Molecular Model, Calcium Sensitivity, and Disease Specificity of a Conformational Thyroperoxidase B-cell Epitope*

While studying the humoral mechanisms involved in thyroid autoimmunity, we located a B-cell autoepitope in the extracellular C-terminal region of human thyroperoxidase. Structural modeling showed that this region encompasses both a Sushi-like and an epidermal growth factor-like domain, the flexible arrangement of which was putatively stabilized by calcium. The recombinant peptide was found to contain the previously identified conformational thyroperoxidase autoepitope. The occurrence of a calcium-induced conformational change was confirmed using a recombinant peptide monoclonal antibody, the decrease of which in binding to calcium-saturated thyroperoxidase was reversed by a chelating agent. The disease specificity of recombinant peptide, which was more frequently recognized by Hashimoto's than by Graves' patients, adds to its potential value as a diagnostic and preventive tool in the context of B-cell autoimmunity.

While studying the humoral mechanisms involved in thyroid autoimmunity, we located a B-cell autoepitope in the extracellular C-terminal region of human thyroperoxidase. Structural modeling showed that this region encompasses both a Sushi-like and an epidermal growth factor-like domain, the flexible arrangement of which was putatively stabilized by calcium. The recombinant peptide was found to contain the previously identified conformational thyroperoxidase autoepitope. The occurrence of a calcium-induced conformational change was confirmed using a recombinant peptide monoclonal antibody, the decrease of which in binding to calcium-saturated thyroperoxidase was reversed by a chelating agent. The disease specificity of recombinant peptide, which was more frequently recognized by Hashimoto's than by Graves' patients, adds to its potential value as a diagnostic and preventive tool in the context of B-cell autoimmunity.
Thyroperoxidase (TPO) 1 plays a key role in the biosynthesis of thyroid hormones by catalyzing both the iodination of tyrosine residues and the coupling of some iodotyrosine residues in thyroglobulin to form tri-and tetraiodothyronines. It is a hemecontaining membrane enzyme which is expressed at the apical pole of thyrocytes facing the colloid space (1). TPO belongs to the mammalian peroxidase family, the members of which include myelo-, lacto-, eosinophil, and salivary peroxidases (2). Human TPO contains 933 amino acids (aa), and shows a large extracellular region consisting of 848 aa and five potential glycosylation sites, a short membrane-spanning region, and a 61-aa cytoplasmic tail. Most of the extracellular region of TPO shows a high degree of homology with myeloperoxidase (aa 1-739). The extracellular region close to the membrane anchorage domain shows homologies with the C4b complement component (aa 739 -794) and EGF (aa 794 -842) (3). The three-dimensional structure of TPO is not yet known, and elucidating this point constitutes an important task for scientists investigating the peroxidase family and for immunologists focusing on autoimmunity questions because TPO is a major thyroid autoantigen.
Autoantibodies (aAb) to TPO are the most sensitive and specific serological markers available for diagnosing autoimmune thyroid diseases. Like most antibodies to exogenous antigens, TPO aAb are produced by B lymphocytes via a T-celldependent mechanism involving specific cellular receptors and soluble cytokines. At the molecular level, however, the specificity of the autoimmune reaction relies on the recognition of TPO epitopes by T and B-cells (4). T-cell receptors recognize linear epitopes from processed TPO associated with class II molecules belonging to the major histocompatibility complex, which are co-expressed at the surface of antigen-presenting cells. By contrast, B-cell receptors, which are membrane-bound immunoglobulins, usually have a more stringent specificity and recognize conformational epitopes, i.e. they are highly dependent on the three-dimensional structure of the TPO molecule (5)(6)(7).
T-cell epitopes are rather cryptic because they are buried in the core of the molecule, whereas B-cell epitopes are surface markers. Consequently, T-cell epitopes may act at the onset of the disease when they are appropriately processed, and B-cell epitopes may provide highly specific targets for pathogenic aAb. Elucidating the molecular structures of these epitopes may help to prevent and cure autoimmune thyroid diseases. The data published so far on T-cell epitopes are unconclusive and cannot be used to draw up a specific therapeutic strategy as previous authors proposed to do (8); whereas in Hashimoto's thyroiditis, which is associated with tissue-destructive events, TPO aAb may fix the complement and mediate aAb-dependent cell-mediated cytotoxicity (9 -12). A specific B-cell epitope therapy that might block the destructive autoimmune process without affecting the immune system as a whole would stop thyroid cell death and do away with the need for conventional substitutive hormone therapy.
TPO aAb are known to be restricted to two immunodominant regions containing different but adjacent surface epitopes (13). However, neither the respective positions of these regions on the molecule nor the epitope structures targeted by the aAb of patients with autoimmune thyroid disease have yet been established. We recently located a conformational B-cell epitope at the C-terminal end of TPO near the membrane anchorage domain of the molecule (aa 742-848); this epitope was found to involve at least one of the three tyrosine residues present in this region (14). In the present manuscript, some insights on the structure of this part of the molecule were provided by * 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.
ʈ performing sequence alignment and modeling studies, which yielded information that was useful when subsequently analyzing the function of this conformational TPO B-cell epitope. After cloning the cDNA that encodes this TPO peptide by deleting the N-terminal part of the molecule showing homologies with myeloperoxidase, we expressed the recombinant peptide (r-pep) in plasmid-transfected Chinese hamster ovary (CHO) cells. The r-pep was found to contain the relevant TPO epitope previously identified in the native protein. In Western blot experiments, this epitope was recognized significantly more frequently by patients with Hashimoto's thyroiditis rather than by those with Graves' disease.

EXPERIMENTAL PROCEDURES
Similarity Search, Sequence Alignment, and Modeling-The two domains of TPO studied here (CCP domain, aa 742-795; calcium binding EGF domain, aa 796 -839) were scanned for similarity against Protein Data Bank sequence entries using the BLAST algorithm (15). The sequences obtained were aligned with the Clustal W multiple alignment algorithm (16). For the molecular modeling, the alignment of the TPO domains with their homologous sequences and the known threedimensional structures of these homologous sequences were used to calculate interatomic distances and dihedral angles. Three-dimensional structures were generated from this set of constraints with the X-PLOR Version 3.85 software program (17) using the default parameter sets, except for some minor modifications performed to increase the duration of the molecular dynamic simulations and the number of energy minimization steps. All the sequence analysis procedures were carried out at the Pôle Bio-Informatique Lyonnais, a WWW server dedicated to protein sequence analysis located in Lyon, France. Structure superimpositions, three-dimensional graphic displays and manipulations were carried out using the ANTHEPROT 2.0 software program (18).
Construction of the ⌬TPO-pcDNA3-The full-length human TPO cDNA (19) was cloned into HindIII and XbaI sites of the transfer vector pcDNA3 (20). A 710 nucleotides cDNA fragment corresponding to a peptide sequence around the 192 C-terminal aa of TPO was amplified by polymerase chain reaction from the TPO-pcDNA3 construction. Oligonucleotides, 5Ј-ACTACCGCTCGAGCGACGACAAGTGTGGCTTCC, and, 3Ј-GATTTACCCGTGTTGGG, were used as primers. The 710-base pair amplified cDNA containing the XhoI and SplI restrictions sites at the 5Ј-and 3Ј-ends of the DNA fragment, respectively, were cleaved by the corresponding enzymes. The resulting 634-nucleotide cDNA fragment (⌬TPO) was cloned into XhoI and SplI sites of the TPO-pcDNA3 construction. This new construction (⌬TPO-pcDNA3) made it possible to conserve the TPO nucleotides coding for the signal peptide. The nucleotide sequence of the ⌬TPO cDNA fragment was determined by performing sequencing to ascertain that the ⌬TPO-pcDNA3 construct was correctly engineered.
Transfection into CHO Cells-⌬TPO-pcDNA3 construct was transfected into the CHO cell line using the LipofectAMINE method (Life Technologies, Gaithersburg, MD). The CHO cells were maintained in Ham's F-12 medium supplemented with 10% fetal calf serum, penicillin (100 IU/ml), and streptomycin (0.1 mg/ml) in a humidified atmosphere under 7.5% CO 2 at 37°C. Stable transfectants were selected with 400 g/ml Geneticin TM G418 sulfate during 6 weeks. Surviving cells were cloned by limiting dilutions and then grown in vitro in the presence of 10 mM n-butyric acid to enhance the expression of TPO r-pep.
Immunodetection of TPO r-pep Expressed in CHO Cells-Stably transfected CHO cells grown in 100 ϫ 20-mm tissue culture dishes were washed three times with phosphate-buffered saline (PBS), pH 7.3. Wild-type CHO cells grown under the same conditions were used in the control experiments. Cells were solubilized by adding 800 l of electrophoresis buffer (Tris-HCl, pH 6.8, containing 30% glycerol, 1% SDS, and 0.02% G-250 Coomassie Brilliant Blue) to each dish. The cells were scraped and homogenized by vortexing and sonication. After centrifugation at 10,000 ϫ g for 10 min, the supernatant was analyzed by performing the Tricine SDS-polyacrylamide gel electrophoresis method (14). Ten l of sample per lane, two lanes per test (transfected and wild-type CHO cells), were electrophoresed either without or with denaturation (2% ␤-mercaptoethanol, heated) on a 16.5% acrylamide, 80 ϫ 100-mm minigel 0.5-mm thick and directly electrotransferred onto a 0.2-m Trans-blot polyvinylidene difluoride membrane (Bio-Rad). Western blot experiments were performed by incubating patients' sera (500 l per test) or TPO monoclonal antibody (mAb) 54 (100 g) in 5 ml of PBS, 3% bovine serum albumin (BSA) overnight at 4°C with constant shaking after previous saturation of the membrane with PBS containing 3% BSA. The membrane was then washed three times for 10 min in PBS. The second, anti-human or anti-mouse antibody labeled with horseradish peroxidase was incubated for 2 h in PBS, 3% BSA at room temperature under shaking. After several additional washes, the blots were developed with 4-chloro 1-naphthol as a substrate.
TPO Monoclonal Antibodies-mAb 47 and 54 were previously produced by immunizing mice with purified human TPO (13). mAb 47 were found to be directed against a linear epitope, aa 713-721, on TPO (21), but the epitopic location of mAb 54 is not known to date. Both TPO mAb were produced in the form of mouse ascitic fluid and then purified by DEAE ion-exchange chromatography.
Enzyme-linked Immunosorbent Assay-Wells of microtiter plates (Nunc, Roskilde, Denmark) were filled with PBS containing 300 ng of human TPO purified as described previously (22). After being incubated overnight at 4°C in a humidified atmosphere, the TPO coated wells were washed and filled with either PBS alone, PBS ϩ CaCl 2 (8 ϫ 10 Ϫ6 M), and PBS ϩ CaCl 2 (8 ϫ 10 Ϫ6 M) ϩ EDTA (1.6 ϫ 10 Ϫ4 M). After a 5-min incubation at room temperature, the wells were washed again, saturated with BSA and filled with various amounts of TPO mAb in PBS, 1% BSA for overnight incubation at 4°C. Wells filled with PBS, 1% BSA without mAb served to evaluate the nonspecific binding. Unbound mAb were then removed by extensive washing and mAb binding was detected by anti-mouse antibodies labeled with alkaline phosphatase. p-Nitrophenyl phosphate was used as the substrate. The optical density was read at 405 nm.
Sera of Patients-Sera were obtained from 24 patients with Hashimoto's thyroiditis and 18 patients with Graves' disease. Diagnosis was based on clinical and laboratory criteria. The pathological sera were positive for TPO aAb as assessed by Dynotest (BRAHMS diagnostica, Berlin, Germany). Twenty sex-and age-matched normal controls had no clinical evidence or past history of thyroid disorder; they were negative for TPO aAb.

RESULTS AND DISCUSSION
The three-dimensional structure of TPO has not yet been elucidated, whereas that of myeloperoxidase has been described (23) and was subsequently used to predict the structure of the part of TPO that showed homologous sequences (24). However, the B-cell autoepitope we previously detected was mapped at the C-terminal end of the extracellular part of TPO (aa 742-848), which is located outside the region homologous to myeloperoxidase. This region was previously reported to consist of two juxtaposed gene modules belonging to the C4b-␤2 glycoprotein and EGF/LDL-receptor gene families (25,26). In light of more recent data, the C4b receptor-like domain was found to be a complement control protein repeat (CCP) which is present in the complement receptor type 1 (27), in the Vaccinia virus complement control protein (VCP) (28), and in the human Factor H (29,30). This short consensus repeat is formed by two disulfide bonds having features characteristic of a Sushi domain, i.e. the first cysteine in the aa sequence is connected to the third one and the second cysteine to the fourth (31). The VCP and the factor H have a known three-dimensional structure that has been deposited in the Protein Data Bank (32). VCP entries 1VVD and 1VVE showed the closest homologies with the Sushi domain of TPO (aa 741 to 795). In addition, the EGF-like domain of TPO, which contains six conserved cysteine residues forming three disulfide bonds (33) also contains five aa that constitute a consensus sequence for calcium binding: D/Nx-D/N-E/Q-xm-D/N*-xn-Y/F, in which the asterisk might be a ␤-hydroxylated residue (34 -36). This domain of TPO acquired a higher score of homology from aa 794 to aa 839 with the homologous domain of human fibrillin-1, a glycoprotein found in connective tissue (37). The functional consequences of calcium binding to proteins containing this structural element are not yet known, but it was suggested that calcium may stabilize the structure of fibrillin-1 in a rod-like arrangement (38) and may protect the molecule from proteolytic degradation (39). The crystal structure of human clotting factor IX which shows a calcium-binding EGF-like domain suggests that calcium may be involved in maintaining the conformation of the N-terminal region of the domain but, more importantly, proves that calcium is able to directly mediate protein-protein interactions (40). The Sushi domain of VCP and the calcium binding EGFlike domain of fibrillin-1 require a specific internal arrangement of the numerous disulfide bonds that we used to model the membrane proximal TPO region. These two TPO domains were aligned with the 1VVD and 1VVE sequences in the case of the Sushi domain and with the 1EMO sequence in that of the EGF domain (Fig. 1). These two alignments show a reasonably high degree of similarity and in particular, all the cysteines involved in disulfide bonds were found to be conserved: C 742 -C 782 and C 768 -C 794 for the Sushi domain; C 800 -C 814 , C 808 -C 823 , and C 825 -C 838 for the EGF domain. In addition, the calciumbinding consensus sequence D/N-x-D/N-E/Q-xm-D/N*-xn-Y/F is conserved in the second alignment.
A molecular model was generated for this peptide based on the three-dimensional structure of 1VVD and 1VVE for the Sushi domain and 1EMO for the EGF-like domain. The coordinates of these proteins were used to calculate the corresponding distance constraints and dihedral angles to generate the three-dimensional model as described under "Experimental Procedures." All the structures generated satisfy these con-straints: there are no violations of the distance or dihedral constraints, no distance deviations Ͼ0.5 Å, and no dihedral deviations Ͼ5°(data not shown). The peptide contains three tyrosine residues at positions 766, 772, and 829. At least one of these tyrosines is involved in the B-cell epitope because we previously observed that the peptide binding to aAb is susceptible to iodination (14). These three tyrosines obtained very different scores with regard to their accessibility to solvent, as calculated with the X-PLOR Version 3.85 software program (17,41): 0.07Å 2 ( ϭ 0.07 Å 2 ) for Tyr 766 , 12.99 Å 2 ( ϭ 3.9 Å 2 ) for Tyr 772 , and 6.37 Å 2 ( ϭ 2 Å 2 ) for Tyr 829 . Tyrosine 829 obtained a high accessibility score, but one should remember that it is located near the transmembrane region beginning at amino acid 849 and that the membrane was not included in the model. As shown in Fig. 2, the most accessible tyrosine was located at position 772, near the linking region between the highly structured cores of the two modules. Consequently, the two modules may belong to and together form a unique B-cell autoepitope. The arrangement of the two domains may be stabilized by calcium ligation of the EGF domain and interdomain hydrophobic packing interactions, as previously found to occur in the case of the EGF domain pair (38) and the TB-EGF domain pair (42). The tyrosine 772, which seems to expose its side-chain at the hinge area, may therefore be involved in interdomain hydrophobic packing interactions. The epitopic specificity of TPO in comparison with other peroxidases such as myeloperoxidase might be because of this original construction, which would explain what purpose these gene modules serve and how they are maintained in evolution. Moreover, the location of a major TPO autoepitope outside the region showing homologies with myeloperoxidase is liable to prevent autoimmune cross-reactions from occurring with this other autoantigen which is implicated in systemic vasculitis and idiopathic necrotizing and crescentic glomerulonephritis (43).
CHO cell lines were transfected with ⌬TPO-pcDNA3 construct, selected with Geneticin TM , and cloned for homogeneity. Clones of CHO cells were grown in the presence of butyrate and screened for TPO r-pep expression in Western blot experiments using a pool of sera from patients previously found to be reactive with the native, proteolytic TPO peptide (14). Several positive clones were obtained, and one of the most productive was selected for further use. Fig. 3A shows that transfected but not wild-type CHO cells contained a peptide of the expected molecular weight (about 20 kDa) which was recognized by aAb present in the pool of sera from the patients. As expected, the targeted B-cell epitope was conformational, i.e. it consisted of juxtaposed aa occurring at intervals in the sequence but brought closely together by the three-dimensional folding of the molecule because the immunoreactivity was lost after the reduction of the disulfide bridges with ␤-mercaptoethanol (Fig.  3B). The antigenic specificity was proved by pre-absorption experiments showing that relevant TPO (Fig. 3C) but not irrelevant thyroglobulin (Fig. 3D) inhibited the binding of TPO aAb to r-pep.
From the 66 murine TPO mAb we previously produced (13), only one, mAb 54, was found to react with the r-pep in Western blot experiments (Fig. 4A). The mAb 54 binding to the r-pep was lost (i) after previous treatment of the r-pep with a reducing agent, confirming the recognition of a disulfide bridge-dependent epitope (Fig. 4B), and ii) after previous absorption of the r-pep with the pool of sera from the patients, indicating that mAb 54 and TPO aAb are directed to the same TPO domain (Fig. 4C). To determine the effects of calcium on the B-cell epitope conformation of r-pep, we tested the binding behavior of mAb 54 to native, purified TPO saturated with calcium. The presence of calcium on TPO was found to greatly decrease its ability to bind to mAb 54. The calcium-induced inhibition was reversed by chelating calcium with EDTA before adding it to TPO (Fig. 5A). The TPO binding of mAb 47, which recognized a linear TPO peptide (aa 713-721) not present in  the r-pep, was not affected by calcium treatment of TPO (Fig.  5B).
Several sera were selected to study the clinical significance of the TPO aAb directed against the epitope present in the r-pep. They came from patients with Graves' disease or Hashimoto's thyroiditis, which are two well characterized autoimmune thyroid diseases with positive TPO aAb. To date, TPO aAb assay does not discriminate between these two diseases. Only a rise in the normal background of TPO aAb is taken to be a sign of thyroid autoimmune attack. The serological diagnosis of Graves' disease may be ascertained by the presence of aAb against the thyrotropin receptor. Diagnosis of Hashimoto's thyroiditis is based on histological examination of diffuse lymphocyte and plasma cell infiltration and fibrosis of thyroid gland rather than aAb determination. However, positive thyroid aAb can alert physicians to the patient's risk of developing thyroid dysfunction associated with Hashimoto's thyroiditis as well as Graves' disease. As shown in Table I, 79.2% of the patients in the Hashimoto's group were positive for TPO r-pep, as compared with only 27.8% of patients with Graves' disease. The latter cases may have reached a transient state in the course of thyroid disease before evolving to overt Hashimoto's thyroiditis. Surprisingly, 6.7% of the normal subjects tested were positive and were therefore candidates for developing thyroiditis. These results are of special interest because this is the first time a single TPO aAb has been found to clearly distinguish between the two main autoimmune thyroid disorders. It is worth noting that the presence of aAb against TPO r-pep was not found to be correlated with the TPO titer of the sera in the two groups of patients (Fig. 6). This is the first direct evidence showing that TPO aAb are directed to at least two different epitopes, in agreement with what was previously suggested by the results of competitive binding studies on TPO between mAb and aAb (13).
In the present study, we propose a structural model for the proximal membrane region of TPO, which provides new insights into the autoimmune role of gene modules assembled and conserved during evolution in large proteins. One particularly noteworthy finding was that the B-cell epitope is coded by two gene modules which are closely juxtaposed after the MPO-like coding sequence and before the membrane spanning region. In the model, this epitope is located near the tyrosine 772 and the calcium binding site. This model provides information which will be of use in further studies, which may lead to the development of a specific therapeutic strategy for preventing the disease from occurring in subjects liable to undergo an autoimmune attack. A recurrent question in the field of autoimmunity relates to the heterogeneity of aAb of the patient, which are often restricted to some epitopic structures. Elucidating the immunodominant B-cell epitopes may help to understand the mechanisms responsible for loss of self-tolerance in organ-specific autoimmune diseases. The proximal membrane region of TPO produced as a peptide in eucaryotic system was used here to characterize a specific B-cell epitope recognized by aAb from Hashimoto's patients. The corresponding TPO region may be critically involved in the development of the disease. Its diagnostic value now requires confirmation by performing large-scale clinical tests.