The human A33 antigen is a transmembrane glycoprotein and a novel member of the immunoglobulin superfamily

Abstract

The mAb A33 detects a membrane antigen that is expressed in normal human colonic and small bowel epithelium and >95% of human colon cancers. It is absent from most other human tissues and tumor types. The murine A33 mAb has been shown to target colon cancer in clinical trials, and the therapeutic potential of a humanized antibody is currently being evaluated. Using detergent extracts of the human colon carcinoma cell lines LIM1215 and SW1222, in which the antigen is highly expressed, the molecule was purified, yielding a 43-kDa protein. The N-terminal sequence was determined and further internal peptide sequence obtained following enzymatic cleavage. Degenerate primers were used in PCRs to produce a probe to screen a LIM1215 cDNA library, yielding clones that enabled us to deduce the complete amino acid sequence of the A33 antigen and express the protein. The available data bases have been searched and reveal no overall sequence similarities with known proteins. Based on a hydrophilicity plot, the A33 protein has three distinct structural domains: an extracellular region of 213 amino acids (which, by sequence alignment of conserved residues, contains two putative immunoglobulin-like domains), a single hydrophobic transmembrane domain, and a highly polar intracellular tail containing four consecutive cysteine residues. These data indicate that the A33 antigen is a novel cell surface receptor or cell adhesion molecule in the immunoglobulin superfamily.

Colorectal cancer is among the most common malignancies of the Western world and is a leading cause of cancer deaths (1). Because of the high resistance of microdisseminated colorectal carcinoma cells to conventional therapies, new treatment methods are needed. Targeted antibody-based therapy is an active area of clinical investigation. The production of less immunogenic, humanized, or chimeric antibodies to reduce the immune reaction in patients, allowing antibodies to be repeatedly administered, and the identification of more suitable antigenic targets offer promising new avenues for exploration (2). One potentially useful target for colon cancer is the surface molecule that interacts with the mAb A33 (3–5). The expression of this antigen is highly restricted to the epithelial cells of the human small and large intestines and 95% of human colorectal tumors (6). Upon mAb binding to the A33 antigen, the antibody–antigen complex is internalized and sequestered in vesicles (7), a property that may underly the marked retention of radioactivity in tumors for up to 6 weeks after administration of ¹²⁵I- or ¹³¹I-labeled A33 (3–5). Other factors contributing to its therapeutic potential are the high number of A33 binding sites per cell (5, 8) and the absence of detectable circulating A33 antigen. Preclinical evaluation of A33 mAb as a therapeutic reagent has been explored in a nude mouse model using xenografts of the human colorectal carcinoma cell line, SW1222. These transplanted tumors undergo regression after treatment with either ¹²⁵I- or ¹³¹I-labeled A33 (9). In phase I and II therapy trials, ¹³¹I- and ¹²⁵I-labeled murine A33 mAb were shown to have anti-tumor effects without bowel toxicity (4, 5). In view of these encouraging data, the mAb A33 was genetically humanized (8) and phase I trials commenced.

Using multidimensional high-resolution chromatography and biosensor detection, the A33 antigen was purified from Triton X-114 extracts of the human colon carcinoma cell line, LIM1215 (10). This protocol yielded sufficient quantities of homogenous 40- to 45-kDa protein to permit N-terminal sequence analysis and identification of two tryptic peptides (10). The delineation of 33 N-terminal residues and subsequent analysis of available data bases suggested that the A33 antigen was a novel cell surface protein (10). Subsequently, affinity purification of the A33 antigen from SW1222 cells extended the N-terminal sequence characterization to 40 residues and identified an additional 101 residues of amino acid sequence from enzymatically cleaved peptide fragments (G.R., R.L.M., H.J., L.C., E.N., B.C., J.K.H., S.J.W., S.W., A.W.B., L.J.O., and R.J.S., unpublished data). The extended N-terminal amino acid sequence was used to design degenerate oligonucleotide primers for use in PCRs with a cDNA library constructed from LIM1215 poly(A)⁺ RNA. One of the amplified products encoded the N-terminal sequence of the A33 antigen and was used to isolate A33 antigen cDNAs from the LIM1215 cDNA library. In this paper we present the cDNA and deduced amino acid sequence of the A33 antigen, demonstrate expression of the protein on the surface of Cos cells, and predict some of its structural characteristics. The A33 antigen sequence suggests a role for the A33 antigen in cell–cell recognition and signaling.

EXPERIMENTAL PROCEDURES

Cloning of A33 Antigen cDNA.

A LIM1215 cDNA library was custom-synthesized in the λZAPII expression vector by CLONTECH using poly(A)⁺ RNA prepared by us from confluent LIM1215 cells by two rounds of enrichment on columns of oligo(dT) cellulose. A cDNA probe encoding the A33 antigen N-terminal sequence was generated using PCR. Antisense primers were designed to hybridize to the carboxyl region of the N-terminal sequence obtained by G.R., et al., (unpublished data). Specifically, the six pools of 17-mer antisense oligonucleotides, each with 8-fold degeneracy, corresponding to amino acid residues 34–39 (LIQWDK) were as follows: primer 1477, 5′-A(AG)(CT)TT(AG)TCCCACTGAAT-3′; primer 1478, 5′-A(AG)(CT)TT(AG)TCCCATTGAAT-3′; primer 1479, 5′-A(AG)(CT)TT(AG)TCCCACTGGAT-3′; primer 1480, 5′-A(AG)(CT)TT(AG)TCCCATTGGAT-3′; primer 5915, 5′-A(AG)(CT)TT(AG)TCCCACTGTAT-3′; and primer 5916, 5′-A(AG)(CT)TT(AG)TCCCATTGTAT-3′.

These were paired with sense primers designed to hybridize to a primer sequence (KS; 5′-TCGAGGTCGACGGTATC) present in the backbone of the λZAPII vector. A 0.3-kb product obtained with primers KS and 1478 in a “touchdown” PCR (11) with an initial primer-annealing temperature of 55°C was gel-purified, sequenced, and found to encode a portion of the A33 N-terminal protein sequence. Precise primers for this A33 antigen cDNA sequence were then used (sense primer, 5′-CCTGTCTGGAGGCTGCCAGT; antisense primer, 5′-AGGTGCAGGGCAGGGTGACA) to amplify a 189-bp PCR product that was radiolabeled with [α-³²P]ATP and [α-³²P]CTP [both at 3000 Ci/mM (1 Ci = 37 GBq); Bresatec, Adelaide, Australia] to a specific activity of >10⁷ dpm/μg DNA using a Megaprime DNA labeling kit (Amersham). Clones (0.8 × 10⁶) were screened according to the Stratagene instruction manual for λZAPII, and 16 positive cDNA clones, ranging in size from 0.4 to 2.8 kb, were automatically excised from the λZAPII vector in the Bluescript plasmid, pBS SK(±), using the Lambda Zap Automatic Excision Process (Stratagene). Both strands of five independent clones were sequenced using an Applied Biosytems automated DNA sequencer. Initially, DNA sequence data were obtained using primers directed to the pBS backbone (KS, SK, T3, T7); the internal DNA sequence was then established by the specific primer-directed method. To establish the relationship of the A33 antigen sequence to known DNA sequences, four data bases [EMBL, GenBank, DDBJ (DNA Data Bank of Japan), and dbEST (data base of expressed sequence tags)] were searched using the data base similarity search algorithms, blast and fasta (12, 13).

Northern Blot Analysis.

For studies of A33 antigen mRNA expression, specimens of colorectal carcinoma (obtained during surgical resection) and confluent layers of colon carcinoma-derived cell lines (LIM1215, LIM1863, LIM1899, LIM2099, LIM2405, and LIM2437; ref. 14) were directly solubilized in denaturing solution (4 M guanidinium isothiocyanate/0.5% Sarkosyl NL30 [BDH]/25 mM sodium citrate/0.1 M 2-mercaptoethanol), and total RNA was prepared according to the method of Chomczynski and Sacchi (15). Samples of normal human colon, counterparts of the tumor samples above, and inflamed colon (from patients with Crohn disease) were first enriched for epithelial cells by incubating in PBS containing 3 mM EDTA and 0.5 mM DTT to release the crypts from the underlying stroma (16) prior to solubilization in denaturing solution. Samples (20 μg) of total cytoplasmic RNA were electrophoresed in 0.4 M formaldehyde and 1% agarose gels and transferred overnight by capillary action to nylon filters (Hybond N; Amersham) and immobilized by exposure to UV light. Filters were prehybridized in a buffer containing 50% formamide, 5× Denhardt’s solution (0.1% Ficoll/0.1% polyvinylpyrrolidone/0.1% BSA), 5× SSPE (1× SSPE = 0.15 M NaCl/0.01 M NaH₂PO₄·2H₂O/1.2 mM EDTA, pH 7.4), 0.5% SDS, and 1% (wt/vol) skimmed milk powder for at least 6 h at 42°C. The filters were then incubated at 42°C overnight in fresh hybridization solution with a 2.6-kb cDNA clone of A33 antigen, gel-purified, and labeled with [α-³²P]ATP (3000 Ci/mM; Bresatec) using a Megaprime DNA labeling kit. Filters were washed in 2× SSPE, 0.1% SDS followed by 1× SSPE, and 0.1% SDS at 42°C, and signals were visualized by autoradiography. To permit quantification of the mRNA signals, the filters were reprobed with an oligonucleotide specific for 18 S rRNA (5′-CGGCATGTATTAGCTCTAGAATTACCACAG), labeled with dATP[γ-³²P] (3000 Ci/mmol; Bresatec) using T4 polynucleotide kinase.

Cos Cell Expression.

A 2.6-kb putative A33 antigen cDNA clone was excised from the pBS (SK±) plasmid using EcoRI and subcloned in the sense orientation into the mammalian expression vector, pcDNA3 (InVitrogen, Leek, The Netherlands). Cos cells were seeded into 15-cm Petri dishes (Nunc) to achieve 50% confluency 24 hr later. The cells were transfected over a 4-hr incubation at 37°C with 15 μg of either pcDNA3/A33 or pcDNA3 (parental vector) using DEAE-dextran in the presence of chloroquine (17). This was followed by dimethyl sulfoxide (DMSO) shock (10% DMSO in PBS) for 90 sec; the cells were then returned to RPMI 1640 medium containing 10% fetal calf serum (FCS), 2 mM glutamine, and 50 μg/ml gentamycin for 3–5 days. The cells were harvested and analyzed for A33 antigen expression by Western blot analysis, flow cytometry, and immunocytochemistry.

Western Blot Analysis.

Transfected cells were solubilized for 30 min at 4°C with 1% (vol/vol) Triton X-100 in 15 mM Tris·HCl (pH 7.4) containing 1 mM phenylmethylsulfonyl fluoride, 1 mM pepstatin, 0.1 mM leupeptin, and 0.01 units per ml of aprotinin. The resulting extracts were centrifuged twice at 4°C for 20 min at 14,000 × g, and 2 μl aliquots were electrophoresed under nonreducing conditions in 8–25% SDS/PAGE Phastgels (Pharmacia) before they were transfered to poly(vinylidene difluoride) membranes and incubated with humanized mAb A33 (2 μg/ml). A33 signals were detected with anti-human IgG conjugated with horseradish peroxidase and visualized by enhanced chemiluminescence (Amersham).

Flow Cytometry and Immunochemistry.

Transfected cells were detached using 10 mM EDTA in PBS for 10 min at 37°C and aspirated gently to produce a single cell suspension. They were then pelleted by centrifugation at 1500 rpm for 5 min and resuspended in 0.5 ml PBS containing 10 mM EDTA and 5% FCS and kept on ice for all remaining procedures to prevent internalization of antigen–antibody complexes. Murine A33 mAb was added to a final concentration of 20 μg/ml for 30 min. The cells were washed twice in PBS/EDTA/FCS and incubated on ice for a further 30 min in 0.5 ml PBS/EDTA/FCS containing fluorescein isothiocyanate-conjugated sheep anti-mouse IgG (Silenus, Hawthorn, Australia) diluted 1:50. Cells were then washed twice more and resuspended in 1 ml PBS/EDTA/FCS for fluorescence-activated cell sorting (FACS) analysis in a Becton Dickinson FACScan and cytospin preparation. Cytospins were prepared in a Shandon cytospin 2 centrifuge (Shandon, Pittsburgh), allowed to air-dry, mounted in glycerol containing antifade, and examined using a Nikon Fluorophot microscope.

RESULTS AND DISCUSSION

Sixteen overlapping putative A33 antigen cDNA clones were obtained, the longest of which (clone 18) was 2793 bp prior to the addition of the poly(A)⁺ tail. Fig. 1 shows the nucleotide sequence and deduced amino acid sequence of clone 18. The longest open reading frame in the cDNA sequence predicts a protein of 319 amino acids. Beginning at amino acid 22 of the predicted translation product, 40 contiguous residues are identical to the established amino-terminal sequence of native A33 antigen (ref. 10; G.R., et al., unpublished data). The predicted sequence also contains regions (hatched boxes) identical with the amino acid sequences of several internal peptides released from the native molecule by enzymatic digestion (Fig. 1).

Figure 1.

cDNA sequence and deduced amino acid sequence of the longest cDNA clone (clone 18; 2.8 kb) encoding the human A33 antigen. The longest open reading frame, encompassing nucleotides 345-1301, contains the known N-terminal sequence of the native A33 antigen and predicts a protein of 319 amino acids. The stop codon at 1302–1304 is boxed, and the amino acid sequences of the internal peptides identified by digestion of the native molecule are shown (shaded areas). A putative signal sequence (bold underline), three potential N-linked glycosylation sites (overline), and a transmembrane domain (second bold underline) are indicated. Adjacent to the transmembrane domain, four consecutive cysteine residues are observed. The spans of the two putative Ig-like domains are enclosed by square brackets with the specific residues conserved in Ig superfamily members shown in circles. Other features of the DNA sequence include a tandem repeat of 25 bp in the 5′ untranslated region (bold overline) and a polyadenylylation signal (AATAAA) 11 bp upstream from the poly(A) tail. The asterisk above the C at position 294 denotes the fact that a C was found in this position in 2 out of 5 independent clones sequenced (including clone 18) and an A in the 3 other clones.

The predicted translation product of the human A33 antigen mRNA is initiated at the ATG positioned 345 bp from the 5′ end of clone 18. This ATG leads off the longest open reading frame and is in a favorable context for initiation of translation by reference to the Kozak consensus sequence, GCC(A/G)CCATGG (18). Following this, a sequence encoding 21 amino acids resembles a hydrophobic signal peptide. The putative cleavage site between alanine and isoleucine, which is required to produce the mature protein with the N-terminal sequence of the native molecule, is consistent with the (−3, −1) rule for signal peptide cleavage (19). The position of the first in-frame stop codon predicts a mature polypeptide chain comprising 298 amino acid residues, M_r 33,276. This M_r is not inconsistent with data demonstrating that the native A33 antigen is a glycoprotein of approximate M_r 40,000–45,000 because the sequence contains three potential N-linked glycosylation sites, one of which (N⁹¹) was strongly suspected from the initial sequence analysis of the peptide fragments. Each of these could be predicted to accommodate an average oligosaccharide chain length of 2.5–3 kDa. Indeed, in experiments where enzymes were used to remove sialic acid and both N- and O-linked glycosides from the A33 antigen polypeptide, there was a reduction of approximately 8000 in the M_r of the A33 antigen (G.R., et al., unpublished data). Based on a Kyte–Doolittle hydrophilicity plot (20) of the sequence (Fig. 2), the molecule appears to have three structural domains: an extracellular region of 213 amino acids containing 6 cysteine residues, a hydrophobic transmembrane domain of 23 amino acids, and a highly polar intracellular tail of 62 amino acids (Fig. 1). Searches of available DNA and protein data bases revealed no direct sequence similarities with any known protein. However, relatively short spans of human A33 antigen nucleotide sequence could be matched with expressed sequence tags derived from the human colonic epithelial-derived cell line, T-84 (92% identity with GenBank accession no. AA055862; length, 344 nt) and the murine teratocarcinoma-derived cell line, F9 (74% identity with EMBL accession no. D28657D28657; length, 249 nt).

Figure 2.

Kyte–Doolittle hydrophilicity plot of the deduced amino acid sequence of the A33 antigen. Several structural features are indicated: the presence of an N-terminal hydrophobic region likely to correspond to a signal sequence, a more distal highly hydrophobic sequence consistent with the presence of a single-span transmembrane domain, and a highly polar C-terminal region consistent with an intracellular location for this part of the molecule.

Manual inspection of the extracellular domain revealed the presence of several residues (circled in Fig. 1) that are conserved in members of the Ig superfamily (21, 22). Specifically, we noted a V-type Ig-like domain at the N terminus, in which the presence of a disulfide bond between the two conserved cysteines (C²², C⁹⁶) was predicted from the microsequence analysis of the peptide fragments generated by enzymatic digestion (Fig. 1). The V-type domain was followed by an Ig-like domain of the C₂-type (22). This combination of a V-type and a C₂-type domain is characteristic of the CD2 subgroup of Ig superfamily members (23). Sequence alignment to proteins in this subgroup with known three-dimensional structure revealed that the A33 antigen shares the highest sequence identity with the D1D2 fragment of human CD4 (16% over 208 residues; ref. 22). The V-type domain is most similar to the VL domains of antibodies (up to 22% sequence identity over 117 residues) and the C₂-type domain to the D4 domain of rat CD4 (18% over 85 residues). The overall similarity with the CD2 group of molecules suggests that the A33 antigen may participate in cell–cell recognition processes, perhaps further implying the existence of a soluble or cell-associated binding partner. The identification of non-antibody binding partners could be addressed using the same biosensor-based technology used recently to identify and purify both the A33 antigen (10) and the ligand for the orphan receptor, HEK (24).

The N terminus of the predicted intracellular region of 62 amino acids begins with four consecutive cysteine residues. Data base analysis has identified 14 mature proteins containing the CCCC motif. Interestingly, three of these, G protein-coupled receptor 3 (25), endothelial-1 receptor (26), and the tachykinin-like peptide receptor (27), are members of the seven transmembrane G protein-coupled receptor family in which palmitoylation of the cysteine residues near the carboxyl terminus has been implicated in receptor coupling to G proteins and in the down-regulation of receptor activity by influencing receptor removal from the cell surface (28). Thus the presence of a CCCC motif adjacent to the transmembrane domain of the A33 antigen may indicate further membrane tethering of the molecule via palmitoylation, and this could play a role in the trafficking of the protein to vesicles.

To verify that the clones we had isolated encoded the A33 antigen, we expressed a 2.6-kb cDNA transiently in Cos cells. The cells were then assayed for A33 antigen expression by Western blot analysis, flow cytometry (Fig. 3 A and B), and immunocytochemistry (data not shown). Only the Cos cells transfected with the expression vector containing a putative A33 antigen cDNA clone produced a protein that was recognized by the mAb A33. Furthermore, the Western blot analysis demonstrated that the recombinant A33 protein expressed by Cos cells was approximately the same M_r (40,000–45,000) as endogenously produced A33 antigen in LIM1215 cells. As expected, FACScan and immunocytochemical analysis of Cos cells transfected with the A33 antigen construct indicated that a large proportion of the A33 antigen was displayed on the cell surface.

Figure 3.

Expression of recombinant A33 antigen by transfected Cos cells. Cos cells were transfected either with the parental vector, pcDNA3, or with pcDNA3 into which a 2.6-kb A33 antigen cDNA had been subcloned. Cells were harvested in this experiment 5 days after transfection and subjected to Western blot analysis (A) and flow cytometry (B). (A) Transfected cells were solubilized in Triton X-100 and electrophoresed into SDS/polyacrylamide gels without reduction and processed for Western blot analysis as described. The signal obtained with Cos cells transfected with pcDNA3 containing A33 antigen cDNA (lane 1) corresponds to approximately the same M_r (43,000) as the signal obtained with LIM1215 cells expressing high endogenous levels of A33 antigen (lane 3). Cells transfected with the parental vector gave no signal (lane 2). The positions of blue prestained standards (lane M) (NOVEX, San Diego) are indicated on the right. (B) FACScan profiles of control and A33 antigen-expressing Cos cells. The profile obtained with A33 antigen-expressing Cos cells (shown in bold) has been superimposed on the profile obtained with Cos cells transfected with pcDNA3 alone to show the shift to the right in fluorescence intensity of cells expressing A33 antigen.

Northern blot analysis demonstrated a strong A33 antigen signal (2.8 kb) in total RNA prepared from A33 antigen positive human colon carcinoma-derived cell lines (LIM1215, LIM1899, and LIM1863). No signal was obtained with total RNA from A33 antigen negative cell lines (LIM2099, LIM2405, and LIM2537; Fig. 4A). Strong expression of A33 antigen mRNA in purified epithelial cells from normal human colon was always observed (Fig. 4B). In comparison, the A33 antigen mRNA signals obtained with RNA extracted from the adjacent tumor tissue were consistently weaker. The Northern blot analysis shown in Fig. 4B has been extended to more than 20 paired samples of normal and transformed colonic tissue with the same results. One explanation for this could be the higher fibroblast content of the tumor samples compared with the normal colonic crypt preparations, which are essentially pure epithelial cells.

Figure 4.

Northern blot analysis of A33 antigen mRNA, indicated by arrows on the left, in (A) cell lines derived from human colorectal carcinoma and (B) normal and diseased human colonic tissue. (A) The six colorectal carcinoma cell lines had been previously analyzed for A33 antigen expression using immunocytochemistry and flow cytometry (unpublished data), and half (LIM1215, LIM1863, and LIM1899) were found to be positive. The remaining half (LIM2099, LIM2405, and LIM2537) were negative. The pattern of A33 antigen mRNA expression shown is consistent with the protein expression data. (B) A33 antigen mRNA expression in samples of normal and diseased human colorectal tissue obtained from patients during surgical resection. The samples are in pairs except for lanes 1–3, which contain RNA extracted from the normal colon, adenomatous polyp, and tumor of the same patient, respectively. Thus, lanes 1, 4, 6, and 8 contain RNA extracted from the crypts of normal colonic mucosa, and the strong hybridization signals correspond to relatively high expression of A33 antigen compared with that in the corresponding tumor preparations (lanes 3, 5, and 7) and the RNA extracted from the large polyp (lane 2), which all produced weaker A33 antigen mRNA signals. Lane 9 contains RNA extracted from the inflamed colonic crypts of a patient with Crohn disease, which produces an A33 antigen signal similar to that of its normal counterpart (lane 8). Lane 10 contains RNA from human peripheral blood buffy coat cells, and lane 11, RNA from LIM1215 cells (positive control). For both blots, the lower panel shows the pattern obtained with a γ-³²P-labeled oligonucleotide probe designed to hybridize to 18 S rRNA. Arrows on the right indicate the positions of RNA markers.

Immunohistochemical studies (6) have demonstrated that the expression of the A33 antigen is essentially restricted to normal intestinal epithelium and 95% of colorectal tumors. Although several other mAbs capable of recognizing determinants on colon cancer cells exist (29–35), none of them matches the restricted tissue specificity of the A33 mAb. Ep-CAM (epithelial cell adhesion molecule) is a cell adhesion molecule expressed by most simple epithelia, and its expression is maintained by a large proportion of colorectal, breast, lung, pancreatic, and gastric tumors (29). Several mAbs to this tumor marker have been used in clinical trials (for example, see ref. 30) and recently chimeric and single-chain versions of two of them, 17–1A and 323/A3, were developed and evaluated in cell-killing experiments (31, 32). Similarly, members of the carcinoembryonic antigen family (33, 34) and the pancarcinoma antigen, TAG-72 (35), have attracted attention as suitable targets for immunotherapy of colorectal, breast, and lung tumors. Though their relatively wide expression offers the opportunity to develop antibodies that can target multiple tumors, experience suggests that optimal localization of antibody to tumors may be impeded by the presence of antigen in a variety of normal tissues as well as by the presence of shed antigen in the circulation. In this respect, it might also have been expected that the strong expression of the A33 antigen in normal colon would have compromised its usefulness as a target in immunotherapy. However, these concerns have been allayed by phase I/II studies that have shown prolonged accumulation of ¹³¹I- and ¹²⁵I-labeled A33 at sites of tumor metastasis for up to 6 weeks after administration, while normal colon was antibody-free by 7–8 days and minimal toxicity to the colonic mucosa was observed (4, 5). The rapid clearance of the antibody from normal colon is not fully understood, but it may be due to rapid transcytosis of radiolabeled A33 mAb from colonocytes (36) and/or the relatively rapid shedding of colonocytes from the top of the crypts (37).

A basic understanding of the mechanisms underlying the tightly regulated, tissue-specific nature of A33 antigen expression as well as insights into its biological function are eagerly awaited. Such insights are likely to advance our understanding of cell–cell interactions and differentiation in the colon. In particular, knowledge of the characteristics of the A33 antigen gene promoter that confers such tissue-specific expression should allow us to develop transgenic mouse models of colon cancer, which could cast light on the role of certain oncogenes and tumor suppressor genes in the tumorigenic process. Indeed, it is conceivable that the tissue specificity of the A33 antigen promoter could also be exploited to target therapeutic genes to the colon.

Clearly the A33 antigen is an exciting target for immunotherapeutic approaches to colon cancer, and our identification of the A33 antigen with its receptor-like structure points toward a new signaling system in the colon. The availability of the A33 antigen should facilitate the development of new immunotherapeutic and ligand-directed approaches to the treatment of metastatic colon cancer.