Monoclonal antibodies for carcinoembryonic antigen and related antigens as a model system: a systematic approach for the determination of epitope specificities of monoclonal antibodies

A systematic approach for the determination of epitope specificities of monoclonal antibodies to a complex antigen system is described. After initial screening to identify antigen-binding monoclonal antibodies, one or more of the clones are isolated by limiting dilution cloning, grown in ascites, and the resulting antibodies secreted into the ascitic fluid are affinity purified on Sepharose-bound protein A, radiolabeled, and cross-compared with antibodies from other clones by a solid-phase competitive immunoassay. In this work, BALB/c mice were immunized with either purified carcinoembryonic antigen (CEA) or the CEA-producing cell line HC 84S. Spleen cells were fused with the mouse myeloma cell line Sp2/0-Ag14. The supernatants from 25 hybrids showed a significant binding of 125I-CEA (greater than or equal to 15%). Nine hybrids were cloned, resulting in 33 different clones. The antibodies produced by the different cloned hybrids and the remaining uncloned hybrids recognized a total of five different epitopes on CEA. All of the epitopes reside on the protein moiety of the molecule as determined by antibody binding to deglycosylated CEA. The monoclonal antibodies with five different epitope specificities were reacted with tissue sections of normal and cancerous tissues and with peripheral blood smears. Each of the five monoclonal antibodies reacted with tissue sections from colonic, gastric, lung, and mammary carcinomas, as well as from a benign colonic polyp and a resection margin from a colonic carcinoma. Four monoclonals reacted with normal liver tissue. Granulocytes in peripheral blood smears bound three antibodies strongly and one antibody weakly, and one antibody was not bound. One monoclonal antibody that reacted with normal liver tissue was not bound by granulocytes. The ability of these five monoclonal antibodies to differentially detect three different CEA-related antigens in normal and malignant tissues may have clinical utility.     

Mixed gel precipitation in the analysis of monoclonal antibodies against determinants of the carcinoembryonic antigen

Mixed gel precipitation technique was used in the study of 5 monoclonal antibodies (MA) to carcinoembryonic antigen (CEA). The method is based on specific inclusion of MA into the precipitate formed by the antigen and polyclonal antibodies to it. The test-system for the determination of nonspecific cross-reacting antigen (NCA) has demonstrated that MA 2D10 are directed to the determinant, common for CEA and NCA, whereas MA 3C12 and 2G10 give no reaction with NCA. The specificity of MA 192 and 35 correlated with the previously established one. Mixed gel precipitation technique is recommended for primary screening and study of MA specificity to any precipitating antigen.     

Immunologic characterization and molecular profile of carcinoembryonic antigen detected by monoclonal antibodies

Four distinct monoclonal antibodies, which reacted with CEA preparations but not with nonspecific cross-reacting antigen or with nonspecific cross-reacting antigen 2, were established. Except for monoclonal antibody AS001 , all of these monoclonal antibodies immunoprecipitated molecular forms of 200K and 180K daltons that are not bridged by disulfide bonds. Immunodepletion experiments and sodium dodecyl sulfate polyacrylamide gel electrophoresis analysis revealed that these monoclonal antibodies recognized the same antigenic structure when 125I-CEA preparation was used. Monoclonal antibody AS001 is of particular interest, because this antibody reacted only with a 200K dalton molecule which is a part of the molecules recognized by the other three monoclonal antibodies. The rosette inhibition assay and the immunoprecipitation experiments suggest that each monoclonal antibody recognizes a different antigenic determinant. The antigenic determinants recognized by monoclonal antibodies YK013 and AS001 may be peptides in nature, whereas the determinants recognized by antibodies YK024 or AS005 might be carbohydrate. The radioimmunoassay with monoclonal antibody AS001 was established, and the results clearly indicate that the incidence of positivity for the sera from digestive tract cancer patients and from lung cancer patients obtained by monoclonal antibody AS001 was higher than that obtained by the polyclonal antibody. Monoclonal antibody AS001 was able to detect the corresponding antigen in the sera, which the polyclonal antibody failed to detect. This study therefore suggests that monoclonal antibodies may enhance and improve the diagnostic value in cancer patients with undetectable or lower CEA levels detected by conventional anti-CEA antibodies.     

Monoclonal antibodies for carcinoembryonic antigen (CEA) as a model system: identification of two novel CEA-related antigens in meconium and colorectal carcinoma tissue by Western blots and differential immunoaffinity chromatography

In a previous study, five monoclonal antibodies against the carcinoembryonic antigen (CEA) with different epitope specificities were delineated. One of these antibodies which exhibits a high affinity for CEA binds to different carcinoma tissues, to liver tissue, and to granulocytes. This antibody was selected for the immunoaffinity purification of CEA and related antigens from colorectal carcinoma tissue, from spleen tissues, from bile, and from meconium. After elution from the immunosorbent, the antigens were separated by SDS-PAGE, were transferred to nitrocellulose, and were incubated with the five different antibodies. Antibody T84.1 bound to the following antigens: 177 kD and 128 kD from colonic carcinoma, 81 kD from bile, 49 kD from spleen, as well as 165 kD and 100 kD from meconium. Two additional antibodies showed a similar binding pattern. The fourth antibody (CEA.11) bound to the 165 kD meconium antigen and to the two colorectal carcinoma antigens. The fifth antibody (T84.66) showed a strong reaction with the 177 kD colorectal carcinoma antigen and a faint reaction with a 183 kD antigen in meconium. As judged from m.w. and immunochemical properties, the 128 kD colorectal carcinoma antigen and the 100 kD meconium antigen are two novel CEA-related antigens. Because antibody CEA.11 did not bind to the 100 kD meconium antigen in Western blots, the 165 kD antigen could be eluted from a CEA.11 immunosorbent without contamination by the 100 kD antigen. Similarly, as predicted from the binding pattern in the Western blots, the two colorectal carcinoma antigens were separated from each other by a T84.66 immunosorbent.     

Sequence and glycosylation site identity of two distinct glycoforms of nonspecific cross-reacting antigen as demonstrated by sequence analysis and fast atom bombardment mass spectrometry

Nonspecific cross-reacting antigen (NCA) is a highly glycosylated membrane protein which is immunologically and structurally related to carcinoembryonic antigen, an important tumor-associated antigen. Two glycoforms of NCA were purified from a single liver metastasis of a colonic carcinoma and characterized with respect to their primary sequence and position of glycosylation sites. The two glycoforms (designated TEX (tumor-extracted antigen), Mr 75,000, and NCA, Mr 45,000) each showed a deglycosylated Mr of 35,000 and yielded identical peptide maps. The structural characterization of TEX and NCA and the assignment of glycosylation sites was performed by fast atom bombardment mass spectrometry and microsequence analysis of the resulting peptides. This approach showed that TEX and NCA were identical with respect to primary sequence and provided direct evidence that 11 of the 12 predicted asparagine-linked glycosylation sites were glycosylated in both TEX and NCA. Indirect evidence was obtained for glycosylation at the other site. Both glycoforms also contain ethanolamine linked to Gly-286, a finding consistent with the conclusion that these proteins are anchored to the plasma membrane through a glycosyl-phosphatidylinositol tail. The large difference in the molecular weights of glycosylated TEX and NCA suggests significant variations in their oligosaccharide structures.     

Mapping of monoclonal antibody-defined epitopes associated with carcinoembryonic antigen,CEA

Summary Six immunoglobulin G monoclonal antibodies reactive with carcinoembryonic antigen (CEA) were evaluated with respect to parameters implicated in their potential diagnostic application and use as tumor targeting agents for cytotoxic drugs or plant or bacterial toxins. Antibody reactivity with surface antigens of the MKN-45 gastric tumor cell line was demonstrated by flow cytofluorimetry. In a subcellular membrane binding assay, each antibody reacted preferentially with membranes isolated from colorectal tumor tissue in comparison with their reaction with membranes from adjacent, apparently normal colonic mucosa. Three of the antibodies (NCRC-23, C228, and 11.285.14) reacted specifically with CEA with little or no reaction with the cross-reacting antigen, NCA. The remaining three antibodies (C24, C161, and C198) were reactive with both CEA and NCA. Analysis of the epitopes defined by these antibodies was performed by competitive binding inhibition assays evaluating the capacity of unlabeled antibodies to compete with ¹²⁵I-labeled antibodies in their binding to CEA. In addition, double determinant or sandwich radioimmunoassays were employed to examine the coexpression of epitopes on CEA molecules. These studies permitted an epitope map to be constructed which describes the coincidence, overlapping, or independent expression of both CEA specific epitopes and epitopes shared between CEA and NCA. The map may be employed for the selection of antibodies for diagnostic and therapeutic use.     

A versatile bifunctional chelate for radiolabeling humanized anti-CEA antibody with In-111 and Cu-64 at either thiol or amino groups: PET imaging of CEA-positive tumors with whole antibodies

Radiolabeled anti-carcinoembryonic antigen (CEA) antibodies have the potential to give excellent images of a wide variety of human tumors, including tumors of the colon, breast, lung, and medullar thyroid. In order to realize the goals of routine and repetitive clinical imaging with anti-CEA antibodies, it is necessary that the antibodies have a high affinity for CEA, low cross reactivity and uptake in normal tissues, and low immunogenicity. The humanized anti-CEA antibody hT84.66-M5A (M5A) fulfills these criteria with an affinity constant of >10 (10) M (-1), no reactivity with CEA cross-reacting antigens found in normal tissues, and >90% human protein sequence. A further requirement for routine clinical use of radiolabeled antibodies is a versatile method of radiolabeling that allows the use of multiple radionuclides that differ in their radioemissions and half-lives. We describe a versatile bifunctional chelator, DO3A-VS (1,4,7-tris(carboxymethyl)-10-(vinylsulfone)-1,4,7,10-tetraazacyclododecane) that binds a range of radiometals including 111 In for gamma-ray imaging and 64Cu for positron emission tomography (PET), and which can be conjugated with negligible loss of immunoreactivity either to sulfhydryls (SH) in the hinge region of lightly reduced immunoglobulins or surface lysines (NH) of immunoglobulins. Based on our correlative studies comparing the kinetics of radiolabeled anti-CEA antibodies in murine models with those in man, we predict that 64Cu-labeled intact, humanized antibodies can be used to image CEA positive tumors in the clinic.     

Apolipoprotein D and cross-reacting human plasma apolipoproteins identified using monoclonal antibodies

We have produced five hybridomas which secreted monoclonal antibodies that reacted with human plasma apolipoprotein D. On analysis by polyacrylamide gel electrophoresis (PAGE) high density lipoproteins and lecithin:cholesterol acyltransferase (EC 2.3.1.43)-enriched fractions of plasma contained many protein bands that reacted with the antibodies. Purified apolipoprotein D had the lowest Mr (29,000), the lowest pI (4.8-5.2), and the greatest migration on alkaline urea-PAGE of all the immunoreactive bands. These characteristics agreed with those described for apolipoprotein D in the literature. The other immunoreactive proteins had apparent Mr from about 39,000 to 98,000, they migrated more slowly than apolipoprotein D on alkaline urea-PAGE, and there were 10 polymorphs on isoelectric focusing. These cross-reacting proteins were present in the high density lipoproteins of each of four individuals sampled on several occasions and in pooled plasma. All of the monoclonal antibodies reacted both with apo-D and the higher Mr cross-reacting proteins. Each of our five monoclonal antibodies bound to one of two distinct antigenic sites on apo-D, determined by antibody competition immunoassays. Neither of these two sites was composed of carbohydrate, but expression of both sites seemed to be influenced by thiol-reducing agents: site 5G10 gained but 4E11 either lost immunoreactivity or was unchanged by reduction according to the conditions. We conclude that apolipoprotein D is only one of several plasma proteins, which contain two homologous polypeptide antigenic sites, recognized by monoclonal antibodies and also by a specific goat antiserum. Apolipoprotein D had the least Mr of these proteins.     

Monoclonal antibodies that inhibit IgE binding

Four monoclonal antibodies were produced that inhibit IgE binding to the high affinity IgE receptor (Fc epsilon R) on rat basophilic leukemia cells. The four monoclonal antibodies (mAb) fall into two groups. The first group was comprised of 3 antibodies (mAb BC4, mAb CD3, and mAb CA5) that reacted with the Fc epsilon R at epitopes close or identical to the IgE-binding site. With 125I-labeled antibodies there was reciprocal cross-inhibition between the antibodies and IgE. The antibodies activated both RBL-2H3 cells and normal rat mast cells for histamine release. The 3 antibodies immunoprecipitated the previously described alpha, beta, and gamma components of the receptor. The number of radiolabeled Fab fragments of 2 of these antibodies bound per cell was similar or equal to the number of IgE receptors. In contrast, the mAb BC4 Fab bound to 2.1 +/- 0.4 times the number of IgE receptor sites. Therefore, the portion of the Fc epsilon R exposed on the cell surface must have two identical epitopes and an axis of symmetry. These 3 monoclonal antibodies recognize different but closely related epitopes in the IgE-binding region of the Fc epsilon R. The fourth monoclonal antibody (mAb AA4) had different characteristics. In cross-inhibition studies, IgE and the other 3 monoclonals did not inhibit the binding of this 125I-labeled monoclonal antibody. The number of molecules of this antibody bound per cell was approximately 14-fold greater than the Fc epsilon R number. This monoclonal antibody caused the inhibition of histamine release and it appears to bind to several cell components.     

Analysis of idiotope variability as a function of distance from the binding site for anti-streptococcal group A carbohydrate antibodies

We have extended our analysis of rat monoclonal anti-idiotopes (anti-Id) specific for previously mapped binding site-associated (distal) and less-or nonbinding site-associated (proximal) idiotopes on a murine monoclonal anti-streptococcal group A carbohydrate (GAC) antibody. By utilizing other monoclonal anti-GAC antibodies and anti-idiotypic antibodies as radiolabeled probands in both competitive and direct radioimmunoassays, we have detected previously unsuspected reactivities of some of the anti-Id. Although the anti-Id recognizing the most proximal idiotopes manifest relatively narrow ranges of binding strengths for anti-GAC antibodies, the anti-Id recognizing the most distal idiotopes display broader, more continuous distributions of binding strengths. These results suggest that mimicry of antigen structure by anti-Id might best be understood from a quantitative perspective, and that idiotopes intimately associated with binding sites display a broader range of variants than those not associated with binding sites. In addition, for one monoclonal anti-Id recognizing a distal determinant, changing the radiolabeled proband in inhibition radioimmunoassays results in dramatic changes in relative inhibitory efficacies for certain anti-GAC antibody inhibitors. This observation suggests the possibility that this anti-Id represents an example of a multispecific (polyfunctional) anti-idiotypic antibody.     

Use of thiol-terminal silanes and heterobifunctional crosslinkers for immobilization of antibodies on silica surfaces

A procedure for covalent immobilization of functional proteins on silica substrates was developed using thiol-terminal silanes and heterobifunctional cross-linkers. Using this procedure, a high density of functional antibodies was bound to glass cover slips and silica fibers. The amount of anti-IgG antibody immobilized was determined to be in the range of 0.66 to 0.96 ng/mm2 using radiolabeled antibody. The relative amount of IgG antigen bound by the immobilized antibody (0.37 to 0.55 mol antigen/mol antibody) was three to five times greater than other investigators have reported. In addition, the amount of protein nonspecifically adsorbed to the antibody-coated surface was further reduced by the addition of blocking agents so that nonspecific adsorption of protein antigens represented only 2-6% of the total antigen binding. With this low background, IgG antigen binding could be measured at levels as low as 150 fmol when an antigen concentration of 3 pmol/ml was applied. The process for antibody immobilization is straightforward, easy to perform, and adaptable for modifying mass quantities of biosensor components.     

Preparation and evaluation of a molecularly imprinted polymer for the selective recognition of testosterone--application to molecularly imprinted sorbent assays

Biomimetic testosterone receptors were synthesized via molecular imprinting for use as antibody mimics in immunoassays. As evaluated by radioligand binding assays, imprinted polymers prepared in acetonitrile were very specific for testosterone because the nonimprinted control polymers bound virtually no radiolabeled testosterone. The polymers present an appreciable affinity, with association constants of K(a) = 3.3 x 10(7) M(- 1) (high-affinity binding sites). The binding characteristics of the polymers were also evaluated in aqueous environment to study their viabilities as alternatives to antibodies in molecularly imprinted sorbent assays. Compared with the testosterone-specific antibodies present in commercial kits, our molecularly imprinted polymers are somewhat less sensitive but show a high selectivity.     

Comparison of CEA distribution in lesions and tumors of salivary glands as determined with monoclonal and polyclonal antibodies

Immunohistochemical localization of carcinoembryonic antigen (CEA) with conventional antibody to CEA (anti-CEA), nonspecific crossreacting antigen (NCA)-absorbed polyclonal antibody to CEA (NCAa-CEA), and monoclonal antibody to CEA (Mono-CEA) have been compared in obstructive lesions and salivary gland tumors. Normal salivary glands gave strong staining of the luminal borders of acinar cells with anti-CEA, whereas no staining occurred with Mono-CEA. Obstructive lesions showed occasionally marked staining with anti-CEA in some acinar cells, but there was no reaction with Mono-CEA. Of 69 pleomorphic adenomas examined, 34 were positively stained with anti-CEA, 18 with NCAa-CEA and 8 with Mono-CEA along the luminal borders of the tumor cells. The frequency of positive staining of material within tubular lumina was similar with all three immunoreagents. Neoplastic cells were positive with Mono-CEA in only three cases, while eight cases were positive with NCAa-CEA and 11 cases with anti-CEA. In salivary gland tumors true CEA is be found mainly at the border of tumor cells, but the frequency of positive reactions is low.     

Development of humanized antibodies as cancer therapeutics

Recent success in the development of monoclonal antibody-based anti-cancer drugs has largely benefitted from the advancements made in recombinant technologies and cell culture production. These reagents, derived from the antibodies of mouse origin, while maintaining the exquisite specificity and affinity to the tumor antigens, have low immunogenicity and toxicity in human. High-level expressing cell clones are generated and used to produce large quantities of the recombinant antibodies in bioreactors in order to meet the clinical demand for therapeutic applications. In this report, the systems and general methodologies developed by us to construct and produce humanized antibodies from the parent mouse antibodies are described. Once the humanized antibodies are available, they can be applied in three principal forms for cancer therapy: (1) naked antibodies, (2) drug- or toxin conjugates, and (3) radioconjugates. Using the humanized anti-CD22 (epratuzumab) and anti-carcinoembryonic antigen (ant-CEA; labetuzumab) antibody prototypes, clinical applications of naked and radiolabeled humanized monoclonal antibodies are described.     

Further characterization of cooperative interactions of monoclonal antibodies

We reported previously that mixtures of some monoclonal antibodies directed against the glycoprotein hormone human chorionic gonadotropin (hCG) had a higher affinity for the antigen than either monoclonal antibody separately. The synergistic interaction could no longer be detected when one of the antibodies was replaced with its F(ab) fragment. This cooperative interaction has now been further characterized. One-half of 10 possible pairs prepared from five IgG1 monoclonal antibodies against hCG result in a synergistic interaction. The addition of an IgG2b monoclonal antibody to one of the IgG1 monoclonal antibodies also induces a cooperative interaction, which shows that the effect is not subclass restricted. Cooperative interactions between antibodies are also not restricted to solution conditions; adsorption of one antibody to a solid support appears to increase the cooperative effect. Indeed, one pair of antibodies that failed to bind hCG synergistically in solution did so when one antibody was bound to a solid surface. The liquid phase antibody also has an effect on the specificity of the solid phase antibody. The sensitivity of the solid phase assay system has enabled us to develop a rapid method of determining if two monoclonal antibodies can bind to an antigen simultaneously. A quantitative theoretical model has been devised that successfully predicts the cooperative behavior observed between antibodies and should be useful in devising conditions that result in sensitive solid phase radioimmunoassays.     

Cell adhesion activity of non-specific cross-reacting antigen (NCA) and carcinoembryonic antigen (CEA) expressed on CHO cell surface: homophilic and heterophilic adhesion

Cell adhesion activity of carcinoembryonic antigen (CEA) and non-specific cross-reacting antigen (NCA) has been analysed by using CHO cells which had been transfected with cDNAs and are ectopically expressing each antigen on their surface. CEA expressing CHO tended to aggregate easily within 30 min after being suspended by trypsinization. Cell adhesion assay between 51Cr labelled cells and monolayered cells showed both homophilic and heterophilic interaction, the extent of which was CEA-CEA much greater than CEA-NCA greater than NCA-NCA. These reactions were completely inhibited by Fab' fragment of anti-CEA antibody. The results strongly suggested that CEA and NCA function as Ca++ independent cell adhesion molecules by homophilic and heterophilic interactions.     

Cancer imaging with CEA antibodies: historical and current perspectives.

This article reviews the history and status of cancer imaging with radiolabeled antibodies against carcinoembryonic antigen (CEA). Although CEA and many other cancer-associated antigens are not distinct for neoplasia, the quantitative increase of these markers in malignant tissues provides a sufficient differential for selective antibody targeting. Animal studies with xenografted human tumors provided the first evidence of the prospects of this technology, followed by initial clinical success with purified goat whole IgG antibodies to CEA, labeled with 131I and with the use of dual-isotope subtraction methods. Subsequently, improved and earlier imaging could be accomplished with monoclonal antibody fragments, which then would permit the use of shorter-lived radionuclides, such as 111In, 123I, and 99mTc. The preferred use of a monoclonal anti-CEA IgG Fab' fragment, labeled with 99mTc by a recently developed, simple and rapid kit, has enabled the detection of small lesions, including those in the liver, within 4 h of injection. By means of SPECT imaging, a high sensitivity and specificity for RAID could be achieved.     

Interaction of LDL, Lp[a], and reduced Lp[a] with monoclonal antibodies against apoB

Five monoclonal antibodies (2A, 9A, 6B, L3, L7) produced in mice against human apolipoprotein B were investigated by competitive and inhibitive electroimmunoassay (EIA) for their reactivity with low density lipoprotein (LDL), lipoprotein[a] (Lp[a]), and reduced Lp[a]. All of the antibodies reacted with apoB of the different lipoproteins indicated by very similar slopes of the binding curves. None of them gave a positive reaction with apolipoprotein[a]. The amount of apoB required for 50% inhibition of antibody binding varied for the different antibodies and lipoproteins. Antibody 9A showed almost the same affinity for LDL, Lp[a], and reduced Lp[a]. Antibodies 2A and 6B bound about twofold better to LDL and reduced Lp[a] than to untreated Lp[a]. Antibodies L3 and L7 needed nearly threefold higher amounts of Lp[a]-apoB for 50% inhibition of antibody binding than of apoB of LDL and reduced Lp[a]. The amount of apoB required for 50% inhibition of antibody binding was somewhat higher in inhibitive assay than in competitive assay. We suggest that apo[a] covers certain epitopes of apoB in native Lp[a] leading to a reduced reaction with the monoclonal antibodies. However, it could also be that the binding of the [a]antigen to apoB via disulfide bridges causes profound conformational changes of the apoB region exposed to the surface.     

An enzyme-linked immunoassay for the detection of antibodies to endoplasmic reticulum

A simple, sensitive solid-phase assay for the detection of antibodies to endoplasmic reticulum is described. The assay is dependent upon the amount of antigen bound to the solid support and upon the amount of antibody bound to the support via the relevant antigen. The assay can be used to measure both polyclonal and monoclonal antibody to endoplasmic reticulum. It has been used to isolate several monoclonal antibodies which can recognize and precipitate specific proteins of the endoplasmic reticulum. In addition, it has been used to probe the membrane orientation of endoplasmic reticulum antigens.     

Preparation and Characterization of Two Human Carcinoembryonic Antigen Family Proteins of Neutrophils, CD66b and c, in Silkworm Larvae

As a step to investigate the cell adhesion mechanism and physiological roles of two CD66 antigens in human neutrophils, carcinoembryonic antigen gene family member 6 (CGM6, CD66b) and nonspecific cross-reacting antigen (NCA, CD66c), we prepared their soluble recombinant forms in silkworm larvae. Each cDNA fragment for CGM6 and NCA was ligated into the transfer vector pBK283 after modification to encode the protein lacking the membrane anchor. The resultant vectors were introduced to theBombyx morinuclear polyhedrosis virus, with which silkworm larvae were infected. Recombinant proteins secreted into the hemolymph of larvae at concentrations up to 1.3 mg/ml were purified by cation exchange followed by gel filtration or antibody affinity chromatography. The smaller apparent masses of the antigens compared with those of the native antigens appeared to be primarily due to incomplete glycosylation. Both recombinant antigens are quite similar to the corresponding native antigens in terms of the antigenic reactivity against a panel of CD66 monoclonal antibodies. In addition, the recombinant CGM6 and NCA exhibited cell binding activity against CHO cells expressing NCA and CGM6, respectively. Thus the two biologically active recombinant CD66 antigens prepared in large quantities in silkworm larvae should be useful for their functional studies, and our present system will be available for the production and purification of other carcinoembryonic antigen family members, whose biological functions are also unknown.