The HOXC11 Homeodomain Protein Interacts with the Lactase-Phlorizin Hydrolase Promoter and Stimulates HNF1α-dependent Transcription*

The lactase-phlorizin hydrolase (LPH) gene is expressed specifically in the enterocytes of the small intestine. LPH levels are high in newborn mammals, but decrease after weaning. We have previously suggested that the promoter element CE-LPH1, located at −40 to −54, plays an important role in this down-regulation, because the DNA binding activity of a nuclear factor that binds to this site is present specifically in small intestinal extracts and is down-regulated after weaning. In an effort to clone CE-LPH1-binding factors, a yeast one-hybrid genetic selection was used, resulting in the isolation of a partial cDNA encoding the human homeodomain protein HOXC11. The full-length HOXC11 sequence was obtained by rapid amplification of cDNA ends. It was shown in a yeast assay and by electrophoretic mobility shift assay that HOXC11 binds to the CE-LPH1 element with similar specificity to the endogenous intestinal factor. Two HOXC11 transcript sizes were identified by Northern blot analysis. The larger transcript (2.1 kilobase pairs) is likely to contain a translational start site in good context and is present in HeLa cells. The shorter 1.7-kilobase pair transcript, present in HeLa and Caco-2 cells, probably encodes a protein lacking 114 amino acids at the N-terminal end. Both forms of HOXC11 potentiate transcriptional activation of the LPH promoter by HNF1α. The expression of HOXC11 mRNA in human fetal intestine suggests a role in early intestinal development.

Lactase-phlorizin hydrolase (LPH) 1 is a membrane-bound small intestinal enzyme that hydrolyzes the lactose present in milk. LPH is highly expressed in the newborn mammal, but is down-regulated after weaning. However, in some humans, mainly of Northern European descent, LPH activity remains high throughout adult life. In other humans, down-regulation of LPH leads to milk intolerance in adolescence and adult life, a clinical condition referred to as adult-type hypolactasia. Several studies have indicated that the down-regulation of LPH activity in mammals is mainly at a transcriptional level, although post-transcriptional mechanisms play a modifying role (1)(2)(3)(4). A 1-kb promoter region of the pig LPH gene is able to drive appropriate tissue-specific and developmental regulation of a linked reporter gene, suggesting that the cis elements required for transcriptional regulation are located within this region (3).
Previous work in our group has sought to identify the cis elements of the lactase gene promoter, and the corresponding DNA-binding factors. A factor present in nuclear extracts prepared from pig small intestine was shown to bind to the CE-LPH1 site at Ϫ40 to Ϫ54 in the promoter, adjacent to the putative TATA-box. This DNA binding activity, termed NF-LPH1, was found to be intestine-specific and was present at higher levels in newborn pigs than in adult animals (5). NF-LPH1 is also present in Caco-2 cells (6), a cell line derived from a human colon carcinoma, which express LPH when allowed to differentiate (7). By using competition electrophoretic mobility shift assays (EMSAs) and photoaffinity-labeling, it was suggested that NF-LPH1 in Caco-2 nuclear extracts is functionally related to a nuclear factor binding to the SIF1 element of the sucrase-isomaltase gene. Both factors recognize a TTTA(T/C) core sequence (6).
The SIF1-binding protein was recently cloned from mouse and identified as a homeodomain protein, Cdx-2 (8). The hamster homologue, termed Cdx-3 or shCdx-2, was previously isolated as a factor binding to an insulin gene promoter element (9). We have shown that Cdx-2 also binds to CE-LPH1 and activates LPH gene expression, dependent on a functional CE-LPH1 element, in co-transfection studies (10). The ability of Cdx-2 to bind to both the CE-LPH1 and SIF1 elements may be important in specifying intestinal expression of the corresponding genes. However, transcription factors in addition to Cdx-2 are probably required for the final developmental expression pattern of LPH and sucrase-isomaltase. Whereas LPH activity is high at birth and is down-regulated in most mammals after weaning, sucrase-isomaltase levels are low at birth but increase significantly after birth.
Previous EMSA analysis suggests that Cdx-2 is an abundant CE-LPH1-binding protein in differentiated Caco-2 cells. Formation of the main CE-LPH1-protein complex with Caco-2 nuclear extract was weakened by antibodies to Cdx-2, although no supershifted complex was generated (10,11). However, this does not preclude that other regulatory factors bind to CE-LPH1 under alternative binding conditions, or if a fractionated nuclear extract were used. The CE-LPH1 element contains two subregions which are both protected in DNase I-footprinting (5). Cdx-2 appears to bind to only the upstream TTTAC subregion. We have suggested that a repressor binds to the other half-site because a mutation here, which does not affect Cdx-2 binding, leads to higher transcriptional activity of a linked reporter gene (10).
A one-hybrid genetic screen in yeast (12) was carried out to identify cDNAs from Caco-2 cells, which encode factors able to bind specifically to the CE-LPH1 element. This approach allows for the detection of DNA binding activity in vivo under physiological conditions, in contrast to EMSA analysis. DNAbinding factors that have previously been isolated using a one-hybrid screen include Olf1, an olfactory-specific transcription factor (12), ORC6, a protein that recognizes the yeast origin of DNA replication (13), and a metal response elementbinding protein (14). Using this screen, a 1.65-kb cDNA was isolated, which was shown to contain the homeobox of the human HOXC11 protein, previously called HOX3H (15). We describe here the isolation of the full-length human HOXC11 sequence from HeLa cDNA using RACE (rapid amplification of cDNA ends). The relative affinity of HOXC11 protein for various DNA elements, the expression pattern of HOXC11 mRNA, and the ability of HOXC11 to activate transcription were analyzed.

EXPERIMENTAL PROCEDURES
Plasmid Construction-Standard molecular biology techniques were used (16), and sequencing of double-stranded DNA was carried out with ThermoSequenase (Amersham Pharmacia Biotech). All inserts generated by PCR (Advantage KlenTaq polymerase, CLONTECH) were sequenced completely to check polymerase fidelity. Oligonucleotides were obtained from Amersham Pharmacia Biotech.
The shCdx-2 insert of pBAT7Cdx-3 (kindly provided by M. German; Ref. 9) was subcloned into the SacI/EcoRI sites of pBluescript and used as template for PCR with the universal M13 primer and a primer annealing to the translational start site (5Ј-GGAAGCTTGTCGACCA-TGTACGTGAGCTACC-3Ј). The 1-kb PCR product obtained was cloned into the SalI/EcoRI site of pPC86 (12), resulting in p86Cdx. The SalI/A-paI insert of pF4 was subcloned into pBluescript and used as template for PCR with the universal M13 primer and a primer annealing to the first in-frame methionine codon 5Ј-TAGCGGCCGCCACCATGAAAAA-CG-3Ј. The 600-bp PCR product was cloned into the NotI/ApaI site of pRC-CMV (Invitrogen), resulting in pRC-HOX⌬N. In order to construct pRC-HOX, the 250-bp BglII/HincII fragment of the 92B1 plasmid (I.M.A.G.E. Consortium; Ref. 18) was first ligated in front of the Hinc-II/ApaI insert of pF4. The resulting 750-bp BglII/ApaI fragment was ligated into pRC-CMV, together with a 200-bp NotI/BglII-digested PCR fragment amplified from a RACE-generated plasmid with the primers 5Ј-TTAGCGGCCGCAACGATGTTTAACTCGGTC-3Ј and 5Ј-GGGTCG-ACGCTGTTCTTGTTGACTG-3Ј.
In order to construct the HOXC11 template for riboprobe synthesis, the 410-bp HincII fragment of the plasmid 92B1 was cloned into pBluescript. The resulting plasmid p92B410 was linearized with XhoI. For the LPH riboprobe template, the 380-bp PstI fragment of the plasmid HULAC (19) was cloned into pBluescript. The resulting plasmid pLPH380 was linearized with BamHI which removed a 100-bp fragment from the LPH insert.
The pGL3-LPH227WT plasmid was constructed by cloning the SstI/ XhoI fragment of pGL2-LPH227, 2 which contains the pig LPH promoter region from Ϫ227 to Ϫ17, upstream of the luciferase reporter gene in pGL3-basic (Promega). The mutations that were generated by PCR overlap extension changed the TTTA(T/C) core sequence of CE-LPH1a and -1c to TTGC(T/C) in the plasmid pGL3-LPH227mut1ac, and the additional mutation of CE-LPH2c from GTTAC to GTCCC in the plasmid pGL3-LPH227mut1ac,2c.
Preparation of the cDNA Library-Total RNA was prepared from differentiated Caco-2 cells (1 week after confluence) using the guanidinium thiocyanate/acid phenol method (20). Poly(A) ϩ mRNA was isolated from total RNA using oligo(dT) 25 Dynabeads (Dynal). The Superscript cDNA synthesis system (Life Technologies, Inc.) was used for the directional synthesis of cDNA containing SalI and NotI overhangs. The size-fractionated cDNA was ligated into pPC86 (12) and electroporated into the Escherichia coli strain DH10B (Life Technologies, Inc.). The bacterial transformants (6 ϫ 10 5 with an average insert size of 1 kb) were scraped from the plates into LB medium, and plasmid DNA was prepared (Qiagen).
Screening of the cDNA Library-Yeast containing the pRS-LPH plasmid were transformed with 6 g of library cDNA and plated onto 6 SD Ϫ His 14-cm plates at an average density of 1.7 ϫ 10 5 transformants/plate. After 4 days of incubation at 30°C, 24 colonies were able to grow. These were colony-purified, and plasmid DNA was recovered and individually re-transformed into reporter plasmid-containing yeast. Transformant colonies (at least three independent ones for each) were resuspended in TE buffer at approximately 1000 viable cells/l. 1 l of each was plated onto selective media plates (SD ϩ His or SD Ϫ His) and grown 2-4 days at 30°C.
Cloning the 5Ј End of HOXC11-In order to clone the 5Ј end of HOXC11 cDNA, a modified RACE protocol was carried out using Marathon-Ready HeLa cDNA (CLONTECH) according to the manufacturer's protocol except that Advantage Tth DNA polymerase (CLONTECH) was used. The gene-specific RACE primer used was F4PR1L (sequence 5Ј-GAAGAAACGGTCGAAGGCTTGAGGCAG-3Ј, see Fig. 3). The reaction mix was overlaid with oil and cycled in a Hybaid thermal cycler where the tube temperature was programmed for 1 cycle of 94°C for 1 min; 5 cycles of 94°C for 30 s and 72°C for 4 min; 35 cycles of 94°C for 30 s and 70°C for 4 min; and 1 cycle of 70°C for 5 min. A portion of the reaction was loaded onto an agarose gel and analyzed by Southern blotting with a 410-bp HincII-digested fragment of 92B1, which was radioactively labeled using the Rediprime kit (Amersham Pharmacia Biotech). The 550-bp RACE product thus identified was cut out and cloned (TA cloning kit, Invitrogen).
Cell Culture, Transfections, Preparation of Whole Cell Extracts, and EMSAs-Caco-2 and HeLa cells were grown in minimum essential medium (Life Technologies, Inc.) containing 10% calf serum. Caco-2 cells were seeded at 2 ϫ 10 6 cells/75-cm 2 flask, and were harvested after 4 days (80% confluent) and 14 days (1 week post-confluence). HeLa cells were transfected by calcium phosphate precipitation with either pRC-HOX⌬N or water (untransfected control) as described previously (10). Whole cell extracts were prepared 48 h later using the protocol of Schöler et al. (24). Protein concentration was determined using Bradford reagents (Sigma Diagnostics). EMSAs were carried out as described previously (10).
Transient transfections of Caco-2 cells by the calcium phosphate method were carried out as described previously (10). All transfections contained 1 g of a ␤-galactosidase expression plasmid as an internal standard for transfection efficiency (PCH110; Amersham Pharmacia Biotech) and 2.5 g of pGL3-LPH227 wild-type or mutated constructs. In addition, 1.25 g of pRC-CMV, pRC-HOX⌬N, pRC-HOX, and/or 0.63 g of RSV-HNF, made up to 2.5 g with pRC-CMV, were co-transfected.
Analysis of RNA Expression-Northern analysis was carried out according to standard techniques (25) using 2 g each of poly(A) ϩ RNA from HeLa cells and small intestine (CLONTECH) and undifferentiated (80% confluent) and differentiated (1 week post-confluent) Caco-2 cells. Poly(A) ϩ RNA was prepared from Caco-2 cells using the Fast-Track kit from Invitrogen. A 400-bp RsaI fragment from the untranslated 3Ј region of the HOXC11 cDNA was purified (Qiaex, Qiagen) and labeled to high specific activity by random priming with the Rediprime labeling kit (Amersham Pharmacia Biotech). Hybridization with 0.5 ϫ 10 6 cpm/ml Express Hyb solution (CLONTECH) was carried out according to the manufacturer's protocol.
For ribonuclease protection assay, the Promega T3/T7 polymerase kit was used to synthesize antisense riboprobes from linearized p92B410 and pLPH380 plasmid templates to high specific activity (1 ϫ 10 9 cpm/g). Both probes were gel-purified according to the protocol provided with the Ambion RNAPII kit. The cyclophilin control antisense riboprobe (Ambion) was labeled to low specific activity (7 ϫ 10 5 cpm/g) and purified by RNAID (Bio101). 1 g of yeast tRNA or poly(A) ϩ RNA, each supplemented with 10 g of yeast tRNA, was hybridized overnight at 50°C with 10 5 cpm HOXC11 riboprobe or with a mixture of 3 ϫ 10 5 cpm LPH riboprobe and 3000 cpm cyclophilin control riboprobe and analyzed after RNase digestion on a non-denaturing polyacrylamide gel, according to the Ambion protocol, together with radioactively labeled double-stranded DNA markers.
PCR with double-stranded cDNA templates was carried out with 10 pmol each of the following primer pairs: glyceraldehyde 3-phosphate dehydrogenase (G3PDH) (5Ј-GACCACAGTCCATGCCATCACT-3Ј and 5Ј-TCCACCACCCTGTTGCTGTAG-3Ј), LPH (5Ј-AGGATACACAGTTT-GGAGTGCGATG-3Ј and 5Ј-GTTGGCTTCGTTGTGTTTTCCCTTG-3Ј), or HOXC11 (5Ј-ACAAATCCCAGCTCGTCCGGTTCAG-3Ј and 5Ј-GCAGCAAGACATTGTCGCCGAGGTG-3Ј) and 0.5 ng of cDNA in each 50 l of reaction mix. The HeLa, fetal intestine (from 18 -28-week Caucasian fetuses), and small intestine (from a 15-year-old Caucasian) cDNAs were Marathon-Ready cDNAs from CLONTECH, whereas the differentiated Caco-2 cDNA was prepared from mRNA from Caco-2 cells 1 week after confluence using the two-hybrid cDNA synthesis kit from CLONTECH. A portion of each PCR reaction was analyzed on an agarose gel. PCR was also carried out under the above conditions on CLONTECH's human fetal multiple tissue cDNA panel. The HOXC11 PCR reactions were analyzed by Southern blotting with the end-labeled oligonucleotide 5Ј-GCGAGAGTCAGATAGAGGGTCCAG-3Ј.

RESULTS
Cloning of a CE-LPH1-binding Protein-In order to clone a factor capable of binding to the CE-LPH1 element of the LPH gene promoter, the CE-LPH1 element was multimerized and placed in front of a HIS3 reporter gene in the construct pRS-LPH. Yeast containing the pRS-LPH reporter plasmid were transformed with a cDNA library prepared from differentiated Caco-2 cells, in which cDNA-encoded proteins are expressed as a fusion (hybrid) with the activation domain of the yeast transcriptional activator GAL4 ( Fig. 2A). All yeast co-transformants grow on minimal medium supplemented with histidine (SD ϩ His); the ability to grow on medium lacking histidine (SD Ϫ His) correlates with the extent of HIS3 reporter gene activation by the GAL4 fusion protein. The plasmid pPC86, which lacks a cDNA insert, is unable to cause activation of the HIS3 gene, so there is no growth on SD Ϫ His (Fig. 2B).
Out of 1 ϫ 10 6 library transformants, 24 were able to grow in the absence of histidine. Two of these grew slowly on selective media, contained insert-less plasmid DNA, and were not studied further. An additional 20 clones encoded proteins that activated the reporter gene equally well in the presence (pRS-LPH) or absence of CE-LPH1 sites (pRS315HIS and p59.7). The p59.7 plasmid contains three copies of an unrelated se-FIG. 2. A, outline of the one-hybrid genetic screen in yeast. The Caco-2 cDNA library is expressed as a fusion protein with the yeast GAL4 activation domain. If the cDNA-encoded peptide binds to the CE-LPH1 elements in the pRS-LPH reporter plasmid, transcription of the HIS3 reporter gene is activated and the yeast can grow on minimal medium, which lacks histidine. B, yWAM2 yeast strains containing the indicated reporter plasmid were transformed with either pPC86, p86Cdx, or pF4 plasmid DNA. Approximately 1000 cells from each co-transformant were plated onto SD ϩ His or SD Ϫ His medium and grown for 2-4 days at 30°C. quence element, which binds an olfactory transcription factor, Olf-1 (12). The significant number of false positives are presumed to encode proteins that bind DNA nonspecifically.
The remaining two yeast positives, pF4 and pF6, did not activate from a reporter gene driven by heterologous Olf-1 binding sites (p59.7). In addition, only the two true positives (pF4 and pF6) were able to grow under increased stringency conditions, i.e. in the presence of 5 mM 3-aminotriazole (Sigma; data not shown). Both plasmids contained identical 1.65-kb cDNA inserts, as judged by restriction digest. The insert of pF4 was sequenced completely on both strands, and shown to contain an open reading frame of 194 amino acids including the homeodomain of human HOXC11, in-frame with the vectorencoded GAL4 activation domain (Fig. 3). The isolated pF4 plasmid DNA was re-transformed into yeast containing either the pRS-LPH reporter plasmid used in the library screen or other constructs designed to test the specificity of binding (Fig.  2B). For comparison, yeast were also transformed with the plasmid p86Cdx, which encodes a fusion protein of the activation domain of GAL4 with full-length shCdx-2 (9). Cdx-2 has previously been shown to bind to the CE-LPH1 element (10).
As shown in Fig. 2B, both the p86Cdx and pF4-encoded fusion proteins activate HIS3 reporter gene expression when driven by SIF1, CE-LPH1, or mut2 LPH1 elements, resulting in growth on SD Ϫ His plates. These elements also bind NF-LPH1 from Caco-2 cells (6). On the other hand, there is no activation, or reduced activation, from mut1 LPH1, mut3 LPH1, mut4 LPH1, and p59.7-driven reporter genes. The mut1 and mut4 LPH1 oligonucleotides do not bind NF-LPH1 from Caco-2 cells, whereas mut3 LPH1 binds weakly, as shown by competition EMSA analysis (6). Although the parental pRS315HIS reporter plasmid lacks CE-LPH1 elements, there is some activation of the HIS3 gene by p86Cdx and pF4, resulting in weak growth on minimal medium lacking histidine (Fig. 2B). This is probably due to the presence of a CE-LPH1like sequence just upstream of the polylinker. A doublestranded oligonucleotide covering this vector sequence, referred to as BSK-17 in Fig. 1, was able to compete for HOXC11 binding to the CE-LPH1-24 probe in EMSA analysis (Fig. 5,  lane 9).
Isolation of Full-length HOXC11 DNA-Hox genes in man and mouse are organized in four chromosomal clusters, which have probably arisen during evolution by gene duplications (reviewed in Ref. 26). It was expected that HOXC11 would show greatest sequence homology with the human HOXA11 and HOXD11 paralogues, which are located at a similar position on the A and D chromosomal clusters; there is no corresponding transcript on the B cluster (27). A putative sequence entry for the N terminus of human HOXC11 was identified by a data base search with the N-terminal sequence of mouse Hoxa11. The corresponding plasmid 92B1 (accession number Z63886) was acquired from the I.M.A.G.E. Consortium (18) and sequenced. The 92B1 genomic insert contains the first 351 nucleotides of the pF4 cDNA insert flanked on the 5Ј side by 325 additional nucleotides and on the 3Ј side by divergent, presumably intronic, sequence (Fig. 3). The presumed intron precedes the homeobox, and corresponds in its location to the intron present in the mouse Hoxa11 and Hoxd11 genes (Fig. 4).
In order to obtain the complete 5Ј nucleotide sequence of HOXC11, RACE was carried out using HeLa cDNA. HeLa cDNA was chosen because it gave a strong PCR product in PCR analysis using HOXC11-specific primers (Fig. 9). The RACE fragment was cloned, and five independent clones were sequenced. The sequence of the longest RACE product is shown in Fig. 3 and is identical to the genomic sequence obtained with the Genome Walker kit from CLONTECH (data not shown). The first methionine of the longest cDNA is in a good translational context (28) and is preceded in the genomic sequence by an in-frame stop codon (data not shown). The deduced fulllength cDNA is 2.0 kb and includes an open reading frame, which encodes a protein with 304 amino acids.
As shown in Fig. 4A, there are scattered regions of homology between human HOXC11 and mouse Hoxa11 and Hoxd11 outside the homeodomain. A glycine/alanine-rich sequence (38 -45%) is also present in the three proteins just before the homeodomain. Similar regions in transcription factors have been implicated in both transcriptional repression (29) and activation (30). The homeodomain and C-terminal extension of human HOXC11 are identical to that of mouse Hoxc11 (31) and are also highly conserved with the mouse paralogous group proteins, Hoxa11 and Hoxd11 (32, 33) (Fig. 4B). Conservation between the homeodomains of Drosophila Abdominal-B and mammalian HOXC11, Hoxa11, and Hoxd11 is especially evident in the recognition helix 3 and N-terminal arm, both of which contact DNA.
Characterization of HOXC11 DNA Binding Activity-In order to analyze HOXC11 DNA binding activity, expression of the pF4 cDNA insert was put under the control of a cytomegalovirus promoter. The resulting plasmid pRC-HOX⌬N, encoding the C-terminal 190 amino acids of HOXC11, was transfected into HeLa cells, which do not contain an endogenous CE-LPH1 binding activity (10). No complex is formed using whole-cell extract from untransfected HeLa cells in EMSA analysis (Fig.  5, lane 1), whereas a specific complex is formed with extract from transfected HeLa cells (Fig. 5, lane 2). HOXC11 is expected to bind to similar DNA sequences as the related Drosophila Abdominal-B protein. Therefore, the binding affinity of HOXC11 to CE-LPH1 and to an optimized Abdominal-B binding site was compared. The CE-LPH1-protein complex formed with extract from pRC-HOX⌬N transfected HeLa cells is competed to a similar extent by cold competitor CE-LPH1-17 (Fig.  5, lanes 3-5) and ABD-B-17 (Fig. 5, lanes 6 -8), a related probe which has three base changes to include the optimal binding site for Abdominal-B (34). An unspecific oligonucleotide did not compete (Fig. 5, lane 10).
The ability of elements present in the yeast reporter plasmids shown in Fig. 2B to bind HOXC11 was analyzed by competition using either CE-LPH1-24 (Fig. 6A) or SIF1-24 (Fig. 6B) as probe in EMSA analysis. The binding affinity of HOXC11 appears to be highest for SIF1-24 and CE-LPH1-24 elements, intermediate for CE-LPH1-17 and mut2 LPH elements, and low for mut1 LPH, mut3 LPH, mut4 LPH, and unspecific elements. This result is consistent with EMSAs carried out with the endogenous NF-LPH1 present in Caco-2 nuclear extracts (6).
Pattern of HOXC11 Expression-We used Northern blot analysis to examine the expression and size of the HOXC11 mRNA transcript (Fig. 7). A probe from the 3Ј-untranslated region of HOXC11 was used to compare levels of HOXC11 expression in HeLa cells, Caco-2 cells in their undifferentiated (80% confluent) and differentiated (1 week post-confluent) state, and small intestine. One-week post-confluent Caco-2 cells have started to express markers of intestinal differentiation, including LPH (7) (Figs. 8 and 9).
HeLa cells contain a substantial amount of a 2.1-kb transcript that hybridizes to the HOXC11 probe. This correlates well in size with the full-length HOXC11 cDNA determined by RACE on HeLa cDNA (2.0 kb, Fig. 3). An additional transcript of 1.7 kb, corresponding to that reported for a hematopoietic cell line (35), is found in HeLa and also in Caco-2 cells but not in the small intestine from a 15-year old (Fig. 7, top panel). The amount of this transcript increases upon differentiation of Caco-2 cells, as judged by hybridization of the same filter with an actin probe (Fig. 7, lower panel).
The same mRNAs used in Northern blot analysis were analyzed by ribonuclease protection assay to determine the 5Ј end of HOXC11 mRNA (Fig. 8). An antisense riboprobe correspond- ing to the 5Ј end of the 92B1 genomic insert gave a protected RNA band in HeLa cells only. For unclear reasons, this protected band appears shorter than the 410 bp expected if the mRNA in HeLa cells corresponded exactly to the riboprobe sequence. There was no protected band corresponding to the 1.7-kb transcript observed in Northern blot analysis, probably because of the small expected size (90 bp), which may not hybridize under the conditions used here. The ribonuclease protection analysis also showed that differentiated (one week post-confluent) Caco-2 cells express LPH, although not to as high levels as small intestine (Fig. 8, compare lanes 11 and 12,  band marked LPH). The cyclophilin probe (Cyc in Fig. 8) hy-bridizes to the ubiquitously expressed cyclophilin RNA.
PCR primers were designed for LPH and HOXC11, which annealed on either side of an intron in the genomic sequence. These were used in qualitative PCR using cDNA templates derived from HeLa cells, human fetal intestine (18 -28 weeks), human small intestine (15 years), and differentiated Caco-2 cells (Fig. 9A). The results confirmed the results from the Northern and ribonuclease protection analyses, and show additionally that human fetal intestine, which does not yet express LPH RNA, expresses higher levels of HOXC11 than small intestine from a 15-year-old Caucasian in which LPH expression is high (Fig. 9A, compare lanes 6 and 7 with lanes 10 and  11). In the human fetal multiple tissue cDNA panel (CLON-TECH), expression of HOXC11 was observed in fetal kidney, skeletal muscle, and small intestine, whereas it was absent in all the other fetal tissues tested (Fig. 9B).
HOXC11 Stimulates HNF1␣-dependent Transcription from the LPH Promoter-We have previously demonstrated that the transcription factor HNF1␣ binds to three sites in the 894-bp upstream flanking region of the LPH gene. 2 The proximal 227 bp of the LPH gene promoter is sufficient to drive differentiation- FIG. 7. Northern blot analysis of HOXC11 expression. The indicated poly(A) ϩ RNA (2 g each) were separated on an agarose/formaldehyde gel. The positions of the 18 S and 28 S ribosomal particles were determined by loading 2 g of total RNA in separate lanes on the same gel. The gel was blotted onto a Hybond N ϩ membrane. In the top panel, the membrane was hybridized with a DNA fragment from the 3Јuntranslated region of HOXC11; the positions of the strongest hybridizing bands are indicated. Caco-2 UD refers to undifferentiated (80%confluent) Caco-2 cells, Caco-2 Diff. refers to differentiated (1 week post-confluent) Caco-2 cells. In the lower panel, the membrane was hybridized with an actin probe. The membrane was exposed to Biomax Kodak MS film at Ϫ70°C for 3 days (top panel) or 2 h (lower panel).

FIG. 8. Analysis of HOXC11 and LPH expression by ribonuclease protection.
In lanes 1-6, 1 g of yeast transfer RNA (tRNA) or the indicated poly(A) ϩ RNA was hybridized overnight with a riboprobe corresponding to the 5Ј region of HOXC11, subjected to RNase digestion, and loaded onto a non-denaturing acrylamide gel. In lanes 7-12, 1 g each of the indicated RNA was hybridized overnight with a mixture of two riboprobes corresponding to the 280-bp BamHI/PstI fragment from the LPH gene and the 103-bp fragment of the cyclophilin gene. The positions of the main protected double-stranded RNAs are indicated by arrows. The bands marked with asterisks (*) are present in the tRNA control lanes and probably correspond to undigested probe. The position of dsDNA marker bands are indicated. The dried gel was exposed to Biomax Kodak MS film at Ϫ70°C for 2 days. FIG. 9. A, PCR analysis of HOXC11 and LPH expression. Each of the indicated cDNAs was subjected to PCR using the indicated primer set. The PCR was carried out in a Robocycler (Stratagene) for 1 min at 94°C, followed by 5 cycles of 30 s at 94°C and 3 min at 72°C, followed by the indicated number of cycles of 30 s at 94°C and 3 min at 70°C. The G3PDH primers amplify a 450-bp fragment of the human G3PDH housekeeping gene. The LPH and HOXC11 primers amplify a 400-bp and a 600-bp fragment, respectively, from regions of the LPH and HOXC11 cDNAs which contain an intron in the genomic sequence. The PCR primers were designed to bind to regions of HOXC11, which are poorly conserved in other HOX gene products. M, 100-bp marker. B, PCR analysis of HOXC11 in human fetal tissues. Each of the indicated cDNAs was subjected to PCR as above, using G3PDH or HOXC11 primer sets. In the bottom panel, a Southern blot of the PCR reactions generated with the HOXC11 primer set was probed with a radioactively labeled HOXC11-specific oligonucleotide. F., fetal. dependent transcription of a reporter gene in differentiated Caco-2 cells, to a level 80% of that seen with the 894-bp LPH promoter construct. 2 Thus, the LPH promoter region from Ϫ227 to Ϫ17, which contains one HNF1␣ binding site CE-LPH2c, and two potential HOXC11 binding sites, CE-LPH1a and -1c, was used in co-transfection studies. Luciferase activity of the pGL3-LPH227WT construct alone in undifferentiated Caco-2 cells is low (1.5 times the level observed with pGL3basic). As shown in Fig. 10, overexpression of the short or long form of HOXC11 gives up to 2.7-fold activation of pGL3-LPH227WT, but no activation of pGL3-LPH227mut1ac, in which potential HOXC11 binding sites have been mutated (compare lanes 1-3 with lanes 7-9). Whereas HNF1␣ alone activates the wild-type LPH promoter 2-fold, co-transfection of pRC-HOX⌬N or pRC-HOX produces a further 7-fold or 19-fold stimulation of transcription, respectively. In the case of pGL3-LPH227mut1ac, a lower stimulatory effect is observed (4-fold and 10-fold, respectively). The stimulation of HNF1␣-dependent transcription by HOXC11 is much reduced for pGL3-LPH227mut1ac,2c in which the potential binding sites for HNF1 and HOXC11 are mutated. The ability of the introduced mutations to abrogate factor binding was confirmed in EMSAs (data not shown).

DISCUSSION
HOXC11 and Cdx-2 Bind to Proximal Regulatory Elements in LPH and Sucrase-Isomaltase (SI)-We have previously dem-onstrated that 1 kb of the lactase gene promoter is sufficient to direct small intestine-specific expression and post-weaning decline in transgenic mice (3). We have focused our attention on a proximal promoter element, CE-LPH1, that is bound by an intestine-specific factor which co-varies with lactase expression (5). Deletion of the CE-LPH1 element in stably transfected Caco-2 cells leads to a 64% reduction in LPH promoter activity in differentiated cells. 2 Using a yeast genetic screen to identify factors capable of binding to CE-LPH1, the HOXC11 cDNA was isolated twice in a screen of 10 6 clones. This suggests that HOXC11 is more abundantly expressed in differentiated Caco-2 cells and/or binds the CE-LPH1 element with higher affinity, compared with other DNA-binding proteins encoded in the cDNA library. However, it should be noted that the cDNA used in the library construction was subjected to a complete digestion with NotI, resulting in the probable exclusion of some cDNAs.
Previously, a factor that binds to the related SIF1 element in the gene for intestinal SI was isolated and identified as the homeodomain protein Cdx-2 (8). In our present work, we have shown that both HOXC11 and Cdx-2, expressed from a singlecopy plasmid in yeast cells, are able to bind to the CE-LPH1 and SIF1 elements under in vivo conditions, in a manner dependent on an intact TTTA(T/C) sequence. Mutation of bases within this 5-nucleotide site, as in mut1 LPH1 and mut4 LPH1, reduces binding of HOXC11 both in the yeast assay and in EMSA analysis. Although binding of HOXC11 to mut3 LPH1 is intermediate between that of mut1 LPH1 and mut2 LPH1, binding of the factors to mut3 LPH1 in yeast does not lead to growth on minimal media, due perhaps to the all-or-none nature of the yeast screen.
HOXC11 and Cdx-2 are both homeodomain proteins. The homeodomain is a 60-amino acid stretch, which contains a helix-turn-helix motif that is involved in DNA recognition. Many homeodomain proteins bind preferentially to a TAAT core sequence. HOXC11 is one of 15 mammalian Hox genes, which are related to the Drosophila homeotic gene for Abdominal-B. The Abd-B-like Hox proteins share several distinguishing features; they are expressed most posteriorly, their homeodomains recognize a divergent TTAT sequence (34), they do not contain a pentapeptide motif conserved in other Hox genes, and they differ in their ability to interact with Pbx proteins (36,37). We have demonstrated similar affinity of HOXC11 for an optimal Abd-B sequence as for the CE-LPH1 element, which contains a related core sequence TTAC. Two residues in the N-terminal arm of the Abd-B homeodomain, K6 and P7, appear to play an essential role in the divergent specificity of Abd-B compared with other Drosophila homeoproteins (34). The Nterminal arm of HOXC11 conserves one of these critical residues, the proline at position 7.
It has recently been demonstrated by binding site selection that the Abd-B-like Hox proteins Hoxb-9, Hoxa-11, Hoxd-12, and Hox-13 bind preferentially to a TTTAC-containing sequence (37). However, not much is presently known about the natural binding sites for mammalian Abd-B-type Hox proteins. The few cases where target genes have been identified involve auto-and cross-regulatory interactions between different Hox genes and their gene products. In the case of the Abd-B-type proteins, HOXD9 and HOXD10 (previously called HOX4C and HOX4D, respectively) have been shown to trans-activate the gene for HOXC5 in cotransfection studies, mediated by a 5Ј-AATTTATGA-3Ј sequence in the HOXC5 promoter (38). In the paralogous mouse Hoxa5 and Hoxb5 gene promoters, the corresponding element resembles the CE-LPH1 element: 5Ј-AATTTACGAC-3Ј. Thus, Abd-B-like Hox proteins recognize a FIG. 10. HOXC11 stimulates HNF1␣-dependent activation of the LPH promoter. Subconfluent Caco-2 cells were transfected with pGL3-LPH227WT, which contains the wild-type LPH promoter from position Ϫ227 to Ϫ17 in front of the luciferase gene, with or without the HNF1␣ expression plasmid RSV-HNF. The LPH227mut1ac construct is mutated in CE-LPH1a and -1c, and the LPH227mut1ac,2c construct has an additional mutation in the HNF1␣-binding site, CE-LPH2c. The effect of co-transfecting expression plasmids for the short or long form of HOXC11 was tested. Luciferase activity was corrected for variation in transfection efficiency and is presented as -fold activation relative to pGL3-LPH227WT alone. The means Ϯ S.D. were calculated from three experiments.
TTTA(T/C) core sequence, which is likely to be physiologically relevant.
There have been previous reports of intestinal factors other than Cdx-2 which bind to CE-LPH1 or SIF1 elements (20, 29, 39 -41). For example, Traber et al. (41) detected a SIF1-binding protein in a fractionated Colo-DM extract that had a lower molecular weight than Cdx-2. We suggest that HOXC11 may account, at least in part, for these DNA binding activities. It is also possible that the other Abd-like Hox proteins bind to CE-LPH1 and SIF1 elements.
HOXC11 mRNA Exists in Two Forms-Caco-2 cells contain a 1.7-kb form of HOXC11 mRNA, which corresponds in size to the mRNA present in hematopoietic cells (35) and to the insert of the Caco-2 cDNA library plasmid, pF4. HeLa cells contain, in addition to the 1.7-kb mRNA, a 2.1-kb form of HOXC11 mRNA. The full-length cDNA determined by RACE on HeLa cell cDNA was 2.0 kb. The cDNA templates for RACE may not extend all the way to the 5Ј end of the mRNA, due to premature pausing of the reverse transcriptase and/or because of the removal of bases from the 5Ј end by T4 DNA polymerase. However, the longest RACE sequence is preceded in the genomic sequence by a stop codon, so the first methionine in Fig. 3, which is in a good translational context, is probably used in HeLa cells. Although HeLa cells are likely to contain the full-length 304-amino acid HOXC11 protein, it is not clear why there is no detectable CE-LPH1 binding activity in untransfected HeLa cell extracts. It is possible that the endogenous factor is unable to bind DNA due to a HeLa-specific modification. Alternatively, the conditions used in the EMSA analysis or the presence of nonspecific DNA-binding proteins in the crude HeLa cell extract may interfere with binding of the endogenous HOXC11 protein to CE-LPH1 DNA.
RACE carried out on Marathon-Ready human 18 -28-week fetal intestinal cDNA suggests that the longer form of HOXC11 mRNA is also present in fetal intestine (data not shown). The PCR data presented in Fig. 9 show that expression of HOXC11 in human fetal tissues is limited in its distribution, being only present in fetal intestine, kidney, and skeletal muscle.
The short form of HOXC11 is expressed in differentiated Caco-2 cells, which also express LPH. Both the LPH and SI proximal promoters contain important regulatory binding sites for HNF1, a transcriptional activator that has previously been demonstrated to be important in hepatic and intestinal gene regulation (42)(43)(44). Although Caco2 cells contain endogenous HNF1␣, transfected pGL3-LPH227 on its own gives a very low luciferase activity in Caco2 cells, indicating the lack of an essential transcription factor and/or the presence of a repressor in the undifferentiated cells. We have shown here that both short and long forms of HOXC11 give only weak activation of a LPH-luciferase construct on their own, but greatly enhance transcriptional activation when co-transfected with an HNF1␣ expression plasmid. This HNF1␣-dependent stimulation requires an intact HNF1␣-binding site. The bifunctional protein DCoH is also known to enhance the transcriptional activity of HNF1␣. Although DCoH does not itself contact DNA, it stabilizes HNF1␣ dimers and HNF1␣/DNA complexes (45,46). The observed stimulatory effect of HOXC11 on HNF1␣ may be due to a similar stabilizing effect on HNF1␣ DNA binding and/or an increase in transcriptional activity of HNF1␣. Although there is a 2-fold drop in stimulation by HOXC11 and HNF1␣ when the HOXC11-binding sites CE-LPH1a and -1c are mutated, it is not clear from our experiments whether HOXC11 binds to DNA in order to stimulate HNF1␣ activity. HOXC11 may be tethered to DNA through an interaction with HNF1␣, or it may bind to low affinity DNA sites in the presence of HNF1␣. Future experiments will address whether HOXC11 and HNF1 interact directly with each other, and whether they co-operate in DNA binding.
It is possible that the intestinal specific expression of LPH is dependent, at least in part, on the overlap in tissue distribution of both HNF1␣ and HOXC11. The presence of HOXC11 in tissues outside the intestine, such as HeLa cells, fetal kidney, and fetal skeletal muscle, suggests that it has additional functions. Homeobox genes typically play an important role in tissue development; for example, pdx-1 (pancreatic and duodenal homeobox gene 1) is essential for pancreatic development in mice (47). Diverse mammalian Abd-B-type HOX genes have an important role in limb and urogenital development (32, 48 -50). The expression of HOXC11 in human fetal intestine suggests that HOXC11 plays an important role in intestinal development.