INTRODUCTION

Top Abstract Introduction Procedures Results Discussion References

Thyrotrophin (TSH)¹ is a heterodimeric glycoprotein indispensable for the control of thyroid structure and function and ultimately metabolism. TSH exerts its effects by binding to a specific receptor (TSHR) on the thyroid follicular cell; this activates the adenylate cyclase pathway and, possibly, the phosphatidylinositol pathway (1-4). The TSHR belongs to a family of G-protein coupled receptors and has clinical significance because it is a frequent target for autoantibody production in humans.

The cDNA sequence (5-7) predicts a polypeptide of M_r 84,500 with six potential glycosylation sites. The receptor is envisaged to form a large extracellular domain (ECD) (polypeptide M_r 45,000 plus carbohydrate), a seven-transmembrane region, and a cytoplasmic tail. It is closely related to the receptors for the other glycoprotein hormones (follicle stimulating hormone, luteinizing hormone, and chorionic gonadotrophin) but has two unique insertions of 8 and 50 amino acids (residue numbers 38-45 and 317-366) in the ECD.

It is generally assumed that the ECD is solely responsible for ligand interaction and that the transmembrane and cytoplasmic regions are involved in signal transduction. However, we showed that the recombinant ECD expressed in isolation as a soluble 60,000 M_r glycoprotein was recognized by autoimmune antibodies but did not bind to TSH with high affinity (8), suggesting that additional components may play a role in ligand interaction. This conclusion is supported by data from other groups (9-11). Evidence for the extracellular loops between the transmembrane regions making an important contribution to TSH binding has been provided by experiments in which mutations or insertions in the first or second extracellular loops resulted in a receptor that was unable to bind TSH (12-14). Conversely, the recombinant ECD expressed on the cell surface, in an ill-defined way by virtue of an 11-amino acid tail derived from a cloning vector, was able to bind TSH with similar affinity to full-length receptor (15), and the isolated recombinant ECD is reported to bind TSH, although less effectively than the full-length molecule (16, 17).

Studies on the ECD are hampered by the fact that, when expressed alone in eukaryotic cells, the recombinant protein is retained within the cell rather than being secreted as expected (8, 9). We have managed to facilitate secretion of ECD material by fusing it to the "hinge" region of rat CD8 or to domains 3 and 4 of rat CD4 (18) and truncated versions of the ECD (down to 261 residues) are secreted (19); however, although these modified ECD are recognized by at least some autoantibodies, they do not bind TSH with high affinity, the same situation as for the ECD retained inside cells. ECD expressed alone in prokaryotes or baculovirus is insoluble and without function (8, 20-25), although there are some reports of producing soluble material from such systems by experimental manipulation (11, 16, 17).

In order to address more clearly the issue of the minimal structural requirements for TSH binding, we have expressed the ECD in isolation on the cell surface in a defined way by means of a glycosylphosphatidylinositol (GPI) anchor and report here characterization of the recombinant product, including binding of TSH and autoantibodies.

	EXPERIMENTAL PROCEDURES

Top Abstract Introduction Procedures Results Discussion References

Construction of TSHR ECD with GPI Anchor-- Polymerase chain reaction was used to add sequences directing attachment of a GPI anchor to the ECD at residue 412, Ile, which is just prior to the first predicted transmembrane sequence of the TSHR. The template used was a construct encoding domains 3 and 4 of rat CD4 attached to the C terminus of the human ECD (18). Domain 3 possesses some amino acids that are acceptable for GPI attachment. These were used together with artificially created sequences to direct a signal for GPI anchoring. The primers used were an 18-base sequence corresponding to nucleotide positions 973-990 of the human TSHR cDNA (sense) 5'-ATCAGAGGAATCCTTGAG-3' and a 55-base sequence containing a signal for GPI attachment, a stop codon, and an artificial BamHI site for subsequent cloning (antisense) 5'-GAGGATCCTAGACGAGCACGAGCAGGAGCAGAAGGATGAGTAGGAAGAAGAACTC-3'. The nucleotide and corresponding amino acid sequence for the added material is shown in Fig. 1, the double underlined serine residue serving as the putative attachment site. The polymerase chain reaction consisted of 30 cycles (95 °C 1 min; 60 °C 2 min; 72 °C 1 min) and produced a fragment encompassing nucleotides 973-1332 of the TSHR (including a NdeI site at position 1265) plus the anchor sequence. The coding region for the extracellular region of the human TSHR has been cloned into the prokaryotic expression vector pGEX-3X (8). This was digested with EcoRI, the overhang filled in with Klenow, further digested with NdeI, and the NdeI-digested polymerase chain reaction fragment was ligated into this. The BamHI fragment containing the ECD plus GPI anchor signal was then excised, ligated into the eukaryotic expression vector pEE14, and the product transfected into CHO-K1 cells, as described (26), to generate stable lines, which were selected by their resistance to methionine sulfoximine and by using slot blotting to identify colonies with a large number of copies of plasmid incorporated into genomic DNA. The sequence of the construct used for transfection was confirmed by dideoxy sequencing using a commercial kit (Pharmacia Ltd.).

View larger version (16K): [in this window] [in a new window]   Fig. 1.   The sequence directing attachment of a GPI anchor. The artificial nucleotide, and corresponding protein, sequence is shown beginning at residue 412 (isoleucine) of the human TSHR. The engineered thrombin cleavage site is underlined and the cleavage position indicated by an arrow. The protein sequence directing attachment of the GPI-anchor is shown in bold and the putative attachment site of the anchor (serine) is double underlined.

Other Recombinant TSHR Constructs-- CHO lines expressing full-length TSHR (designated "FLD4" and "FLE4.2") and isolated ECD (designated "ExG2") have been described previously (8, 26).

Membrane Preparation-- An enriched membrane fraction of recombinant CHO cells was prepared as described (27) using freeze/thaw and hypotonic homogenization.

Immunoblotting-- Samples were sonicated at 100 watts on ice for 3 × 10 s and then heated at 100 °C for 5 min in SDS-PAGE loading buffer (65 mM Tris-HCl, pH 6.8, 2% SDS, 10% glycerol, 5% -mercaptoethanol, 0.001% bromphenol blue). SDS-PAGE was carried out according to the method of Laemmli (28). Western transfer onto nitrocellulose (Electran, 0.45 µm; BDH, Poole, United Kingdom) was as described (29). Immunoblotting was carried out using monoclonal antibodies to TSHR (18) and the Boehringer Mannheim chemiluminescence blotting substrate kit.

Separation of Membrane-associated from Soluble Proteins-- Confluent cells were removed from 9-cm dishes using Versene (Life Technologies, Inc., Paisley, UK) and washed in PBS. An aliquot (2 × 10⁶ cells) was centrifuged at 1,000 × g for 5 min, resuspended in 100 µl of 1% (v/v) Triton X-114 in PBS containing protease inhibitors (2 mM phenylmethylsulfonyl fluoride, 5 µg/ml aprotinin, 10 µg/ml leupeptin), incubated on ice for 10 min and then at 30 °C for 5 min. The sample was then centrifuged at 12,000 × g for 10 min to enhance phase separation. An equal volume of 2 × SDS-PAGE loading buffer was added to the aqueous (containing soluble proteins) and detergent phase (containing membrane proteins) and the samples analyzed by immunoblotting.

Removal of Carbohydrate-- Cells were removed from 9-cm dishes using Versene, washed in PBS and solubilized in 200 µl of 1% Nonidet P-40 in PBS containing protease inhibitors (2 mM phenylmethylsulfonyl fluoride, 5 µg/ml aprotinin, 10 µg/ml leupeptin) for 1 min on ice, insoluble material being removed by centrifugation at 13,000 g for 5 min. One quarter of the supernatant (equivalent to 2 × 10⁵ cells) was analyzed by immunoblotting to check for recovery of protein in the Nonidet P-40 extracts. The remaining supernatant was split into two aliquots (equivalent to 3 × 10⁵ cells) and diluted with an equal volume of 0.1% SDS in PBS, so that the final Nonidet P-40 and SDS concentrations were 0.5 and 0.05%, respectively. To one aliquot was added 10 milliunits of N-glycosidase F (Boehringer) and both aliquots were incubated at 37 °C for 16 h and the resultant digest analyzed by immunoblotting.

Flow Cytofluorimetry-- Cells were removed from 9-cm dishes using Versene and their reaction with an anti-TSHR monoclonal antibody analyzed by flow cytofluorimetry as described (18).

Radioligand Binding Assay-- Cells grown in 24-well plates (approximately 10⁵/well) were washed twice with binding buffer (20 mM Tris-HCl, pH 7.5, 50 mM NaCl, 0.1% bovine serum albumin) and unlabeled TSH or human sera added. Radioiodinated TSH (a kind gift of BRAHMS Diagnostica, Berlin, Germany) was added at approximately 30,000 cpm/well. The final volume was 300 µl and the plates were incubated at 37 °C for 2 h. The cells were then washed rapidly with 500 µl of ice-cold binding buffer, solubilized in 500 µl of 0.5 M NaOH, and their radioactivity determined. Nonspecific binding, determined in the presence of 150 nM TSH, was subtracted from all counts.

Biosynthetic Labeling of Cells and Immunoprecipitation-- Cells (5-6 × 10⁶) were washed twice with PBS and incubated in serum-free, methionine-free, HEPES-buffered Dulbecco's modified Eagle's medium (ICN Flow, Thame, UK) at 37 °C for 30 min. The culture supernatants were then replaced with Dulbecco's modified Eagle's medium supplemented with 2% fetal calf serum and 100 µCi of [³⁵S]methionine or [³H]ethanolamine (Amersham, Bucks, UK) at 37 °C for 3 h. The cells were washed twice with PBS and solubilized on ice with PBS containing 1% Nonidet P-40 and protease inhibitors (2 mM phenylmethylsulfonyl fluoride, 5 µg/ml aprotinin, 10 µg/ml leupeptin). The cell extract was centrifuged at 13,000 × g for 10 min at 4 °C and the Nonidet P-40-soluble material stored at 20 °C until analyzed by immunoprecipitation (8).

	RESULTS

Top Abstract Introduction Procedures Results Discussion References

Following transfection of CHO cells with the GPI construct in the pEE14 vector, successful transfectants were identified and selected by their resistance to methionine sulfoximine; further screening of these stable lines by means of slot blotting DNA identified one clone with the highest copy number of TSHR sequences, which was designated "GPIA6." Western blotting of crude extracts from GPIA6, as well as other clones with lower copy number, using monoclonal antibodies detected a band of approximately 70,000 M_r expressed at high levels, approximately 10,000 larger than the isolated ECD (Fig. 2A, tracks 1 and 3). Digestion with N-glycosidase F decreased the M_r by approximately 10,000 (Fig. 2A, track 4); a similar decrease in M_r was also observed for isolated ECD in ExG2 extracts (Fig. 2A, track 2). N-Glycosidase F does not remove the entire carbohydrate moiety, unlike endoglycosidase F which was used in an earlier study (31) where an M_r of 15,000 was attributed to carbohydrate on the ECD of ExG2 cells. Approximately 50% of the recombinant material partitioned into the detergent phase in Triton X-114-treated cells (Fig. 2B, tracks 3 and 4), indicating that it was membrane bound; in contrast, the vast majority of the ECD construct in ExG2 cells partitioned into the aqueous phase (Fig. 2B, tracks 1 and 2), indicating that it was not membrane bound, as expected (8). In confirmation of this, approximately 30% of TSHR material was released in soluble form after freeze/thawing and homogenization of GPIA6 cells (Fig. 2C, tracks 4-6) compared with greater than 90% released from ExG2 cells (Fig. 2C, tracks 1-3).

View larger version (27K): [in this window] [in a new window]   Fig. 2.   Characterization of recombinant GPI-anchored TSHR by immunoblotting. A, recombinant CHO lines expressing ECD alone (ExG2) (tracks 1 and 2) or GPI-anchored ECD (GPIA6) (tracks 3 and 4) were solubilized in Nonidet P-40 and incubated in the absence (tracks 1 and 3) or presence (tracks 2 and 4) of N-glycosidase F before analysis by immunoblotting. B, recombinant CHO lines expressing ECD alone (ExG2) (tracks 1 and 2) or GPI-anchored ECD (GPIA6) (tracks 3 and 4) were solubilized in Triton X-114 and the mixture separated into aqueous (tracks 1 and 3) and detergent (tracks 2 and 4) phases which were then analyzed by immunoblotting. C, recombinant CHO lines expressing ECD alone (ExG2) (tracks 1-3) or GPI-anchored ECD (GPIA6) (tracks 4-6) were suspended in 10 mM Tris, 1 mM EDTA, pH 8.0, containing protease inhibitors (2 mM phenylmethylsulfonyl fluoride, 5 µg/ml aprotinin, 10 µg/ml leupeptin) and either taken directly for analysis by immunoblotting (tracks 1 and 4) or subjected to two freeze/thaw cycles on dry ice and then passed through a 25-gauge needle 10 times. After centrifugation (12,000 × g, 30 min, 4 °C) to separate soluble from membrane-bound material, the pellets (tracks 2 and 5) and supernatants (tracks 3 and 6) were analyzed by immunoblotting. The equivalent of 2 × 105 cells was loaded on each track. In each case, the blot was incubated with 2C11 monoclonal antibody against TSHR and bound antibody detected using peroxidase-conjugated anti-mouse IgG antibody and a chemiluminescence detection system. The positions on SDS-PAGE of marker proteins of known Mr (× 103) are indicated on the right of each panel.

Analysis of GPIA6 cells using flow cytofluorimetry showed that TSHR material was present on the cell surface (Fig. 3). Using several different monoclonal antibodies, the fluorescence signal was 3-10-fold higher for GPIA6 cells than for cells expressing the full-length receptor (FLE4.2), indicating a substantial increase in surface expression for the GPI-anchored molecule.

View larger version (21K): [in this window] [in a new window]   Fig. 3.   Analysis of cell-surface expression of GPI-anchored TSHR by flow cytofluorimetry. Histograms are presented showing the binding of two monoclonal antibodies against the TSHR (2C11 and 3B12) to CHO lines expressing GPI-anchored ECD (GPIA6) or, for comparison, full-length TSHR (FLE4.2). Each panel shows the reaction of an irrelevant monoclonal antibody (NS) (bold solid line) as well as 3B12 (dashed line) and 2C11 (fainter dotted line). Bound antibodies were detected using a fluorescein-labeled anti-mouse IgG antibody and a FACscan cytofluorimeter.

Immunoprecipitation of detergent-solubilized GPIA6 cells, which had been biosynthetically radiolabeled with methionine, demonstrated that the GPI-anchored material was recognized by autoantibodies in sera from several patients with Graves' disease (Fig. 4A), which also react with the recombinant isolated ECD in ExG2 cells as reported previously (8). The TSHR material from GPIA6 cells, but not from ExG2 cells, also incorporated radiolabeled ethanolamine (Fig. 4B), as expected for a GPI-anchored protein.

View larger version (29K): [in this window] [in a new window]   Fig. 4.   Biosynthetic labeling of GPI-anchored TSHR and recognition by human autoantibodies. Recombinant CHO lines expressing ECD alone (ExG2) (tracks 1-3) or GPI-anchored ECD (GPIA6) (tracks 4-6) were biosynthetically labeled with either [35S]methionine (panel A) or [3H]ethanolamine (panel B), solubilized in Nonidet P-40, and immunoprecipitated with pooled normal human sera (tracks 1 and 4) or serum from Graves' patient S235 (tracks 2 and 5) or S342 (tracks 3 and 6). Following SDS-PAGE fractionation, the radioactive proteins were detected by autoradiography. The positions of marker proteins of known Mr (× 103) are indicated on the right of each panel.

In confirmation of the GPI-anchored structure, incubation of GPIA6 cells with PIPLC decreased the fluorescence signal in flow cytofluorimetry by approximately 70%, whereas this treatment had no effect (or caused a slight increase) on the full-length molecule in FLE4.2 cells (Table I). Similarly, thrombin decreased the fluorescence signal by approximately 50% for GPIA6 cells but had much less effect on FLE4.2 cells, presumably cutting at the thrombin site engineered into the GPI construct just before the GPI attachment signal sequence (see "Experimental Procedures" and Fig. 1). Analysis by immunoprecipitation of the soluble and membrane-bound TSHR content from similar digests of biosynthetically labeled cells demonstrated release of soluble ECD from GPIA6 cells by PIPLC or thrombin treatment (Fig. 5).

View larger version (46K): [in this window] [in a new window]   Fig. 5.   The release of TSHR material from GPI-anchored receptors by thrombin or PIPLC. A, recombinant CHO lines expressing ECD alone (ExG2) (tracks 1-4) or GPI-anchored ECD (GPIA6) (tracks 5-8) were biosynthetically labeled with [35S]methionine and then incubated in the absence (tracks 1, 2, 5, and 6) or presence (tracks 3, 4, 7, and 8) of PIPLC and the supernatants immunoprecipitated with pooled normal human sera (tracks 1, 3, 5, and 7) or a mixture of sera from Graves' disease patients S235 and S342 (tracks 2, 4, 6, and 8). B, recombinant CHO lines expressing GPI-anchored ECD (GPIA6) were biosynthetically labeled with [35S]methionine and then incubated in the absence (tracks 1-4) or presence (tracks 5-8) of thrombin. After this digestion, the supernatants (tracks 1, 2, 5, and 6) and solubilized cell pellets (tracks 3, 4, 7, and 8) were immunoprecipitated with pooled normal human sera (tracks 1, 3, 5, and 7) or a mixture of sera from Graves' disease patients S235 and S342 (tracks 2, 4, 6, and 8). Following SDS-PAGE fractionation, the radioactive proteins were detected by autoradiography. The positions of marker proteins of known Mr (× 103) are indicated on the right of each panel.

Intact GPIA6 cells bound ¹²⁵I-labeled TSH, with a maximum approximately 3-fold higher than that of FLE4.2 cells, which express the full-length receptor (Fig. 6A). When plotted as a percentage of maximum binding, the curves from the two cell lines are very similar (Fig. 6B), indicating that the two recombinant molecules have a similar affinity for TSH; Scatchard analyses of the data from seven independent experiments gave K_D values of 0.209 ± 0.042 nM for GPIA6 cells and 0.281 ± 0.032 nM for FLD4 cells (mean ± S.E.). Membranes prepared from GPIA6 cells also bound ¹²⁵I-labeled TSH (Fig. 7); in contrast, those from ExG2 cells (which have soluble ECD that does not bind TSH) were negative and those from FLE4.2 and FLD4 cells (which express less receptors than GPIA6) were negative or slightly positive.

View larger version (11K): [in this window] [in a new window]   Fig. 6.   Radioligand binding analysis of GPI-anchored TSHR in intact cells. Recombinant CHO lines expressing GPI-anchored ECD (GPIA6, solid squares) or full-length TSHR (FLD4, open circles) in 24-well plates were incubated with a constant amount of radioiodinated TSH together with varying concentrations of unlabeled TSH. The amount of radioactivity bound to the cells was then determined and the nonspecific binding in the presence of 150 nM unlabeled TSH was subtracted. Panel A shows the binding curve from one representative experiment as counts/min bound to the cells (mean ± S.E. of triplicate wells). In panel B the data from seven independent experiments, each in triplicate, are presented as a percentage of the binding in the absence of any inhibitor (mean ± S.E.).

View larger version (25K): [in this window] [in a new window]   Fig. 7.   Radioligand binding analysis of GPI-anchored TSHR in membranes. In two separate experiments (A and B) membranes from recombinant CHO lines expressing full-length TSHR (FLE4.2 and FLD4), ECD alone (ExG2), or GPI-anchored ECD (GPIA6) were incubated with radioiodinated TSH in the absence (cross-hatched bars) or presence (solid bars) of 150 nM unlabeled TSH. The amount of radioactivity bound to the membranes was determined and is presented relative to the protein content of each sample.

The ability of autoantibodies in the sera of Graves' disease patients to inhibit the binding of ¹²⁵I-labeled TSH to intact cells was then investigated (Fig. 8). While most of the sera that inhibited binding to full-length receptors also inhibited binding to the GPI-anchored molecules, there were clear differences in the amount of this inhibition. Thus serum from patients S361, H804, H254, H291, S235, and H569 inhibited binding to the GPI-anchored molecules significantly less than to the full-length receptors, whereas serum from patient S342 or S778 was more inhibitory on the GPI-anchored material.

View larger version (38K): [in this window] [in a new window]   Fig. 8.   Effect of sera from Graves' disease patients on the binding of TSH to GPI-anchored TSHR. Recombinant CHO lines expressing GPI-anchored ECD (GPIA6) (cross-hatched bars) or full-length TSHR (FLE4.2) (solid bars) in 24-well plates were incubated with radioiodinated TSH together with serum from the indicated patient (final concentration 6.7%, v/v, for all except S235, which was 0.2%). The amount of radioactivity bound to the cells was then determined and the nonspecific binding in the presence of 150 nM unlabeled TSH was subtracted. The data are presented as a percentage of the binding in the presence of pooled normal human sera (mean ± S.E. of triplicate wells).

	DISCUSSION

Top Abstract Introduction Procedures Results Discussion References

A GPI anchor consists of phosphatidylinositol (inserted in the outer leaflet of the plasma membrane), a glycan core, and an ethanolamine covalently attached to the C-terminal residue of the protein. No clear consensus sequence exists for directing such anchoring but there are some general requirements for creating a synthetic anchor sequence: (i) a hydrophobic region at the C terminus of the molecule (10-20 amino acids) not followed by a cluster of basic residues; (ii) a "spacer domain" of 7-10 residues preceding the hydrophobic region; (iii) small amino acids after the spacer region, where cleavage of the precursor and attachment of the anchor occurs (32). The GPI anchor is preassembled and added to nascent proteins in the endoplasmic reticulum (33, 34). Concomitant with this step, the initial C-terminal peptide is removed so that the GPI anchor is covalently attached to a new C-terminal amino acid on the protein (35). By introducing sequences conforming to these general rules (Fig. 1), we have directed the attachment of a GPI anchor to the end of the TSHR ECD. This is demonstrated by the biosynthetic labeling of the material with ethanolamine (Fig. 4B) and susceptibility to release from membranes by PIPLC (Fig. 5 and Table I). This achievement allowed studies on the structural requirements of the TSHR for binding TSH, and also demonstrates the general utility of this approach for expressing other proteins with simple membrane anchors.

In addition to the membrane anchored form, a significant amount of recombinant material was present in a soluble, non-membrane bound form (Fig. 2, B and C), presumably retained inside the endoplasmic reticulum. There is no obvious explanation for this. It is possible that the engineered thrombin site is susceptible to partial spontaneous cleavage, as noted for other recombinant molecules containing this site (36), or that an endogenous PIPLC cleaved within the GPI anchor (37); alternatively, the soluble form could be a partially processed product before addition of the GPI anchor. There is a slight difference in M_r of the two forms, the soluble appearing slightly larger (Fig. 2, B and C); a GPI anchor contributes 1500 to the M_r of a polypeptide on SDS-PAGE but such small changes to larger glycosylated proteins are not easily observed (34).

The GPI-anchored material was expressed at much higher levels than the full-length receptor, as demonstrated by its ease of detection using immunoblotting (Fig. 2) and radioligand binding to crude membranes (Fig. 7), neither of which techniques detect full-length molecules in our hands. The GPI-anchored material was present on the cell surface in significantly greater amounts than the full-length receptor (3-10-fold), as demonstrated by flow cytofluorimetric analyses (Fig. 3) and radioligand binding to whole cells (Fig. 6A). This observation supports our earlier conclusion (38) that the similar, limited expression of full-length recombinant molecules in mammalian cells reported by several groups using different expression systems is related to the number of receptor molecules (with a bulky seven-transmembrane region) that a cell can tolerate without impairing its viability, a simpler membrane attachment allows increased cell-surface capacity. However, an increase of this magnitude is unlikely to be sufficient to account for the increased signals observed on immunoblotting, immunoprecipitation, and radioligand binding to membranes. Consequently, we think that the available data indicate that, in addition to the recombinant material expressed on the cell surface, a substantial amount is retained inside the cell, probably in the endoplasmic reticulum, in a membrane-bound form as well as the soluble form discussed above.

The recombinant material from GPIA6 cells was consistently observed as two bands in immunoprecipitation analyses; using immunoblotting only one band was detected, corresponding to the upper band in immunoprecipitation (Figs. 2, 4, and 5). Both bands were also biosynthetically labeled by [³H]ethanolamine (Fig. 4B), suggesting that they both contained a GPI anchor. Presumably, the lower band is a minor component (and hence not easily detectable by immunoblotting) which is synthesized efficiently but which is then degraded or metabolized further thus preventing its accumulation.

The K_D of our full-length receptor for TSH was calculated in an earlier study to be 0.225 nM (26), which is in good agreement with the values determined for the native receptor on thyroid membranes and also with similar full-length recombinant preparations of other groups. In this study, we again obtained a similar K_D value for the full-length receptor and, furthermore, observed that the GPI-anchored material bound TSH with a very similar high affinity (Fig. 6). This is in contrast to the ECD alone which does not bind TSH with high affinity (8-11). Neither of these truncated recombinant molecules contains the extracellular loops and so these cannot be required for TSH binding. It is possible that membrane insertion assists correct folding of the ECD, thus allowing complete formation of the TSH-binding site; this would be in agreement with one earlier report (15). However, the isolated ECD is folded well enough to be recognized by autoantibodies, which do not react with linear epitopes (39). There is no obvious difference in the carbohydrate content of the ECD alone compared with GPI-anchored ECD (Fig. 2) that can be postulated to explain the difference in TSH binding, although we cannot rule out a subtle difference in carbohydrate structure.

The GPI-anchored ECD contains none of the TSHR regions that are predicted to lie within the membrane or cytoplasm. Consequently, this recombinant construct would not be expected to couple to second messenger systems and, in support of this, we found that TSH did not induce cAMP production (data not shown).

Autoantibodies in the sera of some Graves' disease patients recognize the GPI-anchored material, as demonstrated by immunoprecipitation (Fig. 4) and by their inhibition of TSH binding (Fig. 8). However, there was no direct correlation between the two techniques and not every serum that inhibited TSH binding was positive in immunoprecipitation (e.g. patient H804), possibly reflecting a requirement for antibodies of higher affinity in order to be detectable by immunoprecipitation, because of its stringent washes.

Significant differences were noted in the amount of inhibition by individual sera between the GPI-anchored and full-length material (Fig. 8). As an example, it is notable that serum from patient S235 which reacted well with the GPI-anchored material by immunoprecipitation only inhibited TSH binding to that same material by about 25%, in contrast to its 90% inhibition of full-length material. It should be noted that all sera probably contain more than one autoantibody; if some of these react with the extracellular loops, this might explain the greater amount of TSH-binding inhibition of these sera with the full-length construct. This conclusion is somewhat at odds with the data from one other study (19) which found that a truncated version of the ECD of only 261 residues neutralized the majority of autoantibodies from all patients. There is no obvious explanation for the converse situation where a serum inhibits the GPI-anchored material more than the full-length (e.g. patients S342 and S778).

The recombinant GPI-anchored TSHR reported here represents the smallest known fragment of TSHR that has retained high-affinity TSH binding. The high expression of this material by stable recombinant cell lines coupled with its susceptibility to PIPLC allows the production of functional material in both membrane-bound and soluble form. Such material will undoubtedly be useful in defining the structure-function relationships of this important molecule, a task which has proved to be surprisingly arduous.