Shedding of human thyrotropin receptor ectodomain. Involvement of a matrix metalloprotease.

The thyrotropin (TSH) receptor in human thyroid glands has been shown to be cleaved into an extracellular alpha subunit and a transmembrane beta subunit held together by disulfide bridges. An excess of the latter component relative to the former suggested the shedding of the ectodomain. Indeed we observed such a shedding in cultures of human thyrocytes and permanently transfected L or Chinese hamster ovary cells. The shedding was increased by inhibitors of endocytosis, recycling, and lysosomal degradation, suggesting that it was dependent on receptor residency at the cell surface. It was slightly increased by TSH and phorbol esters, whereas forskolin and 8-bromo-cyclic AMP were without effect. Decreasing the serum concentration in cell culture medium enhanced the shedding by an unknown mechanism. The shedding of the TSH receptor alpha domain is the consequence of two events: cleavage of the receptor into alpha and beta subunits and reduction of the disulfide bridge(s). The complete inhibition of soluble TSH receptor shedding by the specific inhibitor BB-2116 indicated that the cleavage reaction is catalyzed probably at the cell surface by a matrix metalloprotease. This shedding mechanism may be responsible for the presence of soluble TSH receptor alpha subunit in human circulation.

The thyrotropin (TSH) receptor in human thyroid glands has been shown to be cleaved into an extracellular ␣ subunit and a transmembrane ␤ subunit held together by disulfide bridges. An excess of the latter component relative to the former suggested the shedding of the ectodomain.
Indeed we observed such a shedding in cultures of human thyrocytes and permanently transfected L or Chinese hamster ovary cells. The shedding was increased by inhibitors of endocytosis, recycling, and lysosomal degradation, suggesting that it was dependent on receptor residency at the cell surface. It was slightly increased by TSH and phorbol esters, whereas forskolin and 8-bromo-cyclic AMP were without effect. Decreasing the serum concentration in cell culture medium enhanced the shedding by an unknown mechanism.
The shedding of the TSH receptor ␣ domain is the consequence of two events: cleavage of the receptor into ␣ and ␤ subunits and reduction of the disulfide bridge(s). The complete inhibition of soluble TSH receptor shedding by the specific inhibitor BB-2116 indicated that the cleavage reaction is catalyzed probably at the cell surface by a matrix metalloprotease. This shedding mechanism may be responsible for the presence of soluble TSH receptor ␣ subunit in human circulation.
Thyrotropin (TSH) 1 is the primary hormone that regulates thyroid cell growth and function via the G protein-coupled thyrotropin receptor (1,2). Members of this receptor family are characterized by their common structural feature of seven transmembrane domains (3). However, luteinizing hormonechoriogonadotropin (LH/CG) (4,5), follicle-stimulating hor-mone (FSH) (6), and TSH receptors (7-10) form a subgroup in this family having a large and glycosylated extracellular domain specialized in high affinity hormone binding (11,12). Interest in the TSH receptor (TSHR) is enhanced by its implication in autoimmune diseases. Autoantibodies directed against this receptor have stimulatory (Graves' disease) or blocking (idiopathic myxoedema) effects on its function (13). From cloning and sequencing of the human TSH receptor cDNA (7)(8)(9)(10), the structure of a preprotein with a calculated molecular mass of 84.5 kDa was deduced. But the characterization of the mature structure of the TSH receptor and of other receptors of that family awaited the generation of high affinity specific antibodies (14,15). Interestingly, while the LH/CG receptor is expressed in target organs as a monomer (14), the TSH receptor is expressed in thyroid membranes as a heterodimer: an extracellular ␣ subunit (ϳ53 kDa) and a membrane-spanning ␤ subunit (ϳ38 kDa) are held together by disulfide bridge(s) (15). This observation is in agreement with one of the models proposed for the structure of the TSH receptor before the cloning (16). The post-translational cleavage in two subunits is almost complete in human thyroid tissue, whereas in L cells stably transfected with the TSH receptor, a small amount of uncleaved mature receptor may still be present. In transfected cells, but not in human thyroid tissue, there is accumulation inside the cells of an unprocessed mannoserich monomeric precursor (17).
Pulse-chase experiments performed in L cells and in human thyrocytes confirmed that the TSH receptor is primarily synthesized as a ϳ95-kDa mannose-rich monomer, which undergoes supplementary glycosylation to yield a ϳ120-kDa monomeric precursor (17). The latter is then cleaved into mature ␣ and ␤ subunits. This processing into two subunits is unique among G protein-coupled receptors.
Precise quantification of each subunit in human thyroid membranes allowed us to observe a 2.5-3-fold excess of ␤ over ␣ subunits (15). This observation led us to postulate that the ␣ subunit might be shed from cell membranes and released into the extracellular space or bloodstream. Such a phenomenon could be important in the context of autoimmune diseases. We report now that spontaneous shedding of the extracellular domain of the TSH receptor occurs in stably transfected L and CHO cell lines as well as in human thyrocytes. This shedding is also a regulated process involving the cleavage of the TSH receptor by a matrix metalloprotease.

EXPERIMENTAL PROCEDURES
Materials-Most reagents were purchased from Sigma. Bovine TSH, monensin, A23187, and ionomycin were from Calbiochem; iodinated bovine TSH (70 Ci/g) was from Eria Diagnostics Pasteur (Marnesla-Coquette, France); iodinated streptavidin was from Amersham Corp. BB-2116, a matrix metalloprotease inhibitor, was kindly provided by British Biotech (Oxford, United Kingdom). * This work was supported by the Institut National de la Santé et de la Recherche Médicale, the Délégation à la Recherche Clinique (Assistance Publique, Hôpitaux de Paris), the Faculté de Médecine Paris Sud, the Association pour la Recherche sur le Cancer, the Ligue Nationale contre le Cancer, and the Fondation pour la Recherche Médicale Française. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  Culture of an L Cell Line Permanently Expressing the Human TSH Receptor-This cell line was cultured as described previously (17). Cell viability tests were performed by the trypan blue exclusion technique.
Primary Cultures of Human Thyrocytes-Human thyroid tissue was obtained by surgery from patients with benign thyroid diseases. Thyroid follicles were prepared as described (18) and cultured for 48 -72 h.
Immunoradiometric Assay of the TSH Receptor-A double-determinant ("sandwich"-type) radioimmunoassay was developed for the specific quantification of the TSH receptor ␣ subunit. This assay uses two additive monoclonal antibodies specific for the extracellular domain of the TSHR (T5-317 and T5-51 (15)). The specificity of these antibodies has been established by a variety of methods: these antibodies interacted with TSH receptor expressed in Escherichia coli or in mammalian cells (COS 7, CHO, and L cells) as tested by enzyme-linked immunosorbent assay, Western blot, and immunocytochemistry. No signal could be be detected using all these methods in mock-transfected cells. Finally antibody 51 has been used to immunopurify 125 I-TSH receptor complexes from the Triton X-100 extract of cell membranes. In the assay T5-317 was used as the "capture" antibody and biotinylated T5-51 in conjunction with 125 I-streptavidin as the "reporter" antibody. Biotinylation of this antibody was performed using the Amersham biotinylation kit (Amersham). The T5-317 antibody (0.5 g/well) was coated onto 96-well plates (Maxisorb, Nunc, Rockilde, Denmark) in 0.05 M potassium phosphate buffer (pH 7.4) for 2 h at room temperature. Plates were washed four times with 300 l of phosphate-buffered saline (PBS) (pH 7.4), bovine serum albumin 0.1% (w/v), Tween 0.05% (PBT) and saturated for 2 h at room temperature with PBS (pH 7.4), bovine serum albumin 1% (w/v), Tween 0.05%. After washing, samples were then incubated, directly or diluted in PBT, for 16 h at 4°C or 2 h at room temperature. The plates were washed with PBT and the biotinylated antibody (T5-51) added at a 0.5 g/ml concentration in PBT for 2 h. After washing, 125 I-streptavidin (75,000 cpm/well) was added for 1 h and then extensively washed with PBT. Finally, 200 l of 0.1 N NaOH was added in each well and an aliquot of 100 l counted. Samples were quantified relatively to a standard curve of TSHR immunopurified from the stably transfected L cell line (17). This cellular TSHR had previously been quantified comparatively to a hTSHR peptide-␤-galactosidase protein fusion standard (15). This assay allowed us to detect routinely as little as 2 fmol/ml of soluble TSHR (sTSHR). We also developed a second assay for the determination of full length heterodimeric (cleaved) and monomeric (uncleaved) forms of the TSHR by replacing T5-317 by an antibody specific for the intracellular part of the TSHR (T3-365) (15).
TSHR Determinations in Culture Medium and Cellular Extracts-For determination of sTSHR concentration, or for the assay of the full-length receptor in the culture medium (␣/␤ receptor), the medium was removed, centrifuged at 1500 ϫ g for 5 min and stored at Ϫ20°C. Cells were washed with ice-cold PBS, scraped, centrifuged at 1500 ϫ g for 5 min, and stored at Ϫ70°C. An aliquot of the Triton X-100 cellular extract (16) was used for cellular TSHR (cTSHR) determinations or protein content quantification using the Pierce BCA protein assay reagent kit (Pierce). 125 I-TSH Binding to Soluble and Cellular TSHR-TSH receptor binding assays were performed using the method described by Petersen et al. (19). Triton X-100 cellular extracts (17) or culture media from stably transfected L cells (concentrated 30-fold using successive Centriprep-10 and Centricon-10 concentrators (Amicon, Beverly, MA)) were used for these experiments. Briefly, 2 pmol of soluble or cellular TSHR were added to iodinated bovine TSH in the absence or in the presence of increasing concentrations of unlabeled bovine TSH (bTSH) human FSH (hFSH) or human CG (hCG). The bound complexes were precipitated with polyethylene glycol (molecular weight: 6000).
Immunopurification of the Soluble TSHR-Immunopurification of sTSHR from the permanently transfected L cells was performed as described previously (17) except that a T5-51 Affi-Gel-10 immunomatrix was used. For purification of sTSHR culture media containing 1% fetal calf serum were concentrated using a Minitan apparatus (Millipore, Bedford, MA) equipped with a filter having a 30-kDa molecular mass cutoff. Deglycosylation of receptors using N-glycanase F and endoglycosidase H and immunoblots were performed as described (17).
LH/CG Receptor Measurements in Culture Medium of a Stably Transfected L Cell Line-A double determinant radioimmunoassay was developed using two additive monoclonal antibodies against the extracellular part of the porcine LH/CG receptor (LHR 38 and biotinylated LHR 436) (14). The sensitivity of the assay was ϳ20 fmol/ml. The methodology of this new assay was similar to the sTSHR assay. LH/CG receptor quantification was performed using an L cell line permanently expressing the porcine LH/CG receptor (12) and an hCG binding assay as described previously.
Quantification of ␣/␤ Heterodimers by Treatment of Cells with Dithiothreitol to Release TSHR ␣ Subunit-At the end of incubation, cells were scraped gently as described above and incubated for 2 h at 4°C in 500 l of a PBS (pH 7.4), 100 mM dithiothreitol solution under agitation. The cells were then centrifuged at 1500 ϫ g for 5 min. An aliquot of the supernatant was diluted 100-fold for the assay of released ␣ subunits. Then the cells were treated with Triton X-100 in order to solubilize the remaining uncleaved TSHR. An aliquot of this cellular extract was used for the assay of the cellular receptor.
Statistical Analysis-Statistical differences between groups were analyzed by one way analysis of variance followed by Fisher's least significant difference test for multiple comparison. p Ͻ 0.05 was considered significant. Statview computer program (Abacus Concepts Inc., Berkeley, CA) was used for calculations.

Identification of a Soluble Form of the TSHR (sTSHR) Released from Transfected
Cells-To detect possible shedding of the extracellular domain of the TSH receptor from cells, we first studied a L cell line stably transfected and expressing high levels of the human TSH receptor. This model system is more amenable to experimental analysis and more reproducible than primary cultures of human thyrocytes. In the latter the concentration of the receptor is known to be very low (15).
A double-determinant (sandwich-type) radioimmunoassay was developed for the specific quantification of the TSH receptor ␣ subunit. This assay uses two additive monoclonal antibodies specific for the extracellular domain (␣ subunit) of the TSHR (T5-317 and T5-51) (15). T5-317 is the capture antibody.
As illustrated in Fig. 1A, the accumulation of a soluble form of the TSHR was detected in the medium of a mouse L cell line stably transfected with the full-length human TSH receptor cDNA (17). This accumulation was time-dependent and reached 50 pM after 48 h of culture (ϳ25% of total receptor present in the cells at this time). The same result was obtained when the medium was filtered through a 0.2-m syringe filter or ultracentrifuged at 15,000 ϫ g to remove cellular debris.
The assay was specific. Competition with an excess of unlabeled T5-51 antibodies suppressed the binding of the reporter antibody, while a nonspecific competitor antibody had no effect. The replacement of one of the TSHR antibodies in the assay by a nonspecific irrelevant monoclonal antibody or the omission of TSHR also suppressed the binding of the 125 I-streptavidin (data not shown).
We also examined the possibility that we were detecting the receptor present in membrane fragments released into the medium. To verify this point we used an alternative assay in which the capture antibody recognized the intracellular domain (␤ subunit) of the receptor (T3-365). This assay detects cleaved or uncleaved whole receptor molecules. No such receptor form was detected in the culture medium (Fig. 1A).
These experiments strongly suggested that spontaneous shedding of a soluble form of the TSH receptor, corresponding to its extracellular domain, occurred in the permanently transfected L cell line. The same determinations were performed on the medium of another TSHR expressing Chinese hamster ovary (CHO)-derived cell line (a gift from C. Maenhaut and G. Vassart) (20). Those cells also produced and released a soluble form of the TSHR into the culture medium in proportions comparable with the stably transfected L cell line (data not shown).
As a control, we also studied L cells stably transfected with the related LH/CG receptor which express high levels of mature receptors (12). The presence of a functional receptor on the surface of these cells has been shown by hormone binding and adenylate cyclase stimulation. A similar assay was developed using two additive monoclonal antibodies specific for the extracellular domain of the LH/CG receptor. No receptor ectodomain could be detected in the cell culture medium even after a 7-fold concentration of the medium (Fig. 1B). This is consistent with the fact that this receptor is expressed as an uncleaved monomer (14).
Thus, the existence of a soluble form of the receptor released into the medium is specific for cells expressing the TSH receptor.
Identification of a Soluble Form of the TSHR in Primary Cultures of Human Thyrocytes-In order to confirm the results obtained with our transfected L cell model, we investigated TSHR ectodomain shedding in a more physiological system, namely in human thyrocytes. We observed an accumulation of an immunoreactive soluble form of the TSHR in primary cultures of human thyrocytes (Fig. 2). The proportion of this soluble TSHR to the total cellular TSHR concentration reached about 17% after 48 h of incubation and more than 22% after 72 h. The presence of sTSHR in thyrocyte culture medium was confirmed in two independent experiments.
Characterization and Purification of the sTSHR from the Stably Transfected L Cell Line-Preliminary experiments had shown that when the medium was filtered on Centricon membranes with a molecular cutoff of 30 kDa, most of the immunoreactivity (Ͼ90%) was retained by the filters (not shown).
To test the ability of sTSHR to bind specifically TSH, concentrated (30 ϫ) culture medium from transfected L cells was incubated with iodinated bovine TSH, in the presence or in the absence of increasing concentrations of unlabeled bTSH, hFSH, or hCG. The bound complexes were precipitated by polyethylene glycol. The same experiment was performed in parallel with cellular TSH receptor present in Triton X-100 extracts. As shown in Fig. 3, sTSHR specifically bound TSH, in a manner similar to the cellular receptor. Competition with unlabeled TSH yielded almost complete displacement, while hFSH and hCG had no significant effect at these concentrations.
Finally, sTSHR was purified from culture medium using immunoaffinity chromatography with the T5-51 immunomatrix (Fig. 4). More than 90% of the receptor present in the medium bound to the matrix and was eluted at acidic pH.
Western blots were performed to compare the immunopurified sTSHR with the cellular TSHR (solubilized from the membranes of the stably transfected L cell line). Antibody T5-317 recognizing the extracellular domain of the receptor was used. As shown in Fig. 5, sTSHR had an apparent molecular mass of ϳ53 kDa (sTSHR, lane C) which is slightly smaller than the ␣ subunit of the TSHR purified from cells (ϳ60 kDa) (cTSHR, lane C). However after digestion with N-glycanase F, which removes all oligosaccharides, both proteins migrated as the same ϳ35 kDa species (cTSHR (lane F) and sTSHR (lane F)). Endoglycosidase H (which removes only mannose-rich moieties of precursor glycoproteins) had no effect on sTSHR or on the ␣ subunit of the cellular TSHR (cTSHR (lane H) and sTSHR (lane H)). These results indicate that sTSHR and the ␣ subunit of the receptor contain the same polypeptide core and differ only slightly in their mature carbohydrate content. This difference may be due to the action of glycosidases during accumulation in the culture medium.
In L cells, a mannose-rich ϳ95-kDa precursor (cTSHR, lane C) was present in high concentration. It was converted into an ϳ80-kDa species after treatment with N-glycanase F or endoglycosidases H (cTSHR (lanes F and H)). A small amount of a ϳ120-kDa monomeric precursor (cTSHR, lane C) which contains mature oligosaccharides was also observed. The latter could be converted into the ϳ80-kDa species by N-glycanase F (cTSHR, lane F) but was resistant to endoglycosidase H (cTSHR, lane H). The amount of the mature ϳ120-kDa species was variable and seemed to be related to cell growth conditions and cell confluence (data not shown). In all conditions, it was markedly less abundant than the cleaved receptor. Pulse-chase experiments had previously confirmed that both forms correspond to precursors of the TSH receptor (17). The high mannose precursor may accumulate inside the transfected cells because the cellular machinery is unable to process to comple-  Fig. 1, except that the cell culture medium was concentrated 15 times on Centricon-10 filters before receptor assay. Ectodomain (␣ subunit) and full-length receptor (␣/␤ receptor) were assayed as described in the legend to Fig. 1. The cellular receptor (cTSHR) was assayed in Triton X-100 membrane extracts. Results are expressed as the mean ratio (n ϭ 2) of soluble TSHR (sTSHR) on the cellular amount of receptor (cTSHR). tion the overexpressed receptor.
Taken together, these experiments allowed us to conclude that there is spontaneous shedding of a functional extracellular domain of the TSH receptor in L cell culture medium, which corresponds to the ␣ subunit of the receptor. The same immunoreactivity was detected in a CHO-derived cell line expressing the TSHR, as well as in the medium of primary cultures of human thyrocytes.
Regulation of sTSHR Shedding-At the beginning of these studies we wondered if the release of soluble receptor could not be an artifact due to proteolytic enzymes present in the serum used for cell culture. To investigate this possibility L cells expressing TSHR were cultured in decreased serum concentrations. However, we observed the opposite effect to that expected. As illustrated in Fig. 6, the accumulation of sTSHR in the medium was markedly increased when the fetal calf serum concentration was reduced in the medium. The ratio of soluble to cellular receptor was increased 1.6-and 3-fold when serum concentration was reduced from 10% to 5% and 1%, respectively (Fig. 6, inset). In 1% serum, sTSHR concentration reached ϳ25% of the cellular receptor content after 24 h of culture. Addition of serum albumin to cell culture medium containing 1% serum to keep protein concentration constant did not prevent the increased shedding (data not shown).
Since serum concentration determines the rate of cell division, we wondered if sTSHR shedding might be linked to cell proliferation. Cells were cultured in presence of reduced serum concentration (1%), but insulin, fibroblast growth factor, or epidermal growth factor were added to the culture medium. All these treatments increased cell proliferation (Fig. 7A) but did not decrease sTSHR shedding (Fig. 7B). Thus the effect of serum is not secondary to changes in cell proliferation rates but is due to the presence of an unidentified component. The latter is not thermolabile, since heating of fetal calf serum for 30 min at 56 or 95°C prior use did not reverse the inhibitory effect.
Effect of TSH and Second Messengers on sTSHR Shedding-The shedding of the extracellular domain of several other membrane receptors is regulated by their ligand or by the activation of protein kinase C (reviewed in Refs. 21 and 22). We thus investigated if sTSHR shedding could be regulated by similar mechanisms. Table I shows the effect of TSH on sTSHR shedding. A limited (ϳ30%), but statistically significant, stimulatory effect was observed. This effect might have been underes-  timated, since TSH is known to promote receptor internalization (17,23). Thus, fewer cell surface receptors were probably available to release their ␣ subunits.
In order to better discriminate the pathways (adenylate cyclase or phospholipase C) implicated in this hormonal effect, we studied forskolin, 8-bromo-cAMP, phorbol 12-myristate 13-acetate (PMA), as well as calcium ionophores (Table II). Only PMA (43% over untreated control) and the calcium ionophores A23187 (46%) and ionomycin (46%) (all p Ͻ 0.01) mimicked the effect of bTSH on sTSHR shedding. Forskolin and 8-bromo-cAMP had no significant effect. These results suggest that the protein kinase C pathway and intracellular calcium concentration are important parameters in the regulation of sTSHR shedding.
Effect of Inhibitors of Intracellular Traffic on sTSHR Shedding-We used various inhibitors to analyze the role of receptor cellular trafficking on the shedding. Cleavage of the receptor might have occurred either during its internalization and its recycling or during its biosynthetic progression through the Golgi apparatus. Initially we analyzed the effect of inhibitors interfering with internalization, recycling, and degradation in lysosomes (Fig. 8). In order to prevent adverse effects on cell viability, incubations with these agents were limited to 8 h. Under these conditions, only small effects on total cellular receptor content were observed (Ͻ10% decrease).
The weak amine chloroquine, which increases the pH of acidic vacuoles, impedes the traffic through acidified compartments and disturbs lysosomal function (24), enhanced sTSHR shedding (50% over untreated cells). Bafilomycin, an inhibitor of the Hϩ-ATPase of acidic vacuoles (25), also enhanced TSHR shedding. Monensin, a monovalent carboxylic ionophore which dissipates transmembrane proton gradient and blocks the recycling pathway (26), had a stimulatory effect on sTSHR shedding. These results suggested that endocytosis and lysosomal proteolysis are not implicated in sTSHR shedding and that all processes which enhance receptor residency on the plasma membrane increase shedding.
Finally brefeldin A, an agent which inhibits Golgi function (27), did not modify TSHR shedding, suggesting that this process is not linked to any event occurring during the progression of the receptor from the Golgi complex to the cell membrane. (The 8-h incubation was insufficient to decrease receptor concentration on the cell surface).
Together, these experiments strongly supported the concept that sTSHR shedding was dependent upon events occurring at or near the cell membrane.
Effect of Protease Inhibitors on sTSHR Shedding-To identify the protease involved in TSHR maturation, a variety of inhibitors was examined for their effect on the accumulation of sTSHR in L cell culture medium (Table III). The majority of the protease inhibitors tested did not modify the accumulation of sTSHR, including inhibitors of serine proteases (aprotinin), FIG. 7. Effects of growth factors on sTSHR shedding. L cells expressing TSH receptor were grown for 24 h in a medium containing 10 or 1% fetal calf serum (FCS). Either insulin (INS, 10 g/ml), fibroblast growth factor (FGF, 50 ng/ml), or epidermal growth factor (EGF, 10 ng/ml) were added to the medium of cells grown in presence of 1% fetal calf serum. Control cells (Cont) were grown in the absence of growth factors. A, the protein content was determined and used to evaluate the cell number. It was expressed as the mean Ϯ S.E. (n ϭ 3). B, the ratio of soluble (sTSHR) to cellular (cTSHR) receptor corresponds to three independent determinations (mean Ϯ S.E.). *, p Ͻ 0.05; **, p Ͻ 0.1 versus 1% FCS control.

TABLE I
Effect of TSH on sTSHR shedding L cells expressing TSHR were grown in medium containing 1% fetal calf serum for 24 h at 37°C in the absence (control) or in the presence of various amounts of bTSH. Cellular and soluble TSHR and total protein content in the homogenate were determined as described under "Experimental Procedures." sTSHR content was standardized per mg of protein in the cell homogenate (this corrects for variations in cell numbers). It was then expressed as percentage of control in the absence of TSH. Mean Ϯ S.E. (n ϭ 3) are indicated.  sTSHR shedding L cells expressing TSHR were grown in a medium containing 1% fetal calf serum for 24 h at 37°C in the absence (control) or in the presence of bTSH, 8-bromo-cAMP, forskolin, PMA, A23187, and ionomycin. Cellular TSHR and total protein content in the homogenate were determined and expressed as described in Table I  cysteine proteases (E64), aspartyl proteases (pepstatin), and aminopeptidase (bestatin). A limited (ϳ24%) inhibition of TSHR shedding was observed only at high concentrations (3 mM) of leupeptin. EDTA and EGTA (3 mM for each) also significantly inhibited sTSHR shedding. Higher concentrations could not be used because of their deleterious effects on cells. Although limited, this inhibition led us to postulate that a metalloprotease might be involved in the maturation of TSHR. Inhibitors of nonmatrix metalloproteinases (captopril, phosphoramidon) had no effect. We thus investigated the effect of a recently described potent inhibitor of matrix metalloproteases, the synthetic hydroxamic acid BB-2116 (28). As shown in Fig. 9A, we observed a dosedependent inhibition of sTSHR accumulation in cell culture medium which became undetectable at 100 g/ml of BB-2116. The concentrations of BB-2116 suppressing TSHR ␣ subunit shedding matched those previously described for the inhibition of pro-tumor necrosis factor ␣ cleavage (28). However the shedding of TSHR ␣ subunit is a two-step process. The first step consists in the cleavage of the receptor and the second step in the reduction of its disulfide bond(s). It was necessary to verify that BB-2116 was indeed acting on the cleavage. We thus measured the concentration of ␣ subunits linked by disulfide bonds to ␤ subunits and which thus remain attached to cell membranes. This was done by treating cell membranes with dithiotreitol and measuring the released ␣ subunits. If inhibition occurs at the level of the cleavage the concentration of "releasable" ␣ subunit should decrease, whereas if inhibition occurs at the level of reduction of disulfide bridge(s) it should increase.
A marked decrease in the concentration of membrane-attached cleaved receptors was produced by incubation with BB-2116 (Fig. 9B). Thus the inhibition occurs at the level of the cleavage of the receptor. However at concentrations of inhibitor

TABLE III
Effect of protease inhibitors on sTSHR accumulation L cells expressing TSHR were cultured for 24 h in 1% serum in the presence of the inhibitors. The soluble TSHR was assayed in the cell culture medium. It was compared with the concentration of soluble receptor in culture medium of cells grown in the absence of inhibitors (mean of three determinations). TLCK, tosyl-L-lysine chloromethyl ketone; TPCK, L-1-tosylamido-2-phenylethyl chloromethyl ketone; TAME, tosylarginyl methyl ester; pCMB, p-chloromercuribenzoate. which totally suppressed ␣ subunit shedding there was only a 55% decrease of cleaved receptors present on the membranes. This suggests that a critical threshold of the ␣/␤ heterodimer must be reached on the cell surface to allow shedding to occur. DISCUSSION Using a quantitative immunoradiometric assay performed with two monoclonal antibodies directed against the extracellular domain of the TSH receptor, we have detected an immunoreactive protein in the culture medium of L cells and CHO cells stably transfected with the TSH receptor. The same immunoreactivity was released from human thyrocytes. This accumulation was time-dependent and reached ϳ25% of the total cellular receptor by 48 h.
This soluble TSHR is a glycosylated protein which is retained by concanavalin A-Sepharose (not shown) and specifically binds its ligand TSH. Immunopurification of sTSHR from the L cell line conditioned media demonstrated that its polypeptide core corresponds to the ␣ subunit (ϳ35 kDa).
These experiments strongly suggested that there is a spontaneous shedding of the extracellular domain of the TSH receptor in stably transfected L and CHO cells and in human thyrocytes. The shedding is specific for TSHR-expressing cells and does not occur in a L cell line stably transfected with the LH/CG receptor. This observation is concordant with the fact that nearly all of the TSH receptors in human thyroid tissue and most of these receptors in L cells (17) have a heterodimeric structure, while the LH/CG receptors are expressed at the cell surface as uncleaved monomers (12,14).
The structure of the TSH receptor has been debated. The hypothesis that artifactual proteolysis occurs during cell homogenization has been proposed by some authors (29 -31). This possibility was not supported by our previous experiments (15,17) and is strongly opposed by the present observation of spontaneous shedding of the extracellular domain of the TSH receptor by intact cells in culture.
In addition, we have detected the presence of sTSHR in normal human serum using the same assay. 2 Complete characterization of this circulating sTSHR will allow us to determine whether it corresponds to the sTSHR detected in the supernatant of cultured cells. There have been some preliminary reports of TSHR related peptide-like immunoreactivity (32) and TSH-binding protein (33) in human serum. Our preliminary data suggest that the sTSHR that we detect in human serum is generated by shedding. Alternatively spliced TSHR mRNAs also have been cloned from normal and Graves' disease thyroids (34,35). Further experiments will allow us to determine whether a correlation can be established between the concentration of sTSHR in the serum and specific pathological conditions. This shed ␣ subunit could be implicated in the pathogenesis of certain autoimmune diseases.
Various cell surface receptors are known to be shed into the extracellular milieu (reviewed in Refs. 21 and 22). In several cases receptor shedding was shown to be increased by its cognate ligand or by phorbol esters (21,22,36). We thus examined the effect of TSH and of various signal mediatory molecules on TSHR shedding. TSH induced a significant increase in the shedding, PMA and calcium ionophores mimicked the effect of TSH, while agents acting on the cAMP pathway had no effect. The effect of TSH might be due to an increase in the concentration or activity of the cleaving enzyme. Ligand binding or cellular activation might also induce changes in the receptor (phosphorylation, conformational changes) that make it more susceptible to enzymatic cleavage. Alternatively TSH might activate another mechanism involved in receptor shedding: possibly the reduction of disulfide bond(s) joining the ␣ and ␤ subunits.
Surprisingly, lowering serum concentration from 10 to 1% greatly enhanced TSHR shedding. This inhibitory effect of serum was not mediated by an effect on cell growth as it could not be reproduced by insulin or other growth factors. The mechanism of this inhibition is not yet understood and needs further study.
The use of various inhibitors of intracellular trafficking of the receptor allowed us to conclude that neither receptor internalization, recycling nor degradation in lysosomes were involved in sTSHR shedding. Moreover all agents that increased the residence time of the receptor at the cell surface increased sTSHR shedding. By contrast, inhibition of Golgi function did not modify receptor shedding. Taken together, these experiments strongly suggested that receptor modifications involved in shedding occur at (or very near) the cell membrane.
We also investigated the nature of the protease involved in TSHR maturation. The best known convertases involved in the maturation of precursor proteins belong to the subtilisin family of serine proteases (37). Their most common cleavage site comprises a pair of basic residues. Such basic sequences are found in the extracellular domain of the TSHR, and we, and others, have previously thought that TSHR convertase might belong to this family (15,38). One member of these proteases, furin, has been shown to be involved in the maturation of the insulin proreceptor (39). This maturation occurs in the late Golgi compartment. In a previous work, pulse-chase experiments indicated that the cleavage mechanism seemed different between TSH and insulin receptors (17).
We examined a variety of protease inhibitors for their effects on sTSHR production. Most of them were ineffective. The divalent ion chelators EDTA and EGTA inhibited TSHR processing (30 and 20% inhibition, respectively). Higher concentrations could not be used in our whole cell system as these compounds had a cytotoxic effect. However these results suggested that the TSHR convertase could be a metalloprotease. Finally we used a potent and specific inhibitor of zinc-dependent matrix metalloproteases. This synthetic hydroxamic acid, BB-2116, inhibits in vitro the maturation of tumor necrosis factor ␣ (28). In our system, this agent totally suppressed sTSHR accumulation in culture medium. This effect was secondary to a strong inhibition of TSHR cleavage into ␣/␤ heterodimers.
BB-2116 inhibits in vitro the activity of collagenases, stromelysins, gelatinases, and PUMP I, which all belong to the matrix metalloprotease family (28). These enzymes are secreted by the cells and their primary role is to degrade extracellular matrix proteins such as collagen, laminin, and proteoglycan. They are zinc-and calcium-dependent enzymes secreted as zymogens and are activated in situ by different mechanisms, particularly proteolysis (reviewed in Refs. 40 and 41). These enzymes are regulated by numerous growth factors, hormones, and tissue inhibitors and are implicated in the physiological remodeling of connective tissue, in destructive inflammatory pathologic processes and in tumour invasion (reviewed in Ref. 42). Very recently matrix metalloproteases have also been implicated in the maturation of tumor necrosis factor ␣, a potent pro-inflammatory and immunomodulatory cytokine produced in inflammatory conditions (27,43). The sites of cleavage are difficult to predict as matrix metalloproteases exhibit broad substrate and sequence specificities (44 -46). The finding that matrix metalloprotease-like enzymes are implicated in TSHR maturation also supports our observation that the cleavage occurs at the cell membrane.
The shedding of TSH receptor is a two-step process: cleavage of the receptor and reduction of disulfide bond(s). It is unknown if the second step may also be regulated. (We cannot, however, eliminate the possibility that in a small fraction of receptor molecules no disulfide bonds are formed and that the cleavage alone is sufficient for the shedding to occur). Further experiments will be necessary to understand the physiological significance of the shedding process. We do not know yet if receptor cleavage is necessary for receptor biological activity. The low concentration of the 120-kDa precursor did not allow to study its hormone binding ability. Purification of the ␤ subunit of the receptor and sequencing of its N-terminal part will allow to define the site of cleavage. Mutagenesis experiments will then lead to the understanding of the role of the cleavage in the function of the TSH receptor. Other questions remain unanswered: is shedding of the ␣ subunit a mechanism limiting the transmission of the signal? Is the shed ␣ subunit able to bind hormone in the plasma and does it influence its biological activity?