Identification of a Novel Epitope in the Thyroid-stimulating Hormone Receptor Ectodomain Acting as Intramolecular Signaling Interface*[boxs]

Glycoprotein hormone receptors (GPHRs) differ from the other seven transmembrane receptors mainly through a complex activation mechanism that requires the binding of a large hormone toward a large N-terminal ectodomain. The intramolecular mechanism of the signal transduction to the serpentine domain upon hormone binding at the ectodomain is not understood. To identify determinants at the GPHR ectodomain that may be involved in signal transduction, we first searched for homologous structural features. Based on high sequence similarity to the determined structures of the Nogo-receptor ectodomain and the intermolecular complex of the Interleukin-8 ligand (IL8) and the N-terminal peptide of the IL8 receptor (IL8RA), the hypothesis was developed that portions of the intramolecular components, Cysteine-box-2 and Cysteine-box-3, of the GPHR ectodomain interact and localize at the interface between ectodomain and serpentine domain. Indeed, point mutations within the D403EFN406 motif at Cysteine-box-3 of the thyrotropin receptor resulted in increased basal cAMP levels, suggesting that this motif may be important for transduction of the signal from the ectodomain to the transmembrane domain. New indications are provided about the tight spatial cooperation and relative location of the new epitope and other determinants at the thyrotropin receptor ectodomain, such as the leucine-rich repeat motif Ser281 and the cysteine boxes. According to the high sequence conservation, the results are of general relevance for the signal transduction mechanism of other glycoprotein hormone receptors such as choriogonadotrophic/luteinizing hormone receptor and follicle-stimulating hormone receptor.

The glycoprotein hormone receptors (GPHRs) 1 CG/LHR, FSHR, TSHR, and the leucine-rich repeat-containing glycopro-tein receptors constitute a subfamily of family A of G-proteincoupled receptors (1). Understanding the specific molecular mechanisms, starting with binding of the heterodimeric hormone to the large ectodomain, and knowledge about the changes in interactions between the ectodomain and transmembrane domain are prerequisites for understanding the specific activation mechanism of GPHRs. Furthermore, it is instrumental in generating new ideas for pharmacological intervention. An x-ray structure of the large ectodomain is not yet available. Based on sequence similarities, the large Nterminal portion of the TSHR seems to consist of five structural components: (i) an N-terminal C-b1 (hTSHR: 1-54), (ii) a 230amino acid-spanning LRR motif (hTSHR: 55-279) with nine repeats back-to-back followed by (iii) the central C-b2 (hTSHR: 280 -316), (iv) the TSHR-specific insertion region (hTSHR: 317-366), and (v) the C-b3 (hTSHR: 370 -410) located close to TMH1 (2). For GPHRs it has been demonstrated that the major binding region for the large hormones hCG/hLH, hTSH, and hFSH is located in the extracellular leucine-rich repeat region (3). The crystal structures of 14 proteins containing LRR motifs show that LRRs fold into a curved shape with a parallel ␤-sheet on the concave face and with various secondary structures, including ␣-helix, 3-10 helix, and pII helix, on the convex face (4). Exposed side chains on the concave face are supposed to contact the ligands and provide selective interaction patterns for hormone receptor binding (5). The structures of hCG (PDB accession number 1QFW (6)) and hFSH (PDB accession number 1FL7 (7)) have been determined. Data of peptide studies and mutagenesis provide established knowledge about direct interaction of the hormone ␤-subunits and the LRRs (8,9), whereas the hormone ␣-subunit is supposed to be oriented toward the serpentine domain and to be involved in ligandinduced signaling (10,11). However, the LRR sequence is only one portion in the middle of the complete N-terminal sequence, and there are additional structural components between the LRRs and the TMH1 of the serpentine domain whose threedimensional structure is unknown. Moreover, it is not completely understood how the intramolecular signal transduction toward the serpentine domain of the GPHR takes place.
The central C-b2 is located back-to-back following the LRR region. Mutants at Ser 281 , directly preceding the first cysteine in C-b2, are known to act as a switch for constitutive activity (12). The structural interrelation between the constitutive active mutant S281N and the necessary presence of the complete LRR motif has been shown by deletion studies for the hCG/ LHR (13). N-terminal truncations of the TSHR ectodomain (14) and mutational studies at corresponding positions of the leucine-rich repeat-containing glycoprotein receptor 2, the CG/ LHR, and the FSHR demonstrated the importance of the Ser 281 region for the maintenance of a conformational switch in all four glycoprotein hormone receptors (15,16). Recent studies suggest that amino acids Pro 276 and Ser 277 in CG/LHR (TSHR: Pro 280 , Ser 281 ) are constituents of a loop-like epitope that can act as an activation switch (15). This is confirmed by substitution of a proline residue, Pro 276 , (TSHR: Pro 280 ) adjacent to Ser 277 (TSHR: Ser 281 ) of the CG/LH receptor with a glycine that exerts less structural restraint and leads to constitutive receptor activation and higher agonist affinity. Mutations of the TSHR cysteines Cys 283 and Cys 284 to serine led to a 2-fold increase of basal TSHR activity (17).
Disulfide bonds are involved in the quaternary structure of the TSHR (18). Although there are currently no direct data to indicate which of the single cysteines form pairs, a number of indirect lines of evidence permit global assignments of disulfide bridges (19,20). Mutations of Cys 283 , Cys 284 (C-b2) and Cys 398 , Cys 408 (C-b3) have dramatic effects on the TSHR structure and TSH binding. The hypothesis that Cys 301 (C-b2) and Cys 390 (C-b3) are paired is supported by recent evidence that intramolecular cleavage of the TSHR releases a "C peptide" located between these two residues (21). Further evidence is provided by the fact that mutation of either Cys 301 or Cys 390 produces the identical effect of reduced hTSH binding affinity (22). Moreover, the CG/LH receptor of New World monkey Callithrix jacchus (23) is missing exon 10 encoding the particular cysteine corresponding to Cys 301 of TSHR. Additionally, this receptor is simultaneously also missing the cysteine corresponding to Cys 390 TSHR. Taken together, the two neighboring cysteines Cys 283 and Cys 284 of C-b2 are very likely paired either to Cys 398 and Cys 408 of C-b3 or in reversed order to Cys 408 and Cys 398 , whereas the Cys 301 (C-b2) is very likely paired with Cys 390 (C-b3). For the large extracellular N-terminal tail of GPHRs, only molecular models for the LRR have been suggested (24,25), based on the first available LRR crystal structure of the ribonuclease inhibitor (26).
To provide support for the molecular understanding of the signal transduction process upon hormone binding at the glycoprotein hormone receptor ectodomain, we decided to determine how the LRR region cooperates with other fragments of the ectodomain, like C-b2 and C-b3. To identify determinants in the ectodomain that may be involved in the intramolecular signal transduction and to obtain indications for their spatial cooperation, we searched for an optimal structural LRR template and for additional homologous structural motifs in the ectodomain. Sequence similarities between the central C-b2 and C-b3 with the IL8/CXCL8 chemokine and IL8RA/CXCR1 in a structure complex (PDB accession number 1ILQ (27)) of an IL8-IL8RA fragment, respectively, led us to propose a tightly packed model of the ectodomain.
In this ectodomain model, C-b3 is located at a prominent interface position between ectodomain and serpentine domain, close to the transmembrane domain, and is therefore very likely a necessary transducer of the intramolecular signal transduction. Based on the proposed model, Asp 403 , Glu 404 , Asn 406 in C-b3 were targeted for mutagenesis and shown to play a role in the signal transduction mechanism. They resulted in the identification of new residues within C-b3 (Asp 403 , Glu 404 , Asn 406 ) that are significantly involved in the intramo-lecular signal transduction. These residues are located in close spatial proximity to Ser 281 . In this study we have demonstrated that the combination of comparative modeling and site-directed mutagenesis at the ectodomain of the TSHR allowed the localization of a new site within C-b3 that is involved in intramolecular signal transduction toward the serpentine domain by acting as a conformational switch for constitutive TSHR activity.
To reveal new homologous structural templates in the ectodomain of the TSHR, in addition to the LRR motif, extensive systematic sequence similarity searches of fragmented sequence portions of different lengths of the ectodomain were conducted in the Protein Data Bank by means of FASTA. Because of the lack of structural templates for the TSHRspecific region containing two cleavage sites, a polyglycine loop was used to link the structural components of C-b2 and C-b3. One putative disulfide bridge between Cys 283 (C-b2) and Cys 408 (C-b3) for the TSHR resulting from the structural template IL8-IL8RA fragment complex (PDB entry 1ILQ (27)) was used as a model constraint. For the TSHR ectodomain model, the following disulfide bridges were formed: Cys 24 / Cys 31 , Cys 29 /Cys 41 , Cys 283 or Cys 284 /Cys 408 , Cys 284 or Cys 283 / Cys 398 , Cys 301 /Cys 390 . Structures of hCG (PDB entry 1QFW (6)) and hFSH (PDB entry 1FL7 (7)) are available. The TSH was modeled by homology using the hFSH structure as template. For the hTSH model, disulfide bridges between cysteines were incorporated based on the template: ␣-subunit, Cys 31 /Cys 55 , Cys 34 /Cys 84 , Cys 52 /Cys 106 , Cys 56 /Cys 108 , Cys 83 / Cys 111 ; ␤-subunit, Cys 22 /Cys 72 , Cys 36 /Cys 87 , Cys 39 /Cys 125 , Cys 47 /Cys 103 , Cys 51 /Cys 105 , and Cys 108 /Cys 115 . Conjugate gradient minimizations were performed until it converged at a termination gradient of 0.05 kcal/(mol*Å). The AMBER 7.0 force field was used. The geometric quality of the models was controlled by the PROCHECK (31) software.
Modeling of Hormone-Receptor Ectodomain Interactions-Components of homologous models were assembled manually or by constrained molecular dynamic simulations using functional data and complementary side chain properties with the biopolymer module or automatically using the docking module FlexX of the Sybyl program package. The electrostatic potentials (calculated by a probe radius of 2.0 Å) were created at the Connolly surface of all three GPHR LRRs and corresponding hormones by the MolCad module of the Sybyl program package. The hormones hTSH, hCG, and hFSH were initially placed close to the corresponding GPHR LRRs according to their complementary electrostatic potentials. The hormones were docked during a constrained molecular dynamics simulation. Putative interaction pairs TSHR Lys 209 -Asp 111 TSH, CG/LHR Glu 206 -Arg 115 CG, FHSR Lys 179 -Asp 108 FSH were used as constraints according to experimental data (5,32), allowing complementary side chain interaction (repulsion and attraction) for the remaining portions. The models were scored according to their consistency to available experimental binding data. All assembling calculations were performed in a vacuum.
The assembled ectodomain-ligand models of the three GPHRs were soaked with water in a periodic boundary box. Initially the ectodomain atoms were kept fixed to relax the water during minimization. Later on, the entire system was studied. Minimizations were performed as described above. Molecular dynamics simulations were performed at 300 K for 1 ns. For both, the AMBER 7.0 force field was used; the geometric quality of the model was controlled by the PROCHECK software (31).
Site-directed Mutagenesis-Mutations were introduced into the human TSHR via a QuikChange site-directed mutagenesis kit (Stratagene). TSHR-pSVL (33) was used as template. PCR products containing the mutations were digested with BspTI and Eco91I (MBI Fermentas) and used to replace the analogous BspTI/Eco91I fragment in the wild type TSHR-pSVL vector. Sequences of mutated TSH receptors were verified by dideoxy sequencing with Big Dye Terminator Cycle Sequencing chemistry (ABI Advanced Biotechnolgies, Inc., Columbia, MD). Sequencing reactions were analyzed on a genetic analyzer ABI 310 (ABI Advanced Biotechnologies).
The mutated receptors were cloned in the expression vector pSVL and transiently expressed in COS-7 cells. Characterization of the constructs was performed by determination of cAMP and phospholipase C-inositol phosphate accumulation, TSH binding, and cell surface expression. The wt receptor and empty pSVL vector were used as controls. Levels for cell surface expression were determined by flow cytometry.
Cell Culture and Transfection-COS-7 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomycin (Invitrogen) at 37°C in a humidified 5% CO 2 incubator. Cells were transiently transfected in 24-well plates (0.5 ϫ 10 5 cells/well) with 0.5 g of DNA/well for cAMP accumulation and TSH binding analysis. For determination of inositol phosphate (IP) formation, cells were transiently transfected in 12-well plates (1 ϫ 10 5 cells/well) with 1 g of DNA/well. COS-7 cells were transfected using FuGENE TM 6 reagent (Roche Applied Science).
FACS Analysis-48 h after transfection, nonpermeabilized cells were detached from the dishes using 1 mM EDTA and 1 mM EGTA in PBS and transferred in Falcon 2054 tubes. Before incubation with the primary antibody, cells were washed once with PBS containing 0.1% bovine serum albumin and 0.1% NaN 3 . For permeabilized cell assay, in the first step cells were fixed with 1% paraformaldehyde for 10 min on ice following an incubation with PBS containing 0.1% bovine serum albumin, 0.1% NaN 3 , and 0.2% saponin for 30 min. Saponin was supplemented in all subsequent buffers. Subsequently, cells were incubated for 1 h with a mouse anti-human TSHR antibody (10 g/ml, 2C11; Serotec, Oxford, UK). Cells were washed twice and incubated for 1 h in the dark with a fluorescein-conjugated F(abЈ) 2 rabbit antimouse IgG (Serotec). Before FACS analysis (FACscan; BD Biosciences), cells were washed twice and fixed with 1% paraformaldehyde. Receptor expression was determined by fluorescence intensity; the percentage of signal positive cells corresponded to transfection efficiency.
Radioligand Binding Assay-Competitive binding studies were performed as previously described (34). Data were analyzed assuming a one-site binding model using the fitting module of SigmaPlot 2.0 for Windows (35).
cAMP Accumulation Assay-Measurement of cyclic AMP (cAMP) accumulation was performed 48 h after transfection as previously described (34).
Stimulation of IP Formation-Transfected COS-7 cells were incubated with 2 Ci/ml myo[ 3 H]inositol (Amersham Biosciences) for 8 h. Thereafter, cells were preincubated with serum-free Dulbecco's modified Eagle's medium containing 10 mM LiCl 2 for 30 min. Stimulation with bTSH for 1 h was performed with the same medium supplemented with 100 milliunits/ml bTSH. Evaluation of basal and TSH-induced increases in intracellular IP levels was performed by anion exchange chromatography as described (36). IP values are expressed as the percentage of radioactivity incorporated from [H]inositol phosphates (IP1-3) over the sum of radioactivity incorporated in IPs and phosphatidylinositols.
Specific Constitutive Activity (SCA)-The assays for determination of B max values and basal cAMP accumulation were performed as follows: COS-7 cells were transiently transfected in 24-well plates (0.5 ϫ 10 5 cells/well) with various concentrations of wt and mutant DNA (50, 100, 200, 300, 400, and 500 ng/well). For radioligand binding assays, cells were incubated in the presence of 180,000 -200,000 cpm of 125 I bovine thyrotropin (bTSH) supplemented with 5 milliunits/ml nonlabeled bTSH. The effect of expression level on basal cAMP accumulation was analyzed according to Ballesteros et al. (37) using the fitting module of Graph Pad Prism 2.01 for Windows.
Confocal Laser Scanning Microscopy-HEK 293 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomycin (Invitrogen) at 37°C in a humidified 5% CO 2 incubator. Cells were seeded on coverslips into 6-well plates (2.5 ϫ 10 5 HEK 293 cells/well). The cells were incubated 36 h before transfection with plasmid constructs (2.5 g of DNA/well) containing the coding sequence of the wt or mutated TSHRs. 48 h after transfection, coverslips were rinsed two times with ice-cold PBS and fixed with 2% paraformaldehyde containing 0.1% Triton X-100 (for permeabilization) for 30 min at 4°C. After two 5-min wash steps with cold PBS, the cells were incubated with the primary antibody for 1 h at 4°C. TSHR was detected using the anti-human TSHR antibody (2C11; Serotec; 1:500 in PBS). The cells were washed two times for 5 min with cold PBS, and the primary antibody was detected by incubation with an Alexa-Fluor® 488-conjugated goat antimouse secondary antibody (Molecular Probes, Eugene, OR; 1:1000 in PBS) for 1 h at 4°C. After four final 5-min wash steps, the coverslips were mounted on glass slides. Confocal analysis was performed on a confocal laser scanning system (TCS SP2; Leica) attached to a microscope (DM IRBE; Leica) with a ϫ100 oil immersion lens (PL Fluotar 1.3; Leica). Sections (0.45 m) were taken, and representative sections corresponding to the middle section of the cells are presented in Fig. 3B. After indirect immunofluorescence staining, no specific fluorescence was observed in untransfected HEK 293 cells or in transfected HEK 293 cells treated only with secondary Alexa-Fluor® 488-conjugated antibody.

RESULTS
Hormone-bound LRR Complex-Studies were carried out to examine the optimal structural template for the LRR domain of the GPHR. The LRR motif from the Nogo-receptor ectodomain (PDB entry 1OZN) (29) was identified from 14 different LRR structures (4) as the best matching template to the LRR features of all three human GPHRs (Table I, hNogo-receptor: hCG/LHR, hFSHR, hTSHR).
Not only did the nine typical repeats match best to the sequence of the three receptors but also the cysteine box of the Nogo-receptor structure, which is N-terminal directly attached to the LRR, showed a best fit to C-b1 of the GPHR sequences. The putative disulfide bridges between the cysteines Cys 24 , Cys 29 , Cys 31 , and Cys 41 in the homologous TSHR model (Fig. 1,  a and b) stabilize an anti-parallel and parallel ␤-strand as integral parts of the LRR, where the latter participates as an additional parallel strand (LRR0) to the convex binding face.
According to the LRR pattern rule, an additional ␤-strand (named ␤-strand X) is adopted at the C-terminal side for all three GPHRs. Based on this match, nine complete LRRs (LRRI-IX) plus one additional ␤-strand (LRR0) at the N terminus plus one at the C-terminal side (LRRX) form 9 ϩ 2 ␤-strands lining the inner convex surface of the hormone-binding region (Fig.  1a). The TSHR LRR sequence not only shows fewer sequence similarities (PAM250 matrix) to the previous RI template (22%) and other LRR templates compared with the new Nogoreceptor template (34%) but the new LRR model for the GPHRs also offers, as the only one, a much larger radius for the inner convex arch of the hormone-binding region. This resulting radius is a shallow deflection like a "scythe blade" shape rather than a "horseshoe" shape, as demonstrated by the superimposition of TSHR models using both templates (Fig. 1b). The LRR structure of the Nogo-receptor ectodomain contains inside the LRR a "Phe spine" (green aromatic rings) for the stabilization of the fold, instead of the missing helices at the concave outer face ( Fig. 1a and Table I). The importance of phenylalanines from such a Phe spine for the overall fold of LRRs, as well as for the GPHRs, is demonstrated by a homozygous mutation at the CG/LHR, where a F194V mutant causes male pseudohermaphroditism by lost trafficking to the membrane (38).
Examination of complementary shapes and properties of electrostatic potentials for LRRs in our docking studies (see supplementary material) and functional data from mutations of all three GPHR subtypes, TSHR, CG/LHR, and FSHR and their hormones (Table II), resulted finally in a refined complex model for LRR-hormone with parallel orientation to each other (Fig. 1c). This is only possible by the widened radius of the scythe blade-shaped LRR. The ␤-subunit of the hormone is thereby oriented with a large face that also includes amino acids of the "determinant loop" (␤-TSH Asn 109 -Asp 114 (25,32,39)) toward the ␤-strands of the LRR (Fig. 1c, TSHR). Residues for hTSH, hCG/hLH, and hFSH and their counterparts at the receptors that mainly take part in complementary intermolecular hormone-LRR recognition in our models are given in Table  II. The models are consistent with previous mutation studies of the hormone (32, 40) and the LRRs of the GPHR (5, 41-43). The sequence from the hNogo-receptor ectodomain (PDB entry 1OZN) best matched the LRR of the GPHR according to sequence similarity, sequence length/repeat, and number of repeats out of 14 LRR structures. An additional N-terminal repeat (named LRR0) including Cys 41 provided by the Cysteine-box-1 (background grey) is a new structural feature. An additional ␤-strand occurs at the C-terminal end (LRRX). The back-to-back following sequence of Cysteine-box-2 (background grey) and the template sequences for malonyl coenzyme and chemokine sequences (Eotaxin2, IL8) homologies are given. Sequence names of crystal structures are underlined. Beside the number of each repeat (repeats 0 -X) the patterns of LRR sequence similarity positions are given by x 1 x 2 Lx 3 Lx 4 x 5 nxxLxaFxx. Highly conserved residues like leucine are marked by upper case letters. Conserved phenylalanine residues forming a Phe spine are marked by bold F. Lower case "n" and "a" sequences contain only less conserved N and A. Additional residues in the loop region occurring in the template and GPHR sequences are marked as x. The ␤-strands are indicated by ␤. The potential binding site comprises residue positions x 1 -x 5 .
This orientation is consistent with several functional data of the hormones previously reported as important features for signaling to the ecto-and serpentine domains, like locations of glycosylation sites (44) and an accessible C-terminus of the ␣-subunit (10,11). Moreover, residues that are highly conserved among hormones and among GPHRs interact in our model. Identical complementary charged residue pairs fitting between the additional repeat LRR0 and hormone are observed for hTSHR-hTSH and hFSHR-hFSH interaction. Furthermore, charged residues of the hormone's determinant loop spatially fit to the LRR position with complementary charges (Table II). Finally, the electrostatic potentials that led to this parallel orientation between hormone and LRR also provide an explanation for the promiscuity of hormone binding caused by some point mutations in LRRs (5, 43) (see also Supplemental Figs. 5-7).
FIG. 1. New LRR structural template allows parallel hormone interaction and complies best with available functional data. a, LRR structure of the Nogo-receptor ectodomain (PDB entry 1OZN). A Phe spine (green aromatic rings) inside the LRR stabilizes the fold instead of the missing helices at the concave outer face. The new template provides the N-terminal flanking C-b1 as an integral structural part of the LRR by contributing with an additional parallel ␤-stand (LRR0 in red) to the convex ␤-sheets. b, comparison of the previous model of TSHR (PDB entry 1XUM, gray) based on ribonuclease inhibitor template (PDB entry 2BNH) and the newly generated homologous LRR structure for the TSHR (colored) based on the x-ray structure of the hNogo-receptor ectodomain. The new LRR model offers a much larger radius and two additional parallel ␤-strands (red) for the inner convex arch of the hormone-binding region compared with the previous model. In our new LRR model TSHR amino acids Thr 273 -Ser 278 arrange the inner region of the additional 11th ␤-strand (LRRX), which is consistent with previously observed functional data. Strong structural importance of these residues in this 11th ␤-strand for expression has also been shown for the mutants T272A, L273A, T274A, Y275A (TSHR: Asp 276 , Leu 277 , Ser 278 , Tyr 279 ) in the CG/LH receptor (50). Chimeras between TSHR and CG/LHR, comprising this region (52), show that substitutions of TSHR residues L 270 HLTRADLS 278 with corresponding amino acids of the CG/LHR resulted in an impaired TSH binding and abolished cAMP response. c, docking model of hTSH (orange, ␣-subunit; violet, ␤-subunit) at the LRR motif (green) of TSHR, which complies best with current available functional data and with complementary shape and electrostatic properties (Table II and "Results"). Shown are residues from ligand and receptor potentially important for the ligand binding process. The three Asn glycosylation sites on hTSH are marked (cyan). Amino acid Asp 111 of the "determinant loop" of hTSH ␤-subunit oriented toward the ␤-strands of the LRR.

Cysteine-box-2 and Cysteine-box-3 Interact and Are Located
Close to the Transmembrane Domain-Next we asked whether additional structural templates exist in the N-terminal ectodomain of the TSHR. Systematic searches with fragmented portions of the ectodomain in the Protein Data Bank for sequence similarities identified a new homologous structural template for C-b2 and C-b3 of the TSHR ectodomain, based on the structure of the complex between chemokine IL8/CXCL8 and chemokine receptor fragment IL8RA/CXCR1 (PDB entry 1ILQ) solved by NMR solution methods (27). Chemokine ligand structures like MIP-II (PDB entry 1HFG), Rantes (PDB entry 1RT0), and Eotaxin-3 (PDB entry 1G2S) show a common extended-like conformation for sequences directly following the two characteristic cysteines. Those are involved in chemokine-ligand-receptor interaction. A sequence similarity to the central C-b2 of the TSHR ectodomain was identified exactly for this region with the consensus sequence pattern C 283 CAFXXQKKI 292 (Fig. 2a).
Moreover, also for the bound peptide, representing a portion of the IL8RA N-terminal tail, a high sequence similarity to the TSH receptor sequence of C-b3 was identified with the consensus sequence pattern P 400 XXDEFNPC 408 (Fig. 2, b and  c). This structural template applied to the TSHR ectodomain model by substituting amino acid side chains from C-b2 (C 283 CAFKNQKKI 292 ) and C-b3 (P400KSDEFNPC 408 ) resulted in homologous model residues with complementary properties matching as interacting side chains like the aromatic interaction of Phe 286 in C-b2 and Phe 405 in C-b3 (Fig. 2e).
Subsequently using this template, Cys 408 of C-b3 is ideally placed close to Cys 283 (or Cys 284 ) of C-b2 to form a disulfide bridge (Fig. 2e). Our findings support the hypothesis of disulfide bridges between cysteines Cys 283 / Cys 284 of C-b2 either to Cys 398 / Cys 408 or to the reverse order, Cys 408 / Cys 398 of C-b3 (21).
The TSHR sequence S 278 YPSHC 283 directly following the last ␤-strand of LRRX and preceding Cys 283 of C-b2 (including Ser 281 ) matched best with sequences of structural fragments containing a turn/loop conformation, such as SVPSHC from the malonyl coenzyme (PDB entry 1MLA) and VLPSHC from a stem cell factor (PDB entry 1EXZ) (Fig. 2, a and b, and Table I). These findings confirm previous suggestions that Ser 281 is an integral part of a turn or loop conformation (15). Adapted to our model, such a turn links the LRR and C-b2 and places the C-terminal portion of the LRR domain directly above the linked C-b2 and C-b3. The function of this turn as a potential spatial pivot or hinge between LRR and C-b2/C-b3 is consistent with the data for mutational activation at Ser 281 (12). The assembled model provides support for a very compact structural arrangement of the C-b1, LRR, Ser 281 turn/loop, C-b2, and C-b3 proximal to the extracellular end of TMH1 (Fig. 2d). Subsequently the residues D 403 EFNPC 408 of the C-b3 are located particularly at prominent interface positions of the ectodomain (Fig. 2e), closest to the transmembrane domain. The hydrophilic residues Asp 403 , Glu 404 , and Asn 406 were therefore hypothesized to very likely participate in the intramolecular signal transduction from the ectodomain toward the serpentine domain.
New Switch for Constitutive Activity Induced by Mutations Identified in the Ectodomain-To provide support for this hypothesis, we experimentally addressed the three amino acids, Asp 403 , Glu 404 , and Asn 406 , in the ectodomain by site-directed mutagenesis. To investigate the functional properties of these residues in the intramolecular signal transduction, the positions of the acidic residues Asp 403 and Glu 404 were replaced by a non-polar and a basic amino acid, alanine and lysine, respectively (Table III). For Asn 406 , a substitution to alanine was characterized.
The cell surface expression of the transfected TSHR constructs compared with the wt TSHR were as follows: wt TSHR set at 100%; D403A, 27.2 Ϯ 0.9%; D403K, 12.3 Ϯ 0.8%; E404A, 95.4 Ϯ 4.6%; E404K, 35.3 Ϯ 1.0%; N406A, 66.4 Ϯ 3. 6% (Fig. 3A and Table III). After permeabilization of the cells, receptors with a strong impaired cell surface expression (D403A, D403K, E404K) showed an increased FACS signal up to 60% compared with the wt TSHR, suggesting an increased accumulation within the cells (Fig. 3B). To confirm our assumption of an intracellular receptor accumulation, we performed confocal laser scanning microscopy in HEK 293 cells. In brief, for mutants D403A, D403K, and E404K a strong intracellular receptor accumulation after permeabilization was observed. In contrast, E404A showed a cell surface expression comparable with the wt TSHR before and after permeabilization. This is in accordance with the results obtained by FACS analysis. However, for mutant N406A with a cell surface expression of 66% (wt TSHR set at 100%) the difference between non-permeabilized and permeabilized cells was not consistent. This is most likely because the reduced cell surface expression for N406A was only very slight compared with the other mutants.
Three new mutations (D403A, E404K, and N406A) in the ectodomain of the TSHR, causing constitutive cAMP activity,  were identified. The mutant with the highest constitutive activity is N406A with a 5-to 6-fold over basal cAMP production (Fig. 3D). Moreover, after stimulation with 100 milliunits/ml of bTSH, a maximal cAMP response of ϳ90% of the wt receptor could be detected (Fig. 3E). Substitution of the TSHR-specific position Glu 404 by lysine resulted in a 4-to 5-fold increase of basal cAMP production. In contrast, the replacement of Glu 404 by alanine had no significant effect on cAMP production. Mutant D403A showed a constitutive activation of the cAMP-signaling cascade (2-fold over wt TSHR basal). The hTSH-induced cAMP response was half of the wild type. A strongly reduced cAMP production was detected by the D403K mutant.
To investigate the effects of these mutations on the basal receptor signal transduction more precisely, we calculated the specific constitutive activity. The calculation was based on a linear increase of basal cAMP accumulation of higher densities of receptors at the cell surface. An increase of basal activity was determined as follows: wt TSHR (slope ϭ 0.0024 Ϯ 0.0002) ϽN406A (slope ϭ 0.031 Ϯ 0.003) ϽD403A (slope ϭ 0.032 Ϯ 0.004) ϽE404K (slope ϭ 0.042 Ϯ 0.004) (Fig. 3H). Stimulated with higher concentrations of hTSH, the activated TSHR also FIG. 2. Additional homologous structural templates identified by sequence similarity assembled to a TSHR ectodomain model are consistent with previous functional data. a, partial sequence alignments for portions with sequence similarity of different chemokines (determined structures) and C-b2 region of the hTSHR upstream (red) of double cysteines Cys 283 Cys 284 (blue). Downstream of the double cysteines, high sequence similarity (cyan) to fragments with turn conformation from stem cell factor (PDB entry 1EXZ) and from malonyl coenzyme (PDB entry 1MLA) was identified. Residues with high similarities are marked in bold. Sequence alignment is shown in reversed order to be consistent with the anti-parallel orientation in the structure in panels b and c. Arrows indicate the corresponding cysteine in the crystal structure. Asterisk marks the templates used in panel b. b, conformational portions based on high sequence similarity adopted to the TSHR receptor are highlighted at the structure complex of IL8 monomer (gray) and its bound peptide (yellow) of the N-terminal tail of the IL8RA receptor fragment (PDB entry 1ILQ). Red, region of sequence similarity to C-b2 region C 284 aFxxQKKI 292 of TSHR. Blue, IL8 cysteines homologous to the TSHR. Yellow, IL8RA fragment of the N-terminal extracellular tail with sequence similarity to the TSHR P 400 XXDEFNPC 408 cysteine (TSHR Cys 408 ) placed in ideal distance to cysteines of the other chain coming from C-b2 (TSHR Cys 283 , Cys 284 ) to form a disulfide bridge. Cyan, turn/loop conformation for a fragment of malonyl coenzyme as an additional template fragment for the TSHR consecutive residues P 280 SxC 283 . c, partial sequence alignments of the IL8RA N-terminal peptide and the TSHR Cysteine-box-3 portion; homologous residues are marked in bold. Arrow indicates the corresponding position in the x-ray structure. d, homology model of the TSHR ectodomain demonstrates a tightly packed structural arrangement of Cysteine-box-1 (orange), LRR (green), Ser 281 turn (cyan), Cysteine-box-2 (red), and Cysteine-box-3 (yellow). The structural components are adopted from the homologous crystal structures of the Nogo-receptor ectodomain and the complex of IL8 and the N-terminal peptide of the IL8RA. For clarity the remaining portions of C-b2 and C-b3 are not visualized. The suggested homologous Ser 281 turn/loop (cyan) adopted from malonyl coenzyme links the LRR and the C-b2 component, which very likely functions as a pivot of the activation mechanism. Irrespective of the question whether the homologous turn (cyan) configuration and LRR are assembled in the "right" orientation (indicated by interrupted backbone and arrows), a turn/loop structure very likely organizes the mutual spatial orientation between the LRR structure and the C-b2/C-b3 component(s). The attached LRR/C-b2 components are linked to the serpentine domain via the disulfide bond between C-b2 and C-b3. e, detail of C-b2 (red) and C-b3 (yellow) interaction, backbone conformations adopted from IL8 and IL8RA. Aromatic interaction of Phe 286 (C-b2) and Phe 405 (C-b3); disulfide bridge between Cys 408 (C-b3) and Cys 283 (C-b2) based on biochemical data (blue) (see "Results"); Ser 281 loop/turn conformation adopted from malonyl coenzyme (cyan). Residues Asp 403 , Glu 404 , Asn 406 are located at an interface position at the ectodomain toward the serpentine domain.
coupled to the phospholipase C-inositol phosphate pathway (IP1-3). No constitutive activation of the IP pathway was observed for any of the mutants (Fig. 3F). After stimulation with 100 milliunits/ml bTSH, the mutations E404K and N406A showed decreased IP1-3 accumulation (Fig. 3G), which correlates with the cell surface expression. The mutant E404A had similar characteristics as the wt receptor. For the mutations D403A and D403K, no IP1-3 response could be detected. These findings were in accordance with the impaired cell surface expression of these two mutants.
In summary, we have identified a complete epitope with three extracellular constitutively activating mutants, D403A, E404K, and N406A, in the N-terminal portion of the TSHR. Aspartate 403 and asparagine 406 are highly conserved in the GPHR family, and the D403A and N406A mutations lead to similar levels of constitutive activity. In contrast, the E404A mutation does not affect basal cAMP accumulation, whereas the E404K mutation leads to an increased basal cAMP accumulation compared with the wt TSHR.

DISCUSSION
Using data from comparative modeling, sequence analysis, and reported site-directed mutagenesis studies, we have provided in this report strong indications explaining the spatial cooperation and relative location of molecular components in the GPHR ectodomain involved in receptor activation. Moreover, we have presented molecular models for the N-terminal TSHR components like the C-b1, the LRR motif, the C-b2 including Ser 281 and a C-terminal epitope of C-b3 based on homologous structural features. By extensive consideration of known functional data, these structural features were assembled to a tightly packed molecular model. This led to the identification of new key residues participating in intramolecular signal transduction at the interface toward the serpentine domain. By a combined modeling/mutagenesis approach, we demonstrated the functional importance of the N-terminal epitope D 403 EFN 406 within the C-b3 component of the TSHR ectodomain for the intramolecular signaling processes in the TSHR.
LRR Oriented in Spatial Proximity to the Serpentine Domain and Linked via Cysteine-box-2 and -3-For the LRR motif of GPHRs a new augmented structural template based on the Nogo-receptor ectodomain with 9 ϩ 2 repeats was generated that resulted in a strongly enlarged radius at the hormone binding site at the inner convex surface of the LRR arch. This allowed for the first time a docking model with parallel orientation of hormone and LRR (Fig. 1, a-c). There are a number of biochemical data available that are consistent with such a type of interaction (see supplemental material). Moreover, the Nterminal-flanked C-b1 is an integral structural part of the LRR by its contribution of an additional parallel ␤-strand (LRR0 (TSHR 37-41 )) to the convex ␤-sheet of the hormone-binding region. The resulting structural LRR models, augmented by C-b1, provide for the first time a structural rationale for the previously observed participation of residues from C-b1 in hormone binding, similar to the three residues at the additional repeat LRR0 of the FSHR 29 -31 (45). Furthermore, a peptide comprising exactly the additional LRR0 ␤-strand region TSHR [37][38][39][40][41] has been reported to inhibit hTSH binding (46). Finally, mutations in the N-terminal epitope from Asp 39 to Pro 47 forming the LRR0 binding site in the CG/LHR model show strongly decreased or no receptor expression and no ligand binding (8) ( Table I).
The next following structural component, C-b2, is attached back-to-back to the 11th ␤-strand of the LRR, very likely via a short turn/loop containing Ser 281 . Irrespective of the question whether the homologous turn configuration suggested by sequence identity is assembled in the right orientation, a turn/ loop structure very likely organizes how the LRR is spatially arranged (e.g. transversal or lateral) relative to C-b2/C-b3 and to the extracellular loops of the serpentine domain. Their mutual orientation obviously has a function as a potential pivot or hinge between them. This is consistent with data based on mutational activation at position Ser 281 (12,47) that indicate that minor conformational changes in the Ser 281 turn/loop activate the receptor. This may support the idea of a mechanism in which a certain spatial adjustment between the LRR motif and the following C-b2 (and remaining components of the ectodomain) via the Ser 281 turn/loop is responsible for the TSHR activation. This is consistent with our model, in which a LRR-bound hormone may also influence the spatial adjustment between LRR and C-b2/C-b3 (Fig. 2d).
Epitope D 403 EFNPC 408 of Cysteine-box-3 Is Localized at the Interface between the Ectodomain and Serpentine Domain-Based on new structural features for the ectodomain, identified by sequence homology with a complex of the IL8 ligand and the IL8RA peptide, we formed the hypothesis that the structural components C-b2 and C-b3 interact and are located close to TMH1. The structural consequence of this model is a potential disulfide bridge between Cys 283 and Cys 408 or Cys 284 and Cys 408 . Our findings are in agreement with suggestions of disulfide bridges between C-b2 and C-b3 based on biochemical data (48,49). Moreover, because the C-terminal end of the LRR is connected to C-b2 via the Ser 281 turn/loop and C-b3 is connected with the transmembrane domain TMH1, we have postulated a tightly packed structure including the C-terminal end of the LRRs and the C-b2 and C-b3 of the ectodomain. Subsequently, the epitope D 403 EFNPC 408 of C-b3 of the ectodomain was structurally localized at the interface between ectodomain and serpentine domain; it was therefore proposed as an important mediator for the intramolecular signal transduction from the ectodomain to serpentine domain (Fig. 2e). In the assembled models the two linked C-b2/C-b3 motifs are arranged across the serpentine domain. Their relative location may ac- count for an interaction with the embedded conformation of ECL2, which fits to data of an ECL2 interaction with the ectodomain reported for CG/LHR (16). Although the linked C-b2/C-b3 motifs are constrained by the fact that they are connected to TMH1, some altered orientations across the extracellular loops should be considered until non-ambiguous contact points are known (Supplemental Fig. 8, arrows). Therefore, interactions of C-b2/C-b3 with the other ECLs, as well as . FACS data are presented as mean Ϯ S.E. of three independent experiments, each performed in duplicate. After permeabilization of the cells, receptors with impaired cell surface expression (D403A, D403K, E404K) showed an increased FACS signal up to 60% compared with the wt TSHR. In confocal laser scanning microscopy experiments, TSHR mutants D403A, D403K, and E404K show strong intracellular receptor accumulation after permeabilization. This is in accordance with the results obtained by FACS analysis. C, B max values were determined by homologous competitive binding experiments. D, the TSHR has been described to be basal active in cAMP signaling. Therefore, elevated basal cAMP levels of cells expressing the wt TSHR in comparison with the pSVL vector alone can be observed after transient transfection of COS-7 cells. Because of the basal activity of the wt TSHR, cAMP levels are expressed as -fold over wt TSHR basal level. E and G, maximum increases in cAMP and IP1-3 levels were determined after stimulation with 100 milliunits of bTSH/ml. simultaneous access of the hormone to C-b2/C-b3 and to one of the ECLs, are feasible and correspond to previous results (50). This would also fit to reported data of ligand (relaxin-3) binding to both the N-terminal ectodomain and ECL2 of leucine-rich repeat-containing glycoprotein receptor 7 (51).
New Intramolecular Switch for Constitutive Activity in the TSHR Ectodomain-Our major finding is a new switch for constitutive activity identified in the TSHR ectodomain. Three amino acids of an epitope in the ectodomain (Asp 403 , Glu 404 , Asn 406 ) within the C-b3 component in proximity to TMH1 can act as switches for constitutive activity of the TSHR induced by point mutations. The residues Asp 403 and Asn 406 are highly conserved in GPHR and show similar constitutive activity for alanine mutants. The TSHR-specific glutamate at position 404 shows constitutive activity for lysine mutation, but not for an alanine substitution. Subsequently, the side chains of consecutive acidic residues Asp 403 and Glu 404 are obviously differently orientated in space and/or are differently tightly packed. Mutation of Asp 403 with a smaller neutral amino acid (alanine) causes a constitutive active mutant, very likely by loss of side chain interaction. In contrast, a lysine mutant at position Asp 403 with a more bulky and basic side chain strongly disturbs the receptor expression, probably by a bulky/electrostatic repulsion, which obviously only has an effect on the overall fold but no effect on basal activity. Therefore it is likely that Asp 403 and Asn 406 constrain the partial inactive receptor state by a side chain interaction via H-bonds toward a tightly packed environment, such as extracellular loops or TMHs. On the contrary, at the TSHR-specific residue Glu 404 , the mutant with a smaller neutral side chain, E404A, behaves as the wt TSHR. However, the mutant with the bulkier basic residue, E404K, behaves as constitutive active mutant, obviously by electrostatic/bulk repulsion effects. Therefore, Glu 404 does not seem to be involved in constraining the partial inactive state by its side chain interaction.
Ser 281 Turn/Loop and the New Epitope Asp 403 -Asn 406 in Cysteine-box-3 Are in Spatial Proximity-The results of our study show that the Ser 281 turn/loop and the epitope Asp 403 -Asn 406 are assembled in close spatial proximity at the interface between ectodomain and serpentine domain (Fig. 2d). Compared with Ser 281 , this epitope to a certain extent reacts differently to mutations and obviously has additional capabilities to induce a constitutive activity. The latter can be seen in previous observations (14) where stepwise N-terminal truncations of the TSHR ectodomain caused a 5-fold specific constitutive activation of the TSHR by deleting the Ser 281 region. An additional increase to 6-fold specific constitutive activation was observed by those fragments in which Asp 403 and/or Asn 406 were also deleted. In conclusion, the stepwise truncation of structural components of the ectodomain led to a graduated constitutive activation (partial activation) of the TSHR cAMP pathway. Single mutations at positon Ser 277 (Ser 281 in TSHR) cause constitutive activation. Their strength depends on size and property of the mutated side chain (47). In contrast to Ser 281 , which seems to be involved in constraining a loop/turn conformation (15), Asp 403 and Asn 406 seem to interact with other partners.
Asp 403 and Asn 406 Are Members of a Constraining Network at the Interface between Ectodomain and Serpentine Domain-From our study we cannot exclude an interaction of Asp 403 and Asn 406 with other N-terminal portions. However, based on our and other data available from reported mutagenesis studies, we suggest that they very likely form a common tightly interacting structural interface together with the extracellular loops and/or TMH residues. Already minor conformational changes (by mutants or by hormone binding) lead to receptor activation.

FIG. 3-continued
We suggest a scenario where this tight structural interface of the ectodomain (C terminus LRR, Ser 281 -turn/loop, C-b2, C-b3) and serpentine domain/extracellular loops constrains the native basal receptor state. Presumably it mediates the signaling upon hormone binding by release/formation of hydrogen bonds between ectodomain and serpentine domain. These predictions follow from the concept that receptor activation may involve the disruption of a constraining hydrogen bond network between the TMHs of the serpentine domain that stabilizes the partial inactive basal state while promoting the formation of a new set of hydrogen bonds that stabilizes the active state. We suggest that this constraining hydrogen bond network in the TSHR has an increased density toward the compact ectodomain. Asp 403 and Asn 406 seem to be members of this constraining network at the interface between ectodomain and serpentine domain.
The known constitutive active mutant switch position at Ser 281 is 100% conserved to a serine in the hCG/LHR (Ser 277 ) and hFSHR (Ser 273 ). According to the 100% conservation of the newly identified residues Asp 403 and Asn 406 in GPHRs, we strongly assume that these residues are of general relevance for the signal transduction mechanism of other glycoprotein receptors such as lutropin and follitropin receptors.