Cell Type-specific Autoregulation of theCaudal-related Homeobox Gene Cdx-2/3 *

The caudal-related homeobox geneCdx-2/3 is a critical “master” control gene in embryogenesis. Mice heterozygous for a null mutation inCdx-2/3 exhibit multiple malfunctions including tail abnormalities, stunted growth, a homeotic shift in vertebrae, and the development of multiple intestinal adenomatous polyps, indicating thatCdx-2/3 is haplo-insufficient. In vitro studies have identified more than a half-dozen downstream target genes expressed in pancreatic and intestinal cells for this transcription factor. We have examined the transcriptional properties of the mouseCdx-2/3 promoter. This promoter could be autoregulated in pancreatic and intestinal cells that express endogenous Cdx-2/3. In contrast, Cdx-2/3 transfection represses the Cdx-2/3promoter in fibroblasts, which do not express endogenous Cdx-2/3. Since Cdx-2/3 activates proglucagon gene promoter in both pancreatic and intestinal cells and in fibroblasts, we suggest that some, yet to be identified, cell type-specific components are required for activating selected target gene promoters of Cdx-2/3, including the Cdx-2/3 promoter itself. Cdx-2/3 binds to the TATA box and another AT-rich motif, designated as DBS, within an evolutionarily conserved proximal element of the Cdx-2/3 promoter. The DBS motif is critical for the autoregulation, whereas the TATA box may act as an attenuating element for the autoregulatory loop. Finally, overexpression of Cdx-2/3 in a pancreatic cell line activated the expression of the endogenous Cdx-2/3. Taken together, our results indicate that the dose-dependent phenotype of Cdx-2/3 expression on its downstream targets in vivo could be regulated initially via a transcriptional network involving cell type-specific autoregulation of the Cdx-2/3 promoter.

identified in mammals. Cdx-2/3 maternal transcripts are expressed at pre-implantation stages and at the time of implantation in trophoectodermal cells. The embryonic expression is first seen at 8.5 days post-coitum (d.p.c.), and its expression increases in the endoderm just before the endoderm intestinal epithelial transition (ϳ15 d.p.c.). In adults, Cdx-2/3 is only expressed in differentiated intestinal epithelial cells, endocrine (L) cells, and in pancreatic A and B cells (10 -16). We and others (13)(14)(15)(16) have demonstrated that Cdx-2/3 is a critical transcriptional activator for proglucagon gene in pancreatic A cells and intestinal L cells. Cdx-2/3 homozygotic null mutant mice die between 3.5 and 5.5 d.p.c., whereas heterozygotic mutants show tail abnormalities, stunted growth, and a homeotic shift in vertebrae (9). In addition, 90% of the heterozygotic mutants develop multiple intestinal adenomatous polyps (9). These extraordinary multiple malfunctions observed in the heterozygotic mutant mice indicate that Cdx-2/3 transcriptional activity must be strongly dosage-dependent since a normal cellular phenotype appears to require expression of both alleles (i.e. Cdx-2/3 exhibits haplo-insufficiency).
In this paper, we have isolated the mouse Cdx-2/3 promoter and studied the transcriptional properties of this promoter. We found that although Cdx-2/3 is able to activate proglucagon gene promoter in both the Cdx-2/3-expressing pancreatic and intestinal cell lines, and the non-expressing fibroblasts, it activates its own promoter in a cell type-specific manner. Cdx-2/3 was shown to be able to bind to its TATA box and another AT-rich motif designated as DBS (downstream binding site) within an evolutionarily conserved proximal element of the Cdx-2/3 promoter. The DBS motif was found to be critical for Cdx-2/3 autoactivation, whereas the TATA box may act as an attenuating element for the autoregulatory loop.

EXPERIMENTAL PROCEDURES
Materials-Radioisotopes were obtained from Amersham Pharmacia Biotech. Oligonucleotides were provided by ACGT Co. (Toronto, Canada). Restriction enzymes and other DNA modification enzymes were molecular biology grade and were purchased from several sources.
GST-Cdx-2/3 Fusion Protein-The full-length hamster Cdx-2/3 cDNA was inserted into a pGEX plasmid (pGEX4T-2, Amersham Pharmacia Biotech). Following transforming a host Escherichia coli stain, the expression of GST-Cdx-2/3 fusion protein was induced with isopropyl-1-thio-␤-D-galactopyranoside. The protein was purified with a glutathione-Sepharose 4B column. The purity of the protein was verified by Coomassie Blue staining to be a single band and confirmed as Cdx-2/3 protein by Western blot analysis.
Cell Culture, Transient Transfection, and RNA Analysis-All the cell lines were grown in Dulbecco's modified Eagle's medium supplemented * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  with serum as described previously (13) or based on ATCC instruction. BHK, Rat-1, NIH3T3, InR1-G9, and Caco-2 cell lines were transfected by the calcium phosphate precipitation method (13). STC-1 cells were transfected by an electroporation procedure (13). Colo205 cells were transfected using LipofectAMINE. The cells were harvested for the LUC reporter gene analysis 14-16 h after the transfection. Methods for RNA extraction and Northern blot analysis were described previously (13).
Nuclear Protein Extraction and EMSA-Nuclear proteins from the cultivated cell lines were prepared as described by Schreiber et al. (18). Electrophoretic mobility shift assay (EMSA) was performed as described previously (13).
Western Blot Analysis-The rabbit polyclonal antibody against a carboxyl-terminal portion of hamster Cdx-2/3 (amino acids 260 -313) was prepared previously (16). For Western blot analysis, 5-10 g of nuclear proteins from cultivated cell lines were size-fractionated by SDS-polyacrylamide gel electrophoresis and transferred to a nitrocellulose membrane (Protran, Schleicher & Schuell). Cdx-2/3 immunoreactive protein was detected with an ECL Western blot analysis system (Amersham Pharmacia Biotech) with the peroxidase-linked anti-rabbit immunoglobulin as the second antibody.

RESULTS
The Structure of the Mouse Cdx-2/3 Promoter-We isolated the mouse Cdx-2/3 gene by screening a mouse phage genomic library using the hamster Cdx-2/3 cDNA as the probe (11). A 12-kb fragment carrying the mouse Cdx-2/3 gene including three exons, two introns, and the 5Ј-and 3Ј-flanking sequences was obtained. It contains five BamHI fragments. Results from partial DNA sequencing analyses and physical mapping indicate that the three internal BamHI fragments, measuring 7.2 kb, correspond to the mouse Cdx-2/3 gene (10). Fig. 1 shows the overall structure of the mouse Cdx-2/3 promoter and the DNA probes utilized for EMSA in this study.
Autoregulation of the Mouse Cdx-2/3 Promoter-A 906-base pair BamHI/SmaI fragment containing the DNA sequence from Ϫ769 to ϩ 137 of the mouse Cdx-2/3 gene was inserted into a promoterless plasmid pBluc (13), immediately adjacent to the coding sequence of the firefly luciferase (LUC) reporter gene. We first examined the transcriptional property of this promoter in three fibroblast cell lines, BHK ( Fig. 2A), Rat1, and NIH3T3 (data not shown). The mouse Cdx-2/3 promoter is very moderately active in fibroblasts, activating the LUC reporter gene only 6 -9-fold. Co-transfecting this promoter with Cdx-2/3 cDNA resulted in a repression of the reporter gene activity ( Fig. 2A). When transfected into the pancreatic cell line InR1-G9 ( Fig. 2A), the intestinal cell lines Caco-2 ( Fig. 2A), HT29, and Colo205, and the mouse intestine cell line STC-1 (data not shown), the promoter was more active, activating the LUC reporter gene 20 -35-fold. Co-transfection of this promoter with the Cdx-2/3 cDNA into the pancreatic and intestinal cell lines resulted in an activation of the reporter gene expression ( Fig. 2A). The activation is dose-dependent at lower dosages (0.5-2.5 g) of Cdx-2/3 transfection but is slightly dropped when 5 g of Cdx-2/3 was utilized. However, co-transfecting the Cdx-2/3 promoter with either an empty expression vector (pBAT7) or a Cdx-2/3 cDNA in the antisense orientation (AS) (13) shows no substantial effect on the reporter gene activity. These results would suggest that the expression of Cdx-2/3 is positively autoregulated in cells that express the endogenous Cdx-2/3. One may argue that the increased activity of Cdx-2/3 promoter in InR1-G9 and Caco-2 cells compared with that in the BHK fibroblasts could be due to a higher background activity of the pBluc, utilized in this study as a control, in the BHK cells. However, we have previously found that Cdx-2/3 is able to activate proglucagon gene promoter in both pancreatic and intestinal cells and in the BHK fibroblasts (13,15). To examine further the phenomenon of cell type-specific autoregulation, we repeated the transfection assay in both the InR1-G9 cells and the BHK fibroblasts against the proglucagon gene promoter and the Cdx-2/3 promoter. For each cell line, these two promoters were transfected in combination with a wider range of dosages of Cdx-2/3 cDNA. Because these two promoters were examined in the same batch experiment, only one set of pBluc served as the control for both promoters. Our results demonstrate that, at all the dosages utilized (0.05-10 g), Cdx-2/3 activates both the Cdx-2/3 and the proglucagon gene promoters in the InR1-G9 cells (data not shown). However, at all the dosages utilized, Cdx-2/3 represses Cdx-2/3 promoter but activates the proglucagon gene promoter in the BHK fibroblasts (Fig. 2B). About 4-fold activation was observed when 0.05 g of Cdx-2/3 cDNA was utilized, and the highest activation was achieved with 0.5 g of Cdx-2/3 cDNA (ϳ8-fold). When 5 and 10 g of Cdx-2/3 cDNA were used, the activation dropped to about 6-and 5-fold, respectively. Western blot analysis shows that comparable amounts of Cdx-2/3 were expressed in the BHK cells transfected with a given dosage of Cdx-2/3 in combination with either Cdx-2/3 or proglucagon promoters (data not shown). Therefore, we suggest that some as yet to be determined cell type-specific component(s) is/are required for activating selected target gene promoters of Cdx-2/3, including the Cdx-2/3 promoter itself.
To locate the cis elements critical for Cdx-2/3 autoregulation, a series of 5Ј deletion Cdx-2/3/LUC fusion genes were generated. Deletion of the DNA sequence from Ϫ769 to Ϫ51 base pairs had no substantial effect on Cdx-2/3 activation (Fig. 3A), indicating that the cis elements critical for the autoregulation are located mainly within the proximal region of the promoter. Fig. 3B shows that transfection of another homeobox gene, HOX11, into a fibroblast, a pancreatic, and an intestinal cell line had no appreciable activation or repression on the Cdx-2/3 promoter.

Cdx-2/3 Expressed by Pancreatic and Intestinal Cells Binds to the Proximal Element of the Cdx-2/3 Promoter-To examine
if Cdx-2/3 binds to its own promoter, EMSA was applied using the nucleotide probes listed in Fig. 1B. We first examined if Cdx-2/3 expressed in pancreatic and intestinal cell lines binds to the proximal element of the Cdx-2/3 promoter. Nuclear extracts from InR1-G9 cells (Fig. 4) and Caco-2 cells (data not shown) were examined using a probe designated as PE (Proximal Element, Fig. 1B), which contains the DNA sequence from Ϫ53 to ϩ 18 of the Cdx-2/3 promoter. We were unable to obtain an autoradiograph with well separated complexes. Furthermore, the addition of either preimmune serum or anti-Cdx-2/3 serum caused increased overall binding activity (lanes 6 -11). This is not unexpected because many different transcription factors may bind to the TATA box and the adjacent regions. However, when the anti-Cdx-2/3 antibody was added, supershifted complexes were observed (lanes 7 and 8, defined as Cdx/Ab), which are comparable to the supershifted complexes obtained using the GC sub-element of G 1 enhancer in the rat proglucagon gene as the probe (lanes 12 and 13; see Ref. 13). Cdx⅐Ab complexes were not observed when a preimmune serum was used (lanes 9 -11), and they were not observed when nuclear extract from BHK fibroblasts was utilized (data not shown). Taken together, we suggest that Cdx-2/3 expressed by FIG. 2. Cdx-2/3 activates its own promoter in a cell type-specific manner. A, Cdx-2/3 activates Cdx-2/3 promoter in pancreatic and intestinal cell lines but represses Cdx-2/3 promoter in fibroblasts. Five g of luciferase reporter (Rep) plasmid was transfected into an indicated cell line with different dosages of pBAT7.Cdx-2/3 (Cdx), empty expression vector (pBAT7), or pBAT7.Cdx-2/3.AS (AS, Cdx-2/3 inserted into pBAT7 in the antisense orientation, see Ref. 13). Total amount of cDNAs in each assay was adjusted to 5 g with the pBAT7 vector. V, pBluc as a promoterless LUC reporter gene. The data are expressed as mean relative luciferase activity (n ϭ 3) Ϯ S.E. normalized to the activity obtained after transfection of pBluc in the same experiment. B, 5 g indicated Cdx-2/3-Luc or Ϫ82 Glu-Luc (13) was transfected into the BHK fibroblasts with indicated amount of Cdx-2/3 cDNA. Total amount of cDNA in each transfection was adjusted to 10 g using the pBAT7 vector. The data are expressed as mean relative luciferase activity (n ϭ 3) Ϯ S.E. normalized to the activity obtained after transfection of pBluc in the same experiment. pancreatic and intestinal cell lines binds to the proximal region of its own promoter, although a specific complex that contains Cdx-2/3 cannot be defined.
Cdx-2/3 Binds to the TATA Box of the Cdx-2/3 Promoter-Cdx-2/3 is able to bind to the TATA box of the calbindin-D9k gene (19). To examine if Cdx-2/3 is also able to bind to its own TATA box, a probe designated as USE (Upstream sub-element of PE) was designed. Again, we were unable to obtain an autoradiograph with well separated complexes when this shorter probe was examined using the nuclear extracts from pancreatic and intestinal cell lines (data not shown). To circumvent this problem, a GST-Cdx-2/3 fusion protein was generated and used in the following assays.
Cdx-2/3 Binds to Another AT-rich Motif in the Proximal Promoter of Cdx-2/3-When the GST-Cdx-2/3 fusion protein was applied to the probe PE, two major complexes and one minor complex were observed, indicating that within the proximal element, there is/are additional binding site(s) for Cdx-2/3 (lane 2, Fig. 6A). The formation of all these three complexes could be supershifted by anti-Cdx-2/3 antibody (data not shown), competed by unlabeled PE (data not shown), or the GC element in the rat proglucagon gene promoter (lanes 9 -11), indicating that all three complexes represent the Cdx-2/3-binding events. One may suggest that C3 represents the monomer binding, and C2 represents the dimer binding, as observed in the binding of Cdx-2/3 to the SIF1 element of the sucraseisomaltase promoter (12). However, such an explanation is not plausible. The TATA probe (USE) only provides one complex, or one binding site. The overall structure of PE is different from that of SIF1, which contains an inverted repeat of ATAAA separated by only two nucleotides (12). In PE, the other binding site must be downstream of Ϫ15; and the TAAT motif from Ϫ6 to Ϫ3 is a good candidate (Fig. 1B). This TAAT motif is separated by 18 base pairs from the TATA box. Thus it is unlikely that a dimer molecule of Cdx-2/3 is able to contact these two well separated AT-rich regions. In addition, if C2 represents the binding of one PE molecule and two separated Cdx-2/3 molecules, reducing the amount of GST-Cdx-2/3 by adding unlabeled USE probe should first interfere with the formation of this slower migrating complex. However, by increasing unlabeled USE probe, we observed gradual disappearance of the faster migrating complex C3 (lanes [3][4][5]. We therefore propose that the TATA box is one binding site for Cdx-2/3, and C2 represents this binding event. A downstream motif designated as DBS is another binding site, and C3 represents this binding event. Although both C2 and C3 consists of one copy of PE and one Cdx-2/3 molecule, binding of Cdx-2/3 to different motifs within the PE may determine the migration positions of the complexes. The minor complex C1, which migrates the slowest, . Lanes 12 and 13 shows the results of EMSA using the same InR1-G9 nuclear extract against the GC subelement of G 1 enhancer in the rat proglucagon gene promoter (13). The major component in the complex C2 was determined to be Cdx-2/3, and the identities of C1, C3, and C4 remain to be explored. FP, free probe; Cdx/Ab, supershifted complex. may represent an event where one copy of PE is bound by two separated Cdx-2/3 molecules.
DBS Is Critical for Cdx-2/3 Autoregulation-Three Cdx-2/3-Luc fusion genes carrying the mutations in the Cdx-2/3-binding site(s) were generated. The fusion genes and the wild type counterpart were transfected into the InR1-G9 cells (Fig. 7) and Caco-2 cells (not shown) with or without Cdx-2/3 cDNA cotransfection. Mutation of the TATA box abolished the activity of the promoter completely. However, activation by Cdx-2/3 cDNA co-transfection was still observed on this mutant (Fig. 7, M TATA ). Mutation of the DBS motif reduced the basal activity of the promoter about 50%, indicating that this motif is important for Cdx-2/3 expression. In addition, mutation of this motif substantially abolished the Cdx-2/3 autoactivation (Fig. 7, M DBS ). M TATA/DBS carries mutations on both the TATA box and the DBS motif. The promoter activity of this mutant is lower than both the wild type (Ϫ50 to ϩ137) and M DBS , but higher than M TATA. Cdx-2/3 is able to activate this mutant promoter slightly higher than 2-fold. At the present time, we are unable to provide an appropriate explanation for this observation.
Cdx-2/3 Activates the Expression of the Endogenous Cdx-2/3 Gene-We demonstrated above that the Cdx-2/3 promoter could be activated by Cdx-2/3 cDNA in pancreatic and intestinal cells. Previously, we also obtained evidence indicating that in the pancreatic InR1-G9 cells, Cdx-2/3 transfection also activates the expression of endogenous Cdx-2/3 mRNA, although it was not appreciated at the time (15). We have generated several Cdx-2/3 overexpression clones after stable transfection of the InR1-G9 cells with Cdx-2/3 cDNA. In those clones, there were two Cdx-2/3 mRNA signals detected by Northern blot analysis (Fig. 8A). Based on their sizes, the signal below the 18 S rRNA (ϳ1.8 kb, Cdx-2/3 T) should represent the mRNA expressed from the transfected Cdx-2/3 cDNA, whereas the signal above the 18 S rRNA (ϳ2.3 kb, Cdx-2/3E) should repre-sent the overexpressed endogenous Cdx-2/3. Therefore, the data indicate that the expression of the transfected Cdx-2/3 up-regulates the expression of the endogenous Cdx-2/3. To extend further this observation to Cdx-2/3 protein level, we inserted the GST-Cdx-2/3 into the pcDNA3 expression vector. InR1-G9 cells were then transiently transfected with different dosages of pcDNA3-GST-Cdx-2/3. In this way, endogenous and exogenous Cdx-2/3 proteins could be differentiated with their sizes in Western blot analysis. As shown in Fig. 8B, the expression of the endogenous Cdx-2/3 was activated, in a dose-dependent manner, by GST-Cdx-2/3. Therefore, in the InR1-G9 cells, Cdx-2/3 not only activates a reporter gene driven by its promoter but also activates the expression of its own gene.
The most exciting observation with respect to Cdx-2/3 was made recently in Cdx-2/3 "knock out" mice (9). The Cdx-2/3 Ϫ/Ϫ embryos die between 3.5 and 5.5 d.p.c., suggesting that maternal transcripts of Cdx-2/3 are essential for implantation (9). Cdx-2/3 ϩ/Ϫ animals, however, show multiple malfunctions, including tail abnormalities, stunted growth, homeotic shift in vertebrae, and the development of multiple intestinal adenomatous polyps. The abnormalities in the Cdx-2/3 ϩ/Ϫ mice vary from one mouse to another, and no obvious abnormality was observed in their pancreas. These results would suggest that the Cdx-2/3 transcriptional activity is dosage-dependent or haplo-insufficient in selected tissues. In the Cdx-2/3 ϩ/Ϫ mutants, Cdx-2/3 is expressed in normal intestinal epithelia but not in the intestinal tumor cells. This is unlikely to be due to mutations or deletions of the remaining wild type copy of the Cdx-2/3 gene because loss of heterozygosity was not observed (9). Therefore the defect is at the expression level.
In this study we isolated the mouse Cdx-2/3 promoter and examined the transcriptional property of this promoter. We found that Cdx-2/3 transfection activates its own promoter in cell lines that express endogenous Cdx-2/3. Therefore, the expression of Cdx-2/3, like many other HD proteins, could be autoregulated (34 -37). However, in fibroblasts, which do not express Cdx-2/3, Cdx-2/3 transfection represses the expression of Cdx-2/3 promoter. This is in contrast to our previous observations in examining proglucagon gene activation where Cdx-2/3 is able to activate proglucagon gene promoter in both the pancreatic InR1-G9 cells (15) and the BHK fibroblasts (13). To investigate further whether Cdx-2/3 autoactivation, indeed, occurs in a cell type-specific manner, we examined the effect of Cdx-2/3 cDNA transfection over a very wide range of dosages on the expression of Cdx-2/3 and the proglucagon gene promoters in InR1-G9 cells and BHK fibroblasts. Our data clearly indicated that at all the dosages utilized, Cdx-2/3 activates both the Cdx-2/3 and the proglucagon gene promoters in InR1-G9 cells. In BHK fibroblasts, Cdx-2/3 transfection activates the proglucagon gene promoter but represses the Cdx-2/3 promoter. Therefore, we suggest that Cdx-2/3 requires some as yet to be determined cell type-specific component(s) to regulate selected downstream target genes, such as Cdx-2/3 itself.
It should be pointed out that although cell type-specific autoregulation of homeobox gene Cdx-2/3 has not been previously studied, cell type-specific activation by Cdx-2/3 on heterologous promoters was recently reported by Taylor et al. (25). Cdx-2/3 was shown, by Suh et al. (12), to activate sucrase-isomaltase promoter via binding to its SIF1 enhancer cassette in the intestinal cells. Taylor et al. (25), however, found that Cdx-2/3 also activated a thymidine kinase promoter fused with one or more copies of the SIF1 enhancer cassette when transfected into intestinal Caco-2 cells. Such an activation, however, was not observed when examined in the NIH3T3 fibroblasts. Furthermore, Taylor et al. (25) found that if the Cdx-2/3 activation domain was linked to the Gal4 DNA binding domain, the chimeric protein was able to activate Gal4 enhancer constructs in the intestinal Caco-2 cell line but not in the NIH3T3 fibroblasts.
Haplo-insufficiency is not common for transcription factors, especially for homeobox genes (34). One explanation is that related homeobox genes may exert overlapping or redundant biological functions. In addition, the expression of many homeobox genes, like Cdx-2/3, could be autoregulated (34 -37). Therefore, via a positive regulatory loop, the expression of a homeobox gene from only one functional allele is able to attain a physiologically required level. Other homeodomain proteins, such as Otx-2 and PAX6, do exhibit haplo-insufficiency for part of their biological functions (38,39). An interesting question, therefore, is why one functional allele for a few homeobox genes is insufficient even though the expression of these genes can be autoregulated? Information from this study has provided one plausible explanation. In Cdx-2/3, some as yet to be defined cell type-specific components (a co-factor, and/or a signal) is critical in determining if the autoregulation will be effectively processed or not. If in an individual cell, autoregulation process failed due to, for example, the combination of loss of one functional allele and lack of a necessary signal, malfunctions including the formation of tumors may occur. It is quite possible for the existence of a critical threshold for triggering the autoregulation. In the Cdx-2/3 ϩ/Ϫ mice, since only one Cdx-2/3 wild type copy exists, the chance for the expression of Cdx-2/3 within some cells not reaching the threshold will be higher. If this hypothesis is correct, it will provide an explanation for why in the Cdx-2/3 ϩ/Ϫ mice, the various malfunctions observed vary, in degree, time, and location, from one mouse to another. To explore this hypothesis, a systematic examination of the mechanisms controlling Cdx-2/3 expression, using both in vivo and in vitro approaches, becomes an essential task.
We identified two Cdx-2/3-binding sites, TATA box and DBS, within the proximal element (PE, Ϫ53 to ϩ18) of the mouse Cdx-2/3 promoter. DNA sequence from Ϫ53 to ϩ4, which covers both the TATA box and the DBS, is 100% conserved between mouse and human (10,40). It is interesting that the autoregulation of the pituitary-specific POU domain transcription factor Pit-1 is also mediated by two Pit-1-binding sites within the proximal region of the promoter (35). These two binding sites are separated by 60 nucleotides. Mutation of the 5Ј Pit-1-binding site (PitB1) abolished the positive autoregulation, whereas mutation of the other binding site that is located immediately 3Ј of the cap site (PitB2) markedly increased the expression of the pit-1 promoter. Therefore, PitB2 is considered as an attenuating component for the autoregulatory loop (35). In the Cdx-2/3 promoter, the overall organization of the two Cdx-2/3-binding sites is different from that of pit-1. The 3Ј DBS motif is critical for autoactivation, whereas the other binding site is the TATA box per se. It is not surprising that mutation of the TATA box almost completely abolished the promoter activity. However, Cdx-2/3-LUC fusion gene with a mutated TATA box still responds to Cdx-2/3 transfection. We suggest that in the Cdx-2/3 promoter, the TATA box could act as an attenuating component in the autoregulatory loop. It should be pointed out that both general transcription factors and Cdx-2/3 bind to the proximal element of the Cdx-2/3 promoter. Based on our observation, one may propose that at a low concentration, Cdx-2/3 will bind to DBS to accelerate transcription via physically interacting with one or more general transcription factors. At a high concentration, Cdx-2/3 may compete with general transcription factors for the TATA box, resulting in repressing the expression of the promoter.