INTRODUCTION

Thyroid-stimulating hormone (TSH)¹ belongs to the glycoprotein hormone family, which also includes the gonadotropins follicle-stimulating hormone (FSH), luteinizing hormone, and chorionic gonadotropin (CG). These hormones exist as heterodimers composed of a common -subunit, which is noncovalently linked to a hormone-specific -subunit (1). Crystallization of hCG had revealed that both subunits have a similar overall structure with a central cystine knot motif (2, 3). Therefore, the glycoprotein hormones are now considered members of the cystine knot growth factor superfamily that includes a variety of structurally related dimeric growth factors, such as nerve growth factor, platelet-derived growth factor, vascular endothelial growth factor, and transforming growth factor- (4, 5). The glycoprotein hormone -subunit is encoded in a single gene and thus identical in the amino acid sequence within a species. In contrast, the -subunits are unique, encoded in distinct genes and responsible for biological specificity (6, 7).

Assembly of the - and -subunits is an early posttranslational event in glycoprotein hormone synthesis occurring in the endoplasmic reticulum (8). Heterodimerization is critical for disulfide bond formation and for hormone-specific posttranslational modifications, such as processing of the carbohydrate side chains, and thus rate-limiting for the secretion of glycoprotein hormones (9, 10). Moreover, dimer formation is essential for hormonal activity, since free subunits have minimal receptor binding affinity (1). In addition, dissociation of heterodimeric glycoproteins into their subunits may be a significant factor in terminating glycoprotein hormone activity in vivo (11).

Therefore, covalent linking of the glycoprotein hormone subunits should overcome assembly-dependent deficiency in secretion and may increase hormone stability and activity. It has recently been pioneered by Boime and colleagues and subsequently shown by the group of Puett that bioactive gonadotropins could be produced as single chains (12-14), but it is not clear whether this approach is applicable for hTSH, or whether such fusion would affect the stability or the in vivo clearance of these hormones. Such fusion should be particularly relevant to TSH, since the free TSH -subunit, in contrast to the free CG -subunit, is unstable in the monomeric form and degraded intracellularly unless stabilized by dimerization with the -subunit (15). Here, we show that it is possible by subunit gene fusion to produce a tethered form of hTSH with comparable in vitro activity to dimeric hTSH. Furthermore, fusion significantly increased the stability and prolonged the in vivo half-life of hTSH. Moreover, the expression of hTSH as a single chain could overcome selected mutagenesis-induced defects in hTSH secretion, and thus, this approach may be used to expand the spectrum of structure-function studies of glycoprotein hormone analogs. Subunit gene fusion therefore appears to be a promising strategy, not only for the generation of long lasting hTSH analogs, but particularly for the development of recombinant mutants with desirable characteristics, the utility of which may be limited by decreased stability.

EXPERIMENTAL PROCEDURES

CHO cells stably transfected with the hTSH receptor (clone JP09) were kindly donated by Dr. G. Vassart, Belgium, and FRTL-5 cells expressing the endogenous rat TSH receptor by Dr. L. D. Kohn, Interthyr Research Foundation (Baltimore, MD). cAMP antibody was generously supplied by Dr. J. L. Vaitukaitis, National Institutes of Health (Bethesda, MD). Cell culture media and reagents were purchased from Life Technologies, Inc., and ¹²⁵I-cAMP and ¹²⁵I-hTSH radiolabeled to a specific activity of 40-60 µCi/µg from Hazleton (Vienna, VA). PCR reagents were obtained from Boehringer Mannheim and New England Biolabs (Beverly, MA).

The construction of the hTSH -subunit bearing the carboxyl-terminal extension peptide of the hCG -subunit (hTSH-CTP) has been described previously (16). To produce single chain hTSH (hTSH-SC), we used overlap extension PCR (17) to fuse the amino terminus of the -subunit cDNA (without the signal sequence) to the carboxyl-terminal end of the hTSH-CTP (Fig. 1). Primers P2 5-CAC ATC AGG AGC TTG TGG GAG GAT CGG and P3 5-ATC CTC CCA CAA GCT CCT GAT GTG CAG span both the carboxyl-terminal end of the hTSH -subunit containing the hCG-CTP (hTSH-CTP) as well as the amino terminus of the coding sequence of the -subunit. In addition, P1 5-CTC GAG TCT AGA ATG ACT GCT CTC TTT CTG ATG was designed to anneal 5 of the hTSH-CTP minigene signal peptide, and P4 5-CGA CGT GGA TCC ATG CTG TAT TCA TTC to anneal 3 of the hTSH -subunit cDNA. Initially, two PCR reactions were performed: P1 and P2 were used with the hTSH-CTP as the template (PCR no. 1), and P3 and P4 using the -subunit cDNA (PCR no. 2). In a third PCR reaction (PCR no. 3), both these overlapping products were used as combined template to generate the single chain hTSH-SC with P1 and P4.

To create Gln⁵²-Gln⁷⁸/TSH-SC in which both -glycosylation recognition sequences were deleted by mutating both Asn⁵² and Asn⁷⁸ to Gln, a previously described -subunit cDNA construct (Gln⁵²-Gln⁷⁸) (18) was used as the template for PCR no. 2. Similarly, to obtain Asp³⁸/TSH-SC, the -subunit cDNA construct Asp³⁸ (19) served as the template in PCR no. 2. Following subcloning of the fused wild type or mutant 2-kilobase pair hTSH-SC constructs into the pLB-CMV expression vector, the entire PCR product was sequenced in each case to rule out any undesired polymerase errors.

CHO-K1 Cells (ATCC, Rockville, MD) were maintained in Ham's F-12 medium supplemented with 5% fetal calf serum, penicillin (50 units/ml), streptomycin (50 µg/ml) and glutamine (4 mM). To obtain dimeric wild-type hTSH (hTSH-wt), cells were cotransfected in 60-mm culture dishes with the -subunit cDNA in pcDNA I/neo and the hTSH minigene in the pLB-CMV vector, using a total amount of 2 µg DNA per dish and a liposome formulation (LipofectAMINE reagent, Life Technologies, Inc.) as described previously (20). The hTSH-SC fusion products in the pLB-CMV vector were transfected with identical amounts of total DNA. On the following day, the transfected cells were transferred to CHO serum-free medium (Life Technologies, Inc.). After an additional 48 h, the supernatants, including control medium from mock transfections using the expression plasmids without gene inserts, were harvested. The collected media were then concentrated using a Centriprep 10 concentrator (Amicon, Beverly, MA) and used for immunoassays, the various activity assays, and clearance studies.

The hTSH constructs were quantified with a panel of different immunoassays, using a total of four different hTSH immunoassays utilizing different monoclonal antibodies, which were described in detail previously (19, 21).

Conditioned media from transiently transfected CHO cells were concentrated, fractionated on ConA-Sepharose columns (Pharmacia), reconcentrated, and denatured by boiling in 0.25% SDS, 0.5% -mercaptoethanol. Samples were then resolved on 14% Tris-glycine polyacrylamide gels, transferred to nitrocellulose membranes, and incubated overnight with a polyclonal rabbit antibody directed against the hTSH -subunit (18). Antigen-antibody complexes were subsequently visualized by chemiluminence using a horseradish peroxidase-coupled anti-rabbit IgG and a luminol substrate (Boehringer Mannheim).

The receptor-binding activity of the various hTSH constructs was determined by their ability to displace ¹²⁵I-bTSH from solubilized porcine thyroid membrane receptor preparations (Kronus, Dana Point, CA) following the manufacturer's instructions. Binding was also studied in whole cells using FRTL-5 cells expressing the endogenous rat TSH receptor, as described previously (20).

CHO cells stably expressing the rhTSH receptor (JP09) were grown in 96-well culture plates in Ham's F-12 medium supplemented as above. Confluent cells were incubated for 2 h at 37 °C, 5% CO₂, with serial dilutions of hTSH constructs or control medium from mock transfections in a modified Krebs-Ringer buffer supplemented with 280 mM sucrose to maintain isotonicity and 1 mM 3-isobutyl-1-methylxanthine. The amount of cAMP released into the medium was determined by radioimmunoassay (20).

FRTL-5 cells were maintained as described elsewhere (20) and, prior to the cAMP assay, grown in 96-well culture plates in the absence of TSH for 6-8 days. cAMP production of hTSH constructs was determined using the protocol for JP09 cells.

FRTL-5 cells were grown in 24-well plates in the presence of TSH to 30% confluence and then cultured in TSH-free medium for 4 days. Subsequently, the cells were incubated with serial dilutions of hTSH constructs or control from mock transfected cells. After 48 h, 1.0 µCi of [³H]thymidine per well (DuPont) was added, and the cells were incubated for an additional 24 h. Subsequently, [³H]thymidine uptake measured as described previously (18).

The clearance rate of the hTSH constructs was determined in the rat after intravenous injection of the different hTSH preparations and subsequent determination of hTSH serum levels at defined intervals from 1 to 120 min. Experimental details of this procedure have been described previously (22, 23).

RESULTS

Genetic Fusion of the hTSH - and -Subunit

Truncation as well as amino acid mutation studies had previously indicated the importance of the -carboxyl terminus for hTSH activity (20). To maintain accessibility of this region, we fused the carboxyl terminus of the TSH -subunit to the amino terminus of the -subunit. We also included the CTP of the hCG -subunit, here defined as the carboxyl-terminal 32 amino acids of the hCG -subunit. The CTP has a high proline/serine content, which lacks significant secondary structure and was previously shown to be suitable as a flexible linker for efficient expression of single chain hFSH (13). In keeping with previous observations, addition of CTP to the hTSH -subunit was predicted not to affect receptor binding or intrinsic activity of hTSH (16). Since addition of the CTP had previously been shown to prolong the half-life of hTSH, the clearance rate of hTSH-SC was compared with both dimeric hTSH-wt as well as hTSH-CTP (see below).

To demonstrate that hTSH-SC was indeed produced and secreted as a single chain, we performed SDS-polyacrylamide gel electrophoresis and subsequent Western blotting of ConA-fractionated conditioned media from CHO cells transiently transfected with either the fusion product or individual hTSH subunits using an antibody against the -subunit. Under reducing conditions, heterodimeric hTSH-wt dissociated into individual subunits, and the free -subunit migrated at the expected 25 kDa. In contrast, the -subunit antibody recognized a 55-kDa band consistent with the covalently linked hTSH fusion protein (Fig. 2). The level of secretion of hTSH-SC from transiently transfected CHO cells, as determined by four different immunoassays, was similar to hTSH-wt (Table I), if individual subunit plasmids were cotransfected at a 3 to 1 molar ratio. Such a 3 to 1 molar excess of the -subunit plasmid led to a higher secretion of dimeric hTSH compared with transfection of both subunits at an equimolar ratio. Addition of the CTP to the hTSH -subunit reduced secretion of dimeric hTSH, whereas fusion of the hTSH subunits with inclusion of the CTP sequence as a linker did not impair subunit folding or expression of the hormone (Table I).

Table I. Secretion of hTSH analogs Hormone levels were determined in conditioned media from CHO cells transiently transfected with the respective hTSH constructs. n denotes number of independent transfections, each performed in triplicate dishes. Free -subunit levels were similar in conditioned media from hTSH-wt and Asp38/hTSH-wt, but not detectable in media from either Gln52-Gln78/hTSH-wt or Gln52-Gln78/hTSH-SC (see text). Analog Subunit plasmid ratio Secretion n ng/ml ± S.E. hTSH-wt  : 3:1 31.2  ± 1.7 5  : 1:1 24.1  ± 1.6 3 hTSH-CTP  : 3:1 25.2  ± 2.8 5  : 1:1 17.1  ± 1.1 3 hTSH-SC 29.6  ± 4.7 11 Gln52-Gln78/hTSH-wt 0.05  ± 0.05 4 Gln52-Gln78/hTSH-SC 5.1  ± 1.8 5 Asp38/hTSH-wt <0.01 3 Asp38/hTSH-SC <0.01 3

Effect of Subunit Fusion on Secretion-deficient hTSH -Subunit Mutants

To test the effects of subunit fusion on mutagenesis-induced defects in hTSH secretion, we studied the secretion of single chain hTSH analogs Gln⁵²-Gln⁷⁸/TSH-SC lacking the two -subunit glycosylation recognition sequences and Asp³⁸/hTSH-SC. These mutations had previously been shown to profoundly decrease or abolish the secretion of dimeric hTSH (18, 19) (Table I). Consistent with findings that non-glycosylated glycoprotein hormone subunits are misfolded and degraded intracellularly, the free Gln⁵²-Gln⁷⁸ subunit was not detectable (<0.01% of hTSH-wt-free -subunit) by an -subunit-specific radioimmunoassay. In contrast, free Asp³⁸ subunit was secreted in levels quantitatively similar to hTSH-wt free -subunit (19), suggesting that the failure of the Asp³⁸ subunit to dimerize with hTSH -subunit was not related to its misfolding or degradation. Fusion of the Gln⁵²-Gln⁷⁸ subunit to the hTSH -subunit increased secretion, indicating that fusion of both subunits can partially overcome the requirement of -subunit carbohydrate chains for hTSH secretion. In contrast, fusion of the Asp³⁸ subunit did not increase the amount of hTSH produced, suggesting that this particular mutation, possibly due to its predicted location at the subunit interface in close proximity to residues forming intersubunit hydrogen bonds (2, 3) prevents subunit association (Table I).

Stability of the different hTSH proteins was tested initially by incubating conditioned media obtained from transient transfections at 37 °C. All three forms of hTSH, hTSH-wt, hTSH-CTP as well as hTSH-SC were stable at this temperature, and there was minimal (<5%) degradation over a period of 21 days, as judged by repeated determinations of hTSH immunoreactivity with an assay specific for heterodimeric hTSH, which does not recognize free subunits. However, incubation at 55 °C showed that the fused hTSH-SC was significantly more stable than dimeric hTSH in that less than 15% of hTSH-SC was degraded after 24 h, compared with more than 50% of dimeric hTSH, either hTSH-wt or hTSH-CTP (Fig. 3).

The receptor binding of the fused hTSH-SC was similar to that of hTSH-wt and hTSH-CTP when tested in porcine thyroid membranes (Fig. 4) or in FRTL-5 cells expressing the endogenous rat TSH receptor (not shown). In addition, the ability of hTSH-SC to induce cAMP stimulation in JP09 cells (Fig. 5a), as well as cAMP stimulation (Fig. 5b) and growth promotion (Fig. 5c) in FRTL-5 cells was comparable to that of hTSH-wt and to that of hTSH-CTP. This indicates that both introduction of the CTP linker as well as subunit fusion did not alter the in vitro characteristics of hTSH.

In accord with previous studies from our laboratory (16), addition of the CTP to the hTSH -subunit significantly prolonged the plasma half-life of dimeric hTSH. 50% of the hTSH-CTP was cleared from the rat circulation after 23.2 ± 7.9 min compared with 8.7 ± 6.1 min for hTSH-wt (p = 0.01). Remarkably, fusion of the individual subunits including the CTP as a linker led to an even further significant prolongation of half-life; 50% of hTSH-SC was cleared after 51.6 ± 14.4 min (p = 0.02 compared with hTSH-CTP) (Fig. 6).

DISCUSSION

The half-life of recombinant analogs can be prolonged by increasing the Stoke's radius of a protein using polyethylene glycolylation or the introduction of new carbohydrate recognition sites, by modification of protease recognition sites to increase stability, or by carbohydrate modification to avoid carbohydrate-specific clearance mechanisms (5, 21, 23). Our present study using a genetically fused, single chain hTSH highlights a novel way by which an increased in vivo half-life may be achieved.

Although it had previously been shown that bioactive hCG and hFSH could be produced as a single chain (12-14), the effect of genetic fusion on glycoprotein hormone stability and plasma clearance rate had not previously been investigated. Further, from the findings on hCG and hFSH, it was not predictable whether a fusion approach would also be feasible for hTSH. In particular, recent mutational analysis of hTSH structure-function relationships has identified common -subunit domains that play strikingly different roles for heterodimer formation, receptor binding, and bioactivity of hTSH compared with hCG and hFSH (18-20). Interestingly, these domains are located in close proximity to the -seat-belt region, which is crucial for hTSH specificity, suggesting that the seatbelt may direct these common domains to function in a hormone-specific fashion (5, 24).

In light of previous observations (16), validated here, that addition of the CTP with its O-linked carbohydrate side chains prolonged hTSH half-life in vivo, the full-length CTP was used as a linker for fusing the hTSH subunits. We anticipated that the linker may synergize with the fusion to prolong the half-life of hTSH in vivo. Indeed, gene fusion significantly decreased the clearance rate of dimeric hTSH even when compared with dimeric hTSH bearing the CTP. This indicates that dissociation of hTSH into its subunits occurs in vivo and contributes to its deactivation, as individual subunits are devoid of in vivo activity and rapidly cleared from the circulation (11).

In addition, fusion of the subunits of hTSH increased its thermostability. It is conceivable that such enhanced stability may become particularly relevant for recombinant glycoprotein hormone analogs with genetically engineered novel features that are less stable than the wild-type hormone. In this respect, hFSH analogs have recently been described, in which site-directed mutagenesis within regions important for activity significantly decreased their stability (25).

Moreover, genetic subunit fusion can overcome certain mutagenesis-induced defects in heterodimer formation. The presence of carbohydrate side chains on both subunits is essential for proper subunit folding and combination, and intracellular assembly of deglycosylated subunits is inefficient (10). Indeed, glycosylation of the -subunit appears necessary to overcome retention of the hTSH -subunit in the endoplasmic reticulum (8), and in contrast to the free hCG -subunit, the free hTSH -subunit is not efficiently secreted (26). Our fusion experiments suggest that the glycosylated hTSH -subunit, if fused to an -subunit devoid of glycosylation recognition sequences, may function as a chaperone inducing -subunit folding despite the absence of carbohydrate chains and thus partially rescue the nonglycosylated -subunit. On the other hand, fusion was not able to induce heterodimer formation with a mutated -subunit Asp³⁸ which, although dimer formation-incompetent, nevertheless appeared to be properly folded and secreted.

It is interesting to consider the dimeric structure of glycoprotein hormones from an evolutionary perspective. The glycoprotein hormones were probably derived from a common ancestor gene, and in less developed organisms, a single primordial monomeric hormone with a corresponding receptor was likely sufficient for the necessary endocrine functions (27). To fulfill the requirements for an increasingly complex organism, adopting a dimeric ligand structure enabled functional diversification and increased flexibility without the need for the development of entirely new mechanisms of receptor activation, albeit perhaps at the expense of reduced protein stability. This diversification appears to have evolved by the emergence of inhibitory domains on both ligand and receptor which impose steric hindrances thus allowing only the intended ligand to interact with the common activation domain (28). Such negative specificity determinants have not only developed in glycoprotein hormones and their receptors, but also in other members of the cystine knot growth factor superfamily, such as neurotropins (29), and also in other G protein-coupled receptors (30). More generally, dimer formation is necessary for the activity and specificity of many, if not all cystine knot growth factors, as well as for other bioactive molecules, such as enzymes and transcription factors. In this respect, fusion of individual protein monomers has recently been used to develop transcription factors and cytokine analogs with defined properties and increased biological activities (31, 32). This approach poses a universal strategy to enhance both stability and bioactivity as well as to control specificity of noncovalently linked oligomers, and may also be used to engineer molecules with novel activities or specificities.