Structural models and considerations on the COA6, COX18 and COX20 factors that assist assembly of human cytochrome c oxidase subunit II

Abstract

The soluble domain of cytochrome c oxidase subunit II (COX2), located in the outer side of the inner mitochondrial membrane, contains a binuclear copper site (CuA) through which electrons flow from cytochrome c to the core of the oxidase where oxygen reduction takes place. Being COX2 encoded in the mitochondrial genome, newly synthesized protein undergoes maturation steps in which it is translocated through and inserted into the inner mitochondrial membrane, and copper ions are loaded to form the CuA site. These steps are ensured by several protein factors in a complex pathway that is not fully understood, including copper-loading and disulfide-reduction proteins plus chaperones that assist proper membrane insertion. While the structure and function of copper-loading and disulfide-reducing proteins Sco1 and Sco2 have been quite studied at atomistic level, the latest biological studies have uncovered roles for other proteins that are not yet much understood at the structural level. In particular, recent experiments showed that membrane protein COX18 is a membrane-protein insertase for COX2, whereas membrane protein COX20 is a chaperone that stabilizes COX2 during translocation through the inner mitochondrial membrane, and soluble protein COA6 is part of the copper-loading pathway in conjunction with Sco1 and Sco2. This work reports structural models for COX18, COX20 and COA6, built judiciously from homology modeling, contact prediction-based modeling and transmembrane helix predictions, while considering the underlying biology. Implications and limitations of the models are discussed, and possible experimental routes to pursue are proposed. All models are provided as PyMOL sessions in the Supporting Information and can be visualized online at

Assessment of hard target modeling in CASP12 reveals an emerging role of alignment‐based contact prediction methods

“…More globally, alignment‐based contact prediction methods are becoming important sources of information in structural biology, used not only to model proteins but also to derive restrains for modeling complexes, to seek out possible alternative conformations, and to understand amino acid variability in proteins . In fact there are already several clear applications to specific proteins in studies where modeling at least overall topologies and interactions through contact prediction is enough to answer certain questions and drive research forward …”

“…[13][14][15][30][31][32][33][34][35][36][37][38][39] In fact there are already several clear applications to specific proteins in studies where modeling at least overall topologies and interactions through contact prediction is enough to answer certain questions and drive research forward. [40][41][42][43][44][45][46][47] We explored more broadly to what extent contact prediction methods might have driven the improvement observed in CASP12, under the premise that any alignment-based contact prediction method should lead to better models for deeper alignment depths. For this we plotted the GDTTS of the best submitted model for each target EU versus the number of sequences retrieved by an HHblits search on Uniprot for that EU, normalized by its length (Neff, actually plotted as log[1 1 Neff] because Neff is 0 for some targets, Figure 7).…”

Definition and classification of evaluation units for tertiary structure prediction in CASP12 facilitated through semi‐automated metrics

“…Upon assessment of the tertiary structure predictions and considering the content of abstracts for the CASP12 meeting, we concluded that contact predictions from alignments might be driving actual improvement in prediction performance, at least for FM targets. Such methods are actually starting to have practical use to explore the topologies of proteins for which no clear templates are available for homology modeling . Strictly speaking, our observation was that for FM targets, alignment depth (Neff, defined as the number of sequences in an alignment divided by the number of residues in the sequence) correlates positively with the GDTTS of top models, which might be for reasons other than improved contact predictions.…”

“…Such methods are actually starting to have practical use to explore the topologies of proteins for which no clear templates are available for homology modeling. [19][20][21][22][23][24][25][26] Strictly speaking, our observation was that for FM targets, alignment depth (Neff, defined as the number of sequences in an alignment divided by the number of residues in the sequence) correlates positively with the GDTTS of top models, which might be for reasons other than improved contact predictions. For example, deeper alignments might facilitate the search of remote PDB templates.…”

The Complexity of Mitochondrial Complex IV: An Update of Cytochrome c Oxidase Biogenesis in Plants

“…Membrane insertion is initiated by Oxa1 and the COX2-specific chaperone Cox20 [ 27 , 40 ]. During this process, an N-terminal extension involved in COX2 translocation is cleaved by the IMM protease Imp1 [ 41 ]. Then, Cox18, an Oxa1-like protein, together with Mss2 and Pnt1, promote the translocation of the globular domain through the IMM [ 40 ].…”