In 2006 our laboratory was able to determine the high resolution (2.55 Å) crystal structure of the TSH receptor (TSHR) leucine rich domain (LRD) in complex with the thyroid stimulating monoclonal autoantibody M22 (1). As a consequence, the structure of the TSHR LRD is now established. Also the molecular interactions between the receptor and M22 are known in detail. These results have important implications for our understanding of glycoprotein hormone interactions with their receptors. Furthermore, there are important implications for the rational design of new molecules which can interact with the TSHR as agonists or antagonists. In addition, the structure provides key insights into the nature of the autoimmune response to the TSH receptor. A brief review of these aspects of the TSHR and how it interacts with M22 and TSH is presented.

TSH receptor structure
The overall structure of the TSHR determined by comparative modelling (2,3) prior to crystal structure analysis is shown in Figure 1. It consists of three major domains: the LRD, cleavage domain (CD) and transmembrane domain (TMD). The concave surface formed by the LRD (with some contribution from the CD) is principally responsible for ligand binding. Obtaining highly purified preparations of the TSHR suitable for crystallisation and analysis by X-ray diffraction has been a major challenge. This was due in part at least to the susceptibility of native and recombinant TSHR preparations to degradation and denaturation. Furthermore this instability is seen in preparations of the isolated LRD.
Early studies (4,5) with preparations of native TSHR extracellular domain (ECD) showed that when the ECD bound patient serum TSHR autoantibodies a stable complex was formed. This complex consisted of one molecule of ECD and one molecule of antibody and it could be purified by various chromatographic procedures and isoelectric focussing. Once the thyroid stimulating human monoclonal autoantibody M22 became available (6,7) it was possible to use a similar approach to these earlier studies to obtain preparations of the TSHR-M22 complex suitable for structural analysis. In particular, a TSHR260 construct (coding for amino acids 1-260 with a 6 His tag at the C-terminus) was expressed in insect cells with M22 Fab included in the culture medium.
As TSHR260 was secreted into the culture medium, it was bound by M22 and a stable complex formed. The complex could then be purified by various chromatographic procedures to give a preparation about 95% pure containing equal proportions of M22 Fab and TSHR260 (1). After deglycosylation, crystals suitable for analysis by X-ray diffraction were obtained and the structure solved at 2.55 Å resolution (1).

TSHR LRD structure
The structure of the TSHR LRD is shown in Figure 2 and the receptor has the shape of a slightly curved helical tube (Figure 3) constructed from leucine-rich repeat motifs. It has opposed concave and convex surfaces with a 10-stranded β sheet located on the concave surface. The inner surface of the tube is lined with hydrophobic residues. The TSHR N-terminal cysteine 31 and cysteine 41 are linked via a disulphide bond. All five glycosylation sites are located on the convex surface of the LRD (Figures 2 and 3), away from the regions involved in ligand binding.

The TSHR-M22 complex
M22 Fab clasps the concave surface of the TSHR LRD at 90° to the axis of the LRD tube and a large part of the LRD surface is involved in interactions with M22 (Figure 3). These extend from the LRD N-terminus to C-terminus and a total of 2500 Ų of solvent accessible surface area is buried in the interface (1). This area is larger than typically observed for antibody-antigen interactions (8). There are a large number of hydrogen bonds and salt bridges in the interface together with non-hydrogen bonding polar interactions, hydrophobic contacts and van der Waals interactions. The heavy chain (HC) of M22 Fab has more amino acids in interaction with the LRD than has the light chain (LC). Both chains form hydrogen bonds and salt bridges with the TSHR, 14 in the case of the HC and 8 in the case of the LC. Most of the residues in M22 which interact with the TSHR are in the hypervariable regions, particularly H2, H3 and L2.

Comparison of receptor bound and free M22
As the crystal structure of M22 Fab is known at 1.65 Å resolution (7), when the TSHR-M22 structure was solved, it was possible to compare receptor bound and free M22 structures in detail. This comparison showed that the majority of M22 variable region residues involved in interactions with the TSHR had almost identical positions in the receptor bound and free structures. The highest deviation of an atom from M22 backbone residues was 1.1 Å. In addition, only six M22 amino acids in the complex showed a deviation of their side chains of greater than 2 Å compared to unbound M22. Consequently, there is essentially no movement of the atoms of M22 when receptor binding occurs. As a result, the free energy loss which would happen as a consequence of a conformational change in M22 does not occur and this is consistent with the high affinity of the M22-TSHR interaction (6,7). The TSHR LRD itself is a relatively rigid structure and a major change in LRD conformation on M22 binding is unlikely. However binding of M22 to the LRD must induce changes in the receptor which cause signal induction but to date the nature of these changes is not known.

The TSHR-TSH and FSHR-FSH complexes
Once the structure of the TSHR LRD was known, it was possible to use the structure, together with the structure of the FSHR LRD (9) to prepare a new comparative model of the TSHR LRD bound to TSH. This involved (a) building a model of the TSHR-TSH complex based on the crystal structure of the FSHR-FSH complex (2.9 Å resolution), (b) replacing the FSHR structure in the model with the TSHR crystal structure (2.55 Å resolution) and (c) minimisation of the interface in order to optimise interactions between TSH and the TSHR. This model of the TSHR-TSH complex is shown in Figure 4 together with the FSHR-FSH structure for comparison.

Comparison of TSHR interactions with hormone and with autoantibody
When the structures of the TSHR LRD-M22 complex and the FSHR LRD-FSH complex are superimposed, the positions of M22 Fab and FSH relative to the concave surface of their respective receptors are remarkably similar (1). As TSH and FSH position themselves on their respective receptors in essentially identical ways (Figure 4) this means that M22 positions itself on the TSHR in a very similar way to TSH. Consequently, two ligands with very different origins and structures show remarkably similar TSH receptor binding features. This is an example of evolutionary convergence and the nature of the evolutionary forces which have resulted in the production of an autoantibody which mimics the actions of the native hormone TSH so well are intriguing.

Comparison of hormone-receptor and autoantibody-receptor interactions
As described above, M22 positions itself on the TSHR LRD in a very similar way to TSH and in a very similar way to FSH on the FSHR LRD. Also, the solvent accessible surface areas buried in the interfaces between the TSHR and M22 and between the FSHR and FSH are similarly large (2500 Ų and 2600 Ų respectively) (1,9). In addition, the overall structures of both complexes are remarkably similar with the rmsd on C_α core atoms of the two receptor LRDs being 1.1 Å (1). M22 and FSH interact with residues contributed by all 10 β-strands on the concave surface of the TSHR or FSHR respectively. However, there are 14 hydrogen bonds in the TSHR-M22 complex but only 6 hydrogen bonds in the FSHR-FSH complex. In addition, M22 interacts with residues in the C terminal part of the TSHR LRD but the equivalent residues in the C terminal part of the FSHR LRD are not involved in FSH binding (1,9). Also, the conformational change in FSH which occurs on binding to the FSHR (9) is not seen in M22 on binding to the TSHR.

Blocking and stimulating TSHR autoantibodies
Although the details of how a thyroid stimulating autoantibody interacts with the TSHR LRD are now known at the atomic level, there is not as yet similar data relating to a TSHR autoantibody which blocks TSH action. The crystal structure of a mouse MAb (RSR-B2) which blocks the stimulating activity of TSH and thyroid stimulating autoantibodies is known (10). Also extensive experiments on the interactions of RSR-B2 with mutated TSHR preparations have been able to establish which amino acids on the LRD concave surface are important for RSR-B2 binding (11). These studies indicate that TSHR residues important for the blocking activity of RSR-B2 are in the N-terminal part of the LRD whereas about half of the TSHR residues important for the stimulating activity of M22 are C-terminal of all the residues important for blocking activity of RSR-B2 ie RSR-B2 interacts with residues in the N-terminal part of the TSHR LRD whereas M22 interacts with amino acids in both the N- and C-terminal parts. Although the TSHR residues important for the activities of the stimulating antibody (M22) and the blocking antibody (RSR-B2) differ, these two antibodies compete effectively with each other for binding to the TSHR and there is evidence for a considerable overlap of their respective binding sites on the TSHR (6,7,10,11). Consequently differences in the regions of the receptor recognised by the two MAbs might be responsible for their different activities.

Mechanism of TSHR activation by TSH and TSHR autoantibodies
The structural information currently available and comparative models derived from it have provided little insight as yet into how ligand binding activates the TSHR. Ligand induced dimerisation has been proposed to be an important step in activation of GPCRs in general (12) and the TSHR in particular (13,14). Analysis of the TSHR LRD-M22 complex however indicates that this does not form dimers (see above) and consequently receptor-receptor interactions involving the first 260 amino acids of the TSHR are unlikely to be important in any dimer formation. In X-ray diffraction analysis of the FSHR-FSH complex, two FSHR-FSH complexes were observed in the asymmetric unit and dimerisation of the FSHR-FSH complex was suggested (9). However this apparent difference between the two complexes may reflect differences in crystal packing.

Implications of recent structural studies
The structures of the regions of the TSHR and of the FSHR which are responsible in the main part for binding activating ligands are now known at 2.55 Å and 2.9 Å resolution respectively. Also, the details of how these receptors interact with their ligands are known at the atomic level. Careful comparison of the TSHR and FSHR structures and how they interact with their respective activating ligands should provide key insights into how the distinct specificity and functions of the two receptor-hormone systems have evolved. In terms of how ligand binding activates the TSHR (and FSHR), structural information on the entire ECD at least and its position relative to the TMD in the 3-domain structure of the receptor (unbound and bound to a ligand) is likely to be needed before this process can be better understood. There is good reason to believe that determining these structures is an attainable goal using a similar approach to that employed to solve the structure of the TSHR LRD in complex with M22. Knowledge of the TSHR-M22 interactions at the atomic level provides a basis for the rational design of small molecules which inhibit the binding of autoantibodies to the receptor. Such compounds are likely to have a higher degree of specificity for the TSHR than the low molecular weight inhibitors of TSHR TMD function which have been described (15,16). This is because the sequence similarities between the glycoprotein hormone TMDs is high (about 70%) whereas that between their LRDs is relatively low (about 40%). Inhibitors of the autoantibody-receptor interaction designed using the TSHR-M22 structure are likely to provide a new generation of drugs which will control thyroid function by targeting the actions of the autoantibodies responsible for Graves' disease with a high degree of specificity.

Acknowledgements
R Núñez Miguel carried out comparison of the TSHR and FSHR structures and their binding arrangements with TSH, M22 and FSH respectively. We thank Carol James for expert preparation of the manuscript.

Conflict of interest statement
RSR Ltd is a developer of medical diagnostics including kits for measuring thyroid autoantibodies.