Identification of the Membrane Receptor Binding Domain of Thyroglobulin

The last stages of thyroglobulin maturation occur in the thyroid follicular lumen and include thyroid hormone formation and glycan completion. In this compartment, newly secreted thyroglobulins interact with a thyrocyte membrane receptor that prevents their premature lysosomal transfer and degradation. Both GlcNAc moieties and thyroglobulin peptide determinants are involved in receptor interaction. Here we used monoclonal antibodies (mAbs) directed against human thyroglobulin either to inhibit (mAb78) or to enhance (mAb240) the thyroglobulin binding and to identify the region of the thyroglobulin involved in the receptor recognition. Peptides containing the mAb epitopes were obtained by immunoscreening cyanogen bromide-derived native human thyroglobulin peptides and a cDNA thyroglobulin expression library. Three peptides, localized in the thyroglobulin N-terminal domain, were obtained. Peptides N1 (Ala1148-Gln1295) and N2 (Ser789-Met1008) were recognized by mAb240 and mAb78, respectively. None of them bound the receptor. The third peptide, N3 (Ser789-Met1172), (i) overlapped all or part of the N1 and N2 peptide sequences and was recognized by both mAbs, (ii) carried two complex glycans at Asn797 and Asn928, of which a subset presented accessible GlcNAc residues, and (iii) inhibited the thyroglobulin binding to FRTL5 cell membrane preparations. The N3 peptide includes tyrosine residues that have been reported to be involved in hormone formation. These results suggest that structural modifications closely associated with hormone formation within this domain act as sensors for the receptor interaction and thus for the intrafollicular retention or lysosomal homing of the prohormone.

The last stages of thyroglobulin maturation occur in the thyroid follicular lumen and include thyroid hormone formation and glycan completion. In this compartment, newly secreted thyroglobulins interact with a thyrocyte membrane receptor that prevents their premature lysosomal transfer and degradation. Both GlcNAc moieties and thyroglobulin peptide determinants are involved in receptor interaction. Here we used monoclonal antibodies (mAbs) directed against human thyroglobulin either to inhibit (mAb78) or to enhance (mAb240) the thyroglobulin binding and to identify the region of the thyroglobulin involved in the receptor recognition.
Peptides containing the mAb epitopes were obtained by immunoscreening cyanogen bromide-derived native human thyroglobulin peptides and a cDNA thyroglobulin expression library. Three peptides, localized in the thyroglobulin N-terminal domain, were obtained. Peptides N1 (Ala 1148 -Gln 1295 ) and N2 (Ser 789 -Met 1008 ) were recognized by mAb240 and mAb78, respectively. None of them bound the receptor. The third peptide, N3 (Ser 789 -Met 1172 ), (i) overlapped all or part of the N1 and N2 peptide sequences and was recognized by both mAbs, (ii) carried two complex glycans at Asn 797 and Asn 928 , of which a subset presented accessible GlcNAc residues, and (iii) inhibited the thyroglobulin binding to FRTL5 cell membrane preparations. The N3 peptide includes tyrosine residues that have been reported to be involved in hormone formation. These results suggest that structural modifications closely associated with hormone formation within this domain act as sensors for the receptor interaction and thus for the intrafollicular retention or lysosomal homing of the prohormone.
Thyroglobulin, the major glycoprotein produced by the thyroid gland, is the substrate for the biosynthesis of the two thyroid hormones, tetra-(thyroxine) and triiodothyronine. Mature thyroglobulin is a dimer with a M r of 660,000, consisting of two apparently identical chains, each containing 67 tyrosyls and 2748 amino acid residues (numbered to exclude a 19-residue leader sequence) in the human (1). The glycoprotein is synthesized and partially matured within thyrocytes (intramonomer disulfide bonding, glycosylation, sulfatation, phosphorylation) and then secreted into the follicular lumen. In this compartment, thyroglobulin is further processed by iodination, hormonogenesis, and completion of some complex type oligosaccharide units (2).
Iodination of thyroglobulin tyrosyls is thought to be mediated by the thyroperoxidase at the thyrocyte apical cell surface. About 10 -50 atoms of iodide are associated with each molecule (3). Hormone formation per se requires thyroperoxidase-mediated coupling of two iodotyrosyls. The coupling reaction leaves iodothyronine and dehydroalanine at the acceptor and donor sites, respectively (4,5). Thyroglobulins from various species have been compared. The hormonogenic iodotyrosyl residues (no more than four per monomer) are in both the N-and C-terminal domains of the prohormone, and their positions are well conserved (6 -11). The hormonogenic 580 C-terminal residue domain is cysteine-poor, presents no repeated sequence, and structure predictions suggest that it is flexible (1). In contrast, the N-terminal domain (encoded by exon 1 to exon 15, residues 1-1209) consists of tandem repeats rich in cysteine residues, called type 1 repeats, interrupted by unrelated sequences (1,12). These cysteine-rich repeats are probably involved in disulfide bonds, suggesting that iodotyrosyl coupling results in a highly coordinate three-dimensional structure with a limited flexibility. In vivo kinetic experiments have shown the following: (i) production of mature prohormones requires intralumenal retention, and (ii) mature thyroglobulins preferentially participate in the secretion of hormones in the venous flow, after thyroglobulin internalization and its proteolytic cleavage in the lysosomal system (2,13,14). These findings and the observation that thyroglobulin iodination is closely associated with glycan completion (15) led us to investigate and demonstrate a quality control system that prevents lysosomal homing and degradation of iodine-poor immature thyroglobulin. This control depends on the recognition of a subset of molecules by an endogenous receptor and suggests that there is a recycling mechanism by which receptor-bound immature thyroglobulins are passed back to the thyroperoxidase iodination site via the Golgi apparatus (16). Thyroglobulin binding to the thyrocyte receptor is optimal at acidic pH (17,18), which is consistent with an interaction in prelysosomal compartments. The thyroglobulin receptor interacts with both carbohydrate and peptide determinants of the prohormone (15)(16)(17), and we have obtained and characterized anti-thyroglobulin monoclonal antibodies wherein some inhibit and others enhance thyroglobulin binding to its receptor (18).
The purpose of the present study was to use anti-thyroglobulin antibodies to identify the receptor binding domain of the * This work was supported by institutional funding from the Center National de la Recherche Scientifique and grants from l'Assistance Publique de Marseille and the Program Hospitalier de Recherche Clinique 1994 (to R. M.). 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 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) 625277.
‡ To whom correspondence should be addressed. human prohormone. We thereby mapped the receptor binding domain to the N-terminal thyroglobulin region encoded by exons 10 to 14. The domain presents incompletely processed oligosaccharide units bearing accessible GlcNAc residues and includes tyrosyl residues reported to be either iodinated or involved in hormonogenesis.

Preparation and Analysis of Human Thyroglobulin Peptides
CNBr Peptides Purification-Peptide fragments from iodine-poor human thyroglobulin (gift from B. Mallet and P-J Lejeune, Faculté de Medecine, Marseille, France) were aliquot fractions of previously described preparations (20). Briefly, thyroglobulin (2.8 atoms of iodine/ mol, traces of T3, 10 mmol T4/mol) was from a patient with a single colloid goiter. Purified thyroglobulin was treated with cyanogen bromide (CNBr), and the resulting digest was eluted by chromatography on a Sephadex G-200 column in 1 M propionic acid. Five fractions (I-V) were collected (void volume to gel-included fractions), dialyzed, and lyophilized.
Preparative Electrophoresis-CNBr peptide fragments of fraction II were further separated using a model 491 Prep Cell preparative electrophoresis apparatus (Bio-Rad). Peptides (20 mg) were solubilized in buffer A (60 mM Tris-HCl, pH 6.8, 2% SDS, 5% ␤-mercaptoethanol, 10% glycerol, 5% bromphenol blue). Samples were heated (5 min at 90°C) and loaded onto a 12% acrylamide running gel on small Prep Cell column (28 mm inner diameter). Electrophoresis was carried out (200 -250 V/40 mA constant current) for 12 h. After elution of the bromphenol blue, 2.5-ml fractions were collected at a flow rate of 1 ml/min. Prestained standards were used to estimate molecular masses. Peptide SDS-Polyacrylamide Gel and Immunoblotting-Fractions II1 to II5 were further characterized by SDS-PAGE (12% acrylamide) and immunoblotting using mAb78 or mAb240 (1-3 g/ml) for 2 h and then alkaline phosphatase-conjugated anti-mouse antibodies for an additional 2 h. Single peptides N2 and N3, corresponding to the major peptide of subfractions II2 and II3, respectively, were separated from minor contaminants either using a C 18 reverse-phase HPLC column in a gradient of trifluoroacetic acid/acetonitrile (peptide N2) or 10% SDSpolyacrylamide gels, excision of the major band, and electroelution (peptide N3).
Peptide Sequencing-Purified N2 and N3 peptides (500 pmol) were concentrated and washed onto a Problott polyvinylidene difluoride membrane (Prospin Sample Preparation Cartridge, Applied Biosystems) and sequenced. Automated Edman degradations were performed using a 476A protein sequencer (Applied Biosystems) with an on-line 140A analyzer for identification of the phenylthiohydantoin-amino acid derivatives separated on a phenylthiohydantoin C 18 reverse-phase HPLC column. The location of sequences within thyroglobulin were found by computer comparison with amino acid sequences deduced from published cDNA sequences (Ref. 1, accession number 625277).
Immunoscreening and Isolation of a Human Thyroglobulin Fusion Peptide-pEX1, pEX2, and pEX3 human thyroglobulin expression libraries (18), which produce hybrid thyroglobulin fragments (19), were plated on nitrocellulose disks (BA85, diameter 85 mm; Schleicher and Schü ll, Dassel, Germany) by filtration to obtain approximately 100 -500 colonies per disk and were grown at 32°C on Luria Bertani/ampicillin medium. After amplification and regrowth, fusion protein synthesis was induced by a 2-h incubation at 42°C. Nitrocellulose disks were treated with chloroform vapor for 15-30 min, incubated in lysis buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM MgCl 2 , 3% bovine serum albumin, 1 g/ml DNase I, 40 g/ml lysozyme) for 4 h at room temperature, and rinsed in Tris-buffered saline. For immunoscreening, disks were incubated for 2 h in the presence of mAb78 or mAb240 (1-5 g/ml). Selected immunoreactive clones were grown in Luria Bertani medium containing 100 g of ampicillin per ml at 32°C to obtain an optical density at 550 nm of 0.4. Protein expression was induced to obtain large quantities of hybrid protein (23). Bacterial extract (5 l) was dissolved in electrophoresis sample buffer, separated by 8% SDS-PAGE, and transferred to nitrocellulose. Immunodetection was performed as described above.
The sequence of the cDNA insert of the positive clone was determined by the dideoxy-mediated chain termination method with T7 DNA polymerase (Sequenase, U. S. Biochemical Corp.).
Iodination-Human thyroglobulin (10 g) was iodinated with IODO-GEN (50 g) and Na 125 I (1 mCi) in 50 l of 50 mM Tris-HCl, pH 8.5, for 15 min at 4°C. The reaction was stopped by dilution with 1 ml of PBS, and Tg was isolated by gel filtration on a Sephadex S-200 column. The specific activity was 95 Ci/g. The N3 peptide (4 g, specific activity 10 Ci/g) was iodinated using the same procedure.
Glycosidase Treatments-Endoglycosidase H treatment was carried out by incubating the reduced N3 peptide (1.5 g) for 20 h at 37°C with 2 milliunits of Endo-H in 100 mM sodium phosphate buffer, pH 6.0, containing 0.1% Triton X-100 and 0.03% SDS. N-Glycanase treatment was also performed on the reduced N3 peptide (1.5 g) by incubation at 37°C for 20 h with 1.5 milliunits of N-glycanase in 50 l of 50 mM acetate buffer, pH 5.0, 10 mM EDTA, 0.5% Triton X-100, 0.1% SDS, 1% ␤-mercaptoethanol. The reaction was stopped by heating for 3 min at 100°C and subjecting the sample to SDS-PAGE under denaturing conditions. The proteins were blotted onto nitrocellulose which was then incubated with mAb78 (diluted 1/100), added for 2 h, and alkaline phosphatase-conjugated anti-mouse antibodies for 2 h. In an additional experiment to assess the extent of deglycosylation at various concentrations of N-glycanase, aliquot fractions of radiolabeled 125 I-N3 peptide (10 ng, 120,000 cpm) were treated in the same medium with N-glycanase at each 25, 2.5, 0.25, 0.025, 0.0025, and 0.00025 milliunits. The samples were resolved by SDS-PAGE under reducing conditions, and the gel was scanned and analyzed with a PhosphorImager (Bio-Rad, Les Ullis, France).
Neuraminidase treatment was performed on aliquots (30 ng) of the radiolabeled N3 peptide by reaction with 25 milliunits of C. perfringens sialidase in 60 l of acetate buffer, pH 5.0, for 20 h at 37°C. The reaction was stopped by heating for 3 min at 80°C, and the sample was then extensively dialyzed against phosphate buffer.
Lectin Affinity Chromatography-Incompletely processed complex type oligosaccharides bearing accessible GlcNAc residues were detected using wheat germ agglutinin (WGA) and Bandeiraea simplicifola II (BSS II)-conjugated agarose columns. A minor modification of established method (18,24) for WGA affinity chromatography was used. Briefly, 20 l of radiolabeled 125 I-N3 peptide (before (250,000 cpm) or after neuraminidase treatment (330,000 cpm)) was applied to WGA columns (150 l) and allow to stand for 1 h. The columns were washed with 15 ml of 10 mM phosphate buffer, pH 7.4, 0.15 M NaCl, 0.2% Nonidet P-40 (wash buffer), and then with 5 ml of a Gal elution buffer (1 M D-(ϩ)-galactose in wash buffer, this step served as a specific control) and 5 ml of a GlcNAc elution buffer (1 M N-acetyl-D-glucosamine in wash buffer). 0.5-ml fractions were collected and counted in a gamma counter. BSS II affinity chromatography was performed as described previously for the native thyroglobulin (15,18). The column (300 l) was equilibrated with loading buffer consisting of 50 mM phosphate buffer, pH 7.4, 0.15 M NaCl, 0.1% bovine serum albumin, and 0.1 mM CaCl 2 , MnCl 2 , and MgCl 2 . After equilibration, 40 l of 125 I-N3 peptide (500,000 cpm) was loaded onto the column and allowed to interact for 1 h. The unbound fraction was recovered by adding 30 ml of loading buffer. The bound fraction was recovered by addition of a buffer containing 0.5 M GlcNAc. 0.5-ml fractions were collected.
Binding Assay-Solid phase assays of thyroglobulin binding to FRTL 5 membranes were performed as follows. Membranes, prepared as described previously (15,18), were added (20 g of total protein in 40 l of PBS buffer) to flat bottom microtiter wells (Falcon 3911, Becton Dickinson, Oxnard, CA) and incubated for 1 h at 37°C and then overnight at 4°C. The wells were washed three times with 150 l of PBS and once with 200 l of PBS containing 2% bovine serum albumin. Thyroglobulin (5 ϫ 10 4 to 10 6 cpm in 100 l of a binding buffer containing 25 mM acetate buffer, pH 5.0, 150 mM NaCl, 5 mM CaCl 2 , and 0.1% bovine serum albumin) was then added and incubated at 4°C for 90 min. The wells were washed four times with the binding buffer, and radioactivity was measured. Reported data are means of at least three separate experiments performed in duplicate or triplicate.

Isolation of a Human Thyroglobulin Fragment-Fusion Pro-
tein Recognized by mAb240 -pEX human thyroglobulin expression libraries were prepared from PstI fragments of thyroglobulin cDNA inserted into the three forms of the pEX plasmid, corresponding to the three reading frames. Between 200 and 700 transformed pOp2136 E. coli cells were plated, corresponding to about seven copies of each fragment in every reading frame. Nitrocellulose culture disk replicas were tested for the presence of peptides encoding the receptor thyroglobulin binding domain by using mAb78 and mAb240 antibodies. One immunoreactive clone recognized by mAb240 was isolated in the initial screening, and only in pEX1 (Fig. 1A). The hybrid protein produced by this positive clone was further characterized. As expected, the length of the Cro-LacZ fusion protein obtained after induction (23) (Fig. 1B, lane 1) was greater than ␤-galactosidase (M r of the fusion protein about 165,000). Moreover, this fusion protein was recognized by mAb240 only (Fig.  1C). The length of the PstI cDNA fragment in this positive clone was 444 base pairs, and the deduced peptide sequence, we named peptide N1, corresponded to Ala 1148 -Glu 1295 on the human thyroglobulin monomeric molecule (Fig. 4).
No other positive clone was detected with either mAb240 or mAb78 antibodies.
Isolation of CNBr Human Thyroglobulin Peptides Recognized by Both mAb78 and mAb240 -The failure to obtain a thyroglobulin fragment-fusion protein encoding the mAb78 epitope prompted us to search for such immunoreactive peptides from human native thyroglobulin. Five fractions previously obtained after Sephadex G-200 filtration of the CNBr digest from iodine-poor thyroglobulin were used as a source of peptide fragments (Ref. 20 and Fig. 2A). After SDS-PAGE and electroblotting, immunoreactive peptides were found only in fraction I (M r 42,000 with mAb78, ranging from 50,000 to 64,000 with mAb240) and fraction II (M r 38,000, and ranging from 50,000 to 80,000 with both mAbs) (Fig. 2B). Because they included the smallest fragment (about 39 kDa) recognized by both mAbs, peptides of fraction II were further separated by preparative electrophoresis. Immunoblot analysis of five fractions, designated II1 to II5 and ranging from 25 to 55 kDa (Fig.  3A), led to the identification of two peptides fractions. One (II2, about 28 kDa) was recognized by mAb78 and the second (II3, about 40 kDa) by both mAb78 and mAb240 (Fig. 3B). The same results were obtained by immunoblot using single peptides N2 and N3, prepared from fractions II2 and II3, respectively, and devoid of minor contaminants after further purification on reverse-phase HPLC column and/or by SDS-PAGE electrophoresis and electroelution (data not shown). Table I presents the results of N-terminal sequencing of the purified N2 and N3 peptides and their deduced position in the human thyroglobulin polypeptide chain (1). After 8 Edman degradation cycles of both peptides, we obtained the same sequence, SYREAASG, corresponding to Ser 789 -Gly 796 on the thyroglobulin polypeptide chain, indicating that both peptides began at Ser 789 . Four additional Edman degradation cycles on peptide N3 failed to identify the residue at position 797. As an Asn residue was expected at this position, but in a context of a potential glycosylation site (NXS), we inferred that Asn 797 was glycosylated. Finally, taking into account both their apparent molecular weights and the possible CNBr cleavage sites within the thyroglobulin corresponding domain (see Fig. 4, lower part), we deduced that the C-terminal residues were Met 1008 and Met 1172 for peptides N2 and N3, respectively. Fig. 4 indicates the complete peptide sequence of the immunoreactive peptides N2 (Ser 789 -Met 1008 ), N3 (Ser 789 -Met 1172 ), and N1 fusion peptides (Ala 1148 -Gln 1295 ) (Fig. 4, lower part)  and their positions on the human thyroglobulin polypeptide chain (Fig. 4, upper part). We have observed that the epitope recognized by mAb240 was present on both peptide N3 and the fusion protein N1 (Fig. 3 and Fig. 1, respectively), and the mAb240 epitope is presumably within the sequence Ala 1148 -Met 1172 , i.e. a type 1 repeat related sequence (25).  Fig. 4), but the N3 peptide contains about 58% unrelated sequences. Second, in contrast to N2, N3 contains two complete type 1 repeats, from amino acid residues 901-909 to 1009 -1055 for the former, and 1056 -1059 to 1084 -1126 for the second (1). Third, among 12 tyrosine residues present in this N3 sequence, 5 (at positions 847, 864, 972, 897, and 1007, respectively) were predicted to be solventexposed and have been reported to be iodinated in vitro or involved in hormone formation in vivo (10,11). Finally, there are as predicted two glycosylation sites, located in hydrophobic (Asn 797 ) and hydrophilic (Asn 928 ) regions, respectively.
Binding Properties of the Isolated N2 and N3 Peptides-To determine whether or not the N2 and N3 peptides contained the domain recognized by the membrane receptor, binding ex-periments using these peptides as competitors were performed. N3, but not N2, completely abolished the binding (Fig. 6). To check whether conserved amino acid residues of type 1 repeats were involved in binding, we used the N-terminal CNBr hormonogenic fragment (Asn 1 -Met 171 ) as competitor. This N-terminal fragment is mainly composed of two type 1 cysteine-rich repeats (about 72% peptide sequence, see Fig. 4 and Refs. 1 and 25). Importantly, neither the N-terminal peptide (Fig. 8), nor the CNBr fragments depleted of the N3 peptide (not shown) inhibited the binding. These data provided evidence that these repeats alone are not directly involved in the thyroglobulin/ receptor interaction.
Glycanase Treatment and Lectin Affinity Chromatography of the N3 Peptide-The peptide N3 presents two potential glycosylation sites, i.e. Asn 797 which is probably glycosylated (see above, Table I) and Asn 928 (Fig. 4 and Fig. 5). GlcNAc-bearing complex type glycans have been implicated in thyroglobulin-   specific binding (17,18), and we therefore determined if such glycans and GlcNAc motifs are present on the N3 peptide that carries the receptor binding domain.
Endoglycosidase H treatment using 2 milliunits of enzyme failed to deglycosylate 1.5 g of peptide N3 (Fig. 7A). By contrast, three forms of the peptide appeared after treatment with 1.5 milliunits of N-glycanase as follows: the fully glycosylated N3 (about 39 kDa), N3 with one glycan chain (about 36 kDa), and the totally deglycosylated N3 (about 34.5 kDa) (Fig. 7B). Even when evaluated in reducing conditions, we failed to obtain completely deglycosylated preparations. As shown in Fig.  7B, deglycosylation of 10 ng of 125 I-labeled N3 peptide, using 25 ϫ 10 Ϫ5 to 25 milliunits glycanase, resulted in the disappearance of only the native 39-kDa peptide and accumulation of both 36-and 34.5-kDa forms. Thus we concluded that N3 was glycosylated at both Asn 797 and Asn 928 and that the monoglycosylated 36-kDa form was partially resistant to deglycosylation, even by up to 25 milliunits of glycanase (Fig. 7B, forms  a-c).
The limited available amount of N3 peptide led us to search for the presence of GlcNAc-bearing glycans by lectin affinity chromatography ( Fig. 8) (18,23). Wheat germ agglutinin binds GlcNAc and, to a lower extent, sialic acid moieties. Fig. 8A shows that a subset of radiolabeled N3 peptide (about 10%) was retained on the column and was selectively released using an elution buffer containing GlcNAc. The same elution pattern was obtained in a control experiment using a neuraminidasetreated N3 peptide (Fig. 8A). This indicates that only GlcNAc moieties were involved in wheat germ agglutinin recognition. These observations were reinforced by the finding that a subset of N3 (about 8.5%) was also selectively retained on a B. simplicifola II affinity column, another GlcNAc-specific lectin (Fig. 8B). In both cases, no additional binding was observed when unbound fractions were analyzed on fresh columns. DISCUSSION We identified the thyroglobulin domain involved in receptor binding. It is localized within a stretch of 383 amino acid residues in the N-terminal region, i.e. within the sequence Ser 789 -Met 1172 . The antigenic domain recognized by mAb78 was located in the region Ser 789 -Met 1008 (peptide N2), a region encoded by the 3Ј end of exon 10 and exon 11 (Fig. 5). mAb78 antagonizes the binding of thyroglobulin to its receptor (18) and recognizes the deglycosylated peptides (present data). Peptide determinants involved in receptor interaction are presumably located in this region. Nevertheless, the observation that complete inhibition of thyroglobulin binding was only obtained with the N3 peptide (Ser 789 -Met 1172 ), but not N2, indicates either the involvement in binding of determinants in the Pro 1009 -Met 1172 sequence, or, alternatively, that this sequence portion is needed for the formation of the three-dimensional structure appropriate for binding. The C-terminal part of N3, namely the Ala 1148 -Met 1172 sequence, is recognized by mAb240, an antibody able to modify the binding domain as judged by its enhancing effect on the thyroglobulin binding (18). This supports the idea that this region is conformationally important.
The thyroglobulin receptor recognizes the Ser 789 -Met 1172 domain consistent with the overall structure and specific determinants of immature molecules being appropriate for intrafollicular retention. These features may then be modified or disappear during maturation to allow the homing of mature prohormone to lysosomes. The Ser 789 -Met 1172 domain fulfills the criteria expected for such a "sensor" domain.
The binding activity of the receptor is partially dependent on the presence of disulfide bonds in thyroglobulin; the reduced prohormone has a higher affinity than the native prohormone for the receptor (18). Flexibility of the binding domain, as for exemple that observed following binding of the mAb240 antibody to thyroglobulin, facilitates the thyroglobulin-receptor interaction (18). Note that the binding domain is in the N terminus of thyroglobulin. Unlike the C-terminal hormonogenic domain, which is cysteine-poor and probably highly flexible, the N-terminal domain contains a sequence repeated 10 times between position 10 and 1177 (in human thyroglobulin), in which the positions of Cys, Pro, and Gly residues are highly conserved (1,25). The presence of these Cys-rich repeats, predicted to form a rigid three-dimensional structure, probably involved in disulfide bonds (1,25,28), may confer a limited and constrained flexibility to this thyroglobulin region. Formation of some disulfide bonds, which leads to the synthesis of dimeric (19 S) or tetrameric (27 S) forms of thyroglobulin, occurs after the release of thyroglobulin into the follicular lumen (29). Formation of such intermonomer disulfide bonds (and also intramonomer bonds) would be facilitated by the H 2 O 2 -generating system involved in the oxidation of iodide and the oxidative coupling of iodotyrosyls into iodothyronines at the apical cell surface (2). Thus, disulfide bonds may form within (or in the vicinity of) the binding domain during intralumenal thyroglobulin maturation. This process may progressively yield folding units with a rigid three-dimensional structure unfavorable for receptor recognition.
Complex-type glycans containing accessible GlcNAc are found in the receptor binding domain; the N3 peptide is glycosylated at Asn 797 and Asn 928 . Although Asn 797 was not predicted as a putative glycosylation site because it is part of a hydrophobic domain (Ref. 1 and Fig. 5), it is flanked by amino acids residues favoring N-glycan grafting (Ser Ϫ2 , Gly Ϫ1 , and Leu ϩ3 (30)). The location of a glycan unit in such a hydrophobic pocket may have consequences on the ability of the glycanase to interact with and cleave the glycan. Deglycosylation of N3 resulted in 2.5-and 4-kDa shifts of the initial molecular mass, corresponding well with the removal of two glycans of 2.5 and 1.5 kDa each. Some glycans were not complete; using lectin affinity columns, we found accessible GlcNAc residues on N3. These GlcNAc residues usually found on immature thyroglobulin (15) may well be those involved in the thyroglobulinreceptor interaction.
Tyrosyl residues may be essential determinants for binding (31). A survey of the literature to identify which tyrosyl residues in the binding domain may be involved in the receptor interaction indicated the following: 1) in human thyroglobulin, Tyr 847 , Tyr 864 , and Tyr 972 are iodinated in vitro by lactoperoxidase, and 2) Tyr 864 is converted to diiodotyrosyl in vivo, whereas Tyr 847 has been considered to be an "attractive candidate for the donor of outer iodothyronyl ring" (2,10). In bovine thyroglobulin, which presents 85% identity with the human thyroglobulin binding domain (Ser 789 -Met 1172 ), the search for potential dehydroalanine residues indicated that Tyr 986 or Tyr 1008 , which correspond to Tyr 985 and Tyr 1007 , respectively, in the human sequence, are available for hormonogenesis as donor sites (11). Thus, at least five tyrosyl residues, from Ser 789 to Met 1172 , are solvent-exposed. As these residues are modified by both iodination or hormone formation, they may act as sensors in the regulation of the thyroglobulin/receptor binding activity.
A recent reinvestigation of the thyroglobulin type 1 repeat by Molina et al. (25,32) showed that it is different than other known cysteine-rich modules. It is found in 32 proteins and is very similar to a cysteine protease inhibitor (33). It is thought to control proteolytic events in some of these proteins. Type 1 modules "could function as binders and reversible inhibitors of the protease involved in the proteolytic processing of thyroglobulin" (32). Binding of thyroglobulin type 1 motifs to proteases in endosomes may thus orient and limit the proteolytic cleavage to hormone-containing domains (32). Note that, in our experimental conditions, thyroglobulin peptides containing the type 1 repeats other than the peptide N3 were unable to bind to the receptor. Therefore, although interaction of type 1 repeats with proteases is by no means excluded, our observations suggest that the binding domain involved in follicular retention requires a domain specifically involved in the prohormone function (iodination or hormone formation) rather than a type 1 repeat module.
In conclusion, we showed that the thyroglobulin domain involved in receptor recognition is localized within a stretch of 383 amino acid residues in the N-terminal region of the prohormone. This domain carries two complex-type oligosaccharide units, up to five tyrosyl residues involved in iodination or hormone synthesis, and could be disulfide-bonded so as to form bridges between internal Cys-rich repeats. These results provide insight into a domain in which coordinate modifications (glycan completion, iodination, hormone synthesis and disulfide bonding) may regulate the receptor/thyroglobulin binding activity and thus the retention of thyroglobulin within the follicular lumen.