COA6 is a mitochondrial complex IV assembly factor critical for biogenesis of mtDNA-encoded COX2

Abstract

Biogenesis of complex IV of the mitochondrial respiratory chain requires assembly factors for subunit maturation, co-factor attachment and stabilization of intermediate assemblies. A pathogenic mutation in COA6, leading to substitution of a conserved tryptophan for a cysteine residue, results in a loss of complex IV activity and cardiomyopathy. Here, we demonstrate that the complex IV defect correlates with a severe loss in complex IV assembly in patient heart but not fibroblasts. Complete loss of COA6 activity using gene editing in HEK293T cells resulted in a profound growth defect due to complex IV deficiency, caused by impaired biogenesis of the copper-bound mitochondrial DNA-encoded subunit COX2 and subsequent accumulation of complex IV assembly intermediates. We show that the pathogenic mutation in COA6 does not affect its import into mitochondria but impairs its maturation and stability. Furthermore, we show that COA6 has the capacity to bind copper and can associate with newly translated COX2 and the mitochondrial copper chaperone SCO1. Our data reveal that COA6 is intricately involved in the copper-dependent biogenesis of COX2.

Introduction

Cytochrome c oxidase (complex IV; CIV), or mitochondrial respiratory chain complex IV, catalyzes the transfer of electrons from intermembrane space cytochrome c to molecular oxygen in the matrix and as a consequence contributes to the proton gradient involved in mitochondrial ATP synthesis. Complex IV dysfunction is a significant cause of human mitochondrial disease (1). In humans, complex IV is composed of 14 subunits, with the core inner-membrane subunits COX1, COX2 and COX3 encoded by mitochondrial DNA (mtDNA). COX1 and COX2 coordinate two heme (a, a₃) and two copper (Cu_A and Cu_B) cofactors to enable electron transfer reactions. The remaining 11 subunits are encoded by nuclear DNA and, following cytosolic translation, are imported into the organelle where they assemble with the complex IV subunits (2–4).

The assembly of mammalian complex IV has been proposed to follow in a largely linear fashion of defined subcomplexes (S1–S4) with COX1 forming the core (S1) and engaging with COX4 and 5a (S2) before assembling with COX2 and 3 and other subunits (S3) prior to formation of the full complex (S4), which then associates with other complexes of the respiratory chain in supercomplexes (5–9). Over the past years, a large number of assembly factors involved in complex IV biogenesis have been identified, particularly in yeast complex IV, which differs from its human counterpart by being composed of only 11 subunits (7,8). Nonetheless, many of the yeast assembly factors have functional counterparts in mammals, and mutations in a number of these have been identified in patients with complex IV deficiency and harboring mitochondrial disease (1).

Mutations in COA6 (c1orf31) were identified in a patient with a complex IV enzyme defect in the heart tissue (10). The patient was diagnosed with hypertrophic obstructive cardiomyopathy at 6 months of age, displayed decreased muscle bulk and tone, was losing weight and had increasing weakness and lethargy, and died at 18 months of age from cardiac failure. Two mutations—p.W59C and p.E87*—were identified, although the role of COA6 in complex IV deficiency was not established. More recently, an additional patient with complex IV deficiency and displaying a neonatal cardiomyopathy was identified to have a mutation in COA6 with a p.W66R mutation. A study in yeast identified COA6 as specifically involved in complex IV biogenesis (11), and this was supported in a zebrafish model (12). Moreover, as COA6 is an intermembrane space protein and contains a conserved CX₉CX_nCX₁₀C motif, it was suggested to play a role in copper homeostasis. Indeed, yeast cells devoid of COA6 were found to display reduced growth defects upon copper supplementation (12).

Here, we investigate the role of COA6 in complex IV assembly using the original patient cells and in HEK293T cells containing a non-functional COA6 protein resulting from transcription activator-like effector nuclease (TALEN)-mediated gene editing. We find that mitochondrial COA6 associates with COX2 and is crucial for its maturation and likewise, complex IV biogenesis. In addition, we show that recombinant COA6 has the capacity to bind copper with high affinity. Moreover, we show that the p.W59C mutation in COA6 does not affect copper binding or import of the protein into mitochondria, but it affects the maturation and stability of the protein. The additional finding of COA6 interacting with the copper chaperone SCO1 indicates that COA6 is intrinsically involved in the copper delivery process for COX2.

Results

COA6 mutation results in a tissue-specific complex IV defect

From next generation sequence analysis of patients with mitochondrial disorders, mutations in COA6 were identified in a patient with neonatal cardiomyopathy and complex IV enzyme deficiency (10). The patient had compound heterozygous mutations in COA6 leading to one allele encoding COA6 with an amino acid substitution (p.W59C) and the other allele encoding an early termination (p.E87*), the latter mutation likely preventing protein biogenesis due to loss of the conserved cysteine residues (13). To investigate whether the mutations in COA6 led to defects in complex IV assembly, mitochondria isolated from heart biopsy and skin fibroblasts from the patient and a control unaffected individual were analyzed by blue native polyacrylamide gel electrophoresis (BN-PAGE) and western blotting. As can be seen in Figure 1A, complex IV was virtually absent from patient heart mitochondria but was present in the patient's fibroblast mitochondria at similar levels to control. The levels of complex I were similar in all cases, although the migration pattern was somewhat affected in patient heart mitochondria, most likely due to the lack of complex IV in its supercomplex forms.

Figure 1.

Loss of COA6 results in a specific complex IV deficiency. (A) Mitochondria were solubilized in digitonin and subjected to BN-PAGE and immunoblot analysis using antibodies against NDUFA9 (CI) and COX1 (CIV). SC, supercomplexes consisting of CI, CIII and CIV. (B) Schematic outlining strategy for TALEN gene editing of COA6 in HEK293T cells. Edited alleles detected in two different clonal cell lines are shown. Gray highlight indicates TALEN binding sites, and underline indicates mutated patient residue p.W59. (C) Mitochondria were solubilized in either digitonin (left panel) or Triton X-100 (right panel) and subjected to BN-PAGE and immunoblotting with antibodies against NDUFA9 (CI), SDHA (CII), CORE1 (CIII) and COX1 (CIV). #, Subcomplex containing COX1; TX-100, Triton X100. (D) Left panel, oxygen consumption was measured over a 170 min period. N = 3, standard error of the mean (SEM). Right panel, equal numbers of HEK293T and ΔCOA6-1 cells were cultured in glucose-containing DMEM for the indicated time. Cell density was measured using a SRB assay (50). N = 6, SEM. (E) Mitochondrial membrane potential in ΔCOA6-1 cells measured by fluorescence-activated cell sorting (FACS). Data are shown as the means ± SEM of the difference between geometric means of TMRM fluorescence in mock- and CCCP-treated cells. N = 3 of 10 000 events per replicate. AFU, arbitrary fluorescence units.

Generation of a human COA6 knockout cell line using TALENs

To obtain a human cell model of a COA6 defect, we used TALENs in HEK293T cells to generate targeted deletions in COA6 at the region encompassing the conserved mutation. Clones deficient in their ability to grow on galactose media, and hence likely to be respiration deficient, were selected for further analysis (14). In genomic analysis, the COA6 alleles in two clones (termed ΔCOA6-1 and ΔCOA6-2) were found to contain deletions leading to loss of residues including the first cysteine residue in the conserved CX₉CX_nCX₁₀C motif and/or deletions that led to early termination of translation (Fig. 1B). Analysis of respiratory chain complexes using digitonin or Triton X-100 solubilization to visualize supercomplex or holoenzyme forms, respectively, revealed that both ΔCOA6 cell lines harbored a specific complex IV deficiency. It should be noted that the presence of a small amount of COX1 was found in a supercomplex form in both ΔCOA6 mitochondria, but the holoenzyme form was not observed following Triton X-100 solubilization (Fig. 1C, compare lanes 11 and 12 with 23 and 24). Similar findings have been reported in the analysis of complex IV deficient patient fibroblasts and cybrids and may represent an alternate mechanism for assembly (15–17). Next, we assessed the respiration defect in ΔCOA6 cells. For these and subsequent measurements, we concentrated on ΔCOA6-1 cells. As the ΔCOA6 cells could not grow on galactose media, they were grown on glucose media only. The oxygen consumption rate (OCR) showed a severe defect in ΔCOA6 cells when compared with controls (Fig. 1D, left panel). ΔCOA6 cells also displayed a significant overall growth defect (Fig. 1D, right panel). This growth defect may be due to the increased lactic acid levels in the growth media of ΔCOA6 cells (data not shown) and also the finding that the mitochondrial membrane potential of ΔCOA6 cells were impaired in comparison with control HEK293T cells (Fig. 1E).

Mitochondrial import and biogenesis of COA6

COA6 is imported into the mitochondrial intermembrane space via the Mia40 pathway (11). The first cysteine residue in Mia40 substrates has been shown to be important for mitochondrial import (13). Given this, we asked whether the introduction of an additional cysteine residue adjacent to the first cysteine of the canonical CX₉CX_nCX₁₀C motif (as seen in the patient) would cause defects in targeting and assembly. In this case, we used the well-characterized Saccharomyces cerevisiae model system for in vitro protein import with yeast Coa6 (yCoa6) as the substrate. Both yCoa6 and yCoa6 with a substitution in the conserved p.W59 residue (yCoa6^W26C) could be imported into a protease-protected location within mitochondria, although the levels of yCoa6^W26C were lower (Fig. 2A). Import of yCoa6 and subsequent blocking of free thiol groups by iodoacetamide (IAA) trapped a transient disulfide-linked Coa6-Mia40 intermediate that is sensitive to addition of the reducing agent β-mercaptoethanol (βME) (Fig. 2B). This intermediate was similar to the previously characterized Tim9-Mia40 intermediate (Fig. 2C) (13). Coa6 with a mutation in either the first cysteine (yCoa6^C25S), the conserved human p.W59 residue (yCoa6^W26C) or both residues (Coa6^C25S/W26C) all formed a protease-protected Mia40 intermediate (Fig. 2D). While accumulation of mature yCoa6 was impaired upon mutation of the first cysteine as expected (Fig. 2D, compare lanes 1–3 with 4–6), we observed a further reduction for both mutants harboring the conserved human p.W59C mutation (lanes 7–12). Taken together, these results point to a defect subsequent to mitochondrial protein import and may underline a potential reason for pathogenesis.

Figure 2.

Mutated COA6 retains partial functionality. (A) In vitro translated [³⁵S]yCoa6 or [³⁵S]yCoa6^W26C was imported into isolated mitochondria for the indicated times, and mitochondrial proteins were analyzed by non-reducing SDS-PAGE and autoradiography. mock, Import reaction lacking mitochondria. (B) In vitro translated [³⁵S]yCoa6 was imported into mitochondria treated with IAA or mitochondria treated with IAA followed by post-import βME treatment, and analyzed as for A. *, Non-specific band. (C) In vitro translated [³⁵S]yCoa6 or [³⁵S]Tim9 were imported into mitochondria treated with IAA. Following proteinase K treatment, samples were treated as for A. ¤, mature Coa6 lanes 1–4 and Tim9, lane 5. (D) In vitro translated yCoa6, yCoa6^C25S, yCoa6^W26C and yCoa6^C25S/W26C were imported into mitochondria treated with IAA, subjected to proteinase K and analyzed by SDS-PAGE and autoradiography. Yeast p.C25 corresponds to human p.C58; yeast p.W26 to human p.W59. (E) Control or ΔCOA6-1 cells expressing COA6-FLAG, COA6^W59C-FLAG or COA6^Δ50–58-FLAG were grown in either glucose- or galactose-containing media. Mitochondria were isolated, subjected to BN-PAGE and immunoblotted with antibodies against COX4 (CIV). Antibodies against ATP5A (CV) and UQCRC2 (CIII) were used as loading controls.

Complementation of COA6 knockout cell line with wild-type and mutant COA6

We next asked whether transient overexpression of the human COA6 mutant could rescue ΔCOA6 cells. For this and other analyses, we utilized the canonical isoform of human COA6, isoform 1. As expected, ectopic expression of COA6 enabled ΔCOA6 cells to grow on galactose media, and restoration of the complex IV holoenzyme was observed (Fig. 2E). Likewise, overexpression of the patient mutation COA6^W59C in ΔCOA6 cells was able to only partially restore complex IV assembly (Fig. 2E, compare lanes 9 and 10, and 12 and 13). The levels of assembled complexes III and V served as loading controls, since proteomic analysis of ΔCOA6-1 cells showed the levels of their subunits to be largely unaltered relative to HEK293T cells (Supplementary Material, Table S1, and see below). It is, therefore, likely that the mutant COA6 seen in the patient retains partial activity but is hypomorphic in tissues of high energy demand. In contrast, overexpression of a mutant COA6 allele generated by our TALEN approach (COA6^Δ58–61, Fig. 1B) was not able to complement the loss of COA6 (Fig. 2E, lane 11).

COA6 associates with newly translated COX2 and is required for COX2 maturation

We next addressed the nature of the complex IV assembly defect. Analysis of digitonin-solubilized mitochondria subjected to BN-PAGE and immunoblot analysis revealed that COX1 and COX4 were absent from the holoenzyme in ΔCOA6 cells but were instead present in similar-sized subcomplexes and, to a lesser extent, in supercomplex forms (Fig. 3A). The subcomplex observed most likely represents the early intermediate termed S2 (5,9,18). To quantitatively analyze changes in the mitochondrial proteome following loss of COA6, we performed quantitative mass spectrometry (MS) using stable isotope labeling by amino acids in culture (SILAC) (19). As can be seen in Figure 3B, the most down-regulated subunits in ΔCOA6 cells relative to the control HEK293T parental cell line were complex IV subunits COX2, COX7A, COX6B and the recently identified complex IV subunit NDUFA4 (3,4). COX3 was not quantified due to the low number of peptides detected. COA6 was also detected at low levels indicating that while the deletions cause a loss of function, it is not rapidly turned over. The levels of COX1, COX4, COX5a and the copper chaperones SCO1 and SCO2 did not significantly alter, while a number of other proteins were up-regulated (Supplementary Material, Table S1) including complex IV assembly factors COA4, COX17 and CMC1.

Figure 3.

COA6 is required for maturation of mtDNA-encoded COX2. (A) HEK293T and ΔCOA6 mitochondria solubilized in digitonin were subjected to BN-PAGE and immunoblot analysis using antibodies against COX1 and COX4. SC, supercomplex containing CI, CIII and CIV; S2, intermediate CIV assembly complex S2. *, Non-specific band. (B) Mitochondria from ‘heavy’ or ‘light’ amino acid labeled HEK293T and ΔCOA6-1 cells were mixed and analyzed by LC-MS as described in the Materials and Methods section. Mean log₂ HEK293T/ΔCOA6-1 ratios (normalized) of proteins quantified in at least two replicates (including a label switch; N = 3) were plotted against their P-values (−log₁₀). The horizontal line corresponds to a P-value of 0.05, the vertical lines, mean HEK293T/ΔCOA6-1 ratios < −1.75 (left) or >1.75 (right). The full dataset can be found in the Supplementary Material, Table S1. (C) MtDNA-encoded subunits were radiolabeled in control or ΔCOA6-1 cells and chased for the indicated times. Isolated mitochondria were analyzed by SDS-PAGE and autoradiography. (D) MtDNA-encoded subunits were radiolabeled in HEK293T cells for 2 h, following which isolated mitochondria were solubilized in Triton X-100 and membrane protein complexes bound to COX1 or COX2 cross-linked Protein A-sepharose. Eluted proteins were analyzed by SDS-PAGE and autoradiography. (E) MtDNA-encoded subunits were radiolabeled in HEK293T and COA6-FLAG cells for 1 h, following which isolated mitochondria were solubilized in 1% digitonin and bound to anti-FLAG affinity gel. Eluted proteins were analyzed by SDS-PAGE and autoradiography. Total, 5%; elution, 100%. (F) MtDNA-encoded subunits were radiolabeled in HEK293T and ΔCOA6-1 cells as for D, following which isolated mitochondria were solubilized in digitonin and analyzed by 2D BN/SDS-PAGE and autoradiography. ◊, Intermediate complex containing COX3.

Next, we assessed the translation and stability of mtDNA-encoded subunits in control and ΔCOA6 cells using pulse-chase assays. Strikingly, the levels of COX2/3 were specifically reduced in ΔCOA6 mitochondria (Fig. 3C). While the slower migrating band was the least stable of the two, it is unclear from the literature whether this represents COX2 or COX3. We, therefore, assessed this by immunoprecipitation (co-IP) of COX2 along with COX1 as a control (Fig. 3D). From this analysis, COX2 was shown to migrate slower than COX3 under our sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) conditions. Next, we sought to determine whether COA6 can associate with newly translated COX2. FLAG-tagged COA6 was stably expressed in ΔCOA6 cells (COA6-FLAG), the cells subjected to a short (1 h) pulse with [³⁵S]-methionine, and mitochondrial protein was immunoprecipitated using FLAG antibodies. As can be seen in Figure 3E, newly translated COX2 was specifically pulled down with COA6-FLAG.

To determine the timing and consequences of COA6 loss on complex IV assembly, we performed pulse-chase analysis. In this case, samples were subjected to BN-PAGE in the first dimension and SDS-PAGE in the second to assess the location of individual subunits within complexes (Fig. 3E) (20). In mitochondria from control cells, COX1 formed the early intermediate (S2) following the 2 h pulse, and after a 24 h chase, it assembled into the holoenzyme and supercomplex forms as expected. COX2 and COX3 were also found in distinct low-molecular-weight subassemblies before migrating with COX1 into fully assembled complex IV. However, in ΔCOA6 mitochondria, COX1 was again observed in the S2 intermediate, but it did not progress any further. Only COX3 was clearly detected at the 2 h pulse (0 h chase), and it appeared to be turned over during the chase period. These results indicate that loss of COA6 affects the early steps in the assembly of complex IV, specifically involving COX2 biogenesis.

COA6 associates with the copper chaperone SCO1

To investigate COA6 interactions, we performed co-IP analysis of FLAG-tagged COA6 expressed in ΔCOA6 cells. In this case, we transiently overexpressed COA6-FLAG and performed SILAC and subsequent quantitative MS of eluted proteins. Flag-tagged COA6^Δ50–58, representing the inactive mutated allele in our ΔCOA6-1 cell line (Figs. 1B and 2E), served as a negative control. A set of enriched proteins were identified (Fig. 4A and Supplementary Material, Table S2) and included complex IV subunits COX4 and COX5A, both of which are found in the S2 intermediate, and the copper chaperone SCO1 which is involved in COX2 biogenesis (21,22). Interestingly, a number of matrix/inner-membrane proteins involved in mitochondrial RNA binding and translation were also pulled down with COA6-FLAG, although the relationship of this to complex IV biogenesis is not clear. To confirm the SCO1 interaction, we repeated the COA6 co-IP and performed SDS-PAGE and immunoblotting. As can be seen in Figure 4B, SCO1 was specifically enriched with COA6-FLAG relative to control HEK293T mitochondria. COX2 and potentially COX4 were also enriched in the COA6-FLAG elution relative to control, but not COX1 or the control complex I subunit NDUFA9. Taken together, these results indicate that COA6 can be found in a complex or complexes with complex IV subunits and SCO1.

Figure 4.

COA6 interacts with copper chaperone SCO1. (A) Mitochondria from ‘heavy’ or ‘light’ amino acid labeled ΔCOA6-1 cells expressing COA6-FLAG or COA6^Δ50–58-FLAG were solubilized in 1% digitonin and bound to anti-FLAG affinity gel. Elutions were mixed and analyzed by LC-MS as described in Materials and Methods. The means of normalized COA6-FLAG/COA6^Δ50–58-FLAG ratios (log₂) of proteins quantified in at least two replicates (including a label switch; N = 3) were plotted against their P-values (−log₁₀). Thresholds were set at P < 0.05 and mean COA6-FLAG/COA6^Δ50–58-FLAG enrichment ratios >2. The full dataset can be found in the Supplementary Material, Table S2. (B) Mitochondria from HEK293T and COA6-FLAG cells were solubilized in 1% digitonin and bound to anti-FLAG affinity gel. Elutions were analyzed by SDS-PAGE and western blotting. Total, 5%; elution, 100%.

COA6 has the ability to bind copper

Given the link between COA6 and copper chaperones, reports that COA6-deficient yeast cells can be rescued by copper supplementation (12), and that levels of complex IV in patient fibroblasts with the different p.W66R mutation could be partially rescued by addition of copper to the growth medium (17), we asked if human cells lacking functional COA6 could also be rescued by copper. However, neither growth rate (Supplementary Material, Fig. S1A) nor complex IV assembly (Supplementary Material, Fig. S1B) was altered upon addition of copper to the growth media. The reason for the discrepancy between yeast and human ΔCOA6 models is unclear; however, this could be due to differences in complex IV assembly between the two organisms (8,9). We sought to test whether COA6 directly binds copper using recombinantly expressed protein in an in vitro assay. As full-length human COA6 and COA6^W59C were insoluble following bacterial expression, we expressed and purified isoform 3, which lacks the N-terminal extension of 46 residues but encompasses the most well-conserved region of COA6 (Fig. 5A). It should also be noted that, during our proteomic analyses, we failed to detect peptides unique to COA6 isoforms 1 and 2 (Supplementary Material, Fig. S2), suggesting isoform 3 to be dominant in HEK293T cells. Importantly, when stably expressed in ΔCOA6 cells, COA6 isoform 3 and the corresponding COA6^W59C mutant efficiently rescued complex IV assembly (Fig. 5B) and growth on galactose media (data not shown). Size-exclusion analysis revealed purified recombinant COA6 co-eluting with copper (Supplementary Material, Fig. S3A). We quantified Cu(I) binding of purified COA6 using competition experiments against the copper binding ligand bathocuproine disulfonate anion (Bcs) (Supplementary Material, Fig. S3B) (23,24). COA6 exhibited strong Cu(I) binding (Fig. 5C; K_D = 10^−17.2). We extended the analysis using up to a 2.4 molar excess of COA6 relative to Cu and found no change in the titration curve (Supplementary Material, Fig. S3C). Interestingly, this dissociation constant is similar to SCO1 (K_D ≈ 10⁻¹⁷) (25). COA6^W59C also bound copper strongly (Fig. 5C) but with a slightly higher dissociation constant (K_D = 10^−17.8) in comparison with COA6.

Figure 5.

COA6 can bind copper. (A) ClustalW alignment of protein sequences for human COA6 isoforms 1–3 against various organisms. A grayscale shade is used to indicate sequence conservation. (B) Mitochondria were isolated from control, ΔCOA6-1 or ΔCOA6-1 cells expressing isoform 1 or 3 of COA6-FLAG and COA6^W59C-FLAG, subjected to BN-PAGE and immunoblotted with antibodies against COX4 (Complex IV) or NDUFA9 (Complex I). (C) Known concentrations (0–26 µM) of recombinant COA6 and COA6^W59C were added to a complex of 24 µM Cu(I)(Bcs)₂, and the absorbance was measured at 483 nm. The data were analyzed using plots of [Cu(I)Bcs₂]³⁻ versus COA6:Cu ratio, and the data fit using Equation (2) (23). Bcs, bathocuproine disulfonate anion. The 350–550 nm absorbance spectrum for COA6 can be found in the Supplementary Material, Figure S3B. (D) Mitochondria isolated from HEK293T, ΔCOA6-1, COA6-FLAG or COA6^W59C-FLAG cells were solubilized in 8 M urea, 50 mm Tris pH 7.4 and incubated as indicated with or without 10 mm TCEP for 30 min at 50°C. Where indicated, free thiol groups were modified with 15 mm AMS by an additional incubation for 30 min at 50°C. Samples were analyzed by SDS-PAGE and western blotting with antibodies specific to the FLAG epitope. COA6-FLAG migrates at ∼16 kDa due to the presence of a linker and tag. (E) Recombinant, untagged COA6, COA6^W59C loaded with Cu(I) or COA6^W59C pre-incubated with βME were analyzed by gel filtration. Approximate molecular weights are indicated based on concurrently run standards. βME, β-mercaptoethanol. (F) Proteins prepared as for E were analyzed by SDS-PAGE in the presence or absence of βME as indicated.

It has been suggested that copper may bind via the CX₉CX_nCX₁₀C motif (12); however, such a scenario would require the presence of reduced cysteine residues for Cu(I) association. We, therefore, sought to assess whether the conserved cysteines in COA6 are present in a reduced or oxidized state in mitochondria. Modification of each free cysteine residue by 4-acetamido-4′-maleimidylstilbene-2,2′-disulfonic acid (AMS) leads to a shift in mobility of the protein by ∼500 Da per reduced cysteine (26–28). Importantly, we performed our experiments using mitochondria isolated from ΔCOA6 stably complemented with isoform 3 of COA6 (Fig. 5B), which contains only the four conserved cysteine residues. As can be seen in Figure 5D, pre-incubation of the mitochondrial lysate with tris(2-carboxyethyl)phosphine (TCEP) prior to AMS treatment results in a mobility shift of COA6 corresponding to modification of all four cysteines (Fig. 5D, compare lanes 4 and 8). In the absence of a reducing agent, AMS modified COA6 resulting in its mobility between totally modified and unmodified COA6 forms (Fig. 5D, compare lanes 7 and 8), consistent with mitochondrial COA6 having two reduced and two oxidized cysteine residues. Finally, we performed AMS modification using mitochondria isolated from ΔCOA6 cells complemented with the W59C mutant. An additional AMS shift of COA6^W59C-FLAG was observed consistent with the presence of the additional free cysteine (Fig. 5, lanes 9 and 10).

How might the presence of an additional free cysteine impact COA6^W59C function and/or stability? Surprisingly, we found that COA6^W59C but not wild-type COA6 spontaneously forms oligomers that are sensitive to treatment with reducing agent βME (Fig. 5E and F) and form independent of the protein being loaded with copper. We conclude that the COA6^W59C mutation most likely impairs optimal folding/maturation leading to defects in complex IV biogenesis.

Discussion

Based on a number of findings, we propose that COA6 is specifically involved in the biogenesis of the mtDNA-encoded complex IV subunit COX2: (1) loss of COA6 causes specific defects in the maturation of the COX2, resulting in its degradation thus blocking complex IV assembly; (2) loss of COA6 results in up-regulation of CMC1 and COX17, both of which are involved in copper delivery to complex IV (29,30); (3) COA6 co-immunoprecipitates with SCO1, a central component mediating copper delivery to COX2; and (4) recombinant COA6 can directly bind copper. From our proteomic analysis and pulse-chase assays, we found that the loss of COA6 does not directly affect the synthesis of COX1, but rather this subunit stalls in an early intermediate subassembly. Furthermore, COX17, which is up-regulated, is required for metalation of COX2 (31). Up-regulation of COX17 has also been observed in SCO1 and SCO2 patients (32). The levels of other complex IV subunits that are integrated into the complex IV assembly pathway at steps subsequent to addition of COX2, including COX7A and COX6B, were also reduced in ΔCOA6 cells, most likely due to their turnover as a result of blocked complex IV assembly. The levels of NDUFA4, a protein originally reported as a subunit of complex I but now known to specifically assemble into complex IV (3,4), were also reduced in ΔCOA6 cells.

How the COA6^W59C mutation found in our patient affects complex IV assembly is unclear, since the mutant binds copper more tightly (K_D = 10^−17.8) than the wild-type protein (K_D = 10^−17.2). We believe that this is an unlikely cause of pathogenesis, however, as overexpression of COA6^W59C complements ΔCOA6 cells. An alternative explanation is the reduced stability or degradation of COA6^W59C, which may be caused by increased oxidation of the mutant protein. Interestingly, the COA6^W59C mutation in COA6 did not lead to defects in fibroblasts but showed a strong loss in complex IV assembly in the heart tissue. This prominent effect resembles that of other patients with complex IV defects in copper metabolism caused by SCO2 mutations (33–36); however, it is difficult to explain given the recent description of a different COA6 mutation (p.W66R) that shows a clear defect in fibroblasts (17). One explanation may be that COA6^W59C is hypomorphic in tissues of high energy demand such as the heart but retains sufficient activity to support cell types with low energy requirements such as the fibroblasts. Our finding that COA6^W59C shows impaired maturation, with capacity to spontaneously oligomerize, suggests that the mutation may cause a reduction in COA6 protein levels in agreement with this hypothesis.

How might COA6 be involved in COX2 biogenesis? COX2 is known to be translated and inserted into the mitochondrial inner-membrane by OXA1 and is stabilized by the newly identified COX20 (37). At this point, SCO1 and SCO2 interact with COX2 to enable copper transfer in the intermembrane space. It has been established that SCO1 and SCO2 undertake distinct process in COX2 biogenesis (21,22,38–41). SCO2 has been suggested to act as a molecular chaperone for COX2, since loss of SCO2 reduces COX2 translation while turnover of any synthesized COX2 is delayed. In contrast, loss of SCO1 leads to newly translated COX2 being rapidly turned over (41), perhaps because SCO1 acts as the main copper donor for COX2 (22). Besides its proposed chaperone function, SCO2 also acts as the thiol-disulfide oxidoreductase for SCO1, and this influences mitochondrial redox signaling for regulation of copper efflux in cells (7,20). Cells lacking functional SCO1 (32,42) or SCO2 (33–36) result in the loss of COX2 levels like that of ΔCOA6 cells and also accumulate COX1-containing S1 and S2 intermediates (41). Our finding that COA6 can interact with both SCO1 and COX2 points to a potential role for COA6 in the copper handling process. Our experiments with recombinant protein demonstrate that COA6 binds copper. However, copper binding has yet to be established for the endogenous mitochondrial protein. It has earlier been proposed that copper binding takes place via the CX₉CX_nCX₁₀C motif (12). However, structural studies of twin CX₉C proteins, including COX17 (43), CHCHD5 and CHCHD7 (44), have shown that they harbor intramolecular disulfide bridges that form a hairpin conformation. In contrast, we find that not all cysteine residues in mitochondrial COA6 are oxidized pointing to a potential mechanism of copper binding within a COA6 dimer. Future structural characterization and mutagenesis approaches of COA6 will aid in this analysis.

During the revision of this manuscript, Pacheu-Grau et al. (45) also reported the functional characterization of COA6. Consistent with our studies, they found that COA6 can bind copper, while loss of COA6 results in a clear defect in COX2 biogenesis. Interestingly, they found that COA6 interacts with SCO2 in both yeast and human cells. Although we were unable to detect SCO2 interacting with COA6, we cannot exclude its presence and further studies are required to clarify these discrepancies. Pacheu-Grau et al. (45) also found that the COA6^W59C mutant could no longer interact with SCO2 and was mistargeted to the mitochondrial matrix. However, our finding that COA6^W59C can rescue COA6 knockout cells indicates that the mutant protein retains some functionality. The increased protease resistance observed for COA6^W59C by Pacheu-Grau et al. (45) may instead be due to the increased aggregation state of this mutant.

Materials and Methods

Patient data

The patient was thought to have been hypotonic from early in life and was investigated at 5 months of age following an upper respiratory tract infection with poor feeding, cough and breathlessness. He had decreased muscle bulk and tone, was losing weight and had increasing weakness and lethargy. He was diagnosed with hypertrophic obstructive cardiomyopathy at 6 months of age and died at 18 months of age from cardiac failure. Patient and control heart biopsy tissues were obtained, and the control fibroblasts used in this study have been reported previously (10).

Cell line generation and screening

TALEN-mediated gene disruption of COA6 was performed essentially as described (14). Briefly, TALENs (NG NN NI NI NN NN NI NI NI NN NI HD NI NN NN NG HD NG and NG NI NI NI HD NI HD NG NG HD HD NI NN NG NI HD NG HD) were assembled according to Reyon et al. (46) and ligated into the appropriate backbones modified to harbor an eGFP or mCherry coding sequence (CDS) positioned 5′ of the TALEN CDS, which is separated by a porcine teschovirus-1 2A (P2A) CDS allowing ribosome skipping (47). Clonally derived HEK293T cells transfected with both TALENs were resuspended in phosphate-buffered saline (PBS) containing 10% (v/v) fetal bovine serum (FBS), and single cells were isolated using a BD FACS Aria III gated on both eGFP and mCherry fluorescence. Clonal populations were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% (v/v) FBS and penicillin/streptomycin with 50 µg/mL uridine at 37°C under an atmosphere of 5% CO₂. To confirm COA6 disruption, the targeted exon was amplified by polymerase chain reaction (PCR) from genomic DNA isolated from individual clones and analyzed by Sanger sequencing.

Cell culture and mitochondrial isolation

Fibroblasts and HEK293T cell lines were grown in DMEM (Invitrogen) supplemented with 10% (v/v) FBS and penicillin/streptomycin and 50 µg/mL uridine at 37°C under an atmosphere of 5% CO₂. For complementation and immunoprecipitation experiments, ΔCOA6 cells were transfected with pCOA6-FLAG, pCOA6^W59C-FLAG and pCOA6^Δ50–58-FLAG constructs using Lipofectamine 2000 (Invitrogen). Cells were allowed to reach confluence and expanded before a second transfection was performed, following which media was refreshed or exchanged with glucose-free DMEM containing 10% (v/v) dialyzed FBS, 5 mm galactose and 50 µg/ml uridine and grown an additional 48–72 h prior to harvesting. For complementation analysis using isoform 3 of COA6, constructs were subcloned into pBABE-puro (Addgene) (48). Retroviral constructs were transfected into HEK293T cells using Lipofectamine 2000 as above, and viral supernatant was collected at 48 h post-transfection and used to infect ΔCOA6-1 cells in the presence of 8 µg/ml polybrene. Cells were selected through growth in galactose-containing media. Mitochondria were isolated as previously described (49), and protein concentration estimated by bicinchoninic acid assay (BCA; Pierce). For growth rate analysis, cells grown for 72 h in glucose-containing media supplemented with the indicated amounts of CuCl₂ were seeded in the same media and cell growth rates were determined using a sulforhodamine B (SRB) assay (50).

Oxygen consumption and membrane potential measurements

OCR was measured in live cells using a Seahorse Biosciences XF24-3 Analyzer according to manufacturer's instructions. Briefly, 50 000 HEK293T cells were plated per well of a poly-d-lysine-treated Seahorse Biosciences culture plate and grown overnight in the above culture conditions. OCR was analyzed in cells in non-buffered media with the following inhibitors: 2 μM oligomycin (Oligo), 0.5 μM carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone (FCCP), and 0.3 μM antimycin A along with 0.5 μM rotenone (AntA + Rot). Four measurement cycles (2 min mix, 2 min wait, 5 min measure) were done for basal conditions, and following each inhibitor injection. For each cell line, 7–10 replicate wells were measured in triplicate plates and measurements normalized to cell number with CyQuant (Life Technologies). Mitochondrial membrane potential was determined essentially as described (51). Cells were incubated for 30 min at 37°C under an atmosphere of 5% CO₂ in DMEM (Invitrogen) supplemented with 10% (v/v) FBS, penicillin/streptomycin, 50 µg/mL uridine and 25 nM tetramethylrhodamine methyl ester (TMRM) in the presence or absence of 10 µM carbonyl cyanide m-chlorophenyl hydrazone (CCCP). Cells were resuspended in PBS containing 2% FBS, 2.5 nM TMRM in the presence or absence of 10 µM CCCP and analyzed on a BD FACSCalibur system (Becton Dickinson). Data were analyzed using the CellQuest software (Becton Dickinson).

Radiolabeling of mtDNA-encoded translation products and protein import

Mitochondrial translation products were labeled with [³⁵S]-methionine/cysteine and either directly separated on 10–16% polyacrylamide Tris–tricine SDS-PAGE gels or membrane protein complexes solubilized in 1% digitonin and separated on 4–10% acrylamide-bisacrylamide BN-PAGE gels followed by two-dimensional (2D)-PAGE as previously described (20). For immunoprecipitation using COX1 and COX2 antibodies, membrane protein complexes solubilized in 1% Triton X-100 were bound to EZview™ Red Protein A Affinity Gel (Sigma) cross-linked to either COX1 or COX2 antibodies (Invitrogen) with dimethyl-pimelimidate. Bound proteins were eluted with 0.1 M glycine (pH 2.5). For immunoprecipitation using the FLAG antibody, mitochondria were solubilized in 1% digitonin, incubated with anti-FLAG affinity gel (Sigma) and eluted with FLAG peptide (Sigma). Samples were trichloroacetic acid precipitated to remove detergent and analyzed by SDS-PAGE and western blotting. For protein import, yeast cells of the strain YPH499 (Mat a, ade2–101, his3-Δ200, leu2-Δ1, ura3-52, trp1-Δ63, lys2-801) were grown on non-fermentable medium [1% (w/v) yeast extract, 2% (w/v) bacto-peptone, 3% (w/v) glycerol] at 24°C to an optical density of 1.0. Mitochondria were isolated by differential centrifugation (52). [³⁵S]-Methionine-labeled precursor proteins were synthesized in rabbit reticulocyte lysate (Promega) in the presence of 10 mm dithiothreitol. Radiolabeled proteins were incubated with isolated wild-type yeast mitochondria as previously described (53), and reactions were stopped on ice with addition of 50 mm iodacetamide (IAA) where indicated. Samples were treated with 50 µg/ml proteinase K prior to separation by SDS-PAGE in the presence of 1% (v/v) βME or 50 mm IAA and autoradiography.

Mass spectrometry

Cells were cultured in SILAC DMEM as described previously (54). Equal amounts of HEK293T and ΔCOA6-1 mitochondrial protein isolated from cells grown in ‘heavy’ (¹³C₆¹⁵N₄-arginine, ¹³C₆¹⁵N₂-lysine) or ‘light’ (¹²C₆¹⁴N₄-arginine, ¹²C₆¹⁴N₂-lysine) amino acids were solubilized in 8 M urea, 50 mm ammonium bicarbonate, acetone precipitated and reduced and alkylated with 5 mm TCEP hydrochloride and 50 mm IAA at 37°C for 30 min. Proteins were diluted to 2 M urea and digested at 37°C overnight with trypsin (Promega). The digest was acidified with 1% (v/v) trifluoroacetic acid and the peptides desalted on SDB-XC (Empore) StageTips as previously described (55). Peptides were analyzed by online nano-HPLC (high-pressure liquid chromatography)/electrospray ionization-MS/MS on an LTQ-Orbitrap Elite Instrument as previously described (54). For data analysis, mass spectrometric raw files were analyzed using the MaxQuant platform (version 1.5.1.2) searching against the human UniProt FASTA database (July 2014; 88 993 entries) and a database containing common contaminants by the Andromeda search engine (56). Default settings were used with modifications. Briefly, cysteine carbamidomethylation was used as a fixed modification, and N-terminal acetylation and methionine oxidation were used as variable modifications. False discovery rates of 1% for proteins and peptides were applied by searching a reverse database, and ‘Re-quantify’ and ‘Match between runs’ options were enabled. Unique and razor peptides were used for quantification, using a minimum ratio count of 1. Data analysis was performed using the Perseus software. Only proteins with a sequence coverage of >5% and quantified based on >1 unique peptide were considered for further data analysis. SILAC ratios (HEK293 T/ΔCOA6-1, normalized) were log₂-transformed and mean log₂ ratios of all proteins quantified in >2/3 replicates were calculated. P-values across replicates were calculated by a two-tailed t-test. For immunoprecipitation experiments, mitochondrial protein isolated from ΔCOA6-1 cells grown in either ‘heavy’ or ‘light’ amino acids complemented with pCOA6-FLAG or control pCOA6^Δ58–61-FLAG constructs was solubilized in 1% digitonin, bound to anti-FLAG affinity gel (Sigma) and eluted with FLAG peptide (Sigma). Equal amounts of elution were mixed, acetone precipitated and analyzed as described above. P-values across replicates were calculated by a one-tailed t-test.

Thiol modification analysis

Mitochondrial proteins were alkylated with AMS as previously described (28) with modifications. Mitochondria isolated from ΔCOA6-1 cells complemented with COA6-FLAG or COA6^W59C-FLAG were solubilized in 8 M urea, 50 mm Tris pH 7.4 and incubated at 50°C for 30 min with or without the addition of 10 mm TCEP (Bond-Breaker, Neutral pH, Thermo Scientific). Where indicated, samples were alkylated with 15 mm AMS for an additional 30 min at 50°C and analyzed by reducing SDS-PAGE and western blotting.

Copper binding

COA6 and COA6^W59C proteins (corresponding to isoform 3) were purified from the soluble cell fraction by glutathione S-transferase (GST) affinity chromatography. Following elution by cleavage of the GST tag with PreScission Protease (GE Healthcare), the eluted protein was concentrated by centrifugal ultrafiltration (Millipore) before final purification by size-exclusion chromatography (HiLoad 16/600 Superdex 200 pg, GE Healthcare). Fractions containing pure protein were analyzed for Cu content by ICP-MS, pooled and concentrated to 10 mg/mL before storage at −80°C. The Cu(I) probe ligand Bcs, which reacts quantitatively with Cu(I) to yield a 1:2 complex [Cu(I)Bcs₂]³⁻ with a characteristic solution spectrum (ϵ = 13 000 M⁻¹cm⁻¹ at λ_max = 483 nm) and a defined formation constant (logβ₂ = 19.8) (23,57), was employed for the quantification of Cu(I) binding to the wild-type and mutant apo-COA6 proteins, as described in (23). Briefly, purified COA6 proteins (in 20 mm Tris-MES pH 8.0) were titrated at various concentrations (1–30 μM) into solutions containing buffer (20 mm Tris-MES pH 8.0), CuSO4 (13 or 24 μM), Bcs (200 μM) and reduced glutathione (400 μM). The exchange of Cu(I) from the [Cu(I)Bcs₂]³⁻ complex to the COA6 proteins [according to Equation (1)] was monitored by measuring the absorbance of the resulting solutions at 483 nm. The data were analyzed using plots of [Cu(I)Bcs₂]³⁻ (as determined from the absorbance values at 483 nm) versus COA6:Cu ratio, and the data fit using Equation (2) (23) to yield K_D = 10^−17.2 and 10^−17.8 at pH 8.0 for wild-type and mutant COA6 proteins, respectively:

$$[\text{Cu}(\text{I})\text{Bc}{\text{s}}_2{]}^{3-}+\text{COA}6\rightleftharpoons \text{Cu}(\text{I})-\text{COA}6+2\text{Bc}{\text{s}}^{2-},$$

(1)

$$\begin{array}{l}\frac{{[P]}_{\mathrm{tot}}}{{[\mathrm{Cu}]}_{\mathrm{tot}}}=1-\frac{[\mathrm{C}{\text{u}}^{\mathrm{I}}{\text{L}}_2]}{{[\mathrm{Cu}]}_{\mathrm{tot}}}+K_{\mathrm{D}}{\beta }_2{\left(\frac{{[L]}_{\mathrm{tot}}}{[\mathrm{C}{\text{u}}^{\mathrm{I}}{\text{L}}_2]}-2\right)}^2[\mathrm{C}{\text{u}}^{\mathrm{I}}{\text{L}}_2]\left(1-\frac{[\mathrm{C}{\text{u}}^{\mathrm{I}}{\text{L}}_2]}{{[\mathrm{Cu}]}_{\mathrm{tot}}}\right)\end{array}.$$

(2)

Purified COA and COA6^W59C proteins were substituted with Cu(I) by first exchanging into 20 mm Tris-MES (pH 8.0) by centrifugal ultrafiltration (Millipore), followed by incubation on ice with CuSO₄ (5 molar equivalents) and reduced glutathione (GSH, 10 molar equivalents) before removal of excess copper and reducing agent by size-exclusion chromatography (Superdex 75 10/300 GL, GE Healthcare). The presence of Cu(I) bound to the eluted proteins was confirmed colorimetrically using the ligand Bcs. The molecular weights of both apo and Cu(I)-substituted samples of COA6 and COA6^W59C proteins were analyzed by analytical size-exclusion chromatography (Superdex 75 3.2/300, GE Healthcare) in PBS. A sample of the COA6^W59C protein was also analyzed after incubation with βME (10 mm). Samples for SDS-PAGE were boiled in loading dye in the presence or absence of 10 mm βME.

Miscellaneous

C-Terminal FLAG-tagged c1orf31 transcript variant 1 (pCOA6-FLAG) was obtained from Origene and used as template for the generation of pCOA6^W59C-FLAG and pCOA6^Δ50–58-FLAG through Quickchange II XL site-directed mutagenesis (Agilent). For protein expression, the COA6 and COA6^W59C cDNAs encoding isoform 3 were subcloned into pGEX-4T1. Standard techniques were used for immunoblotting and chemiluminescence detection. Monoclonal antibodies against SDHA, CORE1 and COX4 (Abcam), COX1 and COX2 (Invitrogen), Tom20 (Santa Cruz, CA, USA) were obtained commercially, while a rabbit polyclonal antibody against NDUFA9 was made in-house. Radiolabeled proteins were detected by PhosphorImager analysis (Molecular Dynamics).

Funding

This work was supported by Australian National Health and Medical Research Council grants (to M.T.R., D.A.S. and D.R.T.) and Fellowships (Early Career Fellowship to D.A.S., Career Development Fellowship to A.E.F., Principal Research Fellowship to D.R.T.), the Australian Mitochondrial Disease Foundation (AMDF), the Australian Research Council (ARC Grant DP140102746 to M.J.M.) and the Victorian Government's Operational Infrastructure Support Program. Research in the BW laboratory is supported by the Deutsche Forschungsgemeinschaft and the Excellence Initiative of the German Federal & State Governments (EXC 294 BIOSS).

Acknowledgements

We thank Christa George, Dinesha Cooray, Alex Lowdin, Ian Potter, Michael Chan and Laura Twigg for technical assistance, and Boris Reljic, Luke Formosa, and Michael Lazarou for reagents, advice and discussions.

Conflict of Interest statement. None declared.