CLIN. CHEM. 36/6, 879-882 (1990) CLINICAL CHEMISTRY, Vol. 36, No. 6, 1990 879 Unique Structure of Glycopeptide from a-Fetoprotein Produced in Human Hepatoma Cell Line, as Determined by Magnetic Resonance Spectroscopy Benedetto Terrana,1 MarIoF. Tecce,1 Roberto Manettl,1 Costante Ceccarlnl,1 Dorlano Lamba,2 and Anna Laura S.gre2 Determination of a-fetoproteinis used in diagnosisoftumors and neural tube defects. A good reliable source of a- fetoprotein would be an obvious advantage to the prepara- tion of diagnostic reagents and their standardization. We have recently developed a method for the production of a-fetoprotein from a human hepatoma cell line. This method, which is suitable for scaling up, allowed us to produce 40 g of a-fetoprotein from culture supematant liquidthrough a simple purification procedure. We have previously shown this pro- tein to be identical to a-fetoprotein produced from other sources. However, because the presence of different glyco- forms has been reported in a-fetoprotein preparations, both from human sources and from other species, it was important to establish the type and extent of glycosylation of a- fetoprotein prepared by our method. By using 1H-NMR spectroscopy we were able to establish that our product contains a single N-linked biantennary, fully sialylated com- plex-type oligosaccharide, typical of human hepatomas. Human a-fetoprotein is the major protein component of embryonal serum; its synthesis decreases late during fetal life, to be replaced by the synthesis of albumin, the major protein of adult serum. a-Fetoprotein is synthesized in the liver and yolk sac of the developing embryo; in adults, a-fetoprotein can be synthesized in patients with hepato- mas and yolk sac tumors (1). The primary sequence of human albumin was determined several years ago by directly sequencing the protein (2); the sequence of human a-fetoprotein has been deduced more recently from that of the cDNA (3). The two proteins are very similar in struc- ture and share 39% amino acid homology. However, al- though a-fetoprotein is a glycoprotein, albumin is not glycosylated. A single potential glycosylation site has been found (asparagine 232) in the predicted amino acid so- quence of human a-fetoprotein (3), whereas none is present in the albumin sequence. The carbohydrate structure of a-fetoprotein from human hepatoma has been determined to be of the biantennary complextype (4). For a-fetoprotein from yolk sac tumors, the structure remains the same except for the addition of a bisecting N-acetylglucosamine (5). However, the protein from both sources is heteroge- neous with respect to the degree of sialylation, the number of sialylated reSiduesvarying from 0 to 2 (4, 5). The cause of this heterogeneity has not been determined. Although the physiological role of a-fetoprotein (or of albumin) has not been fully established, determination of the concentration of this protein in serum is used to evaluate the status of cancer patients or, in serum from gestating women, to detect the presence of fetal malforma- ‘Centro Ricerche Sclavo, Via Fiorentina 1, 53100 Siena, Italy. 2jmi di Strutturistica Chimica “G. Giacomello” Area della Ricerca di Roma, CNR, C.P. 10,00016 Monterotondo Staz., Rome, Italy. Received December 29, 1989; accepted April 2, 1990. tions such as spina bifida (1). Clearly, a good reliable source of a-fetoprotein could be highly beneficial, both for the preparation of diagnostic reagents and their standardiza- tion. The most common sources of a-fetoprotein for prepar- ing diagnostic reagents are cord blood sera, fetal extracts, and either ascites fluids from hepatoma patients or the tumor itself. However, none of these yields large, reproduc- ible amounts of the protein. Recently, we developed a method for producing a-fetoprotein by utilizing a human hepatoma cell line (6), which can be adapted for large-scale production. With this method, one can purify large amounts of protein from an essentially constant and unlim- ited source. We have previously shown that the protein produced by our method is immunologically indistinguish- able from a-fetoprotein prepared from other sources (6), and we have determined that the amino acid sequence of this protein is identical to that predicted from the cDNA sequence (manuscript in preparation). Because the major source of heterogeneity in a-fetoprotein lies in its carbohy- drate moiety, we used ‘H-nuclear magnetic resonance (NMR) spectroscopy to determine the structure of the carbohydrate in our a-fetoprotein preparation. Here we present the results of this analysis, which show that a- fetoprotein obtained by our method is highly homogeneous, containing a single biantennary, fully sialylated complex- type oligosaccharide. These results also indicate that no heterogeneity was induced during the large-scale culture. Materials and Methods Purification of a-fetoprotein and a-fetoprotein glycopep- tide. A human hepatoma cell line was grown in serum-free medium in large bioreactors. The medium supernate, in which a-fetoprotein represents about 35% of the total protein (6), was concentrated by filtration, precipitated with ainmonium sulfate, and stored frozen as a paste until further use. The total amount of a-fetoproteun produced by this method was 40 gin 2000 L of supernate, as determined by an immunoenzymatic assay. After dialysis of the water solution of the protein mixture, we further purified a- fetoprotein by affinity chromatography, using the immune- globulin fraction of a goat antiserum prepared against a homogeneous preparation of the protein (unpublished). The a-fetoprotein thus produced migrated as a single band on sodium dodecyl sulfate-polyacrylamide gel electrophoresis, in the presence and absence of denaturing agents, and it contained no detectable amounts of albumin (6) (recovery was 70% in each purification lot). The glycopeptide portion of the protein was prepared by two different protocols: digestion with proteinase K (EC 3.4.21.14, from Tritirachium album; Boehringer Mann- heim, Mannheim, F.R.G.) or digestion with “Pronase” (pro- tease from Streptomyces griseus; Calbiochem, La Jolla, CA). We digested 130 mg of a-fetoprotein with either a single addition of 100mg of proteinase K in 15 mL of buffer (per liter, 0.1 mol of Tris, pH 8.0, and 2 rnmol of CaC12),at 37 #{176}C for 72 h under a layer of toluene, or with Pronase under the conditions described by Murainatsu et al. (7). All