Isolation and structures of the oligosaccharide units of carcinoembryonic antigen.

Carcinoembryonic antigen was purified from liver metastases of primary colon and breast tumors in high yield by procedures which included gel filtration on Sepharose 4B and affinity chromatography on concanavalin A-agarose. The antigens were homogeneous, contained only a single polypeptide chain, and had identical amino acid sequences for 15 residues from the NHzterminal end of the chain. They had a molecular weight of 180,000 k 20,000 and nearly identical immunological properties. The antigen contained about 40 oligosaccharide chains which are linked to asparagine. The oligosaccharide chains were released by hydrazinolysis or treatment with alkaline borohydride and their structures were determined by sequential hydrolysis with exoglycosidases, periodate oxidation, and methylation analysis. The major oligosaccharide chain, 80%, has the following tetra-antennary structure.

lated in various studies. Different purified preparations of CEA isolated from individual tumors of the human digestive system showed large variations in their carbohydrate composition (4) in spite of the fact that the amino acid compositions of these preparations were very similar (5, 6). Variations in the content of sialic acid cause purified preparations of CEA to appear polydisperse when examined by electrophoresis (6,7). A number of other different antigens with immunological and chemical properties related to those of CEA have also been found in both tumor and normal tissue (8-12).
It is not yet certain what changes occur during neoplastic transformation which lead to the synthesis of CEA. In order to compare the structure of CEA with glycoproteins synthesized by the normal cell, it is necessary to have highly purified preparations of the antigen. The isolation and characterization of homogeneous preparations of the antigen are also required before studies on its biosynthesis in tumor tissue can be initiated. In the present study, isolation procedures including affinity chromatography were used for the purification of sufficient quantities of CEA for structural analysis of the oligosaccharide chains. This communication describes the structural analysis of the N-asparagine-linked oligosaccharide chains of CEA. The sequence of sugars, linkage, and anomeric configuration of the glycosidic bonds in reduced tetra-, tri-, and diantennary oligosaccharide chains isolated from two preparations of CEA purified from liver metastases of colon and breast tumor were established.

EXPERIMENTAL PROCEDURES~
autopsy and froaen at -m a .
f he lrver f r o . a s~n g l e r n d l v l d u l was "red for each preparation, and the malorriy of lzvers selected for there sCYdres had aver 70% neiasint~c L m r d k autopsy m e procedures urea f o r the lsnlatlon of C U f r w liver mkastases of prlirary colon and breast rmor are sunnar~zed ~n T a l e 1. Smrlar results were obtamed by thls pxocedvre wlth more than 10 d~fferenr preparations. The purrflcatron "&E carrled out at 3  The extent Of purlflcatlon of the antrgen *a* est1-Specrfrc rnonunologlcal aCtlYltY.
1N NaOH dnd the combined solurlon~ wexe dlalyred exhaustluely agamsL water at 3and adlusted The dialyzed solution was concentrated 50-fold by blovrng axr at 10-temperature over the dr,ly51s t u b~n g . The "on-xnmuna-reactive preclpledte was remved by cenrrifvqatlon and the supernatant solurxon 1i.ractlon l , Table 11   Purification of carcinoembryonic antigen The protein and immunological activity were measured by the standard procedures described in the text. The specific immunological activity represents the milligrams of antigen determined with anti-CEA antibody divided by protein estimated by the standard spectrophotometric or colorimetric procedures after corrections for carbohydrate content. The data are shown for a liver metastases of colon tumor.

Protein
Spe-    6 (right). Separation of sialylated reduced oligosaccharides isolated from C E A b y ion exchange chromatography on DEAE-cellu1ose.A. about 20pmo1, in 1.2 ml, of partially desialylated, N-glycosidic reduced triantennary chain isolated from fetuin was applied to a DEAE-cellulose column (2.7 X 8 cm). The column was previously equilibrated with 2 M pyridine acetate, pH 5.2, and then washed extensively with water. The oligosaccharides were eluted with a linear gradient made up of 250 ml of water in the mixing chamber and 250 ml of 0.25 M pyridine acetate, pH 5.2, in the reservoir. Fractions of 4.5 ml were collected and aliquots were assayed with the anthrone reagent. Peak I , the oligosaccharide contained 1.05 pmol of sialic acid per triantennary chain. Peak 2, the oligosaccharide contained 2.0 pnol of sialic acid per triantennary chain. Peak 3, the oligosaccharide contained 2.9 p o l of sialic acid per triantennary chain. B , the radioactive oligosaccharides isolated from CEA were applied to a DEAE-cellulose column prepared as described in A and the oligosaccharides were eluted with the same gradient. Peak N, contained the neutral oligosaccharide fraction and no sialic acid was detected in this fraction. Peak 1, the oligosaccharides contained 0.95 pnol of sialic acid per GlcNAc~r residue. Peak 2, the oligosaccharides contained 1.85 pmol of sialic acid per GlcNAc~r residue. N-Acetylneuraminic acid

Molar ratio ofpartially methylated alditol and hexosaminitol acetates in reduced oligosaccharides after treatment with specific plvcosidases
Step 2: after Step 3: after alditol Step  . 8 (left). Elution profiles of the digestion products formed from oligosaccharide B after incubation with specific exoglycosidases. Peak 1, asialo-oligosaccharide B. Peak 2, after digestion with p-galactosidase. Peak 3, after digestion of product in Peak 2 with a-L-fucosidase and a second time with p-galactosidase. Peak 4, after digestion of product in Peak 3 with /3-N-acetylglucosaminidase. Gel fdtration on Bio-Gel P-6 columns was carried out as described in the text. FIG. 9 (center). Elution profiles of the digestion products formed from oligosaccharide AI, after treatment with specific exoglycosidases. Peak 1, the asialo-reduced oligosaccharide. Peak 2, after treatment with @-galactosidase. Peak 3, after treating product in Peak 2 with p-N-acetylglucosaminidase. Peak 4, after treating product in Peak 3 a second time with /?-galactosidase and p-N-acetylglucosaminidase. The separation on Bio-Gel P-6 columns (2.2 X 200 cm) was performed as described in the text. FIG. 10 (right).

DISCUSSION
CEA is usually solubilized and extracted from tissue homogenates with 1 M perchloric acid (6,50). This treatment precipitates most proteins, but CEA and other glycoproteins which have a high carbohydrate content are soluble in acid and remain in the supernatant solution. Subsequent chromatography on Sepharose CL-GB columns remove high molecular weight components, principally blood group glycoprotein and high molecular weight components, principally blood group glycoprotein and high molecular weight mucin, from the included lower molecular weight CEA. In the present studies, CEA was further purified by techniques involving chromatography on ConA-Sepharose 4B columns in order to obtain preparations of the antigen which were less heterogeneous with respect to the carbohydrate moiety. Chu et al. (51) and Rogers et al. (52) separated several glycoprotein fractions with CEA activity from tumor extracts by affinity chromatography on Cod-Sepharose 4B. One fraction bound tightly to the column, while the other showed little or no affinity for ConA. The bound fraction contained more GlcNAc than the unbound sample (53). The purified samples also differed in their ability to inhibit the binding of 1251-labeled CEA to goat anti-CEA. More recently, Slayter and Coligan (54) have also used ConA-Sepharose 4B to purify CEA. They obtained several immunologically active fractions by elution with 2% glucose and 20% mannose. These fractions had very similar carbohydrate and amino acid compositions and it is probable that only one major component was isolated in these studies. The preparations obtained in these earlier studies contained some GalNAc (51-53) and larger amounts of fucose than the samples isolated in the present study. No GalNAc was detected in the purified preparations obtained by the procedure described in this report. These differences may arise from a contamination with blood group or mucin type glycoproteins. These glycoproteins contain GalNAc and a proportionately larger amount of fucose than purified CEA. The presence of ,Bl,3-linked galactose in the CEA isolated in earlier studies (10) and the complete absence of this linkage in the CEA isolated in the present study further support this conclusion. The data clearly show that standard isolation procedures as well as affinity chromatography on lectin columns is required to obtain pure Preparations of the antigen.
The preparations of CEA used in the present study were not contaminated with blood group or mucin glycoproteins. Amino acid analysis and partial sequence determination indicated that most of the preparations had very similar polypeptide chains. The antigenic properties of different preparations of the antigen were also nearly identical. The polydisperse nature of these fractions on high resolution gel electrophoresis and isoelectric focusing further suggests that they contain a mixture of molecules with different types of oligosaccharide chains. The evidence reported in this communication provides an explanation for the polydisperse nature of different preparations of CEA based on the number and degree of sialylation of oligosaccharide chains in this tumor antigen. The general features of the carbohydrate moiety of CEA described in the present studies are somewhat different from those reported earlier for intact CEA by Egan et al. (55). Numerous minor components including 2-linked galactose, 6linked mannose, and 3,4-linked galactose were not detected in our studies. The presence of some GalNAc in the previous preparations (55) suggests that 0-serine-linked glycoproteins were still present. The results of periodate oxidation are similar to those obtained with the intact glycoprotein in earlier studies by Coligan and Todd (56). All of the sialic acid and fucose, about 80% of the galactose, and small amounts of mannose and GlcNAc were lost after treating the intact antigen with periodate.
The microheterogeneity found in the oligosaccharide chains of CEA may account for these differences. Results obtained by isoelectric focusing and immunoelectrophoresis showed that purified preparations of CEA are still polydisperse, in spite of the fact that the polypeptide chains are homogeneous.
Differences in the number of oligosaccharide chains and the extent of sialylation of these chains could account for the detection of numerous ionic species of the antigen. All of these forms of CEA would be expected to react with anti-CEA serum, since earlier studies indicating that periodate oxidation does not destroy immunoreactivity (13,56) and the results obtained in the present studies show that the antibodies raised to CEA are directed against the polypeptide chain of the antigen.
Three different types of N-asparagine-linked oligosaccharide chains were released from CEA by hydrazinolysis or by treatment with alkaline-borohydride. About 80% of the oligosaccharide chains in CEA had a tetra-antennary structure. The rest of the chains had a di-or triantennary structure. The oligosaccharides isolated from CEA which was purified from tumors of a similar histological type from a number of different individuals contained only 1 or at most 2 sialic acid residues. All of the sialic acid in CEA is attached through a-linkages to the 3 and 6 positions of penultimate galactose units. Most of the tetra-antennary chains also contained one or two residues of fucose a-linked to the 3 position of penultimate GlcNAc units or a disaccharide, Gal,B1,4GlcNAc, which was ,B-linked to the 3 position of terminal galactosyl units. These substitutions result in extensive microheterogeneity in the outer portion of the oligosaccharide chains of the antigen. The amounts of fucose, disaccharide, and sialic acid in the terminal chains of the antigen varied considerably in different preparations.
CEA contains a large number of tetra-antennary chains. Most normal plasma glycoproteins have only a few tri-or tetra-antennary chains. Thyroxine-binding globulin contains 3 diantennary chains and one triantennary chain (57) and a,acid glycoprotein contains only 5 to 6 oligosaccharide chains (58). Fetuin, the major glycoprotein in fetal calf serum, contains 3 triantennary chains. The synthesis of N-asparaginelinked carbohydrate requires a specific sequence, Asn-X-Ser/ Thr, in a polypeptide chain before the transfer of precursor oligosaccharides from dolichol phosphate intermediates to nascent polypeptide chains can occur (59, 60). Thus, CEA may contain a very large number of Asn-X-Ser/Thr sequences in its polypeptide chain. All of the glycopeptides isolated from CEA in the present study contained amino acids which are involved in this sequence. At least 120 of the amino acid residues in the antigen may be present at sites with this sequence. The origin of a polypeptide chain with so many of these specific triplet sequences is not presently known. However, some of the properties of CEA resemble those of mucin type glycoproteins. Serine and threonine make UP 25.2%, 25.4%, and 19.5% of the total amino acids in Cowper's gland mucin, ovine submaxillary mucin, and CEA, respectively. These glycoproteins also contain large amounts of aspartic acid and small amounts of lysine, arginine, tryptophan, and methionine. The extensive substitution of asparagine for aspartic acid residues in a mucin type glycoprotein which contains large amounts of serine and threonine could result in the formation of a large number of Asn-X-Ser/Thr sequences. In previous studies, we have shown that amination of a peptide containing the sequence Asp-X-Ser/Thr converts it into an active substrate for peptidyltransferase (28). The type of heterogeneity observed in the distribution of oligosaccharide chains in CEA is also similar to that seen in mucin glycoproteins. Glycoproteins containing N-asparagine-linked chains usually contain a constant number of oligosaccharides which are linked at specifk sites in the polypeptide chain. The presence of a widely varying number of oligosaccharide chains in the polypeptide chain of CEA is characteristic of the type of heterogeneity seen in 0-serine-linked mucin glycoproteins. Many tumors of the adenocarcinoma type originate in the epithelial layer of the respiratory, digestive, and reproductive systems, and cells in these tissues normally form large amounts of mucin glycoproteins.