The Pancreatic and Duodenal Homeobox Protein PDX-1 Regulates the Ductal Specific Keratin 19 through the Degradation of MEIS1 and DNA Binding

Background Pancreas organogenesis is the result of well-orchestrated and balanced activities of transcription factors. The homeobox transcription factor PDX-1 plays a crucial role in the development and function of the pancreas, both in the maintenance of progenitor cells and in determination and maintenance of differentiated endocrine cells. However, the activity of homeobox transcription factors requires coordination with co-factors, such as PBX and MEIS proteins. PBX and MEIS proteins belong to the family of three amino acid loop extension (TALE) homeodomain proteins. In a previous study we found that PDX-1 negatively regulates the transcriptional activity of the ductal specific keratin 19 (Krt19). In this study, we investigate the role of different domains of PDX-1 and elucidate the functional interplay of PDX-1 and MEIS1 necessary for Krt19 regulation. Methodology/Principal Findings Here, we demonstrate that PDX-1 exerts a dual manner of regulation of Krt19 transcriptional activity. Deletion studies highlight that the NH2-terminus of PDX-1 is functionally relevant for the down-regulation of Krt19, as it is required for DNA binding of PDX-1 to the Krt19 promoter. Moreover, this effect occurs independently of PBX. Second, we provide insight on how PDX-1 regulates the Hox co-factor MEIS1 post-transcriptionally. We find specific binding of MEIS1 and MEIS2 to the Krt19 promoter using IP-EMSA, and siRNA mediated silencing of Meis1, but not Meis2, reduces transcriptional activation of Krt19 in primary pancreatic ductal cells. Over-expression of PDX-1 leads to a decreased level of MEIS1 protein, and this decrease is prevented by inhibition of the proteasome. Conclusions/Significance Taken together, our data provide evidence for a dual mechanism of how PDX-1 negatively regulates Krt19 ductal specific gene expression. These findings imply that transcription factors may efficiently regulate target gene expression through diverse, non-redundant mechanisms.


Introduction
The pancreas is a multifunctional organ that comprises an endocrine compartment, which regulates glucose homeostasis mainly through insulin secretion and is responsible for the secretion of other hormones, and an exocrine compartment, where acinar cells produce digestive enzymes that are secreted into the intestine via a network of pancreatic ducts. Each distinct cell type of the pancreas (endocrine, acinar and ductal) originates from a common pool of progenitor cells during development and requires a complex pattern of transcription factors [1] as well as mesenchymal-epithelial interactions [2] for proper lineage specification. In the mouse, dorsal and ventral pancreatic buds start as outgrowths from the foregut endoderm on day E9.5 [3], branch and then fuse at E12 to E13. Functional cells can be observed at day E14.5 [4].
The Pancreatic and Duodenal homeobox protein PDX-1 is critical for pancreatic development. It is expressed in the foregut endoderm at E8.5, prior to the onset of bud formation and embryonic deletion of PDX-1 results in pancreatic agenesis [5]. In addition to its crucial functions during development, PDX-1 is also required for accurate endocrine function by regulating endocrine gene expression [6] and b-cell survival in the adult [7]. Recent three-dimensional cell culture studies have revealed the role of PDX-1 in ductal branching morphogenesis or tubulogenesis, as it appears to be re-expressed in otherwise PDX-1 negative pancreatic ductal cells at the very site of branching [8]. Although PDX-1 has been described as a positive regulator of gene expression in endocrine cells, such as insulin and somatostatin, recent evidence has shown that PDX-1 can be a negative regulator of gene expression in non-endocrine pancreatic cells [9].
Keratins belong to the family of intermediate filament proteins and are critical for proper tissue function and maintenance [10]. The family of keratins consists of 54 genes and can be classified into type I (acidic) or type II (basic to neutral) keratins [11,12]. Typically, one type I keratin forms a heterodimer with a type II keratin. Keratin 19 (Krt19) can be found in various stratified and simple epithelial cells. In the pancreas, its expression is restricted to ductal cells and is absent in acinar or endocrine cells. We have demonstrated previously that Krt19 expression is regulated through KLF4 and Sp1 transcription factors and that differences in distribution of these factors in the pancreas are responsible for ductal specific expression of Krt19 [13]. To investigate the potential interplay between PDX-1 and ductal cell morphogenesis, we have studied the impact of PDX-1 on Krt19 transcriptional activity. In this study, we demonstrated transcriptional repression of Krt19 by PDX-1. In addition, we identified the Hox co-factor family of MEIS proteins as DNA binding partners on the Krt19 promoter [9].
MEIS (Myeloid ecotropic viral integration site) proteins are members of the three amino acid loop extension (TALE) homeodomain transcription factors and are required for the proper DNA recognition by and the correct function of Hox transcription factors [14]. Meis1 was first identified as a common viral integration site in BXH-2 mice, which have nearly a 100% rate of spontaneous leukemia [15]. Thereafter, the genes Meis2 and Meis3 were identified with a homology of 83% and 66%, respectively, at the amino acid level. Additionally, several splice variants exist for each gene. Of note, the Xaenopus laevis homologs Xmeis1-1 and Xmeis1-2 share an amino acid sequence identity of 97% and 100% with MEIS1a, suggesting a highly evolutionarily conserved function [16]. Indeed, Meis1 is essential for embryonic development as Meis1 knockout mice die during embryogenesis [17] due to hematopoietic and vasculature defects. However, functional differences between the isoforms are not well understood and require elucidation. Other members of the TALE homeodomain transcription factors are PBX proteins that interact with the FPWMK motif of Hox transcription factors. This is also true for PDX-1 in pancreatic development as mutation of FPWMK to AAGGQ in PDX-1 abrogates PBX binding and results in pancreatic hypoplasia [18]. A similar phenotype is also observed in Pbx1 knockout mice [19] underscoring the importance of PDX-1/PBX1 interactions for pancreatic development.
Here, we set out (1) to characterize the domains of PDX-1 involved in Krt19 transcriptional repression, (2) to determine how different Meis genes affect Krt19 transcriptional regulation and (3) to elucidate functional interactions between MEIS proteins and PDX-1. We provide evidence for the critical role of the NH 2terminus of PDX-1 in mediating Krt19 repression. In addition, MEIS1 is identified as the main isoform that is functionally relevant in pancreatic ductal cells. MEIS1 binds directly to the Krt19 promoter and knockdown of Meis1 mRNA, but not Meis2 mRNA, decreases Krt19 mRNA levels. In addition, we show coexpression of MEIS1 and KRT19 in pancreatic ductal cells in vivo. Surprisingly, PDX-1 over-expression leads to post-transcriptional down-regulation of MEIS1 protein, which could be prevented by inhibition of the proteasome, although this may potentially be independent of ubiquitination. Taken together, our data suggest a novel, dual mechanism of PDX-1 mediated regulation of Krt19 transcriptional activity, one directly through DNA binding and the other through down-regulation of the Krt19 transcriptional activator MEIS1.

Animals -Ethics Statement
All procedures involving animals were approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania (Protocol #800502). Six to eight week old C57BL/6 mice were purchased at Charles River Laboratories (Wilmington, MA) and housed according to institutional guidelines for another week before experiments were performed.

Plasmids
Deletions of PDX-1 were generated with a PCR based technique using pCMX-PDX-1 as a template [9]. Primers were selected to create full length PDX-1 or to delete amino acids (AA) 1-37, 1-109, 1-144, 260-284, and 235-284 and to add the coding sequence for the V5-epitope in frame with the C-terminus. Restriction sites were 59-XhoI and 39-SacII. Following PCR amplification, coding sequences were subcloned into pIRES2-EGFP. Subsequently, the coding sequence of PDX-1 was removed from pCMX by restriction with XhoI and BamHI and the V5-tagged inserts were subcloned from the pIRES2-EGFP constructs back to pCMX using XhoI and BamHI. Mutation of the pentapeptide motif of PDX-1 FPWMKRAAGGQ was introduced using Quikchange II XL Site Directed Mutagenesis Kit (Stratagene, La Jolla, CA) according to the manufacturer's instructions. MEIS1a was amplified from pCS2-MEIS1a [9] and the coding sequence for FLAG was added in frame with the C-terminus. For this purpose, NheI was used for 59restriction and XhoI for 39-restriction. The amplicon was then inserted into pcDNA3.1(+) (Invitrogen, Carlsbad, CA). pcDNA-MEIS2b-FLAG and pcDNA-MEIS3-FLAG were provided by Dr. Doris A. Stoffers. The pM1, pM1 KOX-KRAB and pGL4.10 Gal4-TK plasmids were generous gifts of Dr. Frank J. Rauscher III and have been described earlier [21]. PDX-1 and PDX-1 1-37 were fused to the GAL4-DNA-binding domain in pM1 by amplification of PDX-1 or PDX-1 1-37 from the PDX-1 vectors described above, using BamHI and SacII restriction sites. For insertion of 56 Gal4 binding motifs into pKrt19 -1970, unique AgeI and SacII restriction sites were introduced at position 2367 bp upstream of the transcriptional start site. Subsequently, the 5xGal4 binding motif was removed from pIDTSMART 5xGal (Integrated DNA Technologies, Coralville, IO) using AgeI and SacII and inserted into pKrt19 -1970. pCMV-HA-Ubiqutin was a generous gift of Dr. J. Alan Diehl and has been described previously [22]. Primer sequences used for cloning, generation of deletions or internal mutations can be found in Table S1.
Transfections, reporter assay, proteasome inhibition and treatment with Trichostatin A (TsA) Quantitative reverse-transcriptase PCR RNA was isolated using the RNeasy kit (Quiagen, Maryland, MD) and 1 mg was transcribed into cDNA (Taqman Reverse Transcription Reagents, Applied Biosystems, Branchburg, NJ). RNA was primed with random hexamers. Quantitative analysis was carried out on an ABIPrism 7000 sequence detection system and the amount of target gene was normalized to the endogenous reference as described. [24,25]. Intron spanning isoform specific primer sequences are listed in Table S2.

Immunoprecipitation-Electrophoretic Mobility Shift Assay (IP-EMSA)
HEK 293T cells were plated in p100 dishes at a density of 5610 6 and transfected with pcDNA3.1(+), pcDNA-MEIS1a-FLAG, pcDNA-MEIS2b-FLAG or pcDNA-MEIS3-FLAG using Lipofectamine 2000 according to the manufacturer's instructions. Fortyeight hrs. post transfection cells were washed twice with PBS. Eight hundred ml lysis buffer (50 mM TRIS pH 8.0, 150 mM NaCl, 1% Nonidet-P 40) containing protease inhibitors (Roche, Indianapolis, IN) were added subsequently. Cells were scraped, transferred to a micro-centrifuge tube and allowed to lyse on ice for 10 min. After clearing by centrifugation (10 min, 14000 rpm, 4uC), protein concentration was determined by the Bradford assay. 1.5 mg protein in a total volume of 500 ml were pre-cleared for 1 h at 4uC by adding 60 ml of a 50% slurry of washed protein agarose G beads (Invitrogen, Carlsbad, CA). Pre-cleared lysates were then incubated overnight at 4uC with 3 mg anti-FLAG (Sigma, St. Louis, MO) or 3 mg control non-immunogenic mouse IgG (Invitrogen, Carlsbad, CA). Samples were prepared for EMSA as described earlier [26]. Briefly, the immunoprecipitates were collected by centrifugation at 3,500 rpm for 1 min at 4uC and washed five times with wash buffer (20 mM HEPES, pH 7.9, 100 mM KCl, 0.2 mM EDTA, 20% glycerol), containing 1% NP-40 and once with wash buffer alone. Elution from beads was performed with 100 ml of 1.6% sodium deoxycholate in wash buffer for 15 min on ice. The sodium deoxycholate was neutralized with 12 ml of 10% NP-40, and the supernatant was directly used for EMSA. 20 ml of the supernatant were used for western blot analysis to confirm the efficiency of the IP.
EMSA 80 fmol of c 32 P labeled oligonucelotide were incubated with the supernatant for 20 min on ice in the presence of 1 mg of anti-FLAG, anti-MEIS1/2/3 (Millipore, Billerica, MA) or mouse control IgG in a reaction mix containing 5% glycerol, 50 mM KCl, 1 mM EDTA, 10 mM TRIS, 50 ng/ml poly(dI-dC) and 1 mM DTT. Complexes were resolved on a 5% polyacrylamid gel in 0.56 TBE buffer for 90 min at 200 V. The gels were then dried and exposed on X-ray film at 280uC. The Krt19 promoter fragment used in this assay was 59-CTAAGGGTGTCAAATTCCTG-39.

Immunoprecipitation
HEK 293T were transfected with pCMV-HA-Ubiquitin, pCDNA-MEIS1a-FLAG and pCMX-PDX-1 in equal amounts. Forty-eight hours post transfection cells were treated with either DMSO or Bortezomib (1 mM). IP and SDS-PAGE were essentially performed as described previously [20]. Briefly, cells were lysed in 800 ml lysis buffer (50 mM TRIS pH 8.0, 150 mM NaCl, 1% Nonidet-P 40) containing protease inhibitors (Roche, Indianapolis, IN). 1.5 mg protein in a total volume of 500 ml was pre-cleared for 1 h at 4uC by adding 60 ml of a 50% slurry of washed protein agarose beads (Invitrogen, Carlsbad, CA). Protein A agarose was used for antibodies produced in rabbit, Protein G agarose was used for all other antibodies. Pre-cleared lysates were then incubated overnight at 4uC with one of the following antibodies: anti-V5, anti-HA (Santa Cruz Biotechnologies, Santa Cruz, CA) or anti-FLAG. Proteins were separated by SDS-PAGE and detected by appropriate HRP-coupled secondary antibodies.

Immunofluorescence staining
Eight to ten week old C/BL6 mice were sacrificed and the pancreas was removed quickly and fixed overnight in 4% paraformaldehyde. Organs were then embedded in paraffin, sectioned at 4 mm and mounted on glass slides. Following standard dewaxing and dehydration procedures, antigen retrieval was performed by microwaving the sections for 12 minutes at 600 W in buffered citric acid (pH 6.0). After permeabilization and blocking, sections were incubated overnight with the following antidodies: Keratin 19 (TROMA III, University of Iowa), MEIS1 (Abcam, Cambridge, MA, ab19867) and/or PDX-1 A-17 (Santa Cruz Biotechnologies, Santa Cruz, CA). Visualization was performed by incubation with Cy2 or Cy3 labeled appropriate secondary antibodies (Jackson Immunoresearch, West Grove, PA, 1:600). Nuclear staining was achieved using 49,6-Diamidino-2phenylindole dihydrochloride (DAPI) (Sigma, St. Louis, MO). Images were obtained using a Nikon E 600 microscope (Nikon Instruments, Inc., Melville, NY) equipped with a Qicam Fast 1394 camera (Qimaging, Surrey, Canada) and iVision software (BioVision Technologies, Exton, PA).

Statistics
A Student's t-test was used for simple comparisons. Analysis of variance (ANOVA) and Dunnett's post-test were performed in experiments that involved multiple comparisons. GraphPad Prizm Software built in analysis tools were used for all statistical analyses. A p,0.05 was considered statistically significant.

PDX-1 represses Krt19 reporter gene activation through the NH 2 -terminus
We assessed the pKrt19-1970 reporter gene activation by cotransfection of pancreatic ductal cells (PDC) with empty vector or full length PDX-1. Krt19 reporter gene activation was reduced by 74.962.1% (p,0.05) in the presence of PDX-1 as compared to mock transfection (Fig. S1). These data are in line with previous findings [9]. To characterize the PDX-1 domains required for Krt19 repression we generated PDX-1 mutants with deletions of the NH 2 -terminus (D1-37, D1-109 and D1-144), deletions of the COOH-terminus (D260-284 and D235-284) and an internal mutation of the amino acid sequence FPWMK to AAGGQ, thereby abrogating possible interaction of PDX-1 with PBX proteins in the latter PDX-1 mutant (Fig. 1A). Proper expression of all constructs was confirmed by western blot analysis (Fig. 1B). Deletion of the NH 2 -terminus resulted in significant relief of Krt19 repression by PDX-1. Interestingly, the deletion of the very first 37  . The NH 2 -terminus of PDX-1 binds to Krt19 promoter DNA. A) MIN6 cells expressing endogenous PDX-1 were lysed and incubated with one of the following biotinylated DNA fragments: Krt19-promoter, scrambled negative control DNA or insulin promoter (Ins1, positive control), followed by precipitation with streptavidin agarose. Precipitated proteins and 2.5% input were separated with SDS-PAGE and blotted onto PVDF membranes. Membranes were probed with anti-PDX-1. PDX-1 specifically bound to Krt19 and Insulin promoter DNA, but not scrambled negative control DNA. B) Extracts of MIN6 cells were incubated with 2 mg of biotinylated Krt19 oligonucleotide and cold unbiotinylated Krt19 oligonucleotide in increasing amounts as indicated [mg]. Separation of precipitated proteins was achieved by SDS-PAGE. Specifically precipitated PDX-1 was detected using anti-PDX-1, non-specifically precipitated proteins served as a loading control and were visualized by Coomassie staining. C) MIN6 cells were transfected with V5-tagged full length PDX-1 or PDX-1 D1-37. DNA precipitation was carried out as described above, and membranes were probed with anti-V5. Deletion of the NH 2 -terminus of PDX-1 resulted in an almost complete loss of PDX-1 binding to the Krt19 promoter. D) PDX-1 negative HEK 293T cells were transfected with V5-tagged full length PDX-1 or PDX-1 D1-37. DNA precipitation was carried out as described above, and membranes were probed with anti-V5. PDX-1 D1-37 bound far less to the Krt19 promoter fragment. E) Quantification of DNA precipitation assays in HEK 293T cells transfected with V5-tagged full length PDX-1 or PDX-1 D1-37. Band intensities were normalized to input levels prior to statistical analysis. Deletion of the NH 2 -terminus resulted in a significant decrease of PDX-1 binding to the Krt19 promoter DNA. *p,0.05, Student t-test. F) PDCs were transfected with either pGal4-Krt19 -1970 or pKrt19 -1970 and pGAL-KOX-KRAB, pGAL-PDX-1 or pGAL-PDX-1 1-37. Reporter gene activity in cells amino acids is sufficient for the relief of repression, whereas deletion of the COOH-terminus or mutation of the PBX interaction site had no effect on Krt19 repression (Fig. 1B).
The NH 2 -terminus of PDX-1 is important for binding to the Krt19 promoter To confirm binding of endogenous PDX-1 to Krt19 promoter DNA, we performed DNA precipitation assays in PDX-1 expressing MIN6 cells employing a Krt19 promoter fragment that has been previously shown to be relevant for Krt19 reporter gene repression [9]. Using this DNA pulldown assay, we were able to demonstrate interaction of endogenous PDX-1 with the Krt19 promoter. Binding of PDX-1 to Krt19 promoter DNA was specific as shown by the complete absence of a signal in the scrambled duplex control (Fig. 2A). As an additional proof of specificity we performed a competition assay using a cold (unbiotinylated) Krt19 promoter fragment. Indeed, adding cold Krt19 olilgonucleotide in increasing concentrations, as indicated, decreased binding of PDX-1 to the biotin-labeled probe, confirming the specificity of the assay (Fig. 2B). Next, we addressed the question whether binding of PDX-1 to Krt19 promoter DNA is reduced upon deletion of the NH 2 -terminus. To that end, MIN6 cells were transfected with full length and PDX-1 D1-37. We found markedly decreased binding of PDX-1 D1-37 to the Krt19 promoter fragment as compared to the full length PDX-1, suggesting that interaction of PDX-1 with the Krt19 promoter requires the NH 2 -terminus of PDX-1 (Fig. 2C). To rule out possible competition of endogenous full length PDX-1 with exogenous D1-37 PDX-1 in this system, full length and D1-37 PDX-1 were transfected in HEK 293T cells, which do not express PDX-1. As expected, PDX-1 D1-37 displayed significantly reduced binding to the Krt19 promoter (69.565.6% (p,0.05, Figs. 2D, E).
To further investigate if the amino acids (AA) 1-37 of PDX-1 have the ability to serve as a portable repressor domain, we fused full length PDX-1 and AA 1-37 to the DNA binding domain of GAL4 and performed co-transfection experiments in PDCs using a luciferase based reporter assay where the reporter is driven by a thymidine kinase promoter containing a 5xGal4 motif. Using this approach, we found that both full length PDX-1 and AA 1-37 increased reporter gene activation (data not shown). These observations are in line with PDX-1 being an activator of multiple genes, i.e. insulin or somatostatin and, more importantly, the fact that the NH 2 -terminus of PDX-1 serves as a transactivating domain on these target genes [27,28]. Interestingly, gene activation through PDX-1 has been reported to require interaction of co-activators like p300 with the NH 2 -terminus of PDX-1 [29]. These data led to the question if there is a potentially sequence specific function of the PDX-1 NH 2 -terminus that mediates the regulation of Krt19 transcription. To that end, we tested whether GAL4-PDX-1 and GAL4-PDX-1 1-37 have an impact on the activities of the wt-Krt19 promoter and a mutant version that harbors a 5xGal4 motif. Surprisingly, both GAL4-PDX-1 and GAL4-PDX-1 1-37 slightly, but still significantly, repressed Gal4-Krt19 transcriptional activation (Fig. 2F), although to a far lesser extent than the established repressor GAL4-KOX-KRAB [21,30]. However, GAL4-PDX-1 and GAL4-PDX-1 1-37 were able to repress efficiently wt-Krt19 transcriptional activation (Fig. 2F), whereas GAL4-KOX-KRAB was less effective. These data suggest that targeting PDX-1 directly to Krt19 DNA impairs PDX-1's repressive function on transcriptional activation. By contrast, when PDX-1 is allowed to choose freely the site of interaction and possible interaction partners, it represses Krt19 transcriptional activation. Most importantly this is also true for the amino acids 1-37 of PDX-1. Taken together, these data support the premise that there is a DNA sequence specific repressive function of the PDX-1 NH 2 -terminus, rather than a universal portable repressor domain.

MEIS proteins bind to Krt19 promoter region in an isoform-specific fashion
We further investigated the mechanism by which PDX-1 may bind to the Krt19 promoter. Although we were unable to identify a consensus TAA(T/T)TAT sequence for the binding of PDX-1 [31] the Krt19 fragment used for precipitation assays harbors a MEIS binding site (Fig. 3A). This raised the question of whether Krt19 expressing PDCs show different levels of MEIS proteins than Krt19 negative PDX-1 positive endocrine MIN6 cells, and also whether there is differential expression of Meis genes. We first assessed mRNA levels of all three Meis genes in PDCs and MIN6 cells (Fig. 3B). We found significant differences in the mRNA levels of Meis1, having a significantly higher expression in PDCs, and Meis3, displaying a significantly higher expression in MIN6. Meis2 mRNA is equally expressed in both cell types. Second, we evaluated expression of MEIS proteins by western blot (Fig. 3C). PDX-1 negative PDCs show higher expression of MEIS proteins, whereas PDX-1 positive MIN6 cells have overall lower levels of MEIS expression. Additionally, both cell lines express different MEIS proteins in line with the mRNA levels.
Due to the high sequence homology between MEIS proteins, similar molecular masses and different splice variants, it is difficult to precisely conclude which MEIS protein is detected by the antibody that recognizes all three isoforms. Therefore, we generated FLAG-tagged isoforms MEIS1a, MEIS2b and MEIS3 and transfected HEK 293T cells with either empty vector or plasmids coding for each of the MEIS proteins. Immunoprecipitations were performed and eluted complexes were incubated with radioactively labeled Krt19 promoter fragment as described in the methods section. In this experiment, binding of MEIS1a and to a lesser extent MEIS2b, but not MEIS3, to the Krt19 promoter fragment was revealed (Fig. 3D). To validate this approach, a supershift with anti-MEIS1/2/3 was performed. Co-incubation of the FLAG-immunoprecipitates with anti-MEIS1/2/3 resulted in a complete shift of MEIS1a and MEIS2b, whereas co-incubation with non-immunogenic control IgG did not alter the mobility of the complexes generated by MEIS1a and MEIS2b. The specificity of the IP was demonstrated by the absence of a signal when the IP was performed with IgG. To rule out possible variations caused by different efficiencies of the IP, a fraction of the IP was separated by SDS-PAGE and the membrane was probed with anti-FLAG antibody (Fig. 3E). The result demonstrated equal efficiencies of transfection and IP, suggesting specific DNA binding of MEIS1a and MEIS2b, but not MEIS3 (Fig. 3D). Noteworthy, if EMSA was performed directly with lysates of transfected cells, a shift could only be observed for MEIS1a, but not MEIS2b, suggesting that MEIS1a is the major MEIS protein that binds to the Krt19 promoter. (Fig. S3).

PDX-1 causes posttranscriptional down-regulation of MEIS1a
Since homeobox proteins are capable of interacting with members of the TALE homeodomain protein family, we evaluated for possible protein-protein interactions between PDX-1 and MEIS1a by co-transfection of HEK 293T cells. Surprisingly, cotransfection resulted in absence of exogenous MEIS1a (Fig. 4A). For visualization of expression kinetics PDX-1 was expressed from a vector that co-expresses EGFP (Fig. 4B). Expression was observed as early as 4 hrs. post transfection and peaked at 24 hrs. post transfection. To rule out possible promoter competition between expression vectors we evaluated PDX-1 expression and endogenous MEIS1 expression at the indicated time points. Indeed, endogenous MEIS1 was down-regulated with increasing amounts of PDX-1 protein levels (Fig. 4C). To further elucidate the nature of MEIS1 down-regulation, Meis1 mRNA levels were measured by quantitative RT-PCR 36 hours post transfection. Meis1 mRNA levels were not changed in the presence of PDX-1, suggesting a possible posttranscriptional mechanism for MEIS1 regulation by PDX-1 (Fig. 4D). One possible mechanism in this setting would be ubiquitination with proteasomal degradation of MEIS1. To test this hypothesis, HEK 293T cells were cotransfected with PDX-1 and MEIS1a and then treated with the proteasome inhibitors MG-132 or Bortezomib. As before, cotransfection of PDX-1 and MEIS1a led to down-regulation of MEIS1a. However, this could be prevented by treatment with MG-132 (Fig. 4E, upper panel) or Bortezomib (Fig. 4E, lower  panel). Stabilization of MEIS1a occurred in a time dependent fashion, and efficiency of drug treatment was proven by stabilization of the proteasomal regulated cell cycle protein p21.
To test the possibility that MEIS1a is degraded through ubiquitination and subsequently the proteasome, we evaluated MEIS1a for increased ubiquitination in the presence of PDX-1 with or without inhibition of the proteasome. Surprisingly, we detected increased laddering for MEIS1a in the absence and the presence of PDX-1, although to a lesser extent when PDX-1 was present (Fig. S4A). We then performed co-immunoprecipitation experiments with PDX-1, HA-tagged ubiquitin and FLAG-tagged MEIS1a in order to validate these findings. Indeed, in the presence of a proteasome inhibitor, HA-ubiquitin and FLAG-MEIS1a interact in the absence of PDX-1. By contrast, in the presence of PDX-1, an interaction could not be observed (Fig. S4B). Taken together, these data suggest a mechanism where MEIS1a is degraded by the proteasome in the presence of PDX-1, and this may happen independently of ubiquitination. Interestingly, this increased turnover of MEIS1a by PDX-1 does not appear to be dependent upon the NH 2 -terminus of PDX-1 (Fig. S4C), thereby indicating functional differences of PDX-1 with regards to binding to the Krt19 promoter and the down-regulation of MEIS1a.
Meis1, but not Meis2, is critical for Krt19 transcriptional activity in vitro and expressed in pancreatic ductal cells in vivo.
Finally, we wanted to know whether Meis1 or Meis2 are functionally relevant for Krt19 transcrtiptional regulation. To address this question, Meis1 and Meis2 were genetically silenced in PDCs using two different siRNAs against Meis1 or Meis2 and RNA was extracted 72 h post transfection. Knockdown of Meis1 and Meis2 was efficient and similar for each siRNA used (Fig. 5A and  S5A). Importantly, depletion of Meis1 (Fig. 5B), but not Meis2 (Fig.  S5B), resulted in significantly reduced Krt19 mRNA levels (56.1163.74% and 65.77611.82%, respectively) indicating that Meis1 is directly involved in the transcriptional control of Krt19 in pancreatic ductal cells. To evaluate this in vivo, we performed immunoflourescence staining on mouse pancreas for KRT19, PDX-1 and MEIS1 using a MEIS1-specific antibody. As expected, MEIS1 was expressed in pancreatic ducts together with KRT19, but not PDX-1 (Fig. 5C). Although there appeared to be some overlapping of KRT19 and MEIS1 (Fig. 5C), KRT19 staining was found in the cytoplasm, whereas MEIS localized predominantly in the nuclei of pancreatic ducts (Fig. 5D).
Taken together, our data indicate two mechanisms by which Krt19 is regulated by PDX-1. The first one involves DNA binding of PDX-1 to the Krt19 promoter that requires its NH 2 -terminus, and the second one acts through down-regulation of MEIS1, which is necessary for the maintenance of Krt19 mRNA levels.

Discussion
The goals of this study were to identify domains of PDX-1 necessary for mediation of Krt19 transcriptional repression and to elucidate the functional interplay of PDX-1 with Hox-cofactors, namely MEIS. We created several NH 2 terminal and COOH terminal deletions as well as an internal mutation of PDX-1. We found that the NH 2 -terminus, but not the COOH terminus or the pentapeptide motif, is critical in mediating Krt19 transcriptional repression. The homeodomain, which was kept intact, harbors the DNA binding site of PDX-1 and also the nuclear translocation signal. Indeed, altered subcellular localization of PDX-1 does not occur upon NH 2 terminal deletion (Fig. S2) and cannot account for the lack of repression of Krt19 by these mutants. In particular, we demonstrated diminished binding of the PDX-1 NH 2 terminal deletion to the promoter DNA of Krt19. Our data obtained with the GAL4 vector transfection studies support the premise that there is a DNA sequence specific repressive function of the PDX-1 NH 2 -terminus, rather than this region of PDX-1 serving as a universal portable repressor domain. Since direct binding of DNA by PDX-1 occurs via the homeodomain, it cannot be excluded that DNA binding mediated by the NH 2 -terminus is of an indirect nature, possibly through protein-protein interactions. This notion is supported by the fact that there is no canonical binding site for PDX-1 within the Krt19 promoter.
Hox type transcription factors are known to bind to DNA either alone or in association with co-factors. This can happen as either heterodimeric or heterotrimeric complexes together with members of the TALE family, such as PBX and MEIS [32,33]. PDX-1 has been reported to interact with PBX proteins via the FPWMK pentapeptide motif, and PBX in turn can recruit co-factors, such as histone deacetylases (HDACs), in order to mediate transcriptional repression [34].
We were able to exclude the involvement of PBX proteins in the repression of Krt19 by PDX-1 through mutation of the PBX interaction motif as described previously [18]. Furthermore, Cells were lysed and immunoprecipitation using anti-FLAG was carried out. Precipitates were used for EMSA. IP against FLAG showed strong occupancy on the Krt19 promoter fragment for MEIS1 and to a lesser extent for MEIS2, but not for MEIS3. Pre-incubation with anti-MEIS1/2/3 resulted in a shift for MEIS1 and MEIS2. Control IP with IgG did not yield a band. E) 20 ml of the input used for EMSA were separated by SDS-PAGE and transferred to PVDF membrane, followed by incubation with anti-FLAG. Western blot analysis showed equal amount of input and efficient pulldown for each FLAG-tagged MEIS protein. doi:10.1371/journal.pone.0012311.g003 . This is consistent with a previous report that the PDX-1 COOH-terminus is capable of recruiting HDACs to regulate gene transcription [35].
Since PBX proteins do not participate in the repression of Krt19 by PDX-1 in a heterotrimeric complex consisting of PBX-PDX-1-MEIS, we explored the possibility of a MEIS-PDX-1 heterodimeric complex. We identified a MEIS binding site in the Krt19 promoter and differential expression patterns of Meis genes with transfection by EGFP. Brightfield and matching fluorescence images are shown at 2006 magnification. C) HEK 293T cells were transfected with pIRES2-EGFP-PDX-1 and lysates were harvested at the indicated time points. Western blot analysis revealed down-regulation of endogenous MEIS1 in a PDX-1 dependent manner. D) RNA was isolated 36 hrs. post transfection from HEK 293T cells that were transfected with PDX-1. Quantitative Real-Time PCR demonstrated stable mRNA levels of endogenous MEIS1. E) HEK 293T cells were transfected with MEIS1 and either empty vector control or PDX-1 and treated with DMSO or MG-132 or Bortezomib, respectively. Cells were lysed at indicated time points and subjected to western blot analysis. Both, MG-132 and Bortezomib were able to stabilize MEIS1 protein levels in the presence of PDX-1. p21 was unaffected by PDX-1 and served as a positive control for the functionality of MG-132 and Bortezomib, respectively. doi:10.1371/journal.pone.0012311.g004 Meis1 and Meis2 being expressed in PDCs and Meis2 and Meis3 expression in MIN6 cells. We confirmed MEIS1 and MEIS2 as the major family members to bind the Krt19 promoter. Interestingly, MEIS3, which displays only very low mRNA levels in PDCs, does not bind to the Krt19 promoter. Meis1 and Meis2 might cooperate to regulate vertebrate retina development through the maintenance of retinal progenitor cells [36]. That Meis1 and Meis2 may be more important for the exocrine ductal lineage is reinforced by the finding that the MEIS family members PREP1 and PREP2 form much stronger binding complexes with PBX1 on the Pax6 pancreatic enhancer during islet cell development than do MEIS1 and MEIS2 [37].
To further understand the possible interaction between PDX-1 and MEIS1, we performed co-transfection experiments and found down-regulation of MEIS1 protein in the presence of PDX-1 while Meis1 mRNA remained at a constant level. Additionally, we confirmed that the level of MEIS1 protein is inversely correlated with PDX-1 expression levels and MEIS1 levels are dependent directly on PDX-1 expression. For these reasons, we conclude that MEIS1 down-regulation occurs on a posttranscriptional or basis. One possible mechanism of how protein levels are regulated posttranscriptionally is by increased turnover and degradation by the proteasome. Indeed, MEIS1 protein could be stabilized in the presence of PDX-1 through inhibition of the proteasome. However, we were unable to demonstrate substantially increased ubiquitination of MEIS1 in the presence of PDX-1. One simple explanation might be that the overall level of MEIS1 is reduced when PDX-1 is present, which makes the detection of ubiquitinated MEIS1 difficult. Another possibility that needs to be explored further is that the degradation of MEIS1 through the proteasome is independent of ubquitination, an exciting mechanism that has become increasingly recognized [38] and has been described for important and wellstudied transcription factors, such as c-Fos, p53 and p73 [39,40,41]. Taken together, our data suggest different ways of how PDX-1 mediates the transcription of Krt19. First, we provide evidence for direct repression through binding to the Krt19 promoter, and second, PDX-1 fosters proteasomal degradation of MEIS1. Third, we have demonstrated that MEIS1 is a regulatory transcription factor that is necessary for Krt19 transcriptional activation in PDCs. Since Meis2 knockdown does not affect Krt19 mRNA, we believe Meis2 does not have a predominant role when compared to Meis1.
The regulation of MEIS1 is poorly understood, although there is evidence for transcriptional regulation by Creb [42]. Here, we show for the first time that MEIS1 appears to be posttranscriptionally regulated by a mechanism that requires PDX-1. One scenario might be the activation of an as yet unidentified MEIS1 interacting factor by PDX-1. We have shown previously that MEIS1 activates the Krt19 promoter, and importantly, that this property could not overcome PDX-1 mediated repression of Krt19 [9]. We now conclude that the posttranscriptional downregulation of MEIS1, both endogenous and exogenous, is responsible for this effect.
A possible mechanism for up-regulation of MEIS target genes is suggested by the recent finding that MEIS proteins increase the histone H4 acetylation status on certain promoters during zebrafish embryogenesis [43]. In this context, MEIS acts by regulating the access of HDAC and CBP to such promoters. Consequently, loss of MEIS1 would result in decreased transcriptional activity of critical target genes. For example, zebrafish embryos lacking Meis1 have a marked reduction in gata1 expression, which is critical for erythroid cell fate, and lack circulating blood cells [44]. This can be observed in the direct regulation of the developmentally important SOX3 by MEIS1 and PBX [45]. Given the overall importance of PDX-1 during development and differentiation of the pancreas, it is tempting to speculate that continued expression of PDX-1 would lead to increased turnover of MEIS1 and prevent expression of Krt19 in the endocrine lineage. Down-regulation of Krt19 transcription may then be further secured by direct interference of PDX-1 with the Krt19 promoter in the ductal lineage; indeed, PDX-1 is down-regulated or absent in differentiated ductal cells. Loss of PDX-1 could lead to stabilization of MEIS1, relief of repression at the Krt19 promoter and subsequent ductal specific Krt19 gene expression (Fig. 6). This model would complement the observation that regulated loss of PDX-1 at mid-pancreatic development may still lead to the formation of a truncated ductal tree, but not of acinar tissue [46].

Supporting Information
Table S1 Sequences of primers that have been used for the generation of PDX-1 mutants and mutation of pKrt19-1970. Primer sequences were selected to generate the indicated truncation or mutation of PDX-1 of for fusion of PDX-1 and  Table S2 Primers for quantitative RT-PCR. Primers for quantitative RT-PCR were designed for the specific mRNA as indicated. Sequences were selected to be intron-spanning to avoid amplification of genomic DNA. and protein transfer to PVDF membranes, visualization was perfomed by HRP-coupled secondary antibodies. Long exposure (3 minutes) revealed markedly increased laddering of MEIS1 in mock transfected samples and slightly increased laddering in the presence of PDX-1 when samples were treated with proteasome inhibitors. p21 served as a control for the efficiency of the drugs. B) HEK 293T cells were transfected with equal amounts of pCMV-HA-ubiquitin, PDX-1 and MEIS1a-FLAG. IP revealed interaction of ubiqutin and MEIS1 in absence, but not in presence of PDX-1, when samples were treated with the proteasome inhibtor Bortezomib (1 mM). Notably, MEIS1 levels were reduced in samples co-transfected with PDX-1, either with or without inhibition of the proteasome as revealed by short exposure and input, respectively. C) The down-regulation of MEIS1 by PDX-1 occurs independently of the C-terminus of PDX-1.