A Conserved Asn in Transmembrane Helix 7 Is an On/Off Switch in the Activation of the Thyrotropin Receptor*

The thyrotropin (TSH) receptor is an interesting model to study G protein-coupled receptor activation as many point mutations can significantly increase its basal activity. Here, we identified a molecular interaction between Asp 633 in transmembrane helix 6 (TM6) and Asn 674 in TM7 of the TSHr that is crucial to maintain the inactive state through conformational constraint of the Asn. We show that these residues are perfectly conserved in the glycohormone receptor family, except in one case, where they are exchanged, suggesting a direct interaction. Molecular modeling of the TSHr, based on the high resolution structure of rhodopsin, strongly fa-vors this hypothesis. Our approach combining site-directed mutagenesis with molecular modeling shows that mutations disrupting this interaction, like the D633A mutation in TM6

The TSH, 1 LH/CG and FSH receptors constitute a subfamily of G protein-coupled receptors (GPCR) characterized by large amino-terminal ectodomains responsible for high affinity binding of their natural agonists, the glycoprotein hormones (1,2). These ectodomains are made essentially of leucine-rich re-peats, a protein fold frequently found to be involved in proteinprotein interactions (3)(4)(5). The serpentine portion of these receptors, responsible for signal transduction, comprises seven transmembrane helices, showing significant similarity with the transmembrane portions of rhodopsin-like GPCRs; they are therefore grouped with them into family 1 of GPCRs.
The glycohormone receptors thus present a clear dichotomy between the agonist-binding and signal-transduction domains, with the mechanism of interaction between the two domains, responsible for receptor function, remaining largely unknown. Sequences displaying strong similarities with the glycohormone receptors have been identified in Anthopleura elegantissima, Drosophila, and Caenorhabditis elegans (6 -9). Related receptors have also been discovered in mammals (10,11). But for all of these cases, the nature of the agonists remains to be identified.
Several characteristics make the TSH receptor an interesting system to study the mechanisms of receptor activation: (i) the wild type receptor has been shown to display significant basal activity (12)(13)(14); (ii) mutations involving more than 20 different residues have been shown to increase its constitutive activity, causing autonomous thyroid adenomas or non-autoimmune hereditary hyperthyroidism (15); (iii) loss of function mutations have also been described, abolishing basal activity (16) or affecting agonist-induced response (17,18).
Interestingly, naturally occurring activating mutations are more rarely found in the LH/CGr (19), and almost never found the FSHr (20). Indeed, no spontaneous activating mutation has been identified in the FSH receptor except for a single case with an unusual phenotype (21). These data reflect a difference in the structural constraints keeping the various glycoprotein hormone receptors in the inactive state. The wild type TSH receptor displaying constitutive activity and being easily activated by mutations would be the less constrained, followed by the LH receptor, devoid of constitutive activity but susceptible to activation by mutations. The FSH receptor would be the most constrained, as it is silent in the absence of stimulation by agonist and difficult to activate by mutations.
The high resolution crystal structure of bovine rhodopsin determined recently (22) provides for the first time a detailed atomic description of a GPCR molecule in an inactive conformation, and represents a solid basis for modeling the structures of other rhodopsin-like GPCRs. Such models can then be used to help rationalize the many observations made on the relations between conserved sequence feature and functional properties. Conservation of functionally important sequence motifs within this receptor family has been interpreted as meaning that the basic characteristics of the rhodopsin fold, as well as the molecular mechanisms leading to receptor activation are similar in the different receptor subtypes. Among such conserved sequence motifs are prolines in transmembrane helices (TM) 6 and 7 and charged residues in TM2 and TM3 (the Asp and Arg of the canonical "DRY" motif). However, considering the extreme diversity of the natural agonists of the different receptors it has also been accepted that different receptor subtypes may have evolved quite specific structural and functional features, probably reflected in the specific sequence signatures of each subtype.
Multiple alignments of the glycohormone receptor sequences reveal one such signature: a conserved Asp residue in TM6 at position 6.44 633 (corresponding to residue 633 in the TSHr sequence; see "Experimental Procedures"). Most other GPCRs harbor a Phe or a Tyr at this position. Natural mutations of residue 6.44 633 in the TSHr and LH/CGr were shown to lead to constitutive activation, causing autonomous toxic thyroid adenomas (12) and male limited pseudoprecocious puberty (23), respectively. This suggested that D6.44 633 might play a role in the mechanism of receptor activation. Recent ex vivo transfection studies suggested that this residue is important for maintaining the inactive conformation of the LH receptor (24), in agreement with earlier molecular modeling studies, according to which the inactive form of the LH receptor would require the formation of H-bonds between TM6 and TM7, involving residues Thr6.43, Asp6.44, Asn7.45, and Asn7.49 (25).
In the present study, we analyzed the multiple alignments of 44 glycohormone and closely related receptors. This lead to the observation that residues D6.44 and N7.49 (the Asn of the conserved NPXXY motif in TM7) are both conserved in glycohormone receptors, except in one of two Drosophila receptors (6), where they are exchanged (this receptor bearing N6. 44 and D7.49). Since such coordinated mutations may reflect spatial interactions between the residues (26,27), our observation was interpreted as indicating the existence of an interaction between D6.44 and N7.49. Considering the activating effect of amino acid substitutions at position 6.44, we formulate the hypothesis that this interaction is required in order to maintain the TSH receptor in an inactive conformation. We then investigated this hypothesis by combining computational and experimental approaches.
Molecular modeling techniques were used to derive the atomic coordinates of the transmembrane region of the TSHr, using as template the corresponding region of the high resolution structure of rhodopsin. To analyze the conformational properties of residues 6.44 and 7.49 and their surroundings in the resulting model, molecular dynamics simulations were carried out in the presence of explicit methane molecules, mimicking the non-polar environment of the membrane. This was done for both the wt receptor and for several substitution mutants of residues that are likely to affect the proposed interaction. In parallel, the same mutants were engineered into human TSHr using site-directed mutagenesis and their effects on receptor expression, hormone binding, and activation were measured.
The results provide compelling evidence that D6.44 and N7.49 do interact in the inactive TSHr and that disruption of this interaction, by mutating either the 6.44 or the 7.49 residues, results in constitutive activation. Furthermore, combining our observations with data from other GPCRs is shown to yield new insights into the general mechanism of GPCR activation

EXPERIMENTAL PROCEDURES
Numbering Scheme of GPCRs-The standardized numbering system of Ballesteros and Weinstein (28) was used throughout to identify residues in the transmembrane segments of different receptors. Each residue is identified by two numbers: the first (1 through 7) corresponds to the helix in which it is located; the second indicates its position relative to the most conserved residue in that helix, arbitrarily assigned to 50. For instance, N7.49 is the asparagine in transmembrane helix 7 (TM7), located 1 residue before the highly conserved proline P7. 50 (22). The sequences of the 2 proteins in this region were aligned so as to equivalence the positions of the following conserved residues: Asn 55 -Asn1.50 432 (the superscripts represent the residue numbering in rhodopsin structure PDB code 1F88, and human TSHr sequence, respectively, and 1.50 is the numbering in the standardized nomenclature), Asp 83 -Asp2.50 460 , Arg 135 -Arg3.50 519 , Trp 161 -Trp4.50 546 , Pro 215 -N5.50 590 , Pro 267 -Pro6.50 639 , and Pro 303 -Pro7.50 675 . The conformations of these side chains were kept as in the rhodopsin crystal structure. Those of the non-conserved amino acids were built using a rotamer library specific for ␣-helices. (29). Ionizable groups in the helices were modeled as uncharged except for Asp2.50, Glu3.37, Glu3.49, Arg3.50, Asp6.30, and Asp6.44.
To relieve residual strain resulting from suboptimal positioning of the side chains, the resulting model was first subjected to energy minimization (1000 steps), and then to a simulated annealing procedure, using molecular dynamics. This involved heating to 600 K for 30 ps (1 ps ϭ 10 Ϫ12 s), equilibration for 120 ps at 600 K, cooling to 300 K for 60 ps, and equilibration for 90 ps at 300 K. During this processes the C ␣ atoms and the side chains of conserved residues were kept fixed at their positions in the rhodopsin crystal structure.
Molecular models for the mutant receptors containing the single substitutions D6.44N, D6.44A, N7.49D, and N7.49A and double substitution D6.44N-N7.45D and D6.44N-N7.49D, were built using, as the starting point the final model derived for the wt TSHr and optimizing the conformations of the substituted side chains using an analogous procedure to that described above. This procedure was deemed reasonable, since the considered substitutions were either isosteric, replacement of Asn by Asp, or vice versa, or side chain deletion (replacement by Ala).
To investigate the structural properties of the modeled conformations, room temperature molecular dynamics simulations were used. In order to mimic the hydrophobic environment of the membrane, these simulations were carried out in the presence of explicit methane molecules, and using periodic boundary conditions. The periodic box was ϳ73 Å ϫ 63 Å ϫ 52 Å in size, and contained between 4219 and 4241 methane molecules in addition to the transmembrane domain. Similar conditions have been recently used to mimic the membrane environment in molecular dynamics simulations of the potassium channel (30).
To start the simulation protein portion was kept fixed while the methane molecules were energy minimized (500 steps), then heated to 300 K for 15 ps and equilibrated for another 35 ps. Following this, the same procedure was repeated on the entire protein-solvent system but with an equilibration run of 250 ps; a 250 ps production trajectory was then generated at constant volume using the Particle Mesh Ewald method for computing electrostatic interactions (31). For analysis purposes, structures were collected every 2 ps. The simulations were performed with the Sander module of AMBER 5 (32), the all-atom force field (33), SHAKE bond constraints on all bonds, a 2 fs integration time step, and constant temperature of 300 K coupled to a heat bath.
Site-directed Mutagenesis of the TSH Receptor-Plasmids encoding the various TSHr mutants were constructed by site-directed mutagenesis using two subsequent PCR amplifications rounds. This procedure requires two partially overlapping complementary primers containing the mutation and two external primers. Two distinct PCRs are performed on TSHr template by using in one tube the direct mutagenic primer and the external reverse primer and in the other tube the reverse mutagenic primer and the external direct primer. One l of each PCR product was mixed and used as template in a subsequent PCR amplification with the two external primers. The resulting amplified fragment contains the mutation and can be cloned after digestion with Bsu36I and BamHI into the pSVL expression vector (Amersham Pharmarcia Biotech, Freiburg, Germany) containing the wild type TSHr (34) using standard procedures. All PCR-generated receptor fragments were verified by sequencing before transfection.
Transfection and Assays-COS-7 cells were grown in Dulbecco's modified Eagle's medium supplemented with fetal bovine serum (10%), sodium pyruvate (1 mM), penicillin (100 IU/ml), streptomycin (100 g/ ml), and fungizone (2.5 g/ml). Cells were seeded at a density of 150,000 or 300,000 cells/3-cm dish. One day later, they were transfected (250 or 500 ng of DNA/dish, respectively) by the DEAE-dextran method followed by a dimethyl sulfoxide shock (35). Two days after transfection, cells were used for flow immunocytofluorometry, cAMP and 125 I-TSH binding studies. Duplicate dishes were used for each assay. Each experiment was repeated at least three times. Cells transfected with pSVL alone were always run as controls.
Flow Immunocytofluorometry-Cells were detached from the plates with phosphate-buffered saline containing EDTA and EGTA (5 mM each) and transferred into Falcon tubes (2052). Cells were centrifuged at 500 ϫ g, at 4°C for 3 min and the supernatant was removed by inversion. They were incubated for 30 min at room temperature with 100 l of phosphate-buffered saline/bovine serum albumin (0.1%) containing the BA8 monoclonal antibody, obtained from genetic immunization with the wild type TSH receptor cDNA (36). The cells were washed with 4 ml of phosphate-buffered saline/bovine serum albumin (0.1%) and centrifuged as above. They were incubated for 30 min on ice in the dark with fluoresceine-conjugated ␥-chain-specific goat antimouse IgG (Sigma) in the same buffer. Propidium iodide (10 g/ml) was used for detection of damaged cells which were excluded from the analysis. Cells were washed once again and resuspended in 250 l of phosphate-buffered saline/bovine serum albumin (0.1%). The fluorescence of 10,000 cells per tube was assayed by a FACScan Flow Cytofluorometer (Beckton Dickinson, Erembodegem, Belgium).
cAMP Determination-Cells were washed with Krebs-Ringer-Hepes buffer (KRH isotonic, pH 7.4). After a preincubation in KRH at 37°C for 30 min, cells were incubated in the same buffer supplemented with Rolipram 25 M (a cAMP phosphodiesterase inhibitor, gift from the laboratory of J. Logeais, Paris, France), in the absence or presence of various bTSH concentrations (Sigma). One hour later, the medium was removed and 0.1 M HCl was added to the cells. The cellular extracts were dried overnight in a vacuum concentrator (Savant) and intracellular cAMP was determined exactly as described previously (37). Basal cAMP was normalized to cell surface expression for each of the constructs. To this end, specific cAMP accumulation (ϭ cAMP of receptortransfected cells Ϫ cAMP of the pSVL-transfected cells) is divided by the specific FACS value (ϭ fluorescence of receptor-transfected cells Ϫ fluorescence of pSVL-transfected cells), which can be summarized as: specific basal activity ϭ (cAMP(receptor) ϪcAMP(pSVL))/(FACS-(receptor) Ϫ FACS (pSVL)). The values are then expressed as percentage of specific basal activity of the wt TSHr.
Binding Assays-Two days after transfection, cells were washed twice with NaCl-free Hank's buffer supplemented with 280 mM sucrose. Cells were incubated at room temperature in this medium (supplemented with low fat milk, 2.5%) containing about 100,000 counts/ml 125 I-TSH (TRAK Assays, BRAHMS Diagnostica, Berlin; 58 Ci/g, 50 -60 units/mg) and various concentrations of unlabeled TSH (Sigma). Four hours later, the cells were rinsed twice with chilled Hank's buffer and solubilized by 1 N NaOH. Bound radioactivity was determined in a ␥-scintillation counter. Under our conditions, the radioactivity nonspecifically bound to COS cells expressing the TSH receptor (defined as the radioactivity bound to the dish in the presence of 100 milliunits/ml unlabeled TSH) was identical to that bound to mock-transfected cells.
In the absence of a consensus about the bioactivity of pure bovine TSH, we have expressed all TSH or TSH receptor concentrations in milliunits/ml, assuming a 1:1 stoichiometry for TSH binding to its receptor. The competition binding curves have been fitted by nonlinear regression, assuming a single receptor site, using the Prism3 program (Graph-Pad Softwares).

Sequence Alignment of the Glycohormone Receptor Subfamily
Multiple sequence alignment of the sixth and seventh transmembrane helices of 44 glycohormone and closely related receptors was performed. Fig. 1 shows the alignment of representative sequences, together with that of bovine rhodopsin. Inspection of this alignment reveals two key sequence motifs characteristic of the rhodopsin-like family: the P6.50 in TM6 and the NPXXY motif in TM7. A noteworthy difference is the presence of Asp at position 6.44 instead of the more common Phe or Tyr, which represent more than 81% of the rhodopsin-like receptors found in the GPCRDB (38). D6.44 is completely conserved throughout the glycohormone family of GPCR, from C. elegans to human, with only a single exception, the Drosophila DLGR1 sequence (accession number U47005 (6)), which contains an Asn residue. Interestingly, this change is correlated with the replacement of the highly conserved Asn at position 7.49 by Asp. Swapped mutations of this kind suggest interaction between the corresponding residues in the three-dimensional structures (26,27). We could indeed verify that in the rhodopsin structure, the C ␣ of Phe6.44 and Asn7.49 (the conserved N of the NPXXY motif) were facing each other at a distance of 11 Å, sufficient to allow direct interaction between the side chains. Given that a series of amino acid substitutions at D6.44 found in thyroid adenomas result in a significant increase in the constitutive activity of the TSHr, we then made the hypothesis that interactions between the D6.44 and N7.49 side chains are important in maintaining the receptor in its inactive conformation.

Molecular Dynamics Simulation of the Transmembrane Region of the TSHr
In order to investigate the proposed interaction between D6.44 and N7.49 in the context of the entire helix bundle, the molecular model of the transmembrane domain of the TSHr was built, based on the high-resolution structure of bovine rhodopsin. This model was then subjected to unrestrained molecular dynamics simulations in the presence of explicit methane molecules mimicking the apolar membrane environment (see "Experimental Procedures"). Having ascertained that the helical segments conserved their secondary structure and that the bundle remained well packed and maintained the rhodopsin fold throughout the trajectory, we analyzed the polar interactions made between TM6 and TM7. In particular, we computed the average hydrogen-acceptor distances in conformations along the trajectory, and the fraction of the conformations in which the bond was formed.
All throughout the simulation D6. 44   the interactions between these side chains and any other residue. Following observations made from our molecular dynamics simulation of the TSHr and other studies on the LH/CG receptor (24) we designed two additional mutants, the single N7.45D substitution and the double substitution D6.44N/ N7.45D. This was done in order to explore the possibility of a direct interaction between D6.44 and N7.45.

Cell Surface Expression of the Mutants
Cell surface expression of the mutated receptors was measured by FACS analysis using the BA8 monoclonal antibody (36). As the epitope of the BA8 antibody is located in the NH 2 -terminal region of the TSHr (1), we do not expect a modification of its affinity following mutations in the transmembrane region. Also, similar results were obtained with another monoclonal antibody (3G4), recognizing a different epitope in the ectodomain (not shown). Hence, it is assumed that observed fluorescence changes will be in direct relation with the number of receptors present at the cell surface. As shown in Fig. 3, mutants D6.44N, D6.44N/N7.49D, and N7.49A are expressed at levels comparable (above 50%) to that of the wild type TSHr. In contrast, mutants N7.49D, D6.44A, and the D6.44N/N7.45D double mutant display a reduced expression, between 15 and 30% of the wild type receptor expression. The level of expression of N7.45D is too low (specific fluorescence of about 3% of wt TSHr) to allow reliable normalization of functional results, it will therefore not be considered further.

Basal and TSH-stimulated cAMP Accumulation in COS-7 Cells Transfected with the Mutants
The constitutive activity of the wt and mutant receptors was characterized by measuring intracellular cAMP accumulation in transiently transfected COS-7 cells. As constitutive activity is linearly dependent on the number of receptors at the cell surface (39), we have normalized the basal activity of each construct using cell surface expression data yielded by FACS analyses (see "Experimental Procedures"). This allows to compute constitutive activity on a per receptor basis, and to compare it to that of the wild type receptor.
Mutants D6.44N, D6.44N/N7.49D, and N7.49A display basal activity similar to that of the wild type (Fig. 4A). Strikingly, the single mutants N7.49D and D6.44A show a dramatic increase in constitutive activity, reaching more than 15 times that of wild type TSHr, when normalized to the level of expression (Fig. 4B). Although the N7.49D mutant, bearing aspartate residues at both positions 6.44 (as in wild type) and 7.49 (introduced by mutation), is among the strongest constitutive mutant of the TSHr ever identified (37), it is remarkable that addition of the D6.44N mutation on the N7.49D background reverses completely the phenotype back to a wild type-like behavior. The D6.44N/N7.45D can also be considered as a constitutive mutant as its basal activity is more than five times that of wt TSHr (Fig. 4B).
We then tested the ability of TSH to stimulate the various constructs. Fig. 5 illustrates typical concentration-response curves and the corresponding EC 50 are summarized in Table I. All mutants except N7.49A are stimulated by TSH in a way similar to the wild type. While starting from a higher basal level, the N7.49D mutant is activated efficiently by TSH, reaching comparable E max (not shown). All EC 50 values were in the range 0.4 -1 milliunits/ml of TSH (Fig. 5, A and B, and Table I). Although the N7.49A mutant displays wt-like constitutive activity, TSH is unable to elicit signal transduction, as only very moderate stimulation if any can be observed at 100 milliunits/ ml, a saturating concentration for wt TSHr (Fig. 5C).

Hormone Binding
The binding properties of the different receptors were tested by homologous competition using 125 I-bovine TSH as tracer. As shown in Fig. 6, all constructs tested were able to bind TSH efficiently.
Mutants D6.44N and D6.44N/N7.49D bind TSH with similar affinity as the wild type receptor, as measured by the IC 50 test (IC 50 of about 3 milliunits/ml, see Table I and Fig. 6A). Noticeably, the various constitutive mutants display a significantly lower IC 50 (below 1 milliunits/ml, see Table I and Fig. 6B). Increase in apparent affinity with no significant modification of hormone potency has been observed in most constitutive mutants of the TSH receptor (40). Interestingly, mutant N7.49A binds TSH with high affinity, even slightly higher than the wt receptor ( Fig. 6A and Table I), although this mutant is completely unable to be activated by the hormone.
It should be noted that the residues identified here as key actors in the activation mechanism of the TSHr have also been studied by several groups on the LH receptor (24,41). Overall these studies are in agreement with the results presented here. Note, however, that these studies did not consider double mutants in different TMs, which in our view are crucial to the identification of TM6-TM7 interactions.

Molecular Dynamics Simulations of the Transmembrane Region of Mutant Receptors
In an effort to interpret, in structural terms, the results of the site-directed mutagenesis experiments, the molecular dy- namics trajectories of the transmembrane region of TSHr featuring the different mutants were analyzed, with focus on the region between TM6 and TM7, surrounding the mutated residues. In particular, we monitored all possible interactions between the various polar groups in the vicinity of 6.44 and 7.49, which were involved in the local hydrogen bond network in the wt TSHr trajectory. Representative structures of the mutant trajectories are shown in Fig. 2.
The D6.44N Mutant Receptor-Notwithstanding the absence of the acid group, the D6.44N mutant was found to maintain polar interactions between TM6 and TM7 (see Fig. 2 (Fig. 2). This interaction too is facilitated by a local rearrangement of TM7 which brings 7.49 closer to 2.50 than in the wt simulation (data not shown).
In summary, the nonconstitutive mutant receptors (D6.44N, D6.44N/N7.49D, and N7.49A) maintain the TM6-TM7 interaction through a complex hydrogen bond network involving the side chains at positions 3. 39, 6.43, 6.44, 7.45, and 7.49. Moreover, as in the wt simulation, the side chain of residue 7.49 is always maintained close to TM6 (except in the N7.49A mutant), essentially through direct partnership with residue 6.44.  (6) an exchange between two highly conserved residues: Asp6.44 and Asn7.49. These two residues are spatially close to each other in the rhodopsin template, suggesting a possible direct interaction. In addition, the observation that a variety of spontaneous mutations affecting D6.44 in both the TSHr and LHr increase their constitutive activity, pointed to a possible role of D6.44 (and thus of the postulated 6.44 -7.49 interaction) in the activation mechanism.
In this context, mutation of N7.49 to Asp is expected to abrogate this putative interaction between residues 6.44 and 7.49, owing to electrostatic repulsion between the 2 negatively charged side chain, which should be rescued by the complementary mutation of D6.44 to Asn. The functional characteristics of these mutants fulfill nicely these expectations when assayed in COS cells: the N7.49D mutant displays a very strong increase in basal activity, reaching more than 15 times that of the wild type receptor, whereas the D6.44N/N6.49D double mutant behaves as the wild type receptor. Together with the reciprocal amino acid exchange observed at these positions in the Drosophila receptor the simplest explanation is that these two residues are close in space and do interact with each other. Similar reasoning and experimental approaches have been used in the study of other receptors (42)(43)(44). Interestingly, they supported the idea that the same N7.49 residue would interact with D2.50, one of the most conserved residues in rhodopsinlike GPCRs. The apparent incompatibility between these two mutually exclusive interactions find an explanation in the study of additional mutants (see below).
Our molecular dynamics study of the helix bundle of the TSHr based on the crystal structure of rhodopsin strongly supports the D6.44-N7.49 interaction. In simulations of the wild type TSHr structure, the D6.44-N7.49 interaction was observed in all conformations of the trajectory, indicating that this putative partnership fits well the rhodopsin template, under our simulation conditions. In agreement with experimental observations, simulation of the D6.44N/N7.49D double mutant shows that the interaction between these amino acids is preserved when the two side chains are exchanged, as found in the Drosophila receptor. It should be noted that our modeling study predicts that a 6. 44-7.49 interaction would be preserved in the D6.44N single mutant, and, indeed, when tested experimentally, this mutant showed a wild type-like behavior. This observation is of particular significance, considering that most if not all other mutations at this position induce constitutive activity in the TSH or LH receptors (35,41). The structural rationale explaining why only Asn is tolerated at position 6.44 in the glycohormone receptors gives support to the local structure proposed here and suggests an explanation as to how the D6.44N/N6.49D double substitution may have occurred stepwise in evolution in the Drosophila receptor.
Our modeling points to additional partners to account for the phenotype of some of the constitutive mutants identified here. The strongest constitutive activity is associated with a complete lack of the 6. 44-7.49 interaction (like the N7.49D and D6.44A mutants). Weakening of this bond as observed in the D6.44N/N7.45D double mutant causes increase in constitutive activity but to a smaller extent, which indicates that this mutant is not completely released from its ground state constraints. In our N7.49D simulation, D6.44 interacts with serine residues S3.36 505 and S3.39 508 , in TM3. Interestingly mutations of S3.36 into Arg or Asn were found to constitutively activate the receptor (40,45). It is likely that the 3. 36-6.44 interaction observed in the N7.49D mutant would be strongly favored in a S3.36 505 R mutant. An ionic R3.36-D6.44 interaction would be more stable than the D6.44-N7.49 (wt situation), therefore redirecting the side chain of D6.44 toward TM3 and releasing N7. 49.
Sequestration of N7.49 Is Implicated in the Inactive Conformation of the TSHr-Overall the modeling and experimental data converge toward the idea that, in the TSHr, a direct D6.44-N7.49 interaction is required to maintain the inactive state of the receptor, and that breaking this link results in high constitutive activity. How does this view agree with current notions of activation of other rhodopsin-like GPCRs?
Despite the fact that residue D6.44 is not conserved throughout the family (in most cases it is a Phe), mutation of 6.44 into Ala leads also to constitutive activation in other rhodopsin-like receptors (46,47). Moreover, in the opsins, 6.44 has been shown to be linked to the activation process, being involved in color sensitivity (48,49). The mechanism by which Asp (in the glycohormone receptors) or Phe (in most of the others) might exert similar functions cannot be determined from the present study.   (42,43). As already mentioned, using a similar experimental strategy as the one described here, it has been shown in these studies that inactivation by a D2.50N mutation could be rescued by an additional N7.49D mutation. Moreover, it has been proposed that D2.50 undergoes a change in H-bonding upon light activation of rhodopsin (50).
A Model of the Early Steps of Activation-We can summarize the different points discussed above in a simple sequence of intramolecular events that would be necessary to trigger the conformational changes leading to activation.
(i) In the inactive state of the receptor, N7.49 is locked away from TM2. In the glycohormone receptor family, this conformational constraint is achieved by a direct interaction with D6.44.
(ii) A necessary step in activation would be the release of this interaction. This could happen either upon ligand binding, or due to mutation of a key residue. For instance, the D6.44A mutation abrogates completely the interaction. Noticeably, this mutation leads to almost maximal activation (2/3 of activity after full stimulation by the hormone).
(iii) N7.49 would then reorient toward the 1.50-2.50 motif, where it could modify the polar equilibrium and subsequently cause a reorganization of the inter-helical H-bond network. This rearrangement would in turn induce the conformational changes of the TM bundle known to be associated with activation (i.e. motions of TM3 and TM6 (51,52)). In the N7.49A mutant, the side chain of A7.49 would not be able to establish the interaction(s) normally fostered by the Asn, and no transition to the active state is observed. Whether this scenario applies to the physiological activation of the receptor by its natural agonist remains to be demonstrated.
The 2.50-7.49 interaction is well supported by numerous studies on various receptors. However, a survey of the GPCR data base yields about 100 sequences of rhodopsin-like receptors with an Asp residue at both the 2.50 and 7.49 positions. In this case a direct 2.50-7.49 interaction is impossible. It has been shown that an Asn to Asp substitution at position 7.49 leads to modification of the signaling characteristics in GPCRs: changing the N7.49PXXY motif for D7.49PXXY leads to changes in the coupling specificity of the GnRH receptor (53). This indicates that the active conformation achieved by receptors bearing an Asp in 7.49 may be somewhat different from that of those bearing an Asn, in agreement with the current concept that GPCRs may achieve multiple active states with the potential to activate different G proteins (52).
When integrated with the current literature, the present results point to the 1.50-2.50-7.49 motif as an important actor in the activation of rhodopsin-like GPCRs. Additional players are likely to be involved as well, which remain to be identified. Among these are positively charged residue(s), participation of structural water molecules and specific protonations. In this context, it was suggested that D2.50 would interact with the conserved R3.50 of the DRY motif at the base of TM3 (54,55). However, since the C ␣ of these two residues are separated by more than 24 Å in the crystal structure of rhodopsin, we do not currently favor such a possibility. Positively charged residues are often found at the cytoplasmic end of TM6 and could pos-sibly interact with the 1.50-2.50-7.49 motif after a rigid body motion of TM6 of the kind which is believed to occur upon activation (56 -58). It is noteworthy that a water molecule is present in close vicinity of D2.50 in one of the two monomers of the rhodopsin crystal structure (22). Also, it is known that protonation events do take place upon activation of GPCRs. One of the protonation site is most probably the highly conserved D3.49 of the DRY motif (59). Recent studies on the ␤ 2 -adrenergic receptor indicate that protonation of an additional negatively charged residue is most probably implicated in activation (60). D2.50 is a good candidate for the "missing" protonated residue in adrenergic receptors and has been proposed to be protonated in rhodopsin (50,61). Protonation of D2.50 could be a necessary step in the activation process, as it has been shown for several receptors that a D2.50N mutation severely impairs the signaling properties, while a D2.50E mutation can be tolerated (44). Protonation of D2.50 would account for the observation that the N1.50-D2.50-N7.49 motif tolerates N7.49D substitutions in many systems, including the TSH receptor. As we were not able to include the various putative additional players in the molecular dynamics simulations, we could not observe a 2. 50-7.49 interaction in our model of the N7.49D mutant. In all fairness, the absence of these players in our simulations and for that matter, also that of the entire extracellular domain and of an explicit lipid environment, are limitations that must be kept in mind also when considering other conclusions deduced from the simulations.
The Active Conformation of the Wild Type Receptor May Be Different from that of D6. 44 Mutants-Of potential interest is the observation that the N7.49A mutant, while it cannot be activated by the hormone, still displays constitutive activity similar to the wild type TSHr. This implies that the interactions achieved by the side chain of N7.49 would be dispensable for the basal activity of the wild type receptor, whereas they are required for that of the ligand induced activity. The existence of multiple active conformations of GPCRs has been well documented in the case of TSHr mutants (16,62) as well as in many other GPCRs (see Ref. 52 for review). The present observation suggests that "noisy" receptors like the unliganded wild type TSHr might achieve an active conformation via a route involving residues distinct from those implicated in full agonistinduced activation.
In conclusion, we have identified a conserved motif central to the activation of the TSH receptor. Our molecular modeling provides a structural framework for understanding how the D6.44-N7.49 link, by sequestrating the side chain of N7.49, would keep it in an "inactive" conformation. Upon activation, this side chain would be released, and free to adopt its "active" conformation, which involves interactions with the N1.50-D2.50 motif. Further rearrangements of the intramolecular network, which remain to be defined, lead to exposure of surfaces allowing interaction between the receptor and its cognate G protein(s). We suggest that in other GPCRs, a similar mechanism might operate, but involving a different type of partner for N7.49 in the inactive state. Now that rhodopsin provides a reliable template for modeling the inactive conformation of GPCRs, it is our hope that further molecular dynamics simulations coupled to site-directed mutagenesis experiments will help in elucidating their active conformation(s).