We have studied glycosylation patterns in glycoprotein hormones (GPHs) and glycoprotein hormone receptor (GPHR) extracellular domains (ECD) from different species to identify areas not glycosylated that could be involved in intermolecular or intramolecular interactions. Comparative models of the structure of the TSHR ECD in complex with TSH and in complex with TSHR autoantibodies (M22, stimulating and K1-70, blocking) were obtained based on the crystal structures of the FSH-FSHR ECD, M22-TSHR leucine-rich repeat domain (LRD) and K1-70-TSHR LRD complexes. The glycosylation sites of the GPHRs and GPHs from all species studied were mapped on the model of the human TSH TSHR ECD complex. The areas on the surfaces of GPHs that are known to interact with their receptors are not glycosylated and two areas free from glycosylation, not involved in currently known interactions, have been identified. The concave faces of GPHRs leucine-rich repeats 3–7 are free from glycosylation, consistent with known interactions with the hormones. In addition, four other non-glycosylated areas have been identified, two located on the receptors’ convex surfaces, one in the long loop of the hinge regions and one at the C-terminus of the extracellular domains. Experimental evidence suggests that the non-glycosylated areas identified on the hormones and receptors are likely to be involved in forming intramolecular or intermolecular interactions.

Binding of a new thyroid-stimulating human monoclonal autoantibody (MAb) K1–18 to the TSH receptor (TSHR) leucine-rich domain (LRD) was predicted using charge–charge interaction mapping based on unique complementarities between the TSHR in interactions with the thyroid-stimulating human MAb M22 or the thyroid-blocking human MAb K1–70. The interactions of K1–18 with the TSHR LRD were compared with the interactions in the crystal structures of the M22–TSHR LRD and K1–70–TSHR LRD complexes. Furthermore, the predicted position of K1–18 on the TSHR was validated by the effects of TSHR mutations on the stimulating activity of K1–18. A similar approach was adopted for predicting binding of a mouse thyroid-blocking MAb RSR-B2 to the TSHR. K1–18 is predicted to bind to the TSHR LRD in a similar way as TSH and M22. The binding analysis suggests that K1–18 light chain (LC) mimics binding of the TSH-α chain and the heavy chain (HC) mimics binding of the TSH-β chain. By contrast, M22 HC mimics the interactions of TSH-α while M22 LC mimics TSH-β in interactions with the TSHR. The observed interactions in the M22–TSHR LRD and K1–70–TSHR LRD complexes (crystal structures) with TSH–TSHR LRD (comparative model) and K1–18–TSHR LRD (predictive binding) suggest that K1–18 and M22 interactions with the receptor may reflect interaction of thyroid-stimulating autoantibodies in general. Furthermore, K1–70 and RSR-B2 interactions with the TSHR LRD may reflect binding of TSHR-blocking autoantibodies in general. Interactions involving the C-terminal part of the TSHR LRD may be important for receptor activation by autoantibodies.

Determination of the full-length thyroid-stimulating hormone receptor (TSHR) structure by cryo-electron microscopy (cryo-EM) is described. The TSHR complexed with human monoclonal TSHR autoantibody K1-70™ (a powerful inhibitor of TSH action) was detergent solubilised, purified to homogeneity and analysed by cryo-EM. The structure (global resolution 3.3 Å) is a monomer with all three domains visible: leucine-rich domain (LRD), hinge region (HR) and transmembrane domain (TMD). The TSHR extracellular domain (ECD, composed of the LRD and HR) is positioned on top of the TMD extracellular surface. Extensive interactions between the TMD and ECD are observed in the structure, and their analysis provides an explanation of the effects of various TSHR mutations on TSHR constitutive activity and on ligand-induced activation. K1-70™ is seen to be well clear of the lipid bilayer. However, superimposition of M22™ (a human monoclonal TSHR autoantibody which is a powerful stimulator of the TSHR) on the cryo-EM structure shows that it would clash with the bilayer unless the TSHR HR rotates upwards as part of the M22™ binding process. This rotation could have an important role in TSHR stimulation by M22™ and as such provides an explanation as to why K1-70™ blocks the binding of TSH and M22™ without activating the receptor itself.

Determination of the structure of the extracellular domain of human thyroid peroxidase (hTPO) by cryo-electron microscopy (cryo-EM) is described. TPO, purified to homogeneity was complexed with the hTPO monoclonal autoantibody 2G4 Fab and also with a mouse monoclonal TPO antibody 4F5 Fab (which competes with autoantibody binding to TPO). Both complexes were analysed by cryo-EM. The two structures (global resolution 3.92 and 3.4 Å for the 2G4 complex and 4F5 complex, respectively) show TPO as a monomer with four domains; the N-terminal domain, the peroxidase domain (POD), the complement control protein (CCP)-like domain and the epidermal growth factor-like domain which are all visible in the structures. The relative positions of the domains are fixed with a disulphide bond between cysteine residues Cys146 in the POD and Cys756 in the CCP domain preventing significant flexibility of the molecule. The entrance to the enzyme active site, the haem group and the calcium binding site are clearly visible on the opposite side of the TPO molecule from the 2G4 and 4F5 binding sites. Extensive interactions are seen between TPO and the two antibodies which both bind to distinct epitopes on the POD domain, including some residues in the immunodominant region B mainly via different residues. However, the epitopes of the two antibodies contain three shared TPO residues. This is the first high-resolution structure of TPO to be reported and it should help guide the development of new inhibitors of TPO enzyme activity for therapeutic applications.

A complex of the TSH receptor extracellular domain (amino acids 22–260; TSHR260) bound to a blocking-type human monoclonal autoantibody (K1-70) was purified, crystallised and the structure solved at 1.9 Å resolution. K1-70 Fab binds to the concave surface of the TSHR leucine-rich domain (LRD) forming a large interface (2565 Ų) with an extensive network of ionic, polar and hydrophobic interactions. Mutation of TSHR or K1-70 residues showing strong interactions in the solved structure influenced the activity of K1-70, indicating that the binding detail observed in the complex reflects interactions of K1-70 with intact, functionally active TSHR. Unbound K1-70 Fab was prepared and crystallised to 2.22 Å resolution. Virtually no movement was observed in the atoms of K1-70 residues on the binding interface compared with unbound K1-70, consistent with ‘lock and key’ binding. The binding arrangements in the TSHR260–K1-70 Fab complex are similar to previously observed for the TSHR260–M22 Fab complex; however, K1-70 clasps the concave surface of the TSHR LRD in approximately the opposite orientation (rotated 155°) to M22. The blocking autoantibody K1-70 binds more N-terminally on the TSHR concave surface than either the stimulating autoantibody M22 or the hormone TSH, and this may reflect its different functional activity. The structure of TSHR260 in the TSHR260–K1-70 and TSHR260–M22 complexes show a root mean square deviation on all C_α atoms of only 0.51 Å. These high-resolution crystal structures provide a foundation for developing new strategies to understand and control TSHR activation and the autoimmune response to the TSHR.

The crystal structures of the thyroid-stimulating hormone receptor (TSHR) leucine-rich repeat domain (amino acids 22–260; TSHR260) in complex with a stimulating human monoclonal autoantibody (M22^TM) and in complex with a blocking human autoantibody (K1-70™) have been solved. However, attempts to purify and crystallise free TSHR260, that is not bound to an autoantibody, have been unsuccessful due to the poor stability of free TSHR260. We now describe a TSHR260 mutant that has been stabilised by the introduction of six mutations (H63C, R112P, D143P, D151E, V169R and I253R) to form TSHR260-JMG55^TM, which is approximately 900 times more thermostable than wild-type TSHR260. These six mutations did not affect the binding of human TSHR monoclonal autoantibodies or patient serum TSHR autoantibodies to the TSHR260. Furthermore, the response of full-length TSHR to stimulation by TSH or human TSHR monoclonal autoantibodies was not affected by the six mutations. Thermostable TSHR260-JMG55^TM has been purified and crystallised without ligand and the structure solved at 2.83 Å resolution. This is the first reported structure of a glycoprotein hormone receptor crystallised without ligand. The unbound TSHR260-JMG55^TM structure and the M22 and K1-70 bound TSHR260 structures are remarkably similar except for small changes in side chain conformations. This suggests that neither the mutations nor the binding of M22^TM or K1-70^TM change the rigid leucine-rich repeat domain structure of TSHR260. The solved TSHR260-JMG55^TM structure provides a rationale as to why the six mutations have a thermostabilising effect and provides helpful guidelines for thermostabilisation strategies of other soluble protein domains.