Characterization of the expression products of recombinant human choriogonadotropin and subunits.

Human choriogonadotropin (hCG) is a placental glycoprotein hormone composed of a 92-amino acid alpha subunit noncovalently linked to a 145-amino acid beta subunit. We report here the expression of biologically active hCG in mouse C127 cells transfected with expression vectors containing the DNA coding for both subunits. In addition, the same cell line was used to express the alpha subunit alone. The expression products were purified by affinity chromatography using specific monoclonal antibodies to hCG or its subunits. The system secreting biologically active hCG also produced a 10-fold or greater molar excess of free beta subunit. The dimeric hormone, as well as the excess beta subunit, resembles the standard urinary hCG and beta subunit by chemical and biological criteria. In contrast, when the vector encoding for the alpha subunit was expressed alone, the alpha subunit had a higher molecular weight than both standard alpha and the alpha found in the expressed dimeric hormone. The molecular weight difference between expressed alpha subunit and standard alpha was found to reside in the alpha peptide consisting of residues 52-91 which contained all of the carbohydrate of the alpha subunit. The N-asparagine-linked carbohydrate moieties in the recombinant alpha were found to be triantennary in contrast to biantennary in urinary alpha, and this hyperglycosylation was responsible for the higher molecular weight of the alpha subunit when it was expressed alone. We found no evidence of O-threonine glycosylation at position alpha 39 reported to be present in free forms of the alpha subunit; however, the companion paper (Corless, C.L., Bielinska, M., Ramabhadran, T. V., Daniels-McQueen, S. Otani, T., Reitz, B. A., Tiemeier, D. C., and Boime, I. (1987) J. Biol Chem. 262, 14197-14203) finds a small quantity of O-glycosylation. Since the excess beta subunit appears to be of normal size and contains the expected complement of sugars, only free alpha subunit seems to be a potential substrate for addition of extra sugar moieties. No large beta subunit forms have been found by others, while large alpha subunits have been described both clinically and in tissue culture systems. These observations imply that the conformation of the free alpha subunit, in the regions of the glycosylation recognition sites, allows easier access for glycosyltransferases than those same sites in the beta subunit. When alpha is combined with beta, the local structures around the alpha glycosylation sites are apparently altered so as to make the synthesis of triantennary chains less favorable.

Human choriogonadotropin (hCG) is a placental glycoprotein hormone composed of a 92-amino acid a subunit noncovalently linked to a 145-amino acid j3 subunit. We report here the expression of biologically active hCG in mouse C127 cells transfected with expression vectors containing the DNA coding for both subunits. In addition, the same cell line was used to express the a subunit alone.
The expression products were purified by affinity chromatography using specific monoclonal antibodies to hCG or its subunits. The system secreting biologically active hCG also produced a 10-fold or greater molar excess of free j3 subunit. The dimeric hormone, as well as the excess j3 subunit, resembles the standard urinary hCG and j3 subunit by chemical and biological criteria. In contrast, when the vector encoding for the a subunit was expressed alone, the a subunit had a higher molecular weight than both standard a and the a found in the expressed dimeric hormone.
The molecular weight difference between expressed a subunit and standard a was found to reside in the a peptide consisting of residues 52-91 which contained all of the carbohydrate of the a subunit. The N-asparagine-linked carbohydrate moieties in the recombinant a were found to be triantennary in contrast to biantennary in urinary a, and this hyperglycosylation was responsible for the higher molecular weight of the a subunit when it was expressed alone. We found no evidence of 0-threonine glycosylation at position aS8 reported to be present in free forms of the a subunit; however, the companion paper (Corless, C. L., Bielinska, M., Ramabhadran, T. V., Daniels-McQueen, S . Otani, T., Reitz, B. A., Tiemeier, D. C., and Boime, I. (1987) J. Biol Chern. 262,[14197][14198][14199][14200][14201][14202][14203]) finds a small quantity of 0-glycosylation.
Since the excess j3 subunit appears to be of normal size and contains the expected complement of sugars, only free a subunit seems to be a potential substrate for addition of extra sugar moieties. No large j3 subunit forms have been found by others, while large a subunits have been described both clinically and in tissue culture systems. These observations imply that the conformation of the free a subunit, in the regions of the glycosylation recognition sites, allows easier access for glycosyltransferases than those same sites in the 8 * This work was supported by National Institutes of Health Grant PO-HD-15454. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U. subunit. When a is combined with 8, the local structures around the a glycosylation sites are apparently altered so as to make the synthesis of triantennary chains less favorable.
Human choriogonadotropin (hCG)' is a glycoprotein hormone, whose primary function early in pregnancy is thought to be stimulation of the corpus luteum to maintain an endometrial environment that is favorable for the implanted fertilized ovum. The hCG molecule is composed of two nonidentical subunits, a and p, each of which is first synthesized as a larger precursor molecule containing a signal peptide (1,2), which is removed by proteolytic cleavage prior to selection. Only dimeric hCG, not the single subunits, possesses biological activity (3). The 92-amino acid (Y chain is glycosylated at asparagines 52 and 78, while the 145-amino acid p chain is glycosylated at asparagines 13 and 30 and also at four serine residues located near the COOH terminus of the polypeptide chain (4,5).
The hormone contains approximately 30% carbohydrate by weight (3), and proper glycosylation is thought to be required for biological function. HF-treated or enzymatically deglycosylated hCG is significantly impaired in its capability to stimulate CAMP formation and steroidogenesis under in vitro assay conditions despite a high binding affinity of the deglycosylated hormone for its receptor (6)(7)(8)(9). This loss of biological activity may reflect a role for carbohydrate in hormone action, or it could be due to perturbations in the tertiary structure caused by the chemical treatment (8) or due to contaminating protease activity when glycosidase enzymes are used (10). The carbohydrate structure also contributes to a prolonged circulating half-life of hCG in plasma (11).
Knowledge of the chemistry and immunochemistry of this glycoprotein hormone permits it to be used as a model to study the regulation and synthesis of complex dimeric glycoproteins. In this report we describe the construction and expression of two vectors: one containing both subunits of hCG and the other containing only the a subunit. The dimeric hCG expression product appears to be chemically, biologically, and immunologically identical to the hCG that is isolated from the urine of pregnant women. By contrast recombinant (Y subunit, when expressed alone, had a significantly higher The abbreviations used are: hCG, human choriogonadotropin; bp, base pair(s); kb, kilobase(s); FPLC, fast protein liquid chromatography; BPV, bovine papillomavirus; HPLC, high performance liquid chromatography; SDS, sodium dodecyl sulfate; PAGE, polyacrylamide gel electrophoresis; endo F, endoglycosidase F; RCM, reduced S-carboxymethylated; PTH, phenylthiohydantoin; TFA, trifluoroacetic acid; THF, tetrahydrofuran. molecular weight than did the cy subunit derived from urinary hCG. It has been noted previously that a free a subunit can be secreted with a larger molecular size than the a subunit isolated from urinary hCG (12)(13)(14)(15)(16)(17). One reported reason for this increased cy subunit size was the addition of an 0-linked oligosaccharide at residue Thr-39 found in the free cy subunit purified from bovine pituitaries (12) and from cell cultures . Blithe and Nisula (18) also reported that a free cy subunit in pregnancy urine exhibited different binding characteristics with lectins and may have had an altered carbohydrate structure as compared to the standard CY subunit derived from the dimeric hormone.
We report here that, when cy subunit is expressed without its subunit complement, the N-linked carbohydrate moieties of the secreted product contain additional sugar residues in contrast to the normal situation where cy and p subunits are expressed together in the same cell. These findings suggest that the several glycosyltransferases involved in the construction of the oligosaccharide chains on glycoproteins are affected by the conformation of the glycosylation site. These results have been obtained by structural analyses of purified expression products. An accompanying communication by Corless et al. (19) shows a similar finding using in uiuo radiolabeling techniques.

DISCUSSION
Human choriogonadotropin is the first two-subunit glycoprotein hormone to be expressed by recombinant technology, and this has been found to result in a biologically active protein very similar to the product isolated from pregnancy urine (41). In this report we demonstrate the unexpected finding that the carbohydrate moieties, when the cy subunit is expressed alone, differ from those in the cy subunit when it is expressed in the presence of the p subunit. Since the state of glycosylation appears to be important to proper folding of the polypeptide chain as it is synthesized (42,43) and also to the in uitro and in uiuo biological activity of the hormone (Refs. 6-9 and 441, factors effecting changes in carbohydrate biosynthesis are important. Construction of Expression Vectors-We have described the construction of expression vectors containing the intact bovine papillomavirus genome, the entire mouse metallothionein gene, and either the cy or both cy and @' hCG DNA sequences in Fig. 1. When the cy and p genes, on separate expression vectors, were cotransferred into mouse recipient C127 cells, bovine papillomavirus transformants were identified and analyzed for production of hCG (Table I). The highest producing clones expressed hCG at a level of about lo-'' mIU/ce11/24 h ( Table I).
The bovine papillomavirus transformants producing and secreting hCG were further characterized. The structures of the DNA of the higher producing lines were basically not rearranged. Clone CMaplh, our best producing line, based on expression levels and growth properties, contains 25-50 copies of both the cy and /3 subunit DNA. The ratio of a/P DNA present within each of the transformants varied widely ( 2). Northern blot analyses demonstrated that the sizes of the major transcripts for both cy and p are about 1.2 kilobases, which is consistent with the size of a transcript using the regulatory signals of the metallothionein gene (results not shown).
Biological Activity of Expressed Materials-In uitro bioassays showed that the recombinant hCG possessed a bioactivity very similar to the urinary hCG standard, including the finding of parallel slopes in the competitive receptor binding assay (Fig. 3). An in vivo bioassay using ascorbic acid depletion also indicated that the expressed hCG had similar biological activity to that of standard hCG (Fig. 4). Since the concentration of recombinant hCG was determined by immunoassay, a precise unitage comparison between standard hCG and the recombinant form is not presented in this report.
Purification of Expressed Materials-Immunoaffinity purification of sufficient quantities of expressed hCG products for structural analyses was accomplished by the use of high capacity IgG-Sepharose columns constructed with 30-59 mg (3-8 mg/ml gel) of purified specific monoclonal IgGs (26). The crude medium was absorbed batchwise. Similarly, washing of the gel with water and elution with 2 M acetic acid was also accomplished batchwise. We elected to elute with acetic acid because of its volatility. A final fast protein liquid chromatography gel filtration step served to remove any contaminant IgG leached from the immunoabsorbant gel as well as to separate expressed products with different molecular weights. Indeed, expressed cy could be segregated into a predominantly high molecular weight cy and a small component (5%) of normal standard sized cy (Fig. 5).

Characterization of Purified Expressed
Products-Each of the purified expressed proteins was examined on SDS-gel. electrophoresis. It was immediately apparent that the subunits of the expressed hCG, as well as the expressed excess p subunit, all migrated exactly as did the standard urinary hCG proteins (Fig. 7). In the system expressing dimeric hormone, a 10-fold or greater production of free /3 subunit was observed which had a size similar to standard urinary p subunit. Structural analyses of the dimeric hormone and the excess free p subunit, produced by the same system, were indistinguishable from urinary standard equivalents (Tables VI and VII).
In contrast, the cy subunit, when expressed in the absence of its complementary p subunit, migrated at a molecular

Characterization of Expressed hCG and Subunits
weight approximately 4000-5000 greater than the standard urinary a subunit (Fig. 7). This higher molecular weight was consistent with its elution volume from the Superose 12 gel filtration column. The possible causes of the higher molecular weight of expressed a were individually explored and are enumerated as follows: 1) extra NHz-terminal or COOHterminal amino acids; 2) 0-glycosylation at Thr-39 as reported by Parsons et al. (12) and Cole et al. (40) for excess pituitary a or a expressed in tissue culture; or 3) extra N-glycosylation, i.e. tri-or tetraantennary oligosaccharide chains. The question of whether extra amino acids were present within expressed a was readily addressed by NHz-terminal sequence analyses and carboxypeptidase digestions. Automated sequence analysis indicated that expressed a contained the expected NHz-terminal structure with none of the heterogeneity that has been observed for the a subunit polypeptide sequence in protein isolated from pregnancy urine (see Table  111). Carboxypeptidase digestions showed that the expected a COOH terminus was intact (Table IV). Amino acid analyses of both the expressed and the urinary a were very similar (Table 11). Therefore, we assumed that carbohydrate differences were responsible for the higher molecular weight of the individually expressed a! subunit. Carbohydrate analysis of the expressed a subunit did, indeed, confirm the presence of carbohydrate differences in comparison to standard urinary a ( Table V).
The strategy employed to define the location of the carbohydrate difference, i.e. 0versus N-glycosylation, was based on earlier data characterizing tryptic digestion of the native a subunit as described by Birken et al. (28). The expressed a subunits, as well as urinary a standards, were each digested with trypsin, reduced and S-carboxymethylated, and applied directly to reverse phase HPLC (Fig. 8). All of the peptide peaks were isolated and analyzed by NHp-terminal sequence analysis (Table VIII; data for noncarbohydrate-containing peptides not shown), SDS-gel electrophoresis (Fig. 9), and carbohydrate analysis (Table V). It was observed that peptide a 3 6 4 2 from expressed a behaved on HPLC exactly the same as this peptide from urinary a. Thus, we concluded there was no 0-glycosylation at Thr-39 (Fig. 8). In addition, both peptides were sequenced through position 39 with recoveries of expected quantities of Thr-39, indicating that the hydroxyl group was not significantly substituted with a carbohydrate. Carbohydrate analyses of both expressed and urinary (~36-42 also confirmed the apparent absence of carbohydrate. By contrast, peptide a62-91 from expressed a appeared as a mixture with (Y46-91 and was shown to be responsible for the extra molecular weight of expressed a as determined by SDSgel electrophoretic analysis (Fig. 9). Sequence analysis indicated an identical primary structure of the recombinant carbohydrate-containing peptides when compared to the standard urinary peptide (Table VIII), but the carbohydrate con-
Parentheses show expected values for the hormone isolated from pregnancy urine. Recombinant al-gel filtration fast protein liquid chromatography (Superose 12 column). Recombinant a2e Sequencing data for these peptides is shown in Table VIII. Note that the recombinant peptides also contain 'Preparation A of (Y62-63 contains a minor sequence of a non-carbohydrate-containing a peptide (see Table IX).
See Table IX for details and sequencing data for preparations A and B. This RCM recombinant a is the parent of preparation B peptides. A clean sequence was obtained for 12 cycles reverse phase HPLC subfractionation (Vydac C4,300A column).
minor sequences.
(data not shown). the carbohydrate-containing peptides and The urinary a subunit profile always displays peptide al-nsas a major and minor peak (28). Each of the peaks was subjected to automated sequence analysis, and the identifications are shown on this figure. The sequence data from the carbohydrate-containing peptides are shown in Table VI11 while the data from the other peptides are not shown to conserve space. tents were significantly different (Table V). ' Carbohydrate analyses indicated a more complex sugar structure on the recombinant peptide a52-91, which contained a higher content of glucosamine and galactose (Table V). Further evidence of the complexity of the recombinant a sugar moiety comes from the detailed evaluation of this peptide, which eluted as a triplet in Fig. 8. Amino-terminal sequencing indicated that the last peak of the triplet began a t cy Thr-46 (Table VIII) instead of Asn-52 as expected from the results of digestion of the standard urinary a. Perhaps the larger sugar structure, present on recombinant a, sterically hinders the trypsin from cleaving as expected between residues Lys-51 and Asn-52.
It was also noted that a46.gl behaved as expected in the Sequencer in that residue 52 was not recovered as is usually observed for N-asparagine-substituted amino acids. However, this was not the case for recombinant peptide as2-91 from which a significant quantity of asparagine was recovered a t the first Sequencer step in contrast to the analogous peptide from the urinary a subunit. A second preparation of recombinant cy yielded less of the peptide a52-91. It is likely that there is heterogeneity in the synthesis of the carbohydrate in the recombinant preparation and that a small quantity may be devoid of carbohydrate a t one of the glycosylation sites. This was investigated further by additional isolations of peptides from the recombinant a subunit and cleavage of the carbohydrate-containing peptides into peptides containing only one carbohydrate moiety. Intact RCM recombinant CY was also cleaved by trypsin to produce such peptides. Each of the newly isolated peptides was sequenced (Table IX) and analyzed for carbohydrate content. The composition results presented in Table V indicate that both carbohydrate moieties in the recombinant a subunit are similar and are likely to be triantennary. This was further supported by the results of periodate oxidation studies (Table  V). Complete loss of a single mannose residue with only a small loss of galactose after periodate oxidation indicates the presence of triantennary chains which are heterogeneously sialylated. The higher sialic acid content of recombinant a (Table V) and the more acidic isoelectric profile ( Fig. 6) is also consistent with a triantennary structure rather than the biantennary branching known to be present in urinary a. Furthermore, recombinant a was resistant to endoglycosidase F cleavage, while the biantennary urinary a was readily digested (Fig. 10).
Carbohydrate analyses of all of the other hCG expression products were similar to the standard urinary preparations (Table V). Since amino acid and structural analyses of all other expressed products showed them to be identical to those of the urinary standards (Tables VI and VII), we concluded that the only difference in the recombinant materials occurs when a subunit is expressed alone.
The sole structural change is the addition of extra sugars to both of the N-asparagine-linked oligosaccharide chains (a Asn-52 or a Asn-78). This is in concurrence with the accompanying paper (19). Blithe and Nisula (18) also found that a preparation of free a subunit isolated from pregnancy urine resembled the expressed free a described in this report by its altered carbohydrate content as well as its inability to combine with p subunit (Fig. 11).
Cole et al. (40) reported that cultured JAr cells secreted hCG as well as free subunits but that only a subunit was hyperglycosylated, while the / 3 subunit was of normal size. In that case a was 0-glycosylated. There are numerous reports in the literature, both concerning clinical situations as well as in vitro cell systems, describing production of free hCG subunits (12-17). In many cases a large a subunit is reported but never an abnormally large / 3 subunit. It is known that the conformation of the a subunit is altered upon combination with p (45,46). Since the hCG subunits start combining prior to leaving the rough endoplasmic reticulum (47-49), it is likely that some of the induced conformational changes alter the environment of the carbohydrate oligosaccharide core structures of the a subunit so that an additional N-acetylglucosamine residue cannot be added, thus blocking triantennary chain formation (50). When a subunit does not combine with its complementary p, the sugar residue can be added resulting in formation of additional antennae. The p subunit does not appear to undergo such changes in its free state.
These results and that of the accompanying communication (19) show the importance of the conformation of a protein to the structure of its post-translation end product. The Nasparagine-linked oligosaccharide branches on the a subunit are biantennary when a combines with its / 3 complement prior to completion of its biosynthesis. In contrast, we have shown that in the same cell system, a expressed without fi subunit contains triantennary instead of normal biantennary carbohydrate structures. While the presence of abnormal carbohydrate structures on the a subunit produced by in vitro or in vivo systems has been known for some time, this is the first demonstration that a is abnormally glycosylated when synthesized in the absence of / 3 subunit.