INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

ABSORPTIVE ENTEROCYTES ORIGINATE from undifferentiated, multipotent stem cells located near the base of small intestinal crypts (20). Through a continuous process of cell division, migration, differentiation, apoptosis, and exfoliation, these cells are renewed every 3-5 days. Absorptive enterocytes comprise over 95% of mucosal cells on small intestinal villi and are responsible for the terminal digestion and absorption of nutrients. To carry out these specialized functions, absorptive enterocytes acquire a differentiated phenotype characterized by the expression of functionally relevant proteins such as digestive enzymes, receptors, transporters, and cytoplasmic carriers. Differences in the panel of genes expressed along the proximal-to-distal (horizontal) axis result in regional differences in intestinal function. The mechanisms that underlie the process of intestinal differentiation are poorly understood.

The effects of specific transcription factors in the regulation of absorptive enterocyte-specific differentiation can be studied by the use of marker genes whose expression parallels that of key events in the differentiation process. Lactase-phlorizin hydrolase (LPH) and sucrase-isomaltase (SI) are intestinal disaccharidases that are expressed in villus enterocytes and display a differentiation-specific pattern of expression. The LPH and SI genes are similarly expressed in jejunum and proximal ileum but demonstrate different patterns during development. LPH gene expression is highest in newborn mammals and declines during weaning, whereas SI gene expression is low or undetectable at birth and increases to adult levels during weaning. In most humans, LPH expression declines, similar to that in other mammals, but in a subpopulation of humans, LPH expression remains high throughout adult life. Previous work in our laboratory has shown that the developmental pattern of lactase activity in rats (8) and the genetic pattern in humans (14) are correlated with the abundance of its mRNA and also that the horizontal and developmental patterns of LPH expression in rats are transcriptionally regulated (32). It is now widely accepted that LPH expression is controlled mainly by the rate of gene transcription.

Studies (33, 40, 41, 64, 68) in transgenic mice indicate that the 5'-flanking regions of both the LPH and SI genes direct transgene expression to villus epithelium. Analysis of the published sequence reveals that the LPH (6, 71) and SI genes (9, 63, 73) contain GATA, hepatocyte nuclear factor-1 (HNF-1), and Cdx-2 consensus binding sites in their 5'-flanking regions (see Fig. 1). These transcription factor families have all been shown to influence the regulation of cellular differentiation. Because GATA, HNF-1, and Cdx-2 factors are expressed contiguously only in intestinal epithelium and their cognate binding sites in the LPH and SI genes are adjacent to the TATA box and to each other, we hypothesized that these transcription factors together are important regulators of LPH and SI gene expression.

View larger version (20K): [in this window] [in a new window]   Fig. 1.   The 100-bp to 20-bp regions of the rat lactase-phlorizin hydrolase (rLPH), human LPH (hLPH), and human sucrase-isomaltase (hSI) 5'-flanking sequence contain binding sites for GATA (WGATAR; W = A or T, R = A or G), hepatocyte nuclear factor-1 (HNF-1) (GTTAATNATTAAC; N = A, C, G, or T), and Cdx-2 (TTTAY; Y = A or T) transcription factors (indicated in italics above the DNA sequence). In the hLPH and rLPH genes, rG1 and hG1 are the 5' GATA sites whereas rG2 and hG2 are the 3' GATA sites. The 3' GATA site in rLPH is indicated by ? because it is not a true 6-base consensus. The hG1 site is the lactase upstream element (LUE) characterized previously using GATA-6 (17). hSI/G refers to the GATA site in the human SI gene. rH and hH are the HNF-1 consensus sites that correspond to CE-LPH2c in the pig LPH gene (59). rC and hC are the CE-LPH1a sites that bind Cdx-2 (7, 16, 22, 53, 65, 66). The HNF-1 (hSIF2 and hSIF3) and Cdx-2 binding sites (hSIF1) in the hSI gene have been previously characterized (60, 72). The TATA boxes are underlined.

The GATA family of zinc finger proteins, defined by a highly conserved DNA-binding domain that interacts specifically with DNA elements containing a consensus WGATAR sequence (W = A or T; R = A or G), has been implicated in cell lineage differentiation during vertebrate development. GATA-1, -2, and -3 are expressed in developing blood cells during hematopoiesis, whereas GATA-4, -5, and -6 are expressed together only in cardiac tissue and small intestine, but individually (or in overlapping patterns) in the stomach, liver, lungs, spleen, ovary, testis, and bladder (3, 29, 36, 50, 51). In both mouse (3) and chicken intestine (18), GATA-4 transcripts are distributed all along the villus, with increasing levels accumulating toward the villus tips. In chicken intestine (18), GATA-5 mRNA is localized to differentiated cells of the villi, whereas GATA-6 transcripts are localized to cells of the upper crypt and lower villus. Mice homozygous for the GATA-4 (35, 48) or GATA-6 (31, 52) null allele die in utero and therefore provide little information regarding maintenance of intestinal differentiation in adults. GATA-5 knockout mice survive and reproduce despite pronounced genitourinary abnormalities in females (49) but show no gross abnormalities in intestinal structure or histology (intestinal gene expression was not reported). Recent studies (15, 17, 18) have suggested that GATA-4, -5, and -6 may modulate the expression of intestinal genes. Gao et al. (18) showed that GATA-4, -5, and -6 can directly stimulate the activity of the promoter for the Xenopus intestinal fatty acid binding protein (IFABP) gene. Fitzgerald et al. (17) demonstrated a 15-fold induction of a human LPH promoter-reporter construct using a chicken GATA-6 expression vector. Fang et al. (15) showed that all three members of this GATA subfamily are capable of activating the rat LPH promoter. Two potential GATA-binding sites are present in the rat and human LPH genes (see Fig. 1), although the 3' site in the rat LPH gene contains only 5 bp of the 6-bp consensus. The human SI gene contains a single putative GATA motif within 100 bp 5' to the transcriptional start site, the characterization of which has not yet been reported.

HNF-1 transcription factors contain a novel DNA-binding homeodomain distantly related to the POU domain and bind as dimers to the consensus sequence GTTAATNATTAAC (42, 67). Two members of this family, HNF-1 and HNF-1, have been described (43, 55) and share highly homologous DNA binding domains but divergent activation domains. Originally thought to be liver specific, these two proteins are also expressed in kidney, stomach, and small and large intestine (5, 34, 43). HNF-1 is also expressed in lung and ovary (43). In mouse small intestine, HNF-1 and HNF-1 mRNAs are expressed at high levels in crypt cells, and expression is gradually reduced until there are low levels at the villus tips (58). Reports (1, 38, 54) of mice lacking HNF-1 gave variable results, including sterility, liver enlargement, renal dysfunction, non-insulin-dependent diabetes mellitus, and reduction in the expression of certain liver genes such as albumin, ₁-antitrypsin, phenylalanine hydroxylase, and liver FABP. The effect of HNF-1 gene disruption on gastrointestinal function has not been described. HNF-1 knockout mice have a disorganized visceral endoderm and die at embryonic day 7.5 (4, 11). In addition to regulating the expression of liver genes, HNF-1 factors also modulate genes that are expressed only in small and/or large intestine, including LPH (59), SI (72), and guanylin (23). In the pig LPH gene, mutational analysis in cotransfection experiments localized an HNF-1-mediated response to an HNF-1 binding site within 227 bp of the transcriptional start site (called CE-LPH2c) (59), which is conserved in position in the human (90 to 70 bp) and rat LPH genes (86 to 66 bp) (see Fig. 1). The human SI promoter has two HNF-1 binding sites, SIF2 (90 to 70 bp) and SIF3 (176 to 159 bp), that interact with both HNF-1 and HNF-1 (72). These two family members, which can form homodimers as well as heterodimers, differentially activate both the LPH and SI promoters (59, 72), suggesting that the relative abundance of HNF-1 isoforms may be important for the regulation of genes with HNF-1 binding sites.

Cdx-2, a homeodomain-containing transcription factor that binds to the consensus sequence TTTAY (Y = A or C), is highly expressed in adult mammalian small intestine and colon and is also detectable in pancreas (19, 25, 26, 60). In mouse small intestine, Cdx-2 is expressed in all epithelial cells in both crypts and villi, with the exception of Paneth cells at the base of crypts (24). Mice heterozygous for a Cdx-2 null mutation develop adenocarcinoma of the small intestine and colon, whereas homozygote null mutant embryos die between 3.5 and 5.5 days postcoitum (10). Cdx-2 has been shown (61) to induce differentiation in IEC-6 cells, suggesting a role in cellular differentiation. Cdx-2 has also been shown to bind and activate the LPH (7, 16, 22, 53, 65, 66) and SI genes (60) via conserved cis elements called CE-LPH1a and SIF1, respectively, both of which are adjacent to their TATA boxes (Fig. 1). CE-LPH1a contains a single binding site, whereas SIF1 contains two adjacent sites on opposite strands. Additional Cdx-2 binding sites, CE-LPH1b (59) and CE-LPH-1c (47), have been identified in the pig LPH gene.

At present, the relative importance of GATA-4, -5, and -6, HNF-1 and , and Cdx-2 on LPH and SI gene expression and possible species differences between rat and human LPH gene expression have not been characterized together under the same experimental conditions. Furthermore, possible interactions among these transcription factors have only recently been examined (47). Mitchelmore et al. (47) have demonstrated that HNF-1 and Cdx-2 physically interact and cooperatively activate the pig LPH promoter. In the present study, we demonstrate for the first time that GATA, HNF-1, and Cdx-2 transcription factors regulate the LPH and SI promoters differently and that GATA-5 and HNF-1 synergistically activate these promoters.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Chemicals and reagents. All chemicals were purchased from Sigma Chemical (St. Louis, MO), GIBCO-BRL (Gaithersburg, MD), or Fisher Scientific (Fair Lawn, NJ) unless indicated otherwise. Enzymes were purchased from GIBCO-BRL, Promega Biotech (Madison, WI), or Pharmacia Biotech (Piscataway, NJ) unless indicated otherwise. Radioactive nucleotides and [¹⁴C]chloramphenicol were purchased from DuPont-New England Nuclear (Boston, MA). [³⁵S]methionine (Redivue) was purchased from Amersham Life Science (Arlington Heights, IL).

Plasmids. The human growth hormone (hGH) gene was used as a reporter in transfection assays. As described previously (33), the hGH gene was subcloned into pBluescript II KS+ (Stratagene, La Jolla, CA). This construct, referred to as Bluescript growth hormone (BSGH), served as a promoterless control for background and the basic construct into which all promoter sequences were subcloned. Sequence from the rat LPH (rLPH) gene (2038 to +15 bp) was isolated (71) and subcloned into BSGH 5' to the hGH reporter, as previously described (33). A segment of the human LPH (hLPH) gene (1025 to +11 bp) was amplified by PCR from human DNA isolated from discarded jejunum of patients undergoing elective gastric bypass surgery. The 5' (5'-ATGGGTACCAAGATACTTATTATAGGAAGAGGA-3') and 3' primers (5'-ATCTAAGCTTTTTCTAGGAACTGTTAGGAGG-3'), designed from the published sequence (6), contain Kpn I and Hind III sites, respectively (underlined), to facilitate subcloning into BSGH. A segment of the human SI (hSI) gene (183 to +54 bp, gift of Dr. P. Traber, University of Pennsylvania) (73) was also amplified (5'-ATGGGTACCTGACAGTACAATTACTAATTAAC-3 and 5'-ACGTAAGCTTAGCCTGTTCTCTTTGCTATG-3) and subcloned into BSGH. Deletions and mutations of all other rat LPH, human LPH, and human SI promoter-reporter plasmids were constructed by PCR using the same Kpn I-Hind III strategy. Mutations were introduced into primer and template oligonucleotides. All constructs were confirmed by sequencing.

Cell culture and transfections. The Caco-2 cell line was the principal cell line used in the present study. Originally derived from a human colon adenocarcinoma, Caco-2 cells differentiate on confluence, exhibiting characteristics of small intestinal absorptive enterocytes, including a microvillus membrane and expression of small intestinal genes. The Caco-2 cells used in the present study have been previously characterized (69) with respect to LPH and SI expression.

Nuclear extracts. Nuclear extracts were isolated from confluent cells grown on 10-cm plates directly (untransfected) or after transfection of specific expression vectors. For transfected cells, 8 µg of expression vector were transfected as described above, and two transfection reactions were combined on 10-cm dishes. Nuclear extracts were prepared by modifying previously published methods (2). Cells were harvested by scraping, and resuspended in ice-cold buffer A [10 mM HEPES, pH 7.9, 10 mM KCl, 0.1 mM EDTA, 0.1 mM EGTA, 1 mM dithiothreitol (DTT), and 0.5 mM phenylmethylsulfonyl fluoride (PMSF)], incubated on ice for 10 min, and lysed by vortexing for 10 s. Nuclei were pelleted in a microcentrifuge for 10 s at full speed and extracted in 100 µl of buffer C (20 mM HEPES, pH 7.9, 0.4 M NaCl, 1 mM EDTA, 1 mM EGTA, 1 mM DTT, and 0.5 mM PMSF) for 20 min on ice. The resulting lysate was cleared of cellular debris by centrifugation for 15 min and stored as a nuclear extract at 80°C. The protein concentration was determined as described previously by Kalb and Bernlohr (28).

In vitro transcription-translation. Labeled ([³⁵S]methionine) and unlabeled proteins were synthesized using a reticulocyte lysate transcription-translation system (TNT, Promega) according to the manufacturer's instructions. Because all expression vectors contain a T7 DNA-dependent RNA polymerase site between the viral promoter and the cDNA insert, a T7-based TNT system was employed. However, because in vitro transcription of the GATA-5 expression vector using a T7 polymerase transcription-translation system was inefficient and PCR could not amplify the GATA-5 coding region, we subcloned a 1.5-kb Sma I-Eco RV fragment containing the GATA-5 coding region into the Sma I site of pGEM-7Zf() and isolated (screened by restriction digests, confirmed by sequencing) plasmids whose orientation enabled correct transcription from the SP6 polymerase site. GATA-5 was efficiently synthesized from this plasmid using an SP6 TNT system.

EMSAs. Oligonucleotides used as probes and competitors in EMSAs are shown in Fig. 2. Probes were made by annealing a 10-fold molar excess of 10- or 11-base reverse-strand oligonucleotide to the forward strand by boiling for 2 min in annealing buffer (100 mM NaCl, 10 mM Tris, pH 8.0, and 1 mM EDTA), then slow cooling for 2 h (<32°C). The annealed oligonucleotides were extended with the large fragment of DNA polymerase I (Klenow) in the presence of [³²P]dATP in extension buffer (50 mM Tris, pH 8.0, 10 mM MgCl₂, and 1 mM DTT) for 1 h at 37°C. After removing an aliquot for specific activity determinations by TCA precipitation, we separated the labeled probe from the short antisense oligonucleotides using a spin column with a 12-base cutoff (STE Select-D G-25, 5 Prime3 Prime, Boulder, CO). Purified probes were stored at 20°C. The specific activity of all probes exceeded 10⁷ cpm/pmol. Competitors were synthesized by annealing equimolar amounts of full-length sense and antisense oligonucleotides as described above.

View larger version (51K): [in this window] [in a new window]   Fig. 2.   Oligonucleotides used as probes and competitors in electrophoretic mobility shift assays (EMSAs). The specific sequences and localization in respective genes are indicated. Binding sites are in bold. An " m" indicates a mutation introduced into the binding site, and the mutation itself is shown in lowercase letters. xIFABP/G and r-Fib/H are previously characterized probes for binding of GATA (18) and HNF-1 factors (12), respectively.

RT-PCR. RNA was isolated from Caco-2 cells grown on six-well plates using TRIzol reagent (GIBCO-BRL) according to the manufacturer's instructions. To synthesize first-strand cDNA, purified RNA was annealed to oligo(dT) and extended with Superscript II (GIBCO-BRL) according to the manufacturer's suggested protocol. PCR was carried out on the first-strand cDNA using the following oligonucleotides: human LPH (5'-GTCCCACTACCGTTTTTCCA-3', 5'-CGTCTGTGGTAGGTCCCAGT-3'), human SI (5'-CTCTCCATCGGTCTTTCCAA-3', 5'-AGAAGGCTCTGGGAGGTGTT-3'), human glyceraldehyde-3-phosphate dehydrogenase (GAPDH; 5'-GGGTCATCATCTCTGCCCCCTCTG-3', 5'-CCATCCACAGTCTTCTGGTGGCA-3'), and mouse GATA-5 (5'-GCCTCTTCTCCCACTCTCCT-3', 5'-GTAGGACCCCACTGAGACCA-3'). After an initial melting step at 95°C for 2 min, the oligonucleotides were annealed to the cDNA template at 55°C for 40 s, extended at 72°C for 60 s, and melted at 95°C for 20 s. After 30 cycles, the reaction was extended for 10 min at 72°C, then cooled to 4°C. The amplification products were analyzed on 1.8% agarose gels stained with ethidium bromide.

Statistics. The t-test or one-way ANOVA was employed in all statistical analyses using InStat software (Graphpad Software, San Diego, CA). The Dunnett's or Tukey multiple comparison test was employed for all statistically significant ANOVA analyses.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Characterization of individual effects of transcription factors on LPH and SI promoters. The individual effects of multiple transcription factors on the LPH and SI promoters have not been compared under the same experimental conditions. Thus transient cotransfection assays in Caco-2 cells were carried out using individual expression vectors for each of the transcription factors under study. As shown in Fig. 3A, GATA factors are capable of activating the LPH and SI promoters. Transcriptional activities of promoter-reporter constructs were increased when cotransfected with expression vectors for GATA-4, GATA-5, or GATA-6 compared with pRC-CMV, a plasmid containing only the CMV promoter. GATA-stimulated transcriptional activities were higher for constructs that contain consensus GATA binding motifs in their promoters (i.e., rLPH108, rLPH2038, hLPH118, hLPH1025, and hSI183). Activation was generally strongest with the GATA-5 expression vector, demonstrating statistically significant increases from pRC-CMV controls for rLPH108, rLPH2038, hLPH118, and hSI183. Transcriptional activities of the basal promoter constructs rLPH37 and hLPH37 and the negative control promoter-reporter construct RSVdE cotransfected with the GATA expression vectors revealed minimal responses, although all were significantly increased over the pRC-CMV controls with GATA-5 cotransfection (P < 0.01 for each). These data demonstrate that GATA family transcription factors activate the LPH and SI promoters, consistent with previous reports for LPH (15, 17). Activation of the SI promoter by GATA-5 has not been previously reported.

View larger version (22K): [in this window] [in a new window]   Fig. 3.   GATA, HNF-1, and Cdx-2 transcription factors individually activate the LPH and SI promoters in transient cotransfection assays in Caco-2 cells. Regions of the rLPH, hLPH, and hSI 5'-flanking sequences (indicated by the number of bases 5' to the transcriptional start site) fused to the human growth hormone (hGH) reporter were cotransfected with individual expression vectors into Caco-2 cells as described in MATERIALS AND METHODS. A: cotransfection with GATA-4, -5, and -6 expression vectors. B: cotransfection with HNF-1 and - expression vectors. C: cotransfection with the Cdx-2 expression vector. pRC-CMV was used as a negative control expression vector, and RSVdE, which does not contain GATA, HNF-1, or Cdx-2 binding sites, was used as a negative control promoter-reporter construct. Transcriptional activity (means ± SE; n = 4) is expressed as total hGH secreted into the medium over 24 h relative to total chloramphenicol acetyltransferase (CAT) activity in cell lysates/well (mg hGH/U CAT activity). * P < 0.05, ** P < 0.01, significantly different from pRC-CMV controls.

View larger version (103K): [in this window] [in a new window]   Fig. 4.   GATA, HNF-1, and Cdx-2 binding sites in rLPH, hLPH, and hSI genes demonstrate different affinities in protein-DNA interactions. A: using xIFABP/G as a test probe for GATA binding (lane 1) and reticulocyte lysates (5 µl) enriched in mouse GATA-5 (TNT, Promega) (lanes 2-24), competition EMSAs were carried out using 20-, 50-, and 200-fold molar excess of double-stranded oligonucleotides containing the putative GATA sites under study (lanes 4-18). Specific (S) and nonspecific (NS) titrations were carried out with xIFABP/G (lanes 19-21) and hSIF2 (lanes 22-24), respectively, for comparison. A supershift complex (SC) with a GATA-5 antibody (G5 Ab; lane 3) demonstrated that the main complex (arrowhead) is authentic GATA-5. Other complexes, which also specifically competed and supershifted, are indicated by *. B: using the r-Fib/H oligonucleotide as a test probe for HNF-1 binding and nuclear extracts (5 µg) from HepG2 cells, which are enriched in HNF-1, competition of the HNF-1 complex (arrowhead, lane 1) was carried out using 10-, 50-, and 200-fold molar excess of oligonucleotides containing the HNF-1 binding sites under study (lanes 4-15). The specific (S) and nonspecific (NS) competitors (r-Fib/H and hSIF1, respectively) were added at 200-fold molar excess. C: using the hSIF1 site as a test probe (lane 1) and nuclear extracts from COLO 320DM cells (lanes 2-12), which are enriched in Cdx-2, competition of the Cdx-2 complexes A and B (lane 2) was carried out using 20-, 100-, and 500-fold molar excess of competitors (lanes 3-11). The nonspecific competitor (NS) was hSIF3 and was added at a 500-fold molar excess (lane 12).

Combined effects of transcription factors on LPH and SI expression. To characterize the combined effects of multiple transcription factors and to identify possible functional interactions among transcription factor families, cotransfection experiments were carried out using all possible combinations of GATA-5, HNF-1, and Cdx-2. These transcription factors were chosen because they were either strong activators (GATA-5 and HNF-1) or the only member (Cdx-2) of the transcription factor families under study. Furthermore, all are expressed on villi coincident with LPH and SI expression. As shown in Fig. 5A, the individual effects of HNF-1 and Cdx-2 produced significantly greater transcriptional activities for hSI183 than for rLPH108 and hLPH118. The combination of GATA-5 plus HNF-1 resulted in mean transcriptional activities that were similar for all three promoter constructs. The combinations of GATA-5 plus Cdx-2, HNF-1 plus Cdx-2, and all three transcription factors together resulted in mean transcriptional activities that were significantly greater for hSI183 than for either of the LPH promoter-reporter constructs (P < 0.05 for each). GATA-5 plus HNF-1 produced the highest activation of all combinations for rLPH108 and hLPH118, whereas all three expression vectors together produced the greatest response for hSI183.

View larger version (22K): [in this window] [in a new window]   Fig. 5.   Combinations of expression vectors reveal differential activation of promoters and functional interactions among transcription factors. A: transcriptional activities (means ± SE, n = 5) after cotransfection of all possible combinations of GATA-5 (G), HNF-1 (H), and Cdx-2 (C) expression vectors were compared among rLPH108, hLPH118, and hSI183. Unstimulated expression (pRC-CMV) is indicated as a control. * P < 0.05, ** P < 0.01, significantly different from rLPH108 and hLPH118. B: interactions among transcription factors were characterized by comparing the activity of multiple expression vectors cotransfected together to that of individual expression vectors cotransfected singly and added together. This relationship, termed interaction response, is depicted as the mean (±SE; n = 5) logarithm of the ratio of combined to added individual effects (e.g., log {(GATA-5 + HNF-1)/[(GATA-5) + (HNF-1)]}). Values of 0.1 to 0.1 were defined as additive effects, greater than 0.1, synergistic effects, and less than 0.1, antagonistic effects. Interaction responses for each combination of expression vectors were compared among rLPH108, hLPH118, and hSI183. Bars with the same lowercase letter within an expression vector treatment group were significantly different: a P < 0.05, b P < 0.001.

View larger version (74K): [in this window] [in a new window]   Fig. 6.   GATA-5, HNF-1, and Cdx-2 activate endogenous LPH and SI gene expression in Caco-2 cells. RT-PCR was carried out on RNA isolated from Caco-2 cells transfected with all combinations of the GATA-5 (G), HNF-1 (H), and Cdx-2 (C) expression vectors as described in MATERIALS AND METHODS. Amplification of endogenous hLPH and hSI mRNAs is indicated by 166- and 201-bp fragments, respectively. Amplification of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA (208-bp fragment) was used as a positive control for all RNAs, and that of mouse GATA-5 mRNA (G5; 226-bp fragment) was used as positive control for transcription of the GATA-5 expression vector. A size marker (lane 1) was included. pRC-CMV was used as the negative control expression vector (lane 2).

Cell-specific activation of human LPH promoter by GATA-5. To determine whether the GATA-5 response is cell specific, cotransfection experiments with the GATA-5 expression vector and the hLPH118 construct were carried out in different cell lines. Table 2 shows a wide range of transcriptional responses (fold change over baseline) in both intestinal (T84, Caco-2, HT-29, COLO 320DM) and nonintestinal cell lines [HepG2, Madin-Darby canine kidney (MDCK), HeLa, 3T3]). The highest responses were found in T84, Caco-2, (both colonic adenocarcinoma), HepG2 (hepatocellular carcinoma), and MDCK (kidney epithelial) cells, whereas the lowest responses were observed in HeLa (cervical carcinoma), COLO 320DM (adenocarcinoma of the colon), and 3T3 (fibroblast) cells. Parallel results were found for rLPH108 and hSI183 (not shown). EMSAs using a probe specific for HNF-1 or Cdx-2 binding (Fig. 7) revealed a close correlation between the presence of HNF-1 binding activity and GATA-5 activation of hLPH118 (Table 2). Those cell lines that had abundant HNF-1 binding activity (Fig. 7A) also displayed high levels of GATA-5 activation responses (Table 2), whereas those cell lines with no evidence of HNF-1 binding activity demonstrated limited GATA-5 responses. There was no correlation between the presence of Cdx-2 binding activity (Fig. 7B) and GATA-5 activation.

View larger version (109K): [in this window] [in a new window]   Fig. 7.   HNF-1 and Cdx-2 binding activity is different in different cell lines as determined by EMSAs. A: using the human HNF-1 binding site probe, EMSAs were carried out using 10 µg of nuclear extract from different cell lines. The main complex (arrowhead) and secondary complex (*) competed specifically and supershifted with HNF-1 and/or HNF-1 antibodies (not shown). B: using the human Cdx-2 binding site probe, EMSAs were carried out using 10 µg of nuclear extract from different cell lines. The main complex (arrowhead) and secondary complexes (*) competed specifically and supershifted with the Cdx-2 antibody (not shown). The secondary complexes become more intense with repeated thawing and refreezing of nuclear extracts (not shown), suggesting that these complexes represent degradation products of Cdx-2. MDCK, Madin-Darby canine kidney cells.

View larger version (16K): [in this window] [in a new window]   Fig. 8.   GATA-5 activates hLPH118 in HeLa cells when HNF-1 is cotransfected. Transient cotransfection assays (means ± SE; n = 6) were carried out in HeLa and Caco-2 cells using hLPH118 and expression vectors for GATA-5 and HNF-1, singly and in combination. ** P < 0.01, significantly different from HeLa cells within the same expression vector group.

Mutational analysis of GATA and HNF-1 binding sites in human LPH promoter. To characterize the effects of disruptions in protein-DNA interactions during individual and combined GATA-5 and HNF-1 cotransfections, mutations were introduced into the GATA and HNF-1 binding sites of the human LPH promoter. To first demonstrate that the introduced mutations disrupt specific protein-DNA interactions, EMSAs were carried out using nuclear extracts from Caco-2 cells transfected with the GATA-5 expression vector and a probe that spans the 5' GATA and the HNF-1 site of the human LPH gene (hG₁H). As shown in Fig. 9, two main complexes were identified. The faster mobility complex was competed with an oligonucleotide containing the wild-type GATA site (hG₁; Fig. 9, lane 3), but not with an oligonucleotide containing the specific mutation in the GATA motif (Fig. 9, lane 4). This complex also supershifted with a GATA-5 antibody (Fig. 9, lane 5) and was therefore indicated as GATA-5 complex. The slower mobility complex competed specifically with an oligonucleotide containing an intact human HNF-1 site (hH; Fig. 9, lane 6), but not with an oligonucleotide containing a mutated site (Fig. 9, lane 7), and supershifted specifically with an HNF-1 antibody (Fig. 9, lane 8). This complex was indicated as HNF-1 complex. Noteworthy is that specific competition with hH revealed partial competition of the GATA-5 complex. An oligonucleotide probe containing the 3' GATA site (hG₂) also binds GATA factors, but a mutation introduced into this site does not compete (not shown). These data demonstrate that mutations introduced into the GATA and HNF-1 sites of the human LPH promoter disrupt specific protein-DNA interactions at these sites.

View larger version (90K): [in this window] [in a new window]   Fig. 9.   Specific mutations disrupt specific protein-DNA interactions. EMSAs were carried out using a probe that spans the 5' GATA and HNF-1 sites of the hLPH gene (hG1H, lanes 1-9) and nuclear extracts from Caco-2 cells transfected with the GATA-5 expression vector (lanes 2-9). Competition (200-fold molar excess) was carried out using oligonucleotides spanning only the GATA (hG1) or HNF-1 (hH) sites as well as oligonucleotides spanning both of these sites, but with specific mutations introduced into the GATA or HNF-1 binding motifs (hG1m and hHm, respectively). Supershifts were carried out using antibodies specific for GATA-5 (G5) or HNF-1 (H1). Nonimmune (NI) serum was used as a negative control. GATA-5-specific complex (G5C), HNF-1-specific complex (H1C), and supershift complexes (SC) are indicated; * denotes a nonspecific complex that neither competes nor supershifts.

View larger version (20K): [in this window] [in a new window]   Fig. 10.   Functional cooperativity between GATA-5 and HNF-1 is dependent on an intact HNF-1 binding site. Transient cotransfections using all combinations of wild-type (wt) and mutated sites in the hLPH118 (h118) construct and the GATA-5 and HNF-1 expression vectors were carried out in Caco-2 cells. h55wt is a promoter-reporter construct containing 55 bp of 5'-flanking sequence of the hLPH gene. Specific mutations were introduced according to the gel shift probes depicted in Fig. 2 (e.g., the h118G1m promoter-reporter construct contains the hG1m mutation). Transcriptional activities (means ± SE; n = 4) were determined as described in MATERIALS AND METHODS. * P < 0.05, ** P < 0.01, significantly different from h118wt (Dunnett's test).

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

This is the first study to extensively compare, under the same experimental conditions, the effect of specific transcription factors on multiple intestinal gene promoters and to identify functional cooperativity between two major classes of transcription factors, the zinc finger GATA proteins and the homeobox HNF-1 family. Studies to date have generally examined regulation of individual intestinal gene promoters, and recently, the effect of multiple transcription factors (46, 47, 72), but only on single promoters. In the present study, we found that GATA, HNF-1, and Cdx-2 transcription factors regulate the rat LPH, human LPH, and human SI promoters differently and that GATA-5 and HNF-1 cooperatively activate these promoters. Together, these data suggest that the specific expression of intestinal genes is modulated by complex interactions involving multiple transcription factors.

Our data are in agreement with previous reports demonstrating individually that members of the GATA-4, -5, and -6 subfamily are capable of activating the human (17) and rat LPH promoters (15), HNF-1 is capable of activating the pig LPH (59) and human SI (72) promoters, and Cdx-2 is capable of activating the pig (65) and rat LPH (16) and human SI promoters (60). Our data also parallel a recent report by Mitchelmore et. al. (47), who showed that HNF-1 and Cdx-2 cooperatively activate the pig LPH promoter, a finding we have also demonstrated for the human LPH promoter. Thus our data are in general agreement with previous reports.

GATA and HNF-1 transcription factors together are key activators of LPH gene expression, whereas Cdx-2 has a comparatively weak effect on the LPH promoters. Although GATA-4, -5, and -6, HNF-1, and Cdx-2 are all capable of activating the LPH promoters individually in cotransfection assays, comparative analyses reveal that the rat and human LPH promoters generally demonstrate greater transcriptional responses to the GATA factors and HNF-1 than to Cdx-2 (Table 1). Similarly, activation of endogenous LPH transcription in Caco-2 cells occurred with overexpression of GATA-5 or HNF-1 but not with Cdx-2 (Fig. 6). In contrast to our data, in which Cdx-2 showed no effect on LPH mRNA abundance, Fang et al. (16) reported a modest twofold increase in LPH mRNA abundance in Cdx-2-transfected Caco-2 cells, which could be due to a more sensitive quantification technique (densitometry of the negative image of ethidium bromide-stained bands in agarose gels) employed in their study. Our results depict the comparative effects of single and multiple transcription factors, and thus relative to GATA-5 or HNF-1 cotransfection, Cdx-2 appears to be a comparatively weak activator of endogenous LPH gene transcription in Caco-2 cells. Finally, the combination of GATA-5 and HNF-1 produces the strongest activation of the LPH promoters, demonstrating functional cooperativity between these two transcription factor families (discussed below). Together, these data suggest that members of the GATA and HNF-1 transcription factor families are significant activators of LPH gene expression, whereas Cdx-2, although perhaps necessary for LPH expression in vivo, has a comparatively lesser role in modulating LPH gene expression.

Although some experiments argue that HNF-1 and Cdx-2 have a greater role than GATA factors in the modulation of the SI promoter, other experiments suggest that all three transcription factor families are important in the activation of SI gene expression. As shown in Table 1, transcriptional responses for the human SI promoter were strongest for HNF-1 and Cdx-2 compared with those for the GATA factors. Furthermore, greater basal expression of the human SI promoter compared with the LPH promoters could be due to the relative abundance of endogenous transcription factors synthesized in Caco-2 cells. Based on the amount of extract used, the intensity of complexes on autoradiographic film, and the length of exposure time, data from EMSAs suggest that Caco-2 cells synthesize abundant amounts of HNF-1 and Cdx-2 proteins but limited amounts of GATA proteins (not shown). Thus it could be argued that HNF-1 and Cdx-2 have a greater role than GATA factors in the activation of the SI promoter (in contrast to the LPH promoters), resulting in a greater basal (unstimulated) expression. HNF-1 and Cdx-2 have greater affinities for SIF3 and SIF1, respectively, compared with their cis element counterparts in the LPH genes (Fig. 4), which could explain the greater basal transcriptional activities of the SI vs. the LPH promoters. Noteworthy is that hSIF1 contains two Cdx-2 motifs, whereas rC and hC (i.e., CE-LPH1a) each have a single binding motif, which likely accounts for the differences in affinity between the two elements.

Cotransfection experiments using combinations of expression vectors, however, suggest that all three transcription factor families cooperatively activate the human SI promoter. For hSI183, cotransfection of the combination of GATA-5, HNF-1, and Cdx-2 expression vectors produced the highest transcriptional activity of any combination of transcription factors (Fig. 5A) and demonstrated a strong synergistic interaction response (Fig. 5B). Furthermore, all three transcription factors were also important for endogenous SI gene expression, because an amplified product corresponding to human SI mRNA was detected only in the RNA isolated from Caco-2 cells transfected with the GATA-5, HNF-1, and Cdx-2 expression vectors (Fig. 6). Activation of SI individually by HNF-1 and Cdx-2 transcription factors has previously been demonstrated (60, 72), but the present study is the first to show activation of the SI promoter by the GATA family and cooperative activation of the SI promoter by members of the three different transcription factor families.

The rat and human LPH promoters, with some exceptions, demonstrated similar responses to the transcription factors under study. Both promoters were significantly activated by GATA, HNF-1, and Cdx-2 transcription factors (Fig. 3), and their patterns with cotransfection of multiple expression vectors were similar to each other (Fig. 5A). A significant difference between the two promoters, however, was revealed by cotransfection of the GATA-6 expression vector, in which the rat LPH promoter demonstrated a greater transcriptional response than the human LPH promoter (Table 1). This suggests a preferential activation of the rat promoter by GATA-6 and may indicate differential affinities of the GATA factors for specific sequence in the binding sites, although this has not been specifically tested. A second difference was found with HNF-1 cotransfection, in which the human LPH promoter demonstrated a significantly greater activation response than the rat LPH promoter (Table 1), suggesting a preferential activation of the human LPH promoter by HNF-1 (although no difference in HNF-1 binding affinity was noted between the rat and human HNF-1 sites; Fig. 4B). Finally, the combination of HNF-1 plus Cdx-2 as well as the combination of all three expression vectors demonstrated antagonistic or additive responses for rLPH108, but strong synergistic responses for hLPH118 (Fig. 5B, discussed further below). Together, these data suggest that the position, configuration, and/or specific sequence within binding motifs of the LPH promoter may influence the magnitude and direction (activation vs. repression) of responses to specific combinations of transcription factors. Implications for the differential regulation of LPH gene expression between humans and other mammals are unknown at this time.

The basal promoter constructs (i.e., rLPH37 and hLPH37), with varying degrees of statistical significance, were activated by many of the transcription factors tested (Fig. 3). It is possible that sequence in the hGH reporter gene may activate reporter expression in response to specific expression vectors. However, the activity of BSGH after specific transcription factor stimulation was subtracted from that of all other constructs stimulated with the same transcription factor, thus eliminating activity due to the reporter itself. In a preliminary report (70), we showed that Cdx-2 is capable of binding and activating the rat and human LPH promoters at their TATA box regions. Thus Cdx-2 may directly activate these promoters at their TATA boxes, whereas GATA and HNF-1 factors, which do not bind to the 37-bp to 1-bp region (not shown), may activate transcription of a basal promoter indirectly (in contrast to direct protein-DNA interactions) or by direct interactions with the transcriptional initiation complex.

In the human LPH gene, we confirm that the two GATA motifs bind GATA-5 with similar affinities (Fig. 4A) and demonstrate that both sites play a role in the activation of the human LPH promoter (Fig. 10). Mutations introduced into either one of the sites had little effect on GATA-stimulated activity, but mutations introduced into both sites resulted in a significant reduction in GATA-5-stimulated transcriptional activity (Fig. 10). In a previous report, Fitzgerald et. al. (17) demonstrated that a mutation in the 5' GATA site of the human LPH promoter resulted in a decrease in GATA-6 activation from the wild-type control. The reason for the apparent difference between our study and that of Fitzgerald et. al. (17) is unknown, but could be due to differences in the ability of chicken GATA-6 compared with mouse GATA-5 to activate the human LPH promoter. In the human LPH gene, the 5' GATA site (hG₁) is on the forward strand whereas the 3' GATA site (hG₂) is on the reverse strand, which could result in possible differences in the ability of different GATA proteins to activate the LPH gene through the hG₂ site. We conclude from our data that GATA-5 is capable of activating the human LPH promoter through either GATA motif and that the two sites together do not activate the LPH gene any more than either of the sites individually.

Binding site affinity experiments suggest that GATA-5 prefers to bind WGATAT. Unbiased analysis of random GATA motifs revealed that the consensus for the GATA-1, -2, and -3 subfamily is WGATAR, where W is A or T and R is A or G, but clear preferences in core and adjacent sequences were noted (30, 45). A similar characterization of the GATA-4, -5, and -6 subfamily has not been reported. Comparison of the relative affinity of GATA-5 for each of the GATA motifs (Fig. 4A), as summarized in Fig. 11, reveals that GATA-5 binds equivalently if A or T (but not G) is in the 5' position and prefers T in the 3' position, although it continues to bind with lesser affinity if A or G are in this position. We therefore propose that the GATA-5 consensus is WGATD, where D is A, G, or T. 

View larger version (15K): [in this window] [in a new window]   Fig. 11.   Sequence analysis demonstrates that GATA-5 preferentially binds a AGATAT motif. Based on affinity experiments in Fig. 4A, the relative affinities of GATA-5 for each of the GATA motifs were assigned either ++ for strong competition, + for intermediate competition, or +/ for weak or lack of competition. The core motif (bold) and adjacent sequence are shown.

In cotransfection experiments, the combination of HNF-1 and Cdx-2 demonstrated synergistic activation of the human LPH promoter (hLPH118), but antagonistic interactions on the rat LPH promoter (Fig. 5B). Our data on the human LPH promoter are in agreement with a previous report by Mitchelmore et al. (47), who demonstrated that HNF-1 and Cdx-2 functionally interact, resulting in a synergistic activation of the pig LPH promoter. Although the reason for the apparent species difference in response to HNF-1 and Cdx-2 is not readily known, structural similarities and differences among the rat, human, and pig LPH promoters were noted. There is greater conservation of sequence identity in the first 100 bp of 5'-flanking region between the human and pig LPH genes (93%) than between the human and rat LPH genes (78%). This includes the HNF-1 binding motif, in which the 3' half is completely conserved between the human and pig sequence and more closely correlates with the HNF-1 consensus than that of the rat LPH HNF-1 (rH) site. Noteworthy, however, is that rH and hH demonstrate similar affinities for HNF-1 binding in competition EMSAs (Fig. 4). Sequence differences also include the TATA boxes, in which the human and pig LPH genes contain a true TATA motif, whereas the rat LPH gene contains a CATA motif, although the effect of such a difference on transcriptional regulation is unknown. The recent demonstration (47) of a second Cdx-2 binding site (called CE-LPH1c) 5' to the HNF-1 site in the pig LPH gene is unlikely to explain the species differences in response to these transcription factors, because CE-LPH1c is not present in the human LPH gene but is present in the rat LPH gene (102 to 97 bp). Because rLPH108 does not demonstrate synergistic activation by HNF-1 and Cdx-2, it is likely that the presence or absence of this site does not play a direct role in the cooperative activation by these two transcription factors. Finally, with regard to binding site position, the distance between the 5' base of the HNF-1 motif and that of the Cdx-2 binding site at CE-LPH1a is 33 bp for the human and pig promoters and 28 bp for the rat LPH promoter, representing a difference of approximately one-half turn of the helix, which could affect the ability of HNF-1 and Cdx-2 to interact. At present, the process underlying the cooperative interactions between HNF-1 and Cdx-2 is unknown, but the findings in the present study imply that the mechanism is more complex than simple protein-protein interactions and the presence of HNF-1 and Cdx-2 binding sites.

In the present report, we demonstrate for the first time functional cooperativity between GATA-5 and HNF-1. This cooperativity was strongest for the human LPH promoter but was present for the rat LPH and human SI promoters (Fig. 5B) as well. Functional cooperativity between zinc finger GATA proteins and homeobox transcription factors has been previously demonstrated (13, 37, 57) for GATA-4 and Nkx-2.5, a cardiac-restricted homeobox protein. GATA-4 was shown to physically associate with Nkx-2.5 and cooperatively activate cardiac-specific promoters. Furthermore, mutational analysis of each transcription factor indicated that physical interaction was required for functional synergy. Deletional analysis in GST pull-down assays revealed that the COOH-terminal zinc finger of GATA-4 was required to interact with Nkx-2.5, and helix III of the homeodomain of Nkx-2.5 was necessary to associate with GATA-4. Both regions are known to bind their cognate DNA binding sites, demonstrating a close association between protein-protein and protein-DNA interactions. Indeed, the homeodomain of HNF-1 was shown to be necessary for interactions with Cdx-2 (47). Based on these models, we hypothesize that the COOH-terminal zinc finger of GATA-5 and the homeodomain of HNF-1 physically interact and that this interaction is necessary for synergistic activation of the human LPH promoter.

Based on the findings in the present study, we hypothesize that the patterns of expression of certain intestinal genes are due, at least in part, to the simultaneous presence of GATA, HNF-1, and Cdx-2 transcription factors. Members of all three transcription factor families are present in the upper crypt and lower villus, which correspond to regions of LPH and SI transcriptional activation during intestinal differentiation. However, the net effect on intestinal gene expression during differentiation is far more complex than the simple presence of these transcription factors and their cognate binding sites. Data in the present study suggest that the number of binding sites and their relative positions on the promoter are important. Furthermore, as shown by transcription from longer promoter constructs (Fig. 3) or from the endogenous genes (Fig. 6), which do not always parallel expression of shorter promoter constructs, other interactions involving sequence further upstream may also influence transcriptional activation. Finally, interaction with other transcription factors, such as HOXC11 (46), as well as with specific cofactors, including friend of GATA-2 (FOG-2) (39, 62), dimerization cofactor for HNF-1 (DCoH) (44), and p300 (27) must be considered.

We conclude that the complex hierarchies of control that govern gene expression during intestinal differentiation and development are determined by the stoichiometry of transcription factors and cofactors within intestinal cells at any given time and the type, number, and arrangement of cis-acting elements in the regulatory regions of intestinal genes. LPH and SI, markers of intestine-specific gene expression and intestinal differentiation, are regulated by complex interactions involving multiple transcription factors. Synergistic activation by combinations of transcriptions factors is likely a mechanism by which specific patterns of tissue-specific expression might be attained by the overlapping expression of specific transcription factors. Characterizing specific interactions will provide insight into the regulatory mechanisms that control the pathways of intestinal differentiation.