Expression of human thyrotropin in cell lines with different glycosylation patterns combined with mutagenesis of specific glycosylation sites. Characterization of a novel role for the oligosaccharides in the in vitro and in vivo bioactivity.

We used a novel approach to study the role of the Asn-linked oligosaccharides for human thyrotropin (hTSH) activity. Mutagenesis of Asn (N) within individual glycosylation recognition sequences to Gln (Q) was combined with expression of wild type and mutant hTSH in cell lines with different glycosylation patterns. The in vitro activity of hTSH lacking the Asn alpha 52 oligosaccharide (alpha Q52/TSH beta) expressed in CHO-K1 cells (sialylated oligosaccharides) was increased 6-fold compared with wild type, whereas the activities of alpha Q78/TSH beta and alpha/TSH beta Q23 were increased 2-3-fold. Deletion of the Asn alpha 52 oligosaccharide also increased the thyrotropic activity of human chorionic gonadotropin, in contrast to previous findings at its native receptor. The in vitro activity of wild type hTSH expressed in CHO-LEC2 cells (sialic acid-deficient oligosaccharides), CHO-LEC1 cells (Man5GlcNAc2 intermediates), and 293 cells (sulfated oligosaccharides) was 5-8-fold higher than of wild type from CHO-K1 cells. In contrast to CHO-K1 cells, there was no difference in the activity between wild type and selectively deglycosylated mutants expressed in these cell lines. Thus, in hTSH, the oligosaccharide at Asn alpha 52 and, specifically, its terminal sialic acid residues attenuate in vitro activity, in contrast to the previously reported stimulatory role of this chain for human chorionic gonadotropin and human follitropin activity. The increased thyrotropic activity of alpha Q52/CG beta suggests that receptor-related mechanisms may be responsible for these differences among the glycoprotein hormones. Despite their increased in vitro activity, alpha Q52/TSH beta, and alpha Q78/TSH beta from CHO-K1 cells had a faster serum disappearance rate and decreased effect on T4 production in mice. These findings highlight the importance of individual oligosaccharides in maintaining circulatory half-life and hence in vivo activity of hTSH.

Thyrotropin (thyroid-stimulating hormone, TSH) 1 is a mem-ber of the glycoprotein hormone family, which also includes chorionic gonadotropin (CG), lutropin (luteinizing hormone, LH) and follitropin (follicle-stimulating hormone, FSH). These hormones are structurally related heterodimers consisting of a common ␣ subunit and a distinct ␤ subunit that confers the biological specificity for each hormone (1). The common ␣ subunit bears two N-linked oligosaccharides, and the ␤ subunit bears either one (in TSH and LH), or two (in CG and FSH) (2,3). The oligosaccharides, which represent 18 -35% of total weight (2)(3)(4), have been shown to play a role in the proper folding, assembly, secretion, metabolic clearance, and biological activity of these hormones (for recent review, see Refs. 3 and 4). In the case of hCG and hFSH, enzymatic or chemical deglycosylation led to a decrease or loss of cAMP production and steroidogenesis, while high affinity binding was maintained (3)(4)(5). Sairam et al. (6) reported that the carbohydrates of the common ␣ subunit rather than those of the ␤ subunits were important for the activity of these hormones. Using site-directed mutagenesis, Matzuk et al. (7) identified the oligosaccharide at position 52 of the ␣ subunit to be critical for the in vitro bioactivity of hCG. Subsequently, this oligosaccharide was shown to be similarly important for the stimulatory activity of hFSH (8,9).
In contrast to hCG and hFSH, the roles of the individual oligosaccharides for the activity of TSH or the more closely related LH are not known. Due to the limited availability of purified pituitary human TSH (phTSH), studies on the role of the carbohydrates for TSH have mostly used pituitary bovine TSH. Similar to the findings for the gonadotropins, these studies have shown by chemical or enzymatic deglycosylation that the oligosaccharides, and particularly those of the ␣ subunit, were required for full in vitro activity of pituitary bovine TSH (10,11). The few studies on phTSH have yielded conflicting results and the role of carbohydrates for hTSH action remained controversial (12,13). The recent availability of recombinant human TSH (rhTSH) (14) allowed further investigation of the role of the carbohydrates in the action of hTSH. rhTSH expressed in CHO-K1 cells contains N-linked oligosaccharides which terminate with Sia␣2-3Gal␤1-4GlcNAc␤1-2Man␣ (14 -16). By comparison, phTSH, which physiologically occurs in a variety of glycosylation isoforms, terminates both in SO 4 -4GalNAc␤1-4GlcNAc␤1-2Man␣ and Sia␣2-3Gal␤1-4GlcNAc-␤1-2Man␣ as described by Green and Baenziger (2). Interestingly, enzymatic removal of terminal sialic acid residues of rhTSH increased the in vitro bioactivity of the hormone (17) similar to findings for recombinant bovine LH (15), but unlike for hCG, in which sequential deglycosylation resulted in a stepwise reduction of activity (18). These findings suggested that the role of the oligosaccharides may be different for hTSH compared with hCG and hFSH.
In the present study we have used site-directed mutagenesis to study the role of individual oligosaccharides of hTSH by selectively inhibiting their cotranslational attachment. We combined site-directed mutagenesis with the expression of the selectively deglycosylated hTSH mutants in different cell lines producing hormones with distinct carbohydrate patterns (19 -21), using our recently developed and optimized transient transfection protocol (22). This novel approach allowed us to identify unique roles for individual oligosaccharides and their terminal sialic acid residues for the in vitro as well as in vivo activity of hTSH.  (Torrance, CA); rhTSH, expressed and purified from stably transfected CHO cells from the Genzyme Corp. (Framingham, MA). Cell culture media and reagents were purchased from Life Technologies, Inc. 125 I-cAMP (specific activity, 40 -60 Ci/g) was from Hazleton (Vienna, VA), polymerase chain reaction reagents were from Boehringer Mannheim and New England Biolabs (Beverly, MA), neuraminidase attached to beaded agarose was from Sigma, immobilized Limax flavus agglutinin was from EY Laboratories (San Maeto, CA), and recombinant N-glycanase was from Genzyme Corp. CHO-K1, CHO-LEC2, CHO-LEC1, and 293 human embryonic kidney cells (293 cells) were obtained from ATCC (Rockville, MD). CHO-LEC2 and CHO-LEC1 cells were deposited there by Dr. P. Stanley, Albert Einstein College of Medicine, New York, NY (20,21,23).
Site-directed Mutagenesis-Mutagenesis of the full-length human ␣ cDNA in the bacteriophage M13 mp18 was performed using M13-based site-directed mutagenesis as described previously (8). After subcloning the wild type and mutant ␣ constructs into pAXNeoRx, the resulting constructs were confirmed by DNA sequencing. Mutagenesis of the hTSH␤ minigene (24) was accomplished by the polymerase chain reaction-based megaprimer method (25). Briefly, a primer encompassing the desired mutation (AAC to CAA, TSH␤ Asn 23 to Gln) was used to amplify a 196-base pair 5Ј fragment of the gene, and this fragment (the megaprimer) was used in a second polymerase chain reaction together with a 3Ј primer to amplify the entire gene. After subcloning into the pLBCMV expression vector (24), the entire polymerase chain reaction product was sequenced to verify the mutation and to rule out any undesired polymerase errors.
Transient Expression-CHO-K1 cells were maintained as described (22), CHO-LEC cell lines 2 and 1 were grown in ␣-modified minimum essential medium supplemented with 5% fetal calf serum, penicillin (50 units/ml), streptomycin (50 g/ml) and glutamine (4 mM). 293 cells were grown in Dulbecco's minimum essential medium containing 10% fetal calf serum supplemented as above. Cells were transiently cotransfected in 60-mm culture dishes with wild type or mutant pAXNeoRx/␣, wild type or mutant pLBCMV/TSH␤ minigene, or pcDNAI/neo/hCG␤ using a liposome formulation (Lipofectamine reagent, Life Technologies, Inc.) in a protocol modified according to the manufacturer's instructions (22). After culture in CHO-serum-free medium (Life Technologies, Inc.) for 48 h, supernatants, including control medium from mock transfections using the expression plasmids without gene inserts, were harvested, concentrated with Centriprep 10 concentrators (Amicon, Beverly, MA), and stored at Ϫ70°C to prevent neuraminidase digestion.
Immunoassays-Wild type and mutant hTSHs were quantified using four different hTSH immunoassays. Two third-generation assays utilizing different monoclonal antibodies (Nichols Institute, San Juan Capistrano, CA, and ICN, Costa Mesa, CA) were used, following the manufacturers' instructions. Further, wild type and mutant hTSHs were measured with two different polyclonal antibodies by radioimmunoassay using polyclonal antibodies NIADDK-anti-hTSH-3 and NIDDK-anti-h␤TSH-IC-2 (22) with rhTSH (Genzyme Corp.) (14) as the standard. Intracellular immunoreactivity in the cell lysates was determined after four freeze-thaw cycles on a methanol/dry ice mix and in a 37°C water bath, respectively. Wild type and mutant hCG was measured with a specific third-generation immunoassay without cross-reactivity to other glycoprotein hormones (Nichols Institute).
Gel Filtration-Conditioned media were chromatographed on a Superdex 75-HPLC column (Pharmacia Biotech Inc.), and eluted at a flow rate of 0.3 ml/min in phosphate-buffered saline (pH 7.4), using hCG as internal standard. hTSH and hCG immunoreactivities were monitored by immunoassay (Nichols Institute).
SDS-Polyacrylamide Gel Electrophoresis and Western Blotting-Conditioned media were concentrated, fractionated on ConA-Sepharose (Pharmacia), reconcentrated, and denatured by boiling in 0.25% SDS, 0.5% ␤-mercaptoethanol. After digestion with N-glycanase, samples were resolved on 14% Tris/glycine polyacrylamide gels, transferred to nitrocellulose membranes, and incubated overnight with a rabbit antibody directed against the hTSH ␣ subunit. Antigen-antibody complexes were subsequently visualized by chemiluminescence using a horseradish peroxidase-coupled anti-rabbit IgG and a luminol substrate (Boehringer Mannheim).
Enzymatic Deglycosylation-Concentrated conditioned media were incubated with 250 microunits of neuraminidase/10 mg of total protein in 100 mM sodium acetate, pH 5.0, for 12 h at room temperature, followed by 1 h at 37°C. After separation of the agarose-beaded neuraminidase by spinning in a microcentrifuge, media were washed, concentrated, and reassayed for hTSH immunoreactivity. Similarly, Nglycanase digestion was performed for 20 h at 37°C with 12.5 units/10 mg of protein in 50 mM sodium phosphate, pH 7.6.
cAMP Production-Confluent CHO cells stably expressing the rhTSH receptor (JP09) (26) and FRTL-5 cells expressing the endogenous rat TSH receptor (27) were incubated in 96-well tissue culture plates for 2 h at 37°C, 5% CO 2 with serial dilutions of wild type and mutant hTSH or hCG as well as control medium from mock transfections. The amount of cAMP released into the medium was assayed by radioimmunoassay (22).
Radioreceptor Assay of hTSH-The receptor-binding activity of wild type and hTSH mutants was determined by their ability to displace 125 I-bTSH from a solubilized porcine thyroid membrane receptor preparation (Kronus, Dana Point, CA) as described previously (11).
In Vivo Bioassay-The in vivo bioactivity of various mutants was determined using a bioassay recently developed and characterized in our laboratory (28). Briefly, male albino swiss Crl:CF-1 mice were given 3 g/ml of T 3 (Sigma) in their drinking water for five to six days to suppress endogenous TSH secretion. Wild type and mutant hTSH as well as mock-concentrated medium at equal volumes was injected intraperitoneally, and blood samples for determination of T 4 (T 4 Kit, Nichols Institute) and hTSH values were obtained from the orbital sinus 6 h later.
Serum Disappearance Rate-The serum disappearance rate of wild type hTSH and selected mutants was determined in the rat by injecting 200 -300 ng of the recombinant hormones intravenously and then measuring serum hTSH concentrations from 1-120 min. Experimental details of this procedure are given elsewhere (16). RESULTS hTSH bears three N-linked oligosaccharide chains at positions Asn 52 and Asn 78 of the ␣ subunit and at Asn 23 of the ␤ subunit (1-3). To prevent the cotranslational attachment of individual oligosaccharides to the hTSH molecule, we disrupted the respective glycosylation recognition sequences NX(T/S), widely considered to be an absolute requirement for glycosylation to occur. In the ␣ subunit (individually and in a composite mutation) and in the TSH␤ subunit we mutated Asn at each position to Gln, thus creating genes coding for the following mutant subunits: ␣Q52, ␣Q78, ␣Q52.Q78, and TSH␤Q23. The conservative mutation of Asn to Gln is unlikely to affect protein conformation and has not been reported to influence the tertiary structure of the related glycoproteins hCG and hFSH, when used to generate mutants lacking individual oligosaccharide attachment sites (7)(8)(9). Thus, the observed effects on the hTSH expression, binding, and activity should be primarily due to changes in the carbohydrate chains.
The shift in the elution profiles of the hTSH mutants observed using gel filtration (Fig. 1A) and the differences in gel migration using Western blotting (Fig. 1B) were consistent with the absence of individual oligosaccharides. After N-glycanase digestion, wild type hTSH, ␣Q52/TSH␤, and ␣Q78/ TSH␤ had similar profiles (Fig. 1, A and B), further demonstrating that the differences prior to N-glycanase treatment were due to differences in glycosylation.
Quantitation of Wild Type and Mutant hTSH-To rule out any potential changes in the specific epitope recognized by an antibody and to avoid inaccurate quantitation of the hTSH mutants, we used four different immunoassays to quantitate wild type and mutant hTSH from the different cell lines (see "Experimental Procedures"). Comparable results were obtained for the selectively deglycosylated hTSH mutants in all four immunoassays. In addition, the slopes of the displacement curves of wild type hTSH were parallel to those for the hTSH mutants using the Nichols immunoassay (data not shown).
hTSH Expression-By cotransfecting wild type and mutant human ␣ and TSH␤ subunits in various combinations into CHO-K1 cells, we generated the following recombinant proteins: ␣Q52/TSH␤, ␣Q78/TSH␤, ␣Q52.Q78/TSH␤, ␣/TSH␤Q23, ␣Q52/TSH␤Q23, and ␣Q78/TSH␤Q23. Elimination of the carbohydrate at a single site reduced expression 2.5-10-fold compared with wild type, and deletion of more than one site decreased levels even further ( Table I). The concomitant decreases of intracellular and secreted immunoreactivity suggest that these reductions were not due to intracellular retention of the hTSH mutants. In fact, Ͼ90% of total hTSH immunoreactivity was detected in the medium in all cases, indicating efficient secretion of the mutants. This is in accord with the proposed role of the oligosaccharide chains in early cotranslational events in promoting proper folding and maintaining the intracellular stability of TSH (4,29). Interestingly, we could not detect any significant production of completely deglycosylated hTSH (␣Q52.Q78/TSH␤Q23). This is in agreement with previous findings on hFSH (8) and emphasizes the essential role of carbohydrates in the biosynthesis of the glycoprotein hormones. Similar reductions of the mutants tested were also observed in 293, CHO-LEC2, and CHO-LEC1 cells (data not shown), indicating that these reductions were independent of differences in the overall carbohydrate pattern.
In Vitro Bioactivity: cAMP Induction-Next, we assessed the ability of wild type and mutant hTSH expressed in CHO-K1 cells that produce highly sialylated carbohydrate chains (15,16) to stimulate cAMP production in CHO cells stably transfected with the rhTSH receptor, JP09 (26). Whereas the maximal cAMP stimulatory activity of the mutants was unchanged, there was a highly significant (p Ͻ 0.001) 6-fold, 2-fold, and 3-fold lower EC 50 of mutants ␣Q52/TSH␤, ␣Q78/TSH␤, and ␣/TSH␤Q23 compared with wild type (Fig. 2, Table II). Further, ␣Q52/TSH␤ was significantly more active than ␣Q78/ TSH␤ and ␣/TSH␤Q23 (p Ͻ 0.001). Due to the dramatically reduced expression levels, it was not possible to obtain full dose-response levels for mutants lacking more than one carbohydrate chain. However, there was no significant cAMP stimulation of mutant ␣Q52.Q78/TSH␤ at 0.5 ng/ml, whereas wild type hTSH at the same concentration induced cAMP 3-fold over base-line levels, indicating that the presence of at least one oligosaccharide of the ␣ subunit is required to activate the TSH receptor. This is in accord with previous findings on pituitary bovine TSH, which suggested that enzymatic deglycosylation of the ␣ subunit led to a decrease of bioactivity (11). At 2 ng/ml, ␣Q52/TSH␤Q23 was twice as active as wild type hTSH (p ϭ 0.02, n ϭ 3), whereas there was no difference between wild type hTSH and ␣Q78/TSH␤Q23.
In FRTL-5 cells expressing endogenous rat TSH receptor (27), we observed, relative to wild type (EC 50 ϭ 12.6 Ϯ 2.1 Conditioned media from transfected CHO-K1 cells were chromatographed using hCG as an internal standard (the peak of hCG immunoreactivity corresponds to fraction 0 in all cases). Fraction size was 0.175 ml. hTSH immunoreactivity was monitored by immunoassay (Nichols Institute). B, Western blotting analysis of ConA-Sepharosefractionated hTSH ␣ deglycosylation mutants before (Ϫ) and after (ϩ) N-glycanase digestion (see "Experimental Procedures"), using an antibody against the hTSH ␣ subunit. rhTSH (Genzyme Corp.) served as the internal standard. In the case of N-glycanase-treated wild type and ␣Q52/TSH␤, presumably due to incomplete enzymatic digestion, two bands were visible, representing hTSH ␣ subunit lacking one or two oligosaccharides, respectively.  (Fig. 3). We next expressed mutants ␣Q52/TSH␤, ␣Q78/TSH␤, and ␣/TSH␤Q23 in 293 cells. In contrast to CHO-K1 cells, 293 cells express N-acetylgalactosaminyl-transferase and GalNAc␤1,4GlcNAc␤1,2Man␣4-sulfotransferase and produce oligosaccharides terminating in Ͼ70% in sulfated N-linked carbohydrate moieties (19). Since we previously found that 293 cells produce small amounts of free ␣ subunit (29), we initially cotransfected the hTSH␤ minigene alone to assess whether significant amounts of wild type hTSH would be produced. This, however, was not the case, indicating that significant contamination of the mutants with wild type hTSH did not occur. Wild type hTSH, expressed in 293 cells had increased cAMP inducing activity, evidenced by a 5.6-fold left shift in the EC 50 compared with wild type expressed in CHO-K1 cells (Fig.  4, Table II) (p Ͻ 0.001). However, in relation to the increased wild type activity, there was no difference in activity between wild type and the selectively deglycosylated hTSH mutants (Fig. 4, Table II).
To further understand which carbohydrate components may be associated with the increase of bioactivity upon expressing the hTSH mutants in CHO-K1 cells, we also transiently transfected CHO-LEC2 and CHO-LEC1 cells. These are CHO glycosylation mutant cell lines selected from the wild type CHO-K1 cell line for resistance to toxic plant lectins (23). CHO-LEC2 cells have a defect in the CMP-sialic acid translocation into the Golgi resulting in the synthesis of glycoproteins with a Ͼ 90% reduction in sialic acid (20), whereas CHO-LEC1 cells lack N-acetylgalactosaminyl-transferase I and hence give rise to oligosaccharides bearing Man 5 GlcNAc 2 intermediates (21). The reduction of sialic acid on the CHO-LEC2-expressed hTSH was evidenced by a decrease in L. flavus agglutinin binding, a lectin from the slug L. flavus that recognizes terminal sialic acids (30) (53 Ϯ 8.3% compared with 94.5 Ϯ 2.7% in the case of CHO-K1expressed hTSH) as well as by an accelerated elution profile from L. flavus agglutinin minicolumns. Further, neuraminidase digestion decreased L. flavus agglutinin binding of hTSH from CHO-K1 cells to levels similar to nondigested CHO-LEC2expressed hTSH. As expected, neuraminidase digestion did not further decrease L. flavus agglutinin binding of hTSH from CHO-LEC2 cells (data not shown).
Wild type hTSH expressed in both CHO-LEC2 and CHO-LEC1 was 7-8-fold more active than wild type from CHO-K1 cells (p Ͻ 0.001) (Fig. 5A, Table II). Compared with this increased wild type activity, there was no difference in the EC 50 of cAMP induction for the various selectively deglycosylated mutants expressed in these cells.
Direct evidence for the inhibitory roles of terminal sialic acids was obtained by testing for cAMP stimulation of wild type and mutant hTSH after neuraminidase digestion (Fig. 5B). As expected, neuraminidase digestion of wild type hTSH from CHO-K1 cells increased its in vitro activity to an EC 50 similar to CHO-LEC2-expressed hTSH. Furthermore, digestion of ␣Q78/TSH␤, which retains the sialylated Asn ␣52 oligosaccharide also increased in vitro activity to similar levels. Conversely, the already maximally increased in vitro activity of ␣Q52/TSH␤, in which the Asn ␣52 sialic acid moieties are absent, was not further increased. As expected, neuraminidase digestion did not further increase the activity of hTSH from CHO-LEC2 cells (data not shown).
In Vitro Bioactivity: Growth Induction of FRTL-5 Cells-Similar to the findings of cAMP stimulation of ␣Q52/TSH␤, the ability of this mutant from CHO-K1 cells to induce growth in FRTL-5 cells was increased compared with wild type hTSH (Fig. 6).
Thyrotropic Activity of Site-specifically Deglycosylated hCG-In a concentration up to 8 g/ml we did not observe any cAMP stimulation of CHO-K1-expressed hCG in JPO9 cells that express the rhTSH receptor (Fig. 9). This is consistent

FIG. 2. cAMP induction in CHO cells expressing the rhTSH receptor (JP09) by the hTSH mutants produced in CHO-K1 cells.
Increasing concentrations of wild type or mutant hTSH were incubated with JP09 cells, and the cAMP concentration in the resulting supernatants was assayed by radioimmunoassay. The amount of cAMP released from the cells in the presence of concentrated medium from mocktransfected cells was not different from base-line levels (buffer only). A representative experiment, repeated at least twice, is shown. Values from triplicate determinations are depicted as mean Ϯ S.E.

TABLE II Cell line-dependent cAMP induction of hTSH mutants
The ability of wild type and mutant hTSH expressed in different cell lines to induce cAMP production in CHO cells stably transfected with the rhTSH receptor was assessed. There were no significant differences in the maximal cAMP stimulatory activity for wild type and mutants in these cell lines. n, the number of independent experiments, each performed in triplicate determinations; values are given as mean Ϯ S.E. *, p Ͻ 0.001 compared with wild type of the same cell line. ND, not determined.
with recent findings of Yoshimura et al. (31), who, using purified hCG in the same system, showed that 50 g/ml was required for significant cAMP production. In contrast, there was a dose-dependent significant stimulation with ␣Q52/hCG␤ in the same concentration range, whereas deletion of the oligosaccharide at Asn ␣78 , tested in a concentration up to 6 g/ml, did not increase the thyrotropic activity of hCG.
In Vivo Bioactivity-Injection of wild type hTSH, ␣Q52/ TSH␤, and ␣Q78/TSH␤ from CHO-K1 cells into T 3 -suppressed mice led to a dose-dependent stimulation of T 4 production that was significant compared with mock at 100 ng/mouse (p Ͻ 0.05) (Fig. 10). Wild type hTSH was significantly (p Ͻ 0.05) more potent than either ␣Q52/TSH␤ or ␣Q78/TSH␤. Determination of hTSH levels in the serum of these mice indicated that, compared with wild type hTSH, the relative amounts of injected ␣Q52/TSH␤ and ␣Q78/TSH␤ remaining were decreased by 60 -70%.
Serum Disappearance Rate-To further establish whether site-specific deletion of the ␣-linked oligosaccharides may lead to a reduced circulatory half-life, we tested the serum disappearance rate of these mutants from CHO-K1 cells after intravenous injection into rats. Whereas the serum disappearance rate of ␣Q52/TSH␤ was slightly increased (p ϭ 0.057), ␣Q78/ TSH␤ was cleared significantly faster than wild type hTSH (p Ͻ 0.05) (Fig. 11). These differences were not observed in the first 15 min after intravenous injection but only after this initial phase, in keeping with previous findings that the carbo-  (Table II) hydrates are not important for the initial phase of tissue distribution but are involved in hepatorenal clearance mechanisms (16,17). DISCUSSION The glycoprotein hormones hTSH, hCG, hLH, and hFSH share a common signal transduction pathway involving interaction with a hormone-specific G protein-coupled receptor and the generation of cAMP as the initial event in the signal transduction cascade (3,4). Previous studies on hCG (7) and hFSH (8,9) had indicated a site-specific requirement of the oligosaccharide at Asn ␣52 for their in vitro activity, but similar studies on hTSH and the more closely related hLH have not been performed. Moreover, previous studies using recombinant DNA techniques to study oligosaccharides in glycoprotein hormones had been limited to using either site-directed mutagenesis of glycosylation attachment sites or glycosylation mutant cell lines to synthesize these hormones. Therefore, in a novel approach to study the role of carbohydrates for hTSH, we combined site-specific deletion of individual oligosaccharides with the expression of recombinant hormones in a variety of cell lines that differentially process the carbohydrate moieties.
In this study we provide evidence for a different role of the oligosaccharides for hTSH bioactivity, distinct from their role for hCG and hFSH. Site-specific disruption of each of the three individual carbohydrate attachment sites led to an increase in the in vitro activity of hTSH expressed in CHO-K1 cells producing sialylated carbohydrates (14 -16). Our findings using a sialic acid-binding lectin (30) as well as enzymatic desialylation show that the hTSH transiently expressed in CHO-K1 cells was indeed sialylated. Further, its in vitro bioactivity as well as its circulatory half-life was indistinguishable from that of rhTSH (Genzyme Corp.), 2 which contains 1.8 -2.2 sialic acid residues/chain (14,16). Interestingly, this increase of in vitro activity upon site-specific deglycosylation was most pronounced upon deletion of the carbohydrate at Asn ␣52 . In contrast, this Asn ␣52 oligosaccharide had previously been shown to be essential for receptor stimulation of hCG and hFSH, which were also expressed in CHO-K1 cells (7)(8)(9). The increase of cAMP-inducing activity in our study occurred to a similar degree with the recombinant hTSH receptor and the endogenous rat receptor, indicating that these effects were neither systemdependent nor species-specific. Furthermore, in addition to the immediate induction of cAMP synthesis, site-specific deglycosylation of hTSH also affected long term effects, including DNA synthesis by and growth of target cells in a similar fashion.
Wild type hTSH expressed in 293, CHO-LEC2, or CHO-LEC1 cells had a higher in vitro activity than wild type from CHO-K1 cells, and this increased activity was not further augmented by site-specific deglycosylation. CHO-LEC mutant cell lines lack a defined glycosylation activity and accumulate specifically truncated carbohydrates typical of intermediates in the biosynthetic pathway (23). In contrast to CHO-K1 cells, which produce terminally sialylated oligosaccharides, Golgi vesicle membranes from CHO-LEC2 cells translocate CMPsialic acid at only 2% of the rate of vesicles from CHO-K1 cells and thus produce carbohydrates with a greater than 90% decrease in sialic acid content (20). CHO-LEC1 cells lack N-acetylgalactosaminyl-transferase I and hence terminate in Man 5 GlcNAc 2 intermediates (21). 293 cells express Nacetylgalactosaminyl-transferase and GalNAc␤1,4GlcNAc␤1, 2Man␣4-sulfotransferase and produce N-linked carbohydrate moieties terminating in sulfate in more than 70% of the chains (19). The fact that the increase of the in vitro activity upon site-specific deglycosylation was confined to hTSH expressed in CHO-K1 cells therefore indicates that the terminal sialic acid residues attenuate signal transduction of hTSH. This was further supported by our present findings after neuraminidase digestion and is in accord with recent reports on sequential deglycosylation of rhTSH (17) and recombinant bovine LH (15). Importantly, our current approach allowed us to identify sitespecific differences in the role of the terminal sialic acid residues and showed that the residues at the Asn ␣52 oligosaccharide attenuated TSH receptor activation to a much higher degree than those at the Asn ␣78 or Asn ␤23 carbohydrate chain. Further, our data demonstrate that hTSH expressing Man 5 GlcNAc 2 intermediates retained full receptor coupling. In contrast to our findings for hTSH, Keene et al. (32) showed that the activity of wild type hCG expressed in these CHO glycosylation mutant cells was lower compared with hCG from CHO-K1 cells. The same group reported that the in vitro activity of hFSH did not change upon expression in these cell lines (33). These studies, however, did not assess the effects of site-specific deglycosylation on the activity of hCG or hFSH expressed in these cells. Taken together, these findings indicate that the terminal sialic acid residues can differently modulate glycoprotein hormone activity in a hormone-dependent manner.
Our study has further demonstrated a unique role of the carbohydrate at Asn ␣52 for hTSH receptor binding. In previous reports for hCG and hFSH, deletion of the oligosaccharide at Asn ␣52 had no effect (7,8) or led to only a minor increase in receptor binding (9). The nonessential role of oligosaccharides Asn ␣78 and Asn ␤23 in hTSH receptor binding observed in this study is similar to that for the gonadotropins (7)(8)(9). In contrast to the findings on bioactivity, the increased binding of ␣Q52/ TSH␤ was independent of the carbohydrate pattern, suggesting that the increase of activity of ␣Q52/TSH␤ from CHO-K1 cells was not a direct result of the increased binding of this mutant. The molecular mechanism by which carbohydrates activate the receptor is unknown, but in accord with our present findings, is believed to occur at a postreceptor binding step (3,4). An indirect mechanism involving a conformational change of the hormone (35) appears more likely than a direct interaction of the oligosaccharide with the receptor, since a lectin-like component identified in the hCG receptor (34) is not present in the hTSH receptor. Recent reports on the crystal structure of hydrofluoric acid-treated hCG (36,37) located the oligosaccharide at position Asn ␣52 to be in a putative receptor binding region. Since the hormone-specific ␤ subunit influences the conformation of the ␣ subunit in a hormone-specific manner (38), differences in spatial orientation of this oligosaccharide may contribute to its differential role in hTSH compared with hCG or hFSH. This speculation, however, awaits confirmation by structural analysis, since, among the glycoprotein hormones, only hCG has been crystallized to date.
It had previously been reported that the weak thyrotropic activity of hCG (for review, see Ref. 39), which has recently been linked to a direct interaction with the rhTSH receptor (31,40,41), increased upon desialylation (42), in contrast to a decrease at its native receptor (18). When we tested the thyrotropic activity of CHO-K1-expressed hCG site-specifically deglycosylated at the ␣ subunit, we observed an increase of activity of ␣Q52/hCG␤ but not of ␣Q78/hCG␤. This leads to the conclusion that the deletion of the oligosaccharide at Asn ␣52 can have opposite effects on hCG activity, depending on the receptor with which the hormone interacts. Therefore, these findings point to the intriguing possibility that the observed differences in the role of individual oligosaccharides for glycoprotein hormone action may be related, at least in part, to differences in glycoprotein hormone receptor structure and/or to receptordependent differences in receptor-ligand interaction.
The carbohydrate moieties are known to be important for the circulatory half-life and hence the in vivo bioactivity of the glycoprotein hormones (for review, see Ref. 43). However, the role of individual oligosaccharides for the in vivo activity of hTSH has not, to our knowledge, been investigated previously. Since the availability of a hTSH superagonist may have potential clinical applications (44), we were interested to determine whether the increased in vitro activity of sialylated ␣Q52/TSH␤ from CHO-K1 cells could be maintained in vivo, despite the known protective effect of terminal sialic acid for the plasma half-life of rhTSH (16). However, we found a significant decrease in the in vivo activity of ␣Q52/TSH␤ at the highest dose tested, as well as a slightly increased serum disappearance rate. By comparison, the significantly greater relative loss in the in vivo activity of the mutant ␣Q78/TSH␤ was correlated with a significant increase in its serum disappearance rate. This significantly greater decrease in circulatory half-life upon FIG. 11. Serum disappearance rate of hTSH mutants from CHO-K1 cells in male rats. After bolus injection of 200 -300 ng of wild type or mutant hTSH into the femoral vein, blood for hTSH determinations was obtained over 120 min at equal time points. An IRMA without cross-reactivity to rat TSH (Nichols Institute), was used. Immunoreactivity was expressed as percentage remaining, and serum concentration at 0 min was defined as 100%. Wild type, n ϭ 4; ␣Q52/ TSH␤, n ϭ 4; ␣Q78/TSH␤, n ϭ 3. Data were compiled from individual experiments. deletion of the oligosaccharide at Asn ␣78 compared with Asn ␣52 may be related to its peripheral, surface-exposed location, whereas the oligosaccharide at Asn ␣52 appears to be buried at the dimer interface (36,37). These results are comparable with recent findings on the site-specific role of the carbohydrates for the in vivo activity of hFSH (45). The lack of correlation between the in vitro and in vivo activities illustrates the importance of carbohydrates in determining hTSH activity in the whole organism. Further, our findings emphasize that the circulatory half-life, and not the in vitro activity, appears to be the primary determinant of the in vivo bioactivity of these hormones (16,45). It will be interesting to study whether further modifications of the ␣Q52/TSH␤ mutant, e.g. fusion of the hCG␤ carboxyl-terminal peptide to the TSH␤ carboxyl terminus (24,46), can compensate for the increased clearance rate and hence help to generate hTSH mutants with increased in vivo bioactivity.
In conclusion, we have demonstrated that the roles of the terminal sialic acids as well as of individual oligosaccharides are different for hTSH compared with hCG and hFSH. Consistent with our recent observations of the unique importance of the ␣ carboxyl-terminal Ser 92 for hTSH action (22), these findings indicate that conserved structures within the context of a given ligand receptor complex may contribute to signal transduction in different ways. In conjunction with previous findings on the glycoprotein hormones (15-18, 32, 33), we conclude that in hCG, which is exclusively sialylated, sialic acid is required for full expression of in vitro activity. In hFSH, which is predominantly sialylated, sialic acid residues appear to be dispensable for in vitro activity. However, in the case of bLH or phTSH, which bear sialic acids and are predominantly sulfated, expression of terminally sialylated oligosaccharides attenuates effective in vitro receptor activation. In particular, whereas the oligosaccharide at Asn ␣52 is necessary for hCG and hFSH action, the same chain, and specifically its terminal sialic acid residues markedly attenuate TSH receptor binding and activation. As posttranslational modifications of carbohydrates regulate glycoprotein hormone activity in normal physiology, modulation of terminal sialylation of the Asn ␣52 oligosaccharide, which appears more heterogeneous than other side chains (47), may hence be important in regulating activity in a hormone-specific manner. Our observation that ␣Q52/CG␤ also had an increased activity at the hTSH receptor, opposite to the effect at its native receptor, points to differences in receptordependent ligand receptor interactions as a possible explanation for these distinct roles that may have evolved during evolution to maintain the specificity of receptor ligand interactions within the glycoprotein hormone family.