Enterocyte-specific Regulation of the Apical Nutrient Transporter SLC6A19 (B0AT1) by Transcriptional and Epigenetic Networks*

Background: Intestinal epithelial cells express genes for nutrient absorption. Results: Expression of the nutrient transporters is suppressed in the crypt and promoted in the villus by opposing transcription factors, changes of DNA methylation, and histone modification. Conclusion: Gene expression in enterocytes is regulated by three different mechanisms. Significance: Removal of repressive transcriptional control elements is required to express enterocyte specific genes in the villus. Enterocytes are specialized to absorb nutrients from the lumen of the small intestine by expressing a select set of genes to maximize the uptake of nutrients. They develop from stem cells in the crypt and differentiate into mature enterocytes while moving along the crypt-villus axis. Using the Slc6a19 gene as an example, encoding the neutral amino acid transporter B0AT1, we studied regulation of the gene by transcription factors and epigenetic factors in the intestine. To investigate this question, we used a fractionation method to separate mature enterocytes from crypt cells and analyzed gene expression. Transcription factors HNF1a and HNF4a activate transcription of the Slc6a19 gene in villus enterocytes, whereas high levels of SOX9 repress expression in the crypts. CpG dinucleotides in the proximal promoter were highly methylated in the crypt and fully de-methylated in the villus. Furthermore, histone modification H3K27Ac, indicating an active promoter, was prevalent in villus cells but barely detectable in crypt cells. The results suggest that Slc6a19 expression in the intestine is regulated at three different levels involving promoter methylation, histone modification, and opposing transcription factors.

The small intestine is a highly specialized organ for nutrient digestion and absorption. Its anatomical structure is adapted to these tasks, including a highly differentiated crypt-villus structure forming a large surface to maximize uptake of digested nutrients. All epithelial cells originate from stem cells within the crypt, resulting in differentiated cell types for mucus production (goblet cells), hormone production (enteroendocrine cells), host defense (Paneth cells), and absorption (enterocytes) (1). Enterocytes make up Ͼ80% of all intestinal epithelial cells. They are highly polarized, possessing a brush border structure at their apical side, across which nutrients are absorbed (1). Therefore, the differentiation of crypt stem cells into fully differentiated absorptive enterocytes located in the villi of the small intestine is an excellent model for studying the transcriptional control mechanisms of protein sets that are required for nutrient absorption.
At the enterocyte brush border of the small intestine, peptides resulting from protein digestion by pancreatic proteases are further hydrolyzed into individual amino acids as well as diand tripeptides (2). The majority of neutral amino acids are taken up through an active transport process involving the neutral amino acid transporter B 0 AT1 (encoded by the Slc6a19 gene; throughout the article, we will use Slc6a19 when referring to the gene and B 0 AT1 to identify the protein). The B 0 AT1 transporter is a Na ϩ amino acid symporter and is located in the apical membrane of the absorptive enterocytes (3). Apart from being expressed in the small intestine, it is also detected in the apical membrane of proximal kidney tubule cells, and to a smaller degree in the pancreas (4) and the skin, but is not detected in any other tissue. B 0 AT1 requires auxiliary proteins for expression at the cell surface, namely collectrin (TMEM27) in the kidney (5,6), and ACE2 (angiotensin-converting enzyme 2) in the small intestine (7). For enterocytes to transport neutral amino acids, it is thus necessary to express both ACE2 and B 0 AT1 simultaneously. Due to its activity as a carboxypeptidase, ACE2 contributes to peptide digestion and feeds, together with aminopeptidase N, amino acids to the B 0 AT1 transporter (8).
The hepatocyte nuclear factors HNF1a, HNF1b, and HNF4a are part of an autoregulatory transcriptional network in mammalian pancreas, liver, kidney, and gut (9). HNF1a and HNF1b are homeobox-containing transcriptional activators. HNF1a was first identified as a tissue-specific regulator of liver function (10) but has been shown to be expressed across a small number of tissues, including kidney, pancreas, and intestine. Mutations in the DNA-binding domain of HNF1a are the cause of type 3 maturity onset diabetes of the young, which is characterized by progressive loss of insulin secretory capacity (11). HNF1a is assumed to act as a master regulator for the transcription of transporters in the kidney because HNF1a deficient mice manifest a general defect in the renal reabsorption of metabolites (12). Additionally, most of the renal amino acid transporters in murines contain HNF1a-binding motifs within their proximal promoter regions (13). Although HNF1a is expressed in three different isoforms in humans, rodents only express a single isoform, most likely resulting in species-specific regulatory differences (9). The absence of epithelial amino acid transporter expression in the liver despite HNF1a expression has been suggested to be due to DNA methylation, thus preventing the interaction of HNF1a with the transport promoter binding sites (13). However, the observed DNA methylation was at a significant distance from the proximal promoter.
Genome-wide expression profiling carried out with liver tissue of HNF1a-deficient mice showed that HNF1a is involved in the regulation of several important hepatic functions such as bile acid and cholesterol metabolism (14). A comparative study with liver and pancreatic island tissue revealed that HNF1a-dependent genes encode a broad range of metabolic functions in both tissues, including those for amino acid transport and steroid, lipid, and xenobiotic metabolism (15). Pancreatic tissue of HNF1a-deficient mice exhibits down-regulation of the non-epithelial amino acid transporter SLC38A4, ACE2, collectrin, and the nuclear factor HNF4a (15,16). In liver, HNF4a appears to control expression of Hnf1a, whereas the opposite is observed in pancreas due to the use of an alternate Hnf4a promoter, which is controlled by HNF1a (16). Both promoters are used in the duodenum. HNF1b is a paralogue of HNF1a with indistinguishable in vitro DNA-binding properties (17). HNF1b is highly expressed in pancreatic islets, yet only expressed at low levels in adult liver (15). Interestingly, HNF1b expression has been shown to be increased in HNF1a Ϫ/Ϫ liver tissue (15). Mutations in HNF1b are the principal cause for the MODY5 phenotype, exhibiting atrophy in the pancreas as well as several forms of renal disease (18).
HNF4a is a highly conserved, zinc finger-containing transcriptional activator of the steroid hormone nuclear receptor superfamily. In contrast to most nuclear receptors however, HNF4a does not appear to be activated by exogenous ligands (19). HNF4a is expressed at high levels in liver, kidney, small intestine, colon, and pancreatic ␤ cells. Loss-of-function mutations in HNF4a result in MODY1, a monogenic autosomal dominant non-insulin-dependent diabetes mellitus type II phenotype, exhibiting a decrease in glucose-induced insulin release (20). As mentioned above, HNF4a controls the transcription of the Hnf1a gene in liver (16), thus indirectly regulating a vast array of genes vital for hepatocyte function (21). It was reported previously that HNF1a can down-regulate HNF4a-mediated activation of transcription via a direct interaction of these transcription factors (22). In contrast, others observed that HNF4a enhances the transcriptional activity of HNF1a independent of its DNA-binding activity, possibly through recruitment of other co-activating regulators (23). HNF4a appears to play an important role early in embryonic development, as indicated by its presence as a maternal component of the egg and its expression in mouse embryonic stem cells (24). During development, HNF4a expression precedes the expression of HNF1a and HNF1b, implying that HNF4a is a transcription factor higher up in the hierarchy of regulatory proteins, thus establishing the importance of HNF4a for proper terminal differentiation of hepatocytes (24).
HNF1a and HNF1b appear to play important roles in terminal differentiation and cell fate commitment in the intestine, particularly for the Paneth cell lineage. HNF1b is expressed throughout development from the primitive gut stage through organogenesis, although HNF1a is expressed later during organogenesis (25). HNF4a has been suggested to be a major transcriptional regulator in the intestine. Genome-wide ChIPchip analysis in the human Caco-2 cell line indicates that HNF4a is associated with regulatory regions of Ͼ1000 genes (26). Although these included some genes involved in digestion and absorption, many key proteins were missing, such as the glucose transporter SGLT1, the cationic amino acid transporter rbat/b 0,ϩ AT, sucrose/isomaltase, B 0 AT1, ACE2 etc. The analysis revealed transcription factor networks, in which HNF4a regulates the expression of CDX2. A model has been proposed in which HNF4a initiates the differentiation-dependent transcription by activating Hnf1a and Cdx2, both encoding transcription factors, which are required simultaneously for the expression of many genes specifically expressed in the intestine (26,27).
To shed light onto the regulatory mechanisms responsible for the co-expression of ACE2 and B 0 AT1 in enterocytes of the small intestine, we have enriched crypt stem cells and absorptive villi enterocytes using a modified Traber method (28) and subjected them to a genome-wide expression analysis, combined with a bioinformatics approach to identify transcription factor candidates for the regulation of ACE2 and B 0 AT1 expression. The involvement of candidate genes was then investigated utilizing both in vitro and in vivo methods, namely reporter gene assays and ChIP.

EXPERIMENTAL PROCEDURES
Molecular Cloning and DNA Manipulation-All primer pairs used in this study are listed in supplemental Table 1 and were purchased from Sigma-Aldrich. The Slc6a19 promoter region was amplified by nested PCR using genomic DNA from mouse kidney. Initially, a fragment ranging from positions Ϫ2639 to position ϩ374 was amplified by standard PCR using Pfu polymerase. This fragment was used as a template for a second PCR, which used primers with an incorporated KpnI and BglII site spanning from positions Ϫ2494 to ϩ57. The product of this reaction was cut with KpnI and BglII and inserted into the reporter gene vector pGL4.12 (Promega). The Ace2 promoter was cloned using a similar strategy. Initially, an 1856-bp fragment was amplified from genomic DNA (positions Ϫ1602 to ϩ254). This fragment was used as a template for a second PCR, which used primers with an incorporated XhoI and KpnI site spanning from positions Ϫ1509 to ϩ170. The product of this reaction was cut with XhoI and KpnI and inserted into the reporter gene vector pGL4.12 (Promega). Promoter deletions were made using the full promoter construct in pGL4.12 by inverse PCR. Reverse and forward primers flanking the desired deletion were used to amplify the complete vector minus the deletion and the resulting fragment was religated. To clone transcription factors, mRNA was isolated from mouse intestine (RNeasy plus, Qiagen) and converted into cDNA using Superscript II reverse transcriptase (Invitrogen). Sox9 cDNA was cloned by nested PCR. Initially, a larger fragment was amplified, followed by amplification of the coding sequence with primers containing a BamHI (sense) and an EcoRI (antisense) site. The fragment was digested with both enzymes and ligated into pcDNA3.1(ϩ) (Invitrogen). Hnf1a cDNA was purchased from Thermo Scientific (clone ID 30471380), the cDNA being inserted into pCMV-SPORT. Hnf4a cDNA was amplified from intestinal cDNA using primers incorporating BamHI and EcoRI cutting sites. After digestion, the cDNA was ligated into the corresponding cites of pcDNA3.1(ϩ).
Putative transcription factor binding sites in the reportergene constructs were verified by site-directed mutagenesis using the QuikChange II site-directed mutagenesis kit (Agilent). Primers for the mutagenesis were designed using the manufacturer's recommendations. All constructs were confirmed by sequencing.
In Situ Hybridization-Tissue specimens of small intestine from C57BL/6, SWR/J, and DBA1/J mice were fixed in 4% paraformaldehyde/0.1 M sodium phosphate buffer (pH 7.2) for 4 h and embedded in paraffin. Five m tissue sections were dewaxed and hybridized as described previously (29). Hybridization probes were generated by in vitro transcription of mB 0 AT1 cDNA cloned into pCR-blunt II-TOPO using SP6 polymerase (antisense) and T7 polymerase (sense). The hybridization mixture (10 mM Tris-HCl, pH 7.4, 50% (v/v) deionized formamide, 600 mM NaCl, 1 mM EDTA, 0.02% polyvinylpyrrolidone, 0.02% Ficoll, 0.05% bovine serum albumin, 10% dextrane sulfate, 10 mM dithiothreitol, 200 g/ml denatured sonicated salmon sperm DNA, and 100 g/ml rabbit liver tRNA) contained either the 35 S-labeled RNA antisense or sense control mB 0 AT1 probe at a concentration of 500 ng/ml. Hybridization with RNA probes proceeded at 42°C for 18 h. Following washing steps, the slide preparations were dipped in NTB2 emulsion (Kodak) and exposed at 4°C for 3 weeks. After development, the slides were stained with hematoxylin/eosin and photographed with a Sony DSC digital camera.
Cell Fractionation of Mouse Intestine-A fractionation method was used to isolate cell populations enriched with crypt and villus cells. This method is based on the inverted intestinal sac method first described by Barnard et al. (30) and further optimized by Traber et al. (28). In brief, the small intestine was cut after the duodenum and before the ileocaecal boundary and flushed with PBS containing 1 mM DTT. Subsequently, 6-cmlong sections of mouse jejunum were inverted onto a plastic rod. The plastic rod was inserted into a hole in the lid of a 12-ml plastic round-bottomed centrifuge tube, containing 10 ml of citrate buffer (96 mM NaCl, 1.5 mM KCl, 27 mM trisodium citrate, 8 mM KH 2 PO 4 and 5.6 mM Na 2 HPO4, pH 7.4). After sealing, the assembly was incubated on a rotatory shaker at 37°C for 10 min. Subsequently, the citrate buffer was changed to 10 ml of PBS containing 0.5 mM DTT, 1.5 mM EDTA, and 1 mg/ml bovine serum albumin at 37°C. Rotary shaking was continued for 10 -20 min until the solution was opaque, indicating detachment of cells. The solution was exchanged against warm (37°C) fresh solution until five to seven fractions were collected. Cell viability was judged by trypan blue exclusion.
RNA Isolation and Microarray Studies-RNA was isolated from intestinal cell fractions using RNeasy (Qiagen). RNA quality was measured on a bioanalyser chip. Gene expression was analyzed using an Agilent SurePrint G3 Mouse Exon Microarray through the Ramaciotti Center at the University of New South Wales. Expression levels of selected mRNAs were confirmed by quantitative PCR (qPCR). 3 qPCR-Quantitative PCR was conducted using Fast SYBR Green Master Mix (Applied Biosystems) with the PCR being performed on an Applied Biosystems 7900 instrument. The reaction mixture (10 l) included 5 l of SYBR Green master mix, 3 l of sample cDNA (1:10 dilution) and 1 l of each primer (concentration of 200 nM). Subsequently, the mixture was pipetted into a 394-well plate. The PCR parameters were 50°C for 2 min, 95°C for 15 s, and 60°C for 1 min, running a 3 The abbreviation used is: qPCR, quantitative PCR. total of 40 cycles. All samples were analyzed in duplicate, and experiments were repeated three times. GAPDH was used as an internal control. Primers were designed using Primer3 program to amplify fragments of 170 -250 bp.
Bioinformatic Promoter Analysis-The Slc6a19 promoter region ranging from Ϫ2494 to ϩ57 was analyzed using the University of California (Santa Cruz) genome browser. Bioinformatic analysis revealed that a region located between Ϫ228 ϩ57 bp of the transcriptional start site was highly conserved among different species (data not shown). A region of ϳ2.5 kb upstream of the transcriptional start site was analyzed using the Genomatix MatInspector (genomatix) program. Relevant transcription factors in the Slc6a19 promoter region are listed in Table 1.
Chromatin Immunoprecipitation Assay-ChIP was performed as per the instructions of the SimpleChIP enzymatic chromatin IP kit (Cell Signaling). Briefly, for each immunoprecipitation, 25 mg of fresh or snap-frozen mouse tissue was minced with a scalpel and placed in 1 ml of PBS containing protease inhibitors. Proteins and DNA were then cross-linked with 1.5% formaldehyde (prepared fresh) for 20 min at room temperature. The reaction was stopped by the addition of a 1/10 of the volume of glycine stop solution (10ϫ). After a 5-min incubation, the tissue was centrifuged for 5 min at 1500 rpm at 4°C and then washed once with 1 ml of PBS containing protease inhibitors. After disaggregating the tissue with a potter homogenizer using 20 -25 strokes, the cell suspension was centrifuged for 5 min at 4°C at 1500 rpm. The supernatant was immediately removed from the cells, and nuclei were prepared as per the instructions of the kit. Chromatin was digested with micrococcal nuclease provided with the kit. This was followed by sonication of the sample twice, for 2 min each, at 80% amplitude with a 15-s pulse on and 15-s pulse off time using a Misonix S4000 sonicator. Lysates were clarified by centrifugation at 10,000 rpm for 10 min at 4°C. The supernatant containing the cross-linked chromatin preparation was incubated overnight in 1-5 g of the specific antibody. Samples were purified according to the instructions of the kit and analyzed using qPCR. Antibodies were purchased as follows: H3K27ac (ab4729), H3K4me3 (ab8580), SOX9 (ab3697), and HNF4a (ab41898) from Abcam; IgG (2729) and H3 (4620) from Cell Signaling; and HNF1a (sc-6547X) from Santa Cruz Biotechnology.
DNA Methylation/Bisulfite Sequencing-Intestinal villus and crypt enterocytes were isolated using the fractionation protocol described above; other tissues were used after dissection. Following tissue extraction, genomic DNA was isolated using a Transcription factors (TFs) with robust expression in either crypt or villus were selected. GATA4 is listed because it has previously been indicated in enterocyte maturation but does not have a predicted binding site. A matrix number of 1.00 indicates an exact match of the binding site consensus sequence. Revised location indicates results from mutagenesis data. A qualitative indicator is provided for the effect on transcription. The ratio of mRNA abundance along the villus-crypt axis was determined by microarray analysis.

RESULTS
At the Protein and mRNA Level, B 0 AT1 Is Highly Expressed in the Villi, but Absent in the Crypt Compartment-In confirmation of previous results (7), B 0 AT1 protein expression in the intestine shows a distinct gradient from the villus tip to the crypt (Fig. 1A). Expression is absent in the crypt and maximal in mature enterocytes at the villus tip. In situ hybridization confirms that this gradient is also observed at the level of mRNA (Fig. 1B). This is a typical feature of proteins involved in nutrient absorption and digestion, including angiotensin-converting enzyme 2 (data not shown), which is required for expression of B 0 AT1 at the cell surface. To understand the mechanisms of Slc6a19 transcriptional control, we analyzed its promoter region. The proximal promoter of the mouse Slc6a19 gene, as judged by conservation of the sequence across mammalian species, appeared to be relatively short, i.e. 150-bp-long. It is a typical class II promoter, where the transcriptional start site is preceded by a TATA-binding motif (31) at position Ϫ23 to Ϫ28 and a reasonably conserved transcription factor IIB recognition motif (Ϫ39 to Ϫ46), indicating the binding site of general tran- scription factors and of polymerase II (31). The transcriptional start site CCACTT is similar to the mammalian consensus initiator sequence (31). Relevant transcription factor binding sites and their position are listed in Table 1.
To investigate the tissue and cell specific expression of B 0 AT1, we generated reporter gene constructs using a 2-kb fragment to account for possible species-specific regulatory elements that might be located outside the proximal promoter. When transfected into several different cell lines we could not detect significant reporter gene activity, suggesting that the promoter does not use common transcription factors found in a variety of cell lines. HNF1a and HNF4a (26) have been implicated as central players in the differentiation-induced expression of epithelial cells. When transfected into HEK293 cells, both transcription factors were sufficient to initiate transcription from the Slc6a19 promoter construct, with HNF4a being a stronger activator than HNF1a (Fig. 2, A and B). Combined transfection of HNF1a and HNF4a did not increase reporter gene activity further (data not shown). The Ace2 promoter, by contrast, could be activated in vitro by HNF1a only (Fig. 2C). To identify the DNA binding sites for HNF1a and HNF4a, we generated promoter truncations and mutated predicted binding sites (Fig. 2). We identified two potential binding sites for HNF1a in the Slc6a19 promoter located within 150 bp of the transcriptional start site (Ϫ126/Ϫ110 and Ϫ153/Ϫ137). Sitedirected mutagenesis was performed, indicating that the more distal site was not used, whereas an active binding site was identified at position Ϫ111/Ϫ114 ( Fig. 2A). Mutation of this binding site abolished the observed transcriptional activation by HNF1a. A more distal site at position Ϫ1205/Ϫ1189 did not contribute to activation of the promoter (Fig. 2A). Using the same constructs, it appeared that HNF4a had two functional binding sites, one located between Ϫ972 and Ϫ1378 and another within 136 bp of the transcriptional start site (Fig. 2B). Mutation of the predicted HNF4a binding site (Ϫ38/Ϫ41) that is proximal to the transcriptional start site abolished all reporter gene activity, indicating that this site is essential for recruitment of general transcription factors. Mutation of the predicted distal site (Ϫ1368/Ϫ1344) did not change the response of the reporter construct to co-transfection of HNF4a (Fig. 2B). Further sequence analysis and mutagenesis of potential binding sites at position Ϫ1108/Ϫ1105 and Ϫ1088/Ϫ1085 did not reveal an obvious HNF4a binding site between Ϫ972 to Ϫ1387. Thus, the reason for the significant drop of HNF4a-dependent reporter gene response after truncation of this region remains unclear. These results nevertheless demonstrate that HNF1a and HNF4a are both necessary and sufficient to initiate B 0 AT1 transcription in the background of HEK293 cells.

Expression of Transcription Factors in Crypt and Villus-enriched Mouse Small Intestine Cell
Fractions-Given the strong activation of B 0 AT1 transcription by both HNF4a and HNF1a, a feasible mechanism to establish the B 0 AT1 expression pattern along the villus-crypt axis would be by way of a villus-specific expression of these key regulators. To test this hypothesis, we used a fractionation method to isolate cell populations along the villus-crypt axis. The method showed strong enrichment of Slc6a19 mRNA at the villus tip fraction (ratio villus tip/crypt ϭ ϳ20-fold, Fig. 3A) and of the well established intestinal stem cell marker Lgr5 (leucine-rich repeat containing G-protein coupled receptor 5) in the crypt-enriched fraction (ratio crypt/villus tip ϳ 100-fold, Fig. 3B). This provided the basis for a genome-wide transcriptome comparison by microarray analysis. We analyzed the crypt/villus ratio of all transcription factors that had significant expression in either crypts or villus tip. (Table 1 lists those that have putative binding sites in the Slc6a19 promoter.) Surprisingly, we found that Hnf1a and Hnf4a mRNAs were expressed in both villus and crypts at similar levels. Hnf4a mRNA was more abundant than Hnf1a mRNA. Expression levels of both mRNAs were confirmed by qPCR (data not shown). Moreover, CDX2 and GATA4, which previously had been implicated in enterocyte-specific expression (26,27,32), were either enriched in the crypt or expressed  (Table 1). Thus, expression of Slc6a19 mRNA in villus cells cannot be explained by a corresponding expression of Hnf1a, Hnf4a, Cdx2, or Gata4. A genome-wide survey showed that only eight transcription factors had robust expression and were significantly higher expressed in the villus than in the crypt (ratio Ͼ 3, Creb3l3, Stat3, Mafb, Tbx3, Max, Fosl2, Foxo6, Trp63). Only two of those (Creb3l3 and Stat3) had putative binding sites in the Slc6a19 promoter. Co-transfection of Stat3 failed to activate reporter gene expression from the Slc6a19 promoter, even when activated by addition of IL-6 ( Table 1 showing qualitative data for all transcription factors). Similarly, Creb3l3 failed to activate transcription when co-transfected (Table 1). We therefore considered negative regulation to explain B 0 AT1 expression. The mRNA of transcription factor Sox9 was highly expressed in the crypt but was barely found at the villus tip fraction (Fig. 3C and Table 1). Co-expression of SOX9 suppressed Slc6a19 promoter activity, regardless of whether it was driven by HNF1a (Fig. 4A) or HNF4a (Fig. 4B). Suppression of reporter gene activity was observed in all but the two shortest constructs, suggesting that the Slc6a19 core promoter contains a functional binding site for SOX9. Moreover, it also suppressed Ace2 promoter activity driven by HNF1a (Fig.  2C), indicating that SOX9 may act in crypts as a general repressor of genes involved in intestinal nutrient absorption. A putative SOX9 binding site was identified at position Ϫ34/Ϫ37 overlapping with the HNF4a binding site. However, mutation of this binding site did not reduce SOX9 inhibition (data not shown). Further inspection of the proximal promoter revealed another putative SOX9 binding site at position Ϫ218/Ϫ215. Mutation of this site increased overall promoter activity and partially abrogated the SOX9 effect regardless whether transcription was driven by HNF1a (Fig. 5A) or HNF4a (Fig. 5B). Additional sequence analysis suggested a possible SOX9 binding sites a position ϩ9/ϩ12, mutation of which did not alter promoter activity. To investigate whether some of the SOX9 effect was independent of DNA binding, we co-transfected cells with the SOX9 mutant H65Y, which is unable to bind DNA (33). Expression of this mutant still partially inhibited transcriptional activity, suggesting that SOX9 can act both as a sequence-specific transcription factor and as an inhibitory cofactor. When SOX9(H65Y) was co-transfected with a promoter containing the mutated SOX9 binding site, transcription was no longer inhibited by SOX9 (Fig. 5, A and B).
HNF1a, HNF4a, and SOX9 Bind to the B 0 AT1 Promoter in Vivo-Transfection and overexpression of transcription factors in cell lines provides a first clue as to whether a potential binding site exists within a regulatory region, however, it does not provide evidence for in situ occupation of this binding site within the environment of a differentiated tissue. To investigate  HEK293 cells were co-transfected with vectors expressing firefly luciferase downstream of the Slc6a19 wild type promoter (open bars) or a Slc6a19 promoter with a mutated SOX9 binding site (black bars). All constructs were expressed together with control vector expressing Renilla luciferase from a constitutive promoter. Firefly luciferase activity is depicted as % of control luminescence generated by Renilla luciferase. To analyze SOX9 action, the Slc6a19 reporter was co-expressed with the following transcription factors: A, HNF1a, HNF1a plus SOX9, HNF1a plus SOX9(H65Y); or B, HNF4a, HNF4a plus SOX9, HNF4a plus SOX9(H65Y). Construct Ϫ473 (see Fig. 2) was used as the base for all constructs. Significant reduction of promoter activity due to SOX9 co-transfection are indicated (*, p Ͻ 0.05; **, p Ͻ 0.01). p values indicate the significance derived from three experiments, each with three separate transfections.
which of the transcription factor candidates identified in the reporter studies interact with the Slc6a19 promoter in vivo, we performed ChIP assays using freshly scraped mucosa (Fig. 6). In agreement with the reporter gene studies, we found binding of both HNF1a and HNF4a in the first 300 bp upstream of the transcriptional start site. An additional binding site for HNF1a was found between Ϫ984 and Ϫ812, which was not observed in the reporter gene assays. No specific second binding site of HNF4a was observed consistent with mutagenesis studies of the putative distal binding site. Binding of SOX9 was also detected in the proximal promoter, demonstrating that it can suppress transcription in situ (Fig. 6).
Epigenetic Elements Affecting Tissue Expression-To investigate whether epigenetic factors could contribute to the onset of B 0 AT1 expression along the crypt-villus axis, we compared vil-lus and crypt fractions (Fig. 7, A and B). Histone modification H3K4me3, indicating open promoter regions, was found both in the crypt and villus fractions. However, H3K27ac, also indicating active promoter regions, was highly abundant in the villus fraction but barely detectable in the crypts. Liver cells showed a similar profile as crypts confirming that lack of H3K27ac is associated with lack of transcription (Fig. 7C). DNA methylation at CpG dinucleotides is widely known to be involved in gene silencing. Interestingly, it appears that CpG dinucleotides in the promoter region of Slc6a19 are highly methylated in the crypt and become demethylated in the villus (Fig. 8). The Slc6a19 promoter was also highly methylated in the liver, where B 0 AT1 is not expressed. The changes of meth-  Chromatin immunoprecipitation was carried out using genomic DNA isolated from enterocytes derived from fraction 2 (villus) and fraction 5 (crypt) and from genomic DNA isolated from liver tissue. Data are given in % relative to a real-time PCR reaction on input DNA. After cross-linking, antibodies specific for histone H3, variant H3K27ac, and H3K4me3 were used to precipitate DNA fragments. Positions of primer sets along the Slc6a19 promoter are indicated. The mean signal Ϯ S.D. of three independent experiments is shown. ylation were particularly apparent around the proximal promoter where HNF1a and HNF4a and polymerase II bind.

DISCUSSION
Here, we present a model for the regulation of expression of proteins relevant for nutrient absorption in mature differentiated enterocytes. The model involves expression of an activating transcription factor in both crypt and villus compartment, which is counteracted by a transcriptional repressor whose expression is high in the crypt and low in the villus. The main activators are either HNF4a and to a lesser extent HNF1a, whereas SOX9, which shows maximum expression levels around the crypt, appears to play the role of a transcriptional repressor.
Genome-wide screens of HNF transcription factor targets have been performed for both pancreas and liver (34), and for HNF4a in intestinal cells (26). However, these datasets have not revealed how enterocytes express a set of apical nutrient transporters. The HNF4a screen in intestinal cells only revealed the proton amino acid transporter PAT1 as a target. However, HNF4a binds to the Hnf1a promoter and vice versa, thereby extending the target genes for any of the two transcription factors (34). Previous studies have found tissue-specific interactions between HNF4a and HNF1a (16). Hnf4a mRNA can be transcribed from two alternate promoters, one of which is HNF1a-dependent. In liver and kidney, Hnf4a transcription is independent of HNF1a, whereas in islets, a different promoter is used resulting in HNF1a dependence. In the intestine, both promoters are used, resulting in an interdependence of Hnf1a and Hnf4a expression. In agreement with these tissue differences, a different model has been suggested in kidney, in which expression of epithelial transporters is driven by HNF1a alone (13). However, in vivo, HNF1a could activate Hnf4a, thereby indirectly activating Slc6a19. Alternatively, expression levels in the kidney are somewhat lower due to the lower efficiency with which HNF1A activates the Slc6a19 promoter.
SOX9 has been shown to be an important regulator of differentiation in various cell types, including cells of the small intes-tine (35,36), where it is integral part of the Wnt signaling pathway (37). In the crypt compartment, it has been shown to act as a repressor cell differentiation, inhibiting the expression of genes involved in maturation such as Cdx2 and Muc2 (37). In analogy to the results presented here for Slc6a19, reduced SOX9 expression in villus cells could allow CDX2 expression, which in turn could then be involved in the regulation of various other transcriptional regulators such as Cdx1, Hnf1a, and Hnf4a (38,39). In agreement with this model, we found that CDX2 expression is higher in the crypt-enriched fraction, thus suggesting that SOX9 is not controlling gene expression indirectly via CDX2. Instead, our results suggest that SOX9 directly prevents the premature expression of genes involved nutrient absorption in undifferentiated intestinal crypt cells, despite the presence of positive regulators in this compartment. The inhibitory action of SOX9 is at least in part mediated by a specific SOX9 binding site in the Slc6a19 promoter. In addition, we found evidence for nonspecific DNA binding or action as an inhibitory cofactor through protein-protein interactions. Upon down-regulation of Sox9 in differentiated enterocytes, transcription of genes involved in nutrient absorption is activated.
This minimal model cannot explain the lack of epithelial nutrient transporter expression in hepatocytes where HNF1a and HNF4a are abundant and SOX9 is missing. Kikuchi et al. (13) suggested that in the liver, specific CpG dinucleotides within promoters of transporter genes would be methylated. For Slc6a19, differential methylation was observed at position Ϫ1080 upstream from the transcriptional start site. Given that the core promoter of Slc6a19 is Ͻ300 bp long, it appears unlikely that the proposed methylation would directly affect the promoter recruitment of activators. Our bisulfite sequencing data, by contrast, demonstrate differential methylation at the proximal promoter where binding of transcription factors and RNA polymerase II occurs. Intriguingly, the methylation pattern of the proximal B 0 AT1 promoter changes rapidly, from fully methylated in the crypt to completely demethylated in mature enterocytes. Due to the thermodynamic stability of CpG methylation, demethylation can only occur by lack of remethylation after mitosis or by a modified DNA excision repair mechanism (40,41). Rapid demethylation has nevertheless been demonstrated in neurons (42) and hormone-sensitive promoters (43). Stem cells in the crypt undergo three to four cell divisions before differentiation (44); thus, demethylation could occur by both mechanisms. The complete methylation of the Slc6a19 promoter observed in liver cells is in line with an absence of expression in this tissue. This suggests that demethylation must occur before HNF1a or HNF4a can bind and activate transcription.
Differential histone acetylation has been reported for the sucrose-isomaltase gene and promoter along the crypt-villus axis (32). Di-acetylated histone H3 (lysine 9 and 14), which is associated with transcriptionally active genes, was found to be very low in the crypts but much higher in the villus. Analysis of the ENCODE database (45) at the Slc6a19 promoter confirmed an active histone signature in the intestine and kidney (Fig. 9). Histone modification H3K4me3 was highly enriched over the promoter, allowing binding of polymerase II. Histone modification H3K36me3 was present over all exons confirming that the gene was actively transcribed. These three elements were absent in other tissues, including the liver, where HNF1a and HNF4a are highly expressed. This suggests that histone modification also should occur before transcription can be initiated.
Our data suggest a model, whereby cell-and tissue-specific regulation of B 0 AT1 expression is mediated by a concurrent combination of three mechanisms: 1) different levels of activating and repressing transcription factors in crypt and villus cells; 2) reversible histone modifications and 3) reversible promoter methylation. Stem cells in the crypt undergo four to five divisions during which differentiation occurs. During these divisions change of histone modification and removal of promoter methylation could occur. In addition, SOX9 could become diluted allowing the onset of B 0 AT1 expression, which gradually increases as cells move to the tip of the villus (Fig. 1).