Engineering a Potential Antagonist of Human Thyrotropin and Thyroid-stimulating Antibody*

Thyrotropin (TSH) and the gonadotropins (FSH, LH, hCG) are a family of heterodimeric glycoprotein hormones composed of two noncovalently linked subunits, α and β. We have recently converted the hTSH heterodimer to a biologically active single chain (hTSHβ·CTPα) by fusing the common α-subunit to the C-terminal end of thehTSH β-subunit in the presence of a ∼30-amino acid peptide from hCGβ ( CTP ) as a linker. The hTSHβ·CTPα single chain was used to investigate the role of theN-linked oligosaccharides of α- and β-subunits in the secretion and function of hTSH. Using overlapping PCR mutagenesis, two deglycosylated variants were prepared: one lacking both oligosaccharide chains on the α-subunit (hTSHβ·CTPα1+2) and the other lacking the oligosaccharide chain on the β-subunit (hTSHβ·CTPα(deg)). The single chain variants were expressed in CHO cells and were secreted into the medium. hTSH variants lacking the oligosaccharide chains were less potent than hTSHβ·CTPα wild-type with respect to cAMP formation and thyroid hormone secretion in cultured human thyroid follicles. Both deglycosylated variants competed with hTSH in a dose-dependent manner. The hTSHβ·CTPα1+2variant blocked cAMP formation and thyroid hormone secretion stimulated by hTSH as well as by the antibody, thyroid-stimulating immunoglobulins, responsible for the most common cause of hyperthyroidism, Graves disease. Thus, this variant behaves as a potential antagonist, offering a novel therapeutic strategy in the treatment of thyrotoxicosis caused by Graves' disease and TSH-secreting pituitary adenoma.

Thyrotropin (TSH) 1 is a member of the glycoprotein hormone family, which includes lutropin (LH), follitropin (FSH), and human chorionic gonadotropin (hCG). These are heterodimers composed of two noncovalent-linked subunits, a common ␣-subunit and a hormone-specific ␤-subunit (1,2). Assembly of glycoprotein subunits is vital to the function of these hormones. The ␣and ␤-subunits contain one (TSH␤ and LH␤) or two (␣, FSH␤, and hCG␤) asparagine-linked (N-linked) oligosaccha-rides (1, 2). These residues have been shown to play a role in determining the biological activity of glycoprotein hormones, including the maintenance of intracellular stability, assembly, secretion, signal transduction, and modulation of plasma halflife (1,3). Deglycosylation of glycoprotein hormones have been utilized using chemical or enzymatic treatments, but these however cannot discriminate between individual sites, are nonspecific, and provide only completely deglycosylated hormones.
Site-directed mutagenesis has become an important tool for studying the structure and function of glycoprotein hormones. However, mutations in either ␣or ␤-subunits can alter the folding and ultimately inhibit subunit assembly and secretion of the hormone (4 -6). To overcome these limitations, the genes encoding the common ␣-subunit and either the hCG ␤-, FSH ␤-, or TSH ␤-subunits have been genetically fused. The resulting polypeptide chains were efficiently secreted and were biologically active (7)(8)(9)(10)(11)(12). These studies presumed that addition of the human CG␤ C-terminal peptide (CTP) as a linker sequence between the subunits would be required for flexibility, hydrophilicity, stability, and successful expression of the single chain forms. The CTP contains several proline and serine residues and thus lacks significant secondary structure. This may permit the appropriate interactions between the subunits. In addition, previous studies showed that ligation of the CTP to hFSH␤ (13), hTSH␤ (14), or hCG␣ (15) did not significantly affect assembly or in vitro biological activity, but was important for the in vivo potency of the chimeras.
Assembly of the hTSH ␤and ␣-subunits is the rate-limiting step in the production of the functional heterodimer (16). Therefore, studying the structure and function of hTSH using site-directed mutagenesis may affect assembly of the subunits and production of functional hormone. Thus, converting hTSH to a single chain form could increase the biological half-life and expand the range of TSH structure-function studies. In the present study, we used site-directed mutagenesis to study the role of the N-linked carbohydrates of the ␣and/or ␤-subunits of the hTSH single peptide chain containing the CTP as a linker between the subunits. The in vitro bioactivity of recombinant hTSH and its derivatives were assayed by cAMP formation and triiodothyronine (T 3 ) secretion in a homologous serum-free culture system of human thyroid follicles (17,18 Construction of hTSH Single Chain Variants-The hTSH single chains were constructed using overlapping PCR mutagenesis, as described previously (7,24). For construction of the hTSH␤⅐CTP␣ 1ϩ2 single peptide chain, which was deglycosylated only on the ␣-subunit, the vectors pM 2 hTSH␤⅐CTP␣ and pM 2 ␣ 1ϩ2 were used as templates for PCR. The vectors, pM 2 hTSH␤⅐CTP␣ and pM 2 ␣ 1ϩ2 , were prepared in our laboratory as previously described (4,11). The following oligonucleotides were used for the chimeric construction: primer 1, 5Ј-GTGGGAT-CAGGGGGATCCTAGATTTCTGAGTTA-3Ј; primer 2, 5Ј-CACATCAG-GAGCTTGTGGGAGGATCGG-3Ј; primer 3, 5Ј-ATCCTCCCACAAGCT-CCTCATGTGCAG-3Ј; and primer 4, 5Ј-TGAGTCGACATGATAATTCA-GTGATTGAAT-3Ј. pM 2 TSH␤⅐CTP␣ was a template for primers 1 and 2 (Fig. 1A). Primer 1 contained the TSH␤ 5Ј-end sequence, which includes a newly formed BamHI site, and primer 2 contained the first four codons of the ␣-subunit and a stretch of the 3Ј-end of CTP sequence. Therefore, the newly synthesized fragment contained the entire TSH␤⅐CTP coding sequence and a part of the ␣ sequence. pM 2 ␣ 1ϩ2 was used as a template for primers 3 and 4 to generate a product containing the 3Ј-end of CTP and the ␣ 1ϩ2 fragment. Primer 3 contained the sequence corresponding to the last four C-terminal codons of CTP and the first five codons of the ␣-subunit, and primer 4 contained some of the flanking sequence of the ␣ exon 4 that also included a newly created SalI site. These fragments were used as overlapping templates to synthesize the single hTSH␤⅐CTP␣ 1ϩ2 gene using primers 1 and 4 ( Fig.  1A).
To construct the deglycosylated hTSH␤⅐CTP␣ single chain on ␣and ␤-subunits (hTSH␤⅐CTP␣(deg)), mutant primers 5 and 6 (primer 5, 5Ј-CAGATGGTGGTGTCGATGGTTAGG-3Ј and primer 6, 5Ј-CCTAAC-CATCGACACCACCATCTG-3Ј) were synthesized for mutagenesis of asparagine (Asn) in position 23 of the ␤ sequence to aspartic acid (Asp). The 5Ј-AAC-3Ј triplet coding sequence for Asn was converted to the 5Ј-GAC-3Ј coding sequence for Asp. pM 2 TSH␤⅐CTP␣ 1ϩ2 was used as template DNA, and three PCR reactions were performed for generating the TSH␤⅐CTP␣(deg) coding sequence. The first reaction included primers 1 and 5, the second reaction included primers 6 and 4, and the final reaction included primers 1 and 4, which resulted in the former fragments containing the mutation in position 23 of the ␤-subunit (Fig. 1A).
Construction of Expression Vectors-The eukaryotic expression vector pM 2 ⅐HA is an expression vector that contains the ampicillin (Amp R ) and the neomycin (Neo R ) resistance genes and a strong promoter of the HaMuSV virus, LTR (25,26). The BamHI/SalI fragments containing the TSH␤⅐CTP␣ 1ϩ2 or TSH␤⅐CTP␣(deg) chimeric genes were inserted at the BamHI/SalI cloning site of pM 2 ⅐HA and used for transfection.
DNA Transfection and Clone Selection-Chinese hamster ovary (CHO) cells (wild-type and/or ldlD), were transfected with pM 2 hTSH␤⅐ CTP␣, pM 2 hTSH␤⅐CTP␣ 1ϩ2 , or pM 2 hTSH␤⅐CTP␣(deg) vectors, according to the calcium phosphate precipitation method (4). Cells were selected for insertion of the plasmid DNA by growth in culture medium containing 0.25 mg/ml of the neomycin analog, G418. Transfected colonies resistant to G418 were harvested and screened for the expression of hTSH variants by metabolic labeling of the cells and immunoprecipitation.
Cell Culture-CHO cells were maintained in Medium 1 (Ham's F-12 medium supplemented with penicillin (100 units/ml), streptomycin (100 mg/ml), and glutamine (2 mM) containing 5% fetal calf serum), at 37°C in a humidified 5% CO 2 incubator. Transfected clones were maintained in the above culture medium supplemented with 0.25 mg/ml active G418 (Medium II). For hormone collection, cells secreting hTSH variants were plated and grown to confluency in T-75 flasks. Cells were washed twice with serum-free medium and 12 ml of Medium III (Medium I without fetal calf serum) were added. Medium was collected every 24 h, clarified by centrifugation, and concentrated using centriprep concentrators (Amicon, Corp., Danvers, MA). Concentrations of hTSH variants were determined by hTSH immunoradiometric assay and a double antibody radioimmunoassay (Diagnostic Products Corp., Los Angeles, CA). In addition, medium from nontransfected CHO cells was collected as described above and used as control.
Metabolic Labeling-On day 0, cells were plated into 12-well dishes (350,000 cells/well) in 1 ml of Medium I. For continuous labeling experiments, cells were washed twice with cysteine/methionine-free Medium IV (Medium I supplemented with 5% dialyzed calf serum) and labeled for 5 h in 1 ml of cysteine/methionine-free Medium IV containing 50 Ci/ml [ 35 S]cysteine/methionine mix. For pulse chase experiments, the cells were washed twice and preincubated for 1.5 h with cysteine-free Medium IV, followed by a 20-min pulse-labeling in cysteine/methioninefree Medium IV containing 100 Ci/ml [ 35 S]cysteine/methionine. Pulsechase experiments using ldlD cells were performed in the presence or absence of 10 M galactose or 100 M N-acetyl galactosamine. The labeled cells were then washed twice with Medium IV containing 1 mM unlabeled cysteine/methionine and incubated in this chase medium for the indicated time. Media and cell lysates were prepared, immunoprecipitated using monoclonal antisera against the ␣-subunit and resolved on 15% SDS-polyacrylamide gels as described previously (27).
Tunicamycin Treatment-Cells were plated into 12-well dishes in 1 ml of Medium I. On the second day, medium was changed with medium I containing 2 g/ml tunicamycin. Cells were incubated at 37°C for 1.5 h. At the end of the incubation time, the medium was exchanged with Medium II containing 2 g/ml tunicamycin and 50 Ci/ml of [ 35 S]cysteine/methionine mix. Further analysis proceeded with metabolic labeling as described above.
In Vitro Bioassay-The bioactivity of hTSH variants were determined by measuring their ability to stimulate cAMP formation and T 3 secretion from cultured human thyroid follicles as described previously (17,18). Essentially, human thyroid cells were prepared from colloid tissue obtained at thyroidectomy from patients with benign nodules. 200 ϫ 10 3 plated onto 24-well microtiter plates and incubated with 0.5 ml of serum-free medium (DCCM-1, which contains insulin (1 g/ml), and antibiotics), in the presence or absence of the hTSH variants and cultured for 7 days at 37°C, in an atmosphere of 5% CO 2 in a water- FIG. 1. Construction of hTSH variants. A, mutant forms of hTSH were engineered using overlapping PCR mutagenesis. B, hTSH␤⅐CTP␣ wild type contains the N-linked glycosylation sites at Asn 52 and Asn 78 of the ␣-subunit and at Asn 23 of the ␤-subunit. hTSH␤⅐CTP␣ 1ϩ2 contains one N-linked glycosylation site on Asn 23 of the subunit. hTSH␤⅐CTP␣(deg) has no N-linked glycosylation recognition sites. saturated incubator. For T 3 measurements, potassium iodide (0.1 M) was added to the medium at the start of the culture period. For cAMP measurements, 1-methyl-3-isobutylxanthine (0.5 mM), which inhibits cAMP degradation, was added to the medium. At the end of the culture period, the cAMP and T 3 secreted into the medium (concentrations remaining in the cells were negligible) were measured by radioimmunoassay as described previously (17).
Statistical Analysis-Each experiment was repeated at least three times, and results are presented as the mean Ϯ S.E. of at least three replicate determinations. Statistical analysis of the data was performed using Student's t test and analysis of variance. p Ͻ 0.05 was considered significant.

RESULTS
Oligonucleotide-directed mutagenesis was chosen to examine the functional importance of N-linked oligosaccharides in hTSH single chain (hTSH␤⅐CTP␣) bioactivity. The hTSH␤⅐ CTP␣ contains three N-linked glycosylation sites: two on the ␣and one on the ␤-subunit (Fig. 1B). Mutagenesis of the Asn in the Asn-X-Thr/Ser recognition sequence for Asn-linked glycosylation is sufficient to prevent transfer of the carbohydrate to the protein (4). The coding sequence of Asn (5Ј-AAC-3Ј) at positions 52 and 78 of the ␣-subunit was converted to the coding sequence for Asp (5Ј-GAC-3Ј). This variant was deglycosylated only on the ␣-subunit (pM 2 hTSH␤⅐CTP␣ 1ϩ2 ). To construct deglycosylated variants on the ␣and ␤-subunits (hTSH␤⅐ CTP␣(deg)), the coding sequence for Asn (5Ј-AAC-3Ј) on the ␤-subunit of the hTSH␤⅐CTP␣ 1ϩ2 variant was converted to the coding sequence for Asp (5Ј-GAC-3Ј) (Fig. 1B). The coding sequence of hTSH␤⅐CTP␣ variants was sequenced to verify the mutations and the absence of other sequence alterations.
Stable clonal cell lines expressing hTSH␤⅐CTP␣, hTSH␤⅐ CTP␣ 1ϩ2 , and hTSH␤⅐CTP␣(deg) were selected. Cells were labeled in the presence of [ 35 S]methionine/cysteine mix for 7 h, media and lysates were immunoprecipitated with polyclonal human anti-␣ antiserum, and the proteins were resolved by SDS-polyacrylamide gel electrophoresis. Intracellular (lysate) forms of hTSH␤⅐CTP␣ wild-type and its mutants migrated faster than corresponding extracellular (medium) forms ( Fig.  2A). This is because of the differences in terminal processing of the N-linked oligosaccharides and the addition of the O-linked oligosaccharides prior to secretion. The secreted mutant forms migrated faster than wild-type (19). This is because of their lower content of oligosaccharide chains. To confirm this assumption, CHO cells were treated with tunicamycin. Because tunicamycin prevents the addition of N-linked oligosaccharides to the protein, a difference in mobility is expected between proteins secreted from cells treated with tunicamycin compared with those untreated. The results showed that hTSH␤⅐ CTP␣ and hTSH␤⅐CTP␣ 1ϩ2 variants secreted from cells treated with tunicamycin have, as expected, the same mobility as hTSH␤⅐CTP␣(deg) secreted from treated and untreated cells (Fig. 2B). This indicates that N-linked oligosaccharides are present in the ␣and ␤-subunits of hTSH␤⅐CTP␣ as well as in the ␤-subunit of hTSH␤⅐CTP␣ 1ϩ2 and absent in the hTSH␤⅐ CTP␣(deg) variant.
The secretion kinetics of hTSH variants was determined by pulse-chase analysis and immunoprecipitation with anti-␣ antiserum (Fig. 3). Whereas the hTSH␤⅐CTP␣ (Fig. 3A) and hTSH␤⅐CTP␣ 1ϩ2 (Fig. 3B) were secreted efficiently with a similar t1 ⁄2 of ϳ2 h, the secretion rate of hTSH␤⅐CTP␣(deg) (Fig. 3C) was significantly slower with a t1 ⁄2 of ϳ17 h. These differences were not because of the O-linked oligosaccharides associated with the CTP, because similar results were detected using lDld cells (results not shown), which had a reversible defect in the synthesis of O-linked oligosaccharides (20). The data thus indicate that conversion of hTSH into a single peptide chain together with CTP as a linker allows the heterodimeric-like configuration of the mutated ␣and ␤-subunits, as shown by secretion of the protein into the medium.
The biological activity of hTSH variants was examined by their ability to stimulate cAMP formation and T 3 secretion from cultured human thyroid follicles. Treatment of the cells with increasing concentrations (1-100 microunits/ml) of hTSH single chain resulted in a dose-dependent increase in cAMP formation (Fig. 4A) and triidothyronine (T 3 ) secretion (Fig. 4B). The maximal effect on cAMP formation and T 3 secretion was seen at concentrations of 50 and 5 microunits/ml, respectively.

DISCUSSION
The present study indicates that deglycosylated variants of hTSH single chains are expressed and secreted from CHO cells. These variants are less potent than wild-type hTSH with regard to cAMP formation and T 3 secretion in cultured human thyroid follicles. Moreover, hTSH␤⅐CTP␣(deg) reduced the biological activity of hTSH or hTSI, whereas hTSH␤⅐CTP␣ 1ϩ2 significantly blocked the activity of hTSH and hTSI.
The Asn-linked oligosaccharides have been implicated in several physiologic functions such as maintenance of intracel-lular stability, secretion, biological activity, and modulation of the plasma half-life (1,3). Site-directed mutagenesis has become an effective method to study the role of individual carbohydrate chains on multiglycosylated proteins. However, sitedirected mutagenesis may affect assembly of heterodimer subunits. It has been reported that mutating the hTSH␤-subunit significantly reduced TSH dimer formation (14). Other studies indicate that loss of oligosaccharides from the ␣-subunit reduced assembly and/or stability of hCG (4). To bypass the problem of dimerization of deglycosylated subunits, the subunits ␣ and ␤ were genetically fused in a single chain hormone. Single chains of hCG (7), hFSH (8), and hTSH (11,12) retained a biologically active conformation similar to that of the heterodimer (11,12). Therefore, fusion of ␣and ␤-subunits in a single chain bypasses the assembly of the subunits, which is a rate-limiting step for hormone secretion and bioactivity. It is apparent that despite the single chain structure, correct conformation of the subunits occurs, and the single chains have a normal biological activity (11,12).
For studying the role of N-linked oligosaccharides on hTSH function, we used the single chain of hTSH that contains the CTP as a linker between ␣and ␤-subunits (11). Previous studies indicated that fusing the CTP to the C-terminal end of hFSH␤ (13,21), to the N-terminal of hCG␣ (15), or to hTSH␤ (22) does not affect the assembly, secretion, and signal transduction of the respective dimers compared with the wild type. In addition, it was reported that CTP and associated O-linked oligosaccharides in hCG are not important for receptor binding or in vitro signal transduction but are critical for in vivo biological response (23). Moreover, it has been shown that the secretion of the hFSH single chain increased in the presence of CTP as a linker between the subunits (8). The fact that deglycosylated variants of hTSH are secreted efficiently from CHO and ldlD cells (ldlD cells have a reversible defect in synthesizing oligosaccharide chains, data not shown) indicates that Nlinked and O-linked oligosaccharides are not vital for the secretion of the hTSH single chain. Therefore, the signal for the secretion of the hormone exists in the single chains itself.
The bioactivity of the variants was examined in an in vitro system of human thyroid follicles cultured in suspension, under serum-free conditions, in which the follicular three-dimensional structure is retained (17,18). This bioassay has several advantages over the current methods used for testing thyroid biological activity. First, it allows the measurement of T 3 secretion, the physiologically relevant hormonal end-point response, which is very seldom measured when testing for thyroid biological activity. Second, the cells are of human origin, which is important in view of wide species variation in thyroid response to TSH agonists. The results indicated that deletion of the N-linked oligosaccharides from the single chain diminished biological activity. However, a difference in bioactivity between the variants was apparent. hTSH␤⅐CTP␣(deg) was more potent than hTSH␤⅐CTP␣ 1ϩ2 , and this may be related to the difference in conformation of the variants.
The competition experiments indicated that hTSH␤⅐ CTP␣ 1ϩ2 markedly blocked cAMP formation and T 3 secretion induced by hTSH and hTSI. Therefore, this variant can be considered as a potential antagonist to both hTSH as well as hTSI. It is worth noting that we tested, to the best of our knowledge for the first time, not only TSH but TSI as well, i.e. the factor responsible for the most common cause of hyperthyroidism, Graves' disease, thus substantiating considerably the clinical implications of our study. Therefore, the hTSH␤⅐ CTP␣ 1ϩ2 variant, behaves as a potential antagonist that may offer a novel therapeutic strategy of thyrotoxicosis because of Graves' disease and TSH-secreting pituitary adenoma. The existence of CTP as a linker between the subunits in the hTSH␤⅐CTP␣ 1ϩ2 may prevent rapid degradation in vivo and increase their half-life in the circulation. The therapeutic efficacy of this analog needs to be established in in vivo studies and clinical trials.
Acknowledgments-We thank Dr. Irving Boime (Washington University, St. Louis, MO) for his constructive comments regarding the manuscript. We would also like to thank Orit Sadeh and Dr. Ronit Heinrich for assistance in the bioassays.