal cells do form (20). This might mean that PDX1 needs to be degraded or sequestered from ductal cells during development, thereby permitting MEIS1a and/or MEIS1b as well as KLF4 to exert their positive regulatory effects upon the ductal lineage, such as in K19. Future investigations are geared to understand how this transcriptional regulatory machinery exerts its effects in mouse models.