Homology Modeling of Human Thyroid Peroxidase

Template sequences were selected by querying the TPO protein sequence against the Protein Databank (PDB) using the NCBI BLAST web server (http://blast.ncbi.nlm.nih.gov/Blast.cgi). The MPO-like, EGF-like, and CCP-like domains were modelled using for templates Protein Data Bank (PDB) entries 1CXP [16 (link)], 1EMO [29 (link)], and 1VVD [30 (link)], respectively. Specifically, we used the following residues for modelling: MPO-like domain (residues 142–738 modelled using residues 167–744 from template 1CXP, 47% sequence identity), CCP-like domain (residues 740–795 modelled using residues 147–202 from template 1VVD, 35% sequence identity), and EGF-like domain (residues 798–839 modelled using residues 2162–2205 from template 1EMO, 39% sequence identity). The cholinesterase-like (ChEL) domain of thyroglobulin was modelled using the crystal structure of recombinant human acetylcholinesterase in the apo state (PDB ID: 4EY4) [31 (link)] as a template. Transmembrane regions of TPO were identified using the membrane protein prediction server TMHMM [32 (link)], which found a transmembrane helix immediately after the EGF-like domain. A putative helix-helix interaction motif was identified as reported [33 (link)]. For the generation of the transmembrane helices dimer, the structure of a GxxxG motif dimer (PDB ID: 2L2T)[34 (link)] was used as a template. 200 homology models of human TPO (residues 142–880) were built using MODELLER v9.12 [35 (link)] and sorted by their Discrete Optimized Protein Energy (DOPE) score. The inter-dimer disulfide bridge (residues 153 in 1CXP) was modelled in the TPO dimer model, as was the iron-protoporphyrin IX (heme) group. Symmetry was maintained using MODELLER symmetry restraints between the two chains. Model plausibility was assessed by visual inspection of domain juxtaposition, and by determining the quality of the models using three separate assessments of model quality: (1) VERIFY 3D [36 (link)], which determines the compatibility of an atomic model with its own amino acid sequence by assigning a structural class based on its location and environment (alpha, beta, loop, polar, nonpolar etc.) and comparing the results to good structures. All models pass the VERIFY 3D test; (2) MolProbity [37 (link)], a widely used method (used by the PDB, for example) to asses the stereochemical quality of structures; (3) QMEAN Server for model quality estimation [38 (link)] (http://swissmodel.expasy.org/qmean), a method for the estimation of the absolute quality of individual protein structure models which is independent of protein size.