Cytochrome c oxidase (CIV) deficiency is one of the most common respiratory chain defects in patients presenting with mitochondrial encephalocardiomyopathies. CIV biogenesis is complicated by the dual genetic origin of its structural subunits, and assembly of a functional holoenzyme complex requires a large number of nucleus-encoded assembly factors. In general, the functions of these assembly factors remain poorly understood, and mechanistic investigations of human CIV biogenesis have been limited by the availability of model cell lines. Here, we have used small interference RNA and transcription activator-like effector nucleases (TALENs) technology to create knockdown and knockout human cell lines, respectively, to study the function of the CIV assembly factor COX20 (FAM36A). These cell lines exhibit a severe, isolated CIV deficiency due to instability of COX2, a mitochondrion-encoded CIV subunit. Mitochondria lacking COX20 accumulate CIV subassemblies containing COX1 and COX4, similar to those detected in fibroblasts from patients carrying mutations in the COX2 copper chaperones SCO1 and SCO2. These results imply that in the absence of COX20, COX2 is inefficiently incorporated into early CIV subassemblies. Immunoprecipitation assays using a stable COX20 knockout cell line expressing functional COX20-FLAG allowed us to identify an interaction between COX20 and newly synthesized COX2. Additionally, we show that SCO1 and SCO2 act on COX20-bound COX2. We propose that COX20 acts as a chaperone in the early steps of COX2 maturation, stabilizing the newly synthesized protein and presenting COX2 to its metallochaperone module, which in turn facilitates the incorporation of mature COX2 into the CIV assembly line.
Defects in the assembly and function of cytochrome c oxidase or mitochondrial respiratory chain complex IV (CIV) are a frequent cause of oxidative phosphorylation disorders in humans. Patients suffering from CIV-related mitochondrial diseases present with heterogeneous clinical phenotypes ranging from encephalomyopathy, hypertrophic cardiomyopathy and liver disease to Leigh's syndrome (1–4). Over the last 30 years, progress made on the characterization of CIV structure, biogenesis and function (5–11) has yielded considerable insight into the molecular basis of many of these human disorders.
CIV is an electron-driven proton pump that catalyzes the reduction of dioxygen to water (12). As the terminal oxidase of the mitochondrial respiratory chain, it participates in the formation of an electrochemical proton gradient that is subsequently used by the mitochondrial F1Fo ATPase to drive ATP synthesis (13). Human CIV is formed by 13 subunits of dual genetic origin. The mitochondrial genome encodes the three core, catalytic subunits (COX1, 2 and 3), whereas the nuclear genome encodes the remaining subunits. COX1 and COX2 coordinate metal prosthetic groups essential for electron transfer. COX1 coordinates one redox center with a heme a and a second binuclear center containing a heme a3 paired with a Cu(I) atom (CuB site). The third redox center of the enzyme resides in COX2 and is formed by two copper ions of mixed valences (CuA site). During catalysis, electrons are transferred from ferrocytochrome c to the CuA active site and subsequently to the heme a site. From heme a, electrons are transferred to the heme a3-CuB binuclear center where bound molecular oxygen is ultimately reduced to water.
CIV biogenesis has been described as a linear process with the different subunits being added in an ordered manner (14). Additionally, data obtained in Saccharomyces cerevisiae indicate that at least the catalytic subunits COX1 and COX2 form associations with assembly factors in individual assembly lines prior to their incorporation into the general CIV assembly line (15,16). CIV assembly is initiated around a seed formed by COX1, and proceeds with the formation of subassemblies of progressively increasing complexity (14). Maturation of COX1 seems to be the first step in the process, with the insertion of its metal co-factors only occurring once COX1 has been inserted into the membrane (17). COX4 and COX5a are then added to mature COX1 to form the S2 intermediate. Incorporation of COX2 and most of the remaining structural subunits results in the formation of the S3 intermediate. Finally, insertion of COX6a and COX7a/b produces the CIV monomer (14).
Progression through all stages of the CIV assembly process requires a large number of nucleus-encoded proteins termed COX assembly factors, many of which were first identified in S. cerevisiae (18,19). For example, yeast Cox20 is involved in Cox2 biogenesis by acting as a Cox2-specific chaperone. Both Cox20 and Cox2 have two transmembrane domains, with their N- and C-termini protruding into the intermembrane space. Cox20 binds to newly synthesized Cox2, which in yeast is synthesized as a precursor protein, and presents it to the inner membrane peptidase (IMP) complex to facilitate the cleavage of its N-terminal leader sequence (16,20). Cox20 also fulfills an essential role in the translocation of the C-terminus of Cox2 through the membrane. The metallochaperone Sco1 then interacts with Cox2 to facilitate the maturation of its CuA site (21). Over the last 15 years, in silico approaches and the analysis of tissues and cultured cell lines derived from patients suffering from CIV deficiencies have allowed for the identification and characterization of both evolutionarily conserved and mammalian-specific COX assembly factors (22–26). Human SCO1 and SCO2 are homologues of yeast Sco1 and are both required for COX2 metallation (27), with mutations in either SCO gene causing mitochondrial encephalocardiomyopathies or hepatopathies (28,29). Recently, a mitochondrial disorder associated with CIV deficiency was reported in two families carrying the same homozygous mutation in FAM36A (23,30) the open reading frame encoding the human homologue of yeast Cox20 (31). Human COX20 was found to be important for the assembly of CIV, probably by acting on COX2 (23), but its specific role in holoenzyme biogenesis has yet to be fully described. We therefore sought to further elucidate how human COX20 stabilizes COX2, to allow for its efficient maturation and insertion into assembling CIV. By using siRNA and transcription activator-like effector nucleases (TALENs) technology (32–34) to silence COX20 expression in human cell lines, we demonstrated that human COX20 plays a role as a COX2 chaperone in the early steps of COX2 biogenesis. We also showed that a transient complex composed of newly synthesized COX2, COX20, and the copper chaperones SCO1 and SCO2 is formed and that it facilitates COX2 membrane insertion, maturation and ultimately assembly of the holoenzyme.
RNAi silencing of human COX20 causes an isolated CIV deficiency by destabilizing newly synthesized COX2
To analyze the role of COX20 in mitochondrial metabolism, we started by conducting siRNA-specific knockdown of endogenous COX20 expression in cultured HEK293T cells. Compared with the mock control, transient transfection of two different COX20-specific siRNA duplexes (Fig. 1A) led to an ∼80% reduction in COX20 mRNA levels at Days 1, 2 and 3 (Fig. 1B and C). Immunoblot analysis confirmed the efficacy of knockdown at the protein level. Even though our affinity-purified anti-COX20 antibody cross-reacts strongly and non-specifically with larger proteins, it clearly detects a protein that migrates at the predicted molecular weight of COX20 (15 kDa) whose steady-state levels are markedly attenuated in COX20-silenced cells (Fig. 1D). The net effect of COX20 silencing on mitochondrial metabolism was a decrease in endogenous cell respiratory capacity (Fig. 1E). The abundance of COX2 and COX3 subunits was significantly decreased in COX20-silenced cells, whereas the levels of COX1 or of subunits of other oxidative phosphorylation complexes were not affected (Fig. 1F). Consistent with these observations, Blue Native-polyacrylamide gel electrophoresis (BN-PAGE) analyses showed that COX20 silencing produced an isolated CIV deficiency (Fig. 1G) and led to the accumulation of assembly intermediates containing COX1 alone (S1) or at least COX1 and COX4 (S2 and S3) (Fig. 1H). These subassemblies correspond to the intermediates observed in cell lines derived from patients with mutations in the COX2 metallochaperone SCO1 or in the CIV assembly factor SURF1 (Fig. 1I). The CIV assembly defect in COX20-silenced cells is not due to a gross defect in COX2 synthesis as assessed by in vivo pulse-labeling of mitochondrial translation products. Rather, chase-labeling experiments showed that when COX20 levels are limiting, newly synthesized COX2 is unstable and most of it is rapidly degraded within 2 h of the chase (Fig. 1J). ATP6 was also found to be labile in this chase assay, although we did not observe any effect on the steady-state levels of the F1Fo ATPase (Fig. 1G). These results are consistent with the recently reported phenotype in fibroblasts derived from a COX20 patient (23), and suggest that COX2 is less stable upon depletion of COX20 because of a defect in its membrane insertion, maturation of its CuA site and/or a block in its incorporation into the S2 assembly intermediate.
Generation of a human COX20 knockout cell line using TALENs in HEK293T cells
Working with RNAi-silenced cells has its practical limitations because one cannot exclude the possibility of residual COX20 expression, and because it is difficult to obtain the number of cells that are required to isolate enough mitochondria for extensive biochemical analyses. Therefore, to further understand the exact function of COX20 in CIV biogenesis, we turned to a gene-editing approach and used transcription activator-like effector nucleases (TALENs) technology to create a stable human COX20 knockout line in HEK293T cells. TALENs are sequence-specific nucleases created by the fusion of the catalytic domain of an endonuclease (the FokI restriction enzyme) to TALE proteins, a large group of bacterial plant pathogen proteins that have emerged as alternatives to zinc finger proteins because of their very high fidelity DNA-binding capacity (32,34). We used the services of Cellectis Bioresearch (Paris, France) to design and generate the ‘best hit’ TALEN binding pair to disrupt COX20 expression in HEK293T cells, which target a region within the first exon of the gene that is directly downstream of the translation start codon (Fig. 2A). Although TALENs have a high level of specificity, due in part to the 15–20 base pair length of their recognition site, we further increased the specificity of the TALENs reaction by using two TALEs fused to two endonuclease monomers that are only active upon dimerization.
Co-transfection of the TALEN pair into HEK293T cells was expected to generate double strand DNA breaks in the first exon of the COX20 gene and, since the non-homologous end joining repair system often yields small deletions or insertions (35), an out-of frame mutation and a nonsense transcript that leads to the absence of COX20 protein. Two days after transfection, single HEK293T cells were isolated by fluorescence-activated cell sorting, amplified and screened for CIV activity immunohistochemically (not shown). Twenty of 300 clones exhibited reduced COX staining, and were selected for secondary screening. Although our COX20 antibody is suitable for immunoblotting of mitochondrial fractions, it yields an extremely poor signal in whole cell extracts. We therefore screened the steady-state levels of COX2 in these 20 clones, since its abundance is markedly decreased in COX20 siRNA-silenced cells. Mitochondria were subsequently isolated from one of the clones with the lowest steady-state levels of COX2, and the COX20 protein was found to be undetectable by immunoblotting (Fig. 2B). Sequencing of the COX20 gene in this clone revealed that both alleles contained mutations within the TALEN-targeted region, with a complete recombination on one allele and three point mutations on the second allele (C6A, G7T and C10G) that resulted in two amino acid changes (Ala3Ser, Pro4Ala). We presume that these point mutations affect the import or stability of the protein, since COX20 is undetectable in this clone (Fig. 2B), henceforth referred to as KO-COX20.
The KO-COX20 cell line shows a profound decrease in CIV levels and activity
Knockout of COX20 in HEK293T cells reduced the respiratory capacity to ∼50% of the parental line (Fig. 2C), an effect that was a direct consequence of a severe, isolated CIV deficiency (Fig. 2D). Consistent with these observations, and like COX20 siRNA-silenced cells (Fig. 1), the steady-state levels of COX2 and COX3 were markedly decreased in KO-COX20 mitochondria, while those of COX1 and COX4 were unchanged (Fig. 2E). The abundance of COX6B, which is readily detected in isolated mitochondria, was also significantly reduced (Fig. 2E). BN-PAGE analysis using CII as a loading control further demonstrated very low residual levels of fully assembled CIV in the KO-COX20 cell line, whereas the other OXPHOS complexes were unaffected (Fig. 2F). We additionally observed a marked accumulation of the CIV S1 and S2 assembly intermediates that contain COX1 and COX4 (Fig. 2G), again reflecting a block in the incorporation of COX2 into the S2 assembly intermediate (27,36,37).
The CIV assembly defect we observed in the KO-COX20 cell line was fully restored by stable expression of C-terminal, FLAG-tagged COX20 (Fig. 2H and I), eliminating the possibility that our results were attributable to off-target effects that arose during the gene disruption procedure. Next, we investigated the effect of knocking out COX20 on the ability of respiratory complexes to form supercomplexes, by analyzing digitonized whole cell extracts using BN-PAGE. The KO-COX20 cell line showed normal levels of I-III2 complexes but supercomplexes containing CIV (III2 + IV and I + III2 + IVn) were too scarce to be detected (Fig. 2J). In summary, our molecular and biochemical analyses of the KO-COX20 cell line support a model whereby COX20 functions to promote the assembly of CIV but not of other respiratory chain enzyme complexes. Our results also indicate that COX20 is not essential for CIV biogenesis but that it dramatically enhances the efficiency of the process.
Human and yeast COX20 are orthologs but do not function in the heterologous system
To gain further insight into the role of human COX20, we tested the reciprocal heterologous complementation of S. cerevisiae and human COX20. A knockout strain or cell line from each species was transfected with either ScCox20 or HsCOX20. The yeast ScCox20 is a Cox2 chaperone, whose absence leads to lack of Cox2 proteolytic cleavage by IMP and rapid Cox2 degradation. The sequence conservation between ScCOX20 and HsCOX20 is low but two hydrophobic regions are particularly conserved (Fig. 3A). Similar to ScCox20, HsCOX20 is an integral inner membrane protein that likely contains two transmembrane domains, with its C- and N-termini localized to the intermembrane space [(23) and Supplementary Material, Fig. S1]. Notably, ScCox20 has a 58 amino acid N-terminal extension that is missing in HsCOX20.
We expressed HsCOX20-HA, ScNΔ58Cox20-HA and ScCox20-HA from either episomal or integrative vectors in a Δcox20 yeast strain. ScCox20-HA complemented the respiratory defect of Δcox20 yeast, but the human protein did not (Fig. 3B), even though it was expressed and imported into mitochondria (Fig. 3C). Although the truncated NΔ58 version of ScCox20 was also unexpectedly imported into mitochondria (Fig. 3C), and markedly complemented the growth defect of Δcox20 yeast on respiratory media, the Cox2 precursor was inefficiently processed in this strain (Fig. 3D) and the steady-state levels of mature Cox2 were only ∼30% of wild-type. Therefore, the N-terminus of ScCox20 may be particularly important for Cox2 processing, and may in fact be involved in the interaction with the IMP peptidase, which would explain why this N-terminal region is not conserved in humans where there is no proteolytic processing of COX2.
In the reciprocal complementation studies, the CIV assembly defect in the human KO-COX20 cell line was not rescued by stably expressing yeast Cox20 or NΔ58Cox20 (Fig. 3E and F). The lack of reciprocal complementation suggests that the sequence divergence between ScCox20 and HsCOX20 (Fig. 3A) prevents each protein from interacting with functional partners in the heterologous context.
Human COX20 stabilizes COX2 by interacting with the newly synthesized protein
To test the effect of HsCOX20 on COX2 biogenesis, we examined the synthesis and stability of newly synthesized COX2 in the KO-COX20 cell line. Pulse-chase [35S]methionine labeling showed that COX2 and all other mitochondrial proteins were synthesized at normal rates in the KO-COX20 cell line (Fig. 4A). However, like COX20 siRNA-silenced cells, newly synthesized COX2 and COX3 but not COX1 were unstable following a 1 h chase (Fig. 4A). Nascent ATP6 was also found to be labile, although its steady-state levels (Fig. 2H) and the abundance of the F1Fo ATPase (Fig. 2F and I) were unaffected. The stability of newly synthesized COX2 and of other mitochondrial polypeptides in KO-COX20 cells was restored to wild-type levels by overexpressing COX20-FLAG (Fig. 4B).
To address if HsCOX20 acts as a COX2 chaperone like its yeast counterpart, we next investigated whether COX20 physically interacts with COX2 or any other newly synthesized mitochondrial proteins. For this purpose, we labeled these proteins with [35S]methionine in whole KO-COX20 cells stably expressing functional COX20-FLAG, and adsorbed a mitochondrial extract onto anti-FLAG-agarose beads. We observed that of the [35S]methionine-labeled proteins that were recovered from the beads, COX2 was selectively enriched when compared with the parental line (Fig. 4C). COX20-FLAG was even able to co-immunoprecipitate (IP) a small fraction of the total COX2 in mitochondrial extracts from unlabeled cells (Fig. 4D). Taken together, these observations argue that the stabilizing effect of COX20 on CIV subunits is due to a direct interaction of COX20 with newly synthesized COX2.
Human COX20 is part of a ∼90 kDa complex whose stability is independent of mitochondrial protein synthesis
To better understand the function of COX20 in COX2 maturation, we analyzed its native molecular mass. BN-PAGE analysis of KO-COX20 + COX20-FLAG whole cell extracts using an anti-FLAG antibody detected a complex of ∼90 kDa (Fig. 5A). A comparable result was obtained by analyzing the sedimentation properties in sucrose gradients of COX20 extracted from KO-COX20 + COX20-FLAG mitochondria under the same conditions used for the immunoprecipitation experiments (Fig. 5B). These results indicate that the ∼15 kDa COX20 protein forms a stable ∼90 kDa complex in mitochondria by interacting with unidentified partners or by forming a homohexamer.
According to proteomics studies [see Protein Abundance Across Organisms (http://pax-db.org)], the molar amount of COX20 is 10–20 fold lower than the amount of CIV subunits, including COX2 (Supplementary Material, Table S1). Although COX20 is a low abundance protein and the COX20–COX2 interaction, we described earlier is expected to be transient in nature, the possibility nonetheless existed that the ∼90 kDa COX20-containing complex includes newly synthesized COX2 or represents a COX2-containing CIV assembly intermediate. However, sucrose gradient analyses failed to detect co-sedimentation of COX20-FLAG and COX2 (Fig. 5B), which accumulated in fractions containing the CIV monomer. Because the sucrose gradient analysis is less sensitive than the co-immunoprecipitation assays, where eluates were concentrated several fold (Fig. 4D), we further explored the relevance of COX2 to the native size of COX20 by investigating whether the integrity of the ∼90 kDa complex depends on the presence of newly synthesized COX2. For this purpose, KO-COX20 + COX20-FLAG cells were treated overnight with doxycycline to inhibit organelle protein synthesis and hence deplete the cells of newly synthesized mitochondrial polypeptides, including COX2 (Fig. 5C). Doxycycline treatment did not affect the stability of COX20-FLAG (Fig. 5C), which still sedimented in sucrose gradients as a ∼90 kDa complex (Fig. 5B). These results indicate that the majority of COX20 exists in a ∼90 kDa complex whose stability is independent of COX2.
COX20 interacts with SCO1 and SCO2 in a COX2-dependent manner
To gain further insight into how the chaperone function of COX20 assists with COX2 stability, we sought to assess whether other known CIV assembly factors critical to the maturation of COX2 transiently interact with COX20. For this purpose, we took a candidate approach, and concentrated on OXA1L, COX18, SCO1 and SCO2. Yeast Oxa1 is required for co-translational translocation and insertion of Cox2 into the inner membrane (38–40), and human OXA1L can functionally substitute for yeast Oxa1 (41). Human COX18 is the uncharacterized putative ortholog of S. cerevisiae Cox18, which is necessary for the export of the C-terminal domain of yeast Cox2 (42). Human SCO1 and SCO2 code for essential metallochaperones with cooperative functions in the biosynthesis of the CuA site in COX2 (27). SCO2 is required for efficient COX2 synthesis, and also acts as a thiol-disulfide oxidoreductase to oxidize the copper-coordinating cysteines in SCO1 during COX2 maturation (43). While COX18 could not be included in our co-immunoprecipitation experiments due to the poor quality of the commercially available antibody, OXA1 was not specifically pulled down with COX20-FLAG (Fig. 6A). In contrast, a small portion of both SCO1 and SCO2 was specifically and reproducibly co-IP with COX20-FLAG (Fig. 6A and D). The interaction between COX20 and SCO1 was subsequently confirmed by reciprocal immunoprecipitation with a SCO1 antibody (Fig. 6B). To evaluate whether the interaction between COX20 and SCO, proteins was direct or whether it was mediated through COX2, the immunoprecipitation experiment was repeated using cells that had been pretreated overnight with doxycycline to inhibit mitochondrial protein synthesis. The steady-state levels of COX2 were markedly decreased in doxycycline-treated cells, whereas the abundance of COX20 and, to a lesser extent, of SCO1 and SCO2 remained more stable (Fig. 6C). Importantly, SCO1 and SCO2 were no longer co-IP with COX20-FLAG (Fig. 6D), indicating that these interactions depend on the presence of COX2. Finally, the sedimentation of SCO1 and SCO2 in a sucrose gradient is not affected by the absence of COX20 (Fig. 6E), supporting the transient nature of their interaction and suggesting that only a small fraction of the total SCO1 or SCO2 protein pool interacts with COX20. Consistent with this idea, although siRNA knockdown of COX20 did not affect COX2 abundance in control fibroblasts, it reduced the steady-state levels of COX2 in SCO patient fibroblasts (Fig. 6F), which exhibit very low levels of residual SCO1 or SCO2 protein function. These results strongly suggest that COX20 binds to newly synthesized COX2 to stabilize the protein and present it to SCO1 and SCO2 to promote the maturation of its CuA site.
This study presents several lines of evidence demonstrating that human COX20 functions as a CIV assembly factor by stabilizing newly synthesized COX2 and promoting the maturation of its CuA site. First, silencing of COX20 expression by RNAi or gene knockout (KO-COX20) in HEK293T cells resulted in a severe, isolated CIV deficiency that was caused by marked instability of newly synthesized COX2. Overexpression of FLAG-tagged COX20 eliminated the accumulation of early assembly intermediates in KO-COX20 cells and rescued the CIV deficiency, confirming that the observed phenotypes were indeed a direct consequence of manipulating COX20 expression. Second, COX20 interacts with and stabilizes newly synthesized COX2, presumably by facilitating its co-translational insertion into the mitochondrial inner membrane. Third, COX20 also physically interacts with the copper metallochaperones SCO1 and SCO2, which are known to fulfill independent but cooperative roles in the biosynthesis of the CuA site of COX2. Finally, siRNA knockdown of COX20 reduces the abundance of COX2 in fibroblasts derived from SCO1 and SCO2 patients, but not from controls, providing functional genetic evidence that further supports the importance of the physical interactions between these three proteins to the maturation of COX2.
Our current understanding of COX2 synthesis and its subsequent maturation stems from functional studies of yeast mutants and cell lines derived from patients, all of which exhibit an isolated CIV deficiency. In yeast, nascent pCox2 interacts with the Oxa1 machinery, which facilitates the insertion of the first transmembrane domain and concomitant export of the soluble N-terminal domain of pCox2 across the inner membrane (38,44). ScCox20 most likely interacts with pCox2 after export of its N-terminus (16), and facilitates both the proteolytic removal of the N-terminal presequence and the export of the C-terminal tail of pCox2 (20,45,46). Consistent with the existing yeast data (16), our immunoprecipitation results indicate that human COX20 does not physically interact with OXA1L. Although Cox18 acts on pCox2 after the cleavage of its leader sequence (47), and a Cox2-dependent interaction between Cox20 and Cox18 was recently reported in yeast (16), we were unable to address the potential conservation of this interaction in the human system because of the unavailability of high-quality COX18 antibodies.
It has been proposed that yeast Cox20 remains bound to mature Cox2 once the leader sequence of Cox2 has been cleaved, stabilizing the protein and facilitating its eventual incorporation into early CIV assembly intermediates (20). Our results demonstrate that although the yeast and human proteins are not functional in the heterologous system, human COX20 fulfills an evolutionarily conserved function by interacting with newly synthesized COX2 to promote its stability. We also show that COX20 physically interacts with the COX2 metallochaperones SCO1 and SCO2 in a COX2-dependent manner, strongly suggesting that the CuA site of human COX2 is matured while it is bound to COX20. This contention is indirectly supported by the fact that KO-COX20 cells and fibroblasts from COX20, SCO1 and SCO2 patients all accumulate the S2 assembly intermediate, which contains COX1-COX4-COX5A but lacks COX2. Consistent with the idea that physical interactions between COX20 and SCO1 and SCO2 are important to the maturation of the CuA site, the steady-state levels of COX2 were reduced upon siRNA knockdown of COX20 in SCO1 and SCO2 patient fibroblasts, which harbor pathogenic alleles that code for highly unstable mutants or allelic variants that have little residual function. Future studies will be required to glean mechanistic insight into the ordered protein–protein interactions that promote the stability and metallation of COX2, and will likely benefit from our efforts to identify whether the stable ∼90 kDa complex containing COX20 is a homohexamer or a hetero-oligomeric stabilization complex.
We feel it is important to emphasize that while the findings of our study and of those focused on COX20 patients (23,30) support a role for COX20 as a COX assembly factor, human COX20 is unlike its yeast counterpart in that it is not essential for CIV assembly. While it is conceivable that the Thr52Pro mutant expressed by both COX20 pedigrees retains sufficient residual function to adequately stabilize COX2 and promote its maturation (23), even HEK293T cells lacking COX20 retain roughly 20% of the CIV activity present in the parental line. However, because the CIV deficiency is much more severe in COX20 patient muscle when compared with patient fibroblasts (23), an intriguing possibility is that COX20 function may be more important in cell types in which the kinetics of CIV assembly are faster because the holoenzyme has a shorter half-life. Fibroblasts may also be less sensitive to loss of COX20 function because they possess alternate, COX20-independent CIV assembly mechanisms. While cell types from a greater number of COX20 pedigrees will need to be characterized prior to drawing any definitive conclusions about the tissue-specific requirement of COX20 function, the clinical course of disease in those COX20 patients identified to date is undeniably mild when compared with that in patients with mutations in other COX assembly factors. In summary, we used the innovative TALENs technology to generate a COX20 knockout cell line, which allowed us to demonstrate that human COX20 is a chaperone that stabilizes COX2 and interacts with the COX2-specific metallochaperones SCO1 and SCO2 to promote CuA site maturation, thereby facilitating COX2 insertion into the assembling CIV holoenzyme.
MATERIALS AND METHODS
Human cell lines, transfection and culture conditions
HEK293T, HeLa and fibroblast cell lines were cultured in DMEM containing 10% FBS, 1 mm pyruvate and 50 µg/ml uridine at 37°C under 5% CO2. Cells were transfected with lipofectamine 2000 (Invitrogen), mixed with plasmids in OPTIMEM-I media (Gibco). Stable lines were established by transfection of HEK293 T or KO-COX20 cell lines with pIRESpuro2 empty vector or pIRESpuro2 containing variants of COX20. Two days after transfection, the media was supplemented with 2.5 µg/ml puromycin and drug selection was maintained for at least 1 month. Patient cell lines included a SURF1-deficient line carrying a homozygous frameshift mutation (48), and several SCO-deficient cell lines all compound heterozygotes, carrying a nonsense mutation on one allele and one of three missense mutations on the second allele (SCO1–1, P174L; SCO1-2, M294V; SCO2, E140 K) (28).
Two Stealth RNAi duplexes against human COX20 (HSS151056 siRNA1, HSS151057 siRNA2, Invitrogen) were used to knock down COX20 expression. Cells were transfected using Lipofectamine RNAiMAX (Invitrogen) according to the manufacturer's specifications on Days 0, 3 and 6. HEK293T cells (10 nm) and fibroblasts (12.5 nm) were transiently transfected with each siRNA, a scrambled control [BLOCK-iT Alexa Fluor (Invitrogen)] or mock transfected.
TALEN plasmids were obtained from Cellectis (Paris, France). The left and right TALENs were designed to bind the TAGTCCTCATGGCCGCC and the GAGCCCGAGGAGAGGAA DNA sequences, respectively, at the COX20 locus. The left TALEN expresses a TALE with the variable amino acids: NI-NN-NG-HD-HD-NG-HD-NI-NG-NN-NN-HD-HD-NN-HD-NG and the right: NG-HD-HD-NG-HD-NG-HD-HD-NG-HD-NN-NN-NN-HD-NG-NG.
HsCOX20-1HA and HsCOX20-1FLAG were amplified by PCR using the COX20 cDNA (Sequence Genbank BC0118519.1) as a template. ScCOX20-HA and NΔ58ScCOX20-HA were amplified by PCR from genomic DNA from the yeast W303 strain. All four amplicons were then cloned into pIRESpuro2 using NheI and BamHI sites. HsCOX20-1HA, ScCOX20-HA and NΔ58ScCOX20-HA were cloned as HindIII and XbaI fragments into the yeast multicopy expression plasmid YEplac112 and the integrative plasmid YIplac128 (XbaI and BamHI), under the control of the ScTEF1 promoter. The oligonucleotides used for cloning in this study are listed in Supplementary Material, Table S2.
An affinity-purified rabbit polyclonal antibody against a C-terminal human COX20 peptide (RYNYAKQRIQERIAREEIKK) was generated by Open Biosystems/Thermo Scientific (Huntsville, AL, USA). A human SCO1 antibody was generated by injecting rabbits with a synthetic peptide (CASIATHMRPYRKKS) that had been conjugated to KLH via a reactive sulfhydryl group (Pierce) and emulsified in Titermax (Pierce) (27). The crude polyclonal antiserum was purified by its sequential incubation with glutathione-S-transferase (GST) and a GST-SCO1 antigen fusion protein, each of which had been crosslinked to AminoLink (Pierce) columns according to the manufacturer's instructions. Antibody bound to the GST-SCO1 column was eluted with acetic acid, rapidly neutralized with unbuffered Tris and then buffer exchanged and concentrated in PBS containing 50% glycerol.
The following commercial antibodies were used in this study; FLAG, VDAC, HSP70 and SCO2 were from Sigma-Aldrich; ATP6, COX1, COX2, COX3, COX4, COX6b, ATP5α, NDUFA9, CORE2 and h70 kDa were from Abcam; HA was from Covance and OXA1 for LSBio. Secondary antibodies coupled to horseradish peroxidase were obtained from Santa Cruz Biotechnology. Secondary Mouse TrueBlot ULTRA (Rockland) was used for immunoblot after immunoprecipitation to eliminate interference of the heavy and light chains of the immunoprecipitating antibody.
Mitochondria from human cells were isolated from at least ten 80–100% confluent 175 cm2 flasks as described previously (49). Briefly, the cells were resuspended in ice-cold T-K-Mg buffer (10 mm Tris–HCl, 10 mm KCl, 0.15 mm MgCl2, pH 7.0) and disrupted with 10 strokes in a homogenizer (Kimble/Kontes, Vineland, NJ, USA). Using a 1 m sucrose solution, the homogenate was brought to a final concentration of 0.25 m sucrose and a postnuclear supernatant was obtained by centrifugation of the samples twice for 3 min at 1500 g. Mitochondria were pelleted by centrifugation for 10 min at 8000 g and resuspended in 0.6 m Sorbitol 10 mm Hepes pH 7.4.
Protein concentration was measured with the folin phenol reagent (50). For immunoblotting studies, 20–60 µg of mitochondrial proteins were separated by SDS–PAGE in the Laemmli buffer system (51). After transfer, membranes with immobilized proteins were decorated with antibodies against the appropriate proteins followed by a second reaction with anti-mouse or anti-rabbit IgG conjugated to horseradish peroxidase (Sigma, St Louis, MO, USA). Chemiluminescence was used for the final detection.
Mitochondrial protein synthesis
Mitochondrial protein synthesis was determined in human cells by pulse-labeling 80% confluent cell cultures growing in a 6-well plate in media supplemented with 100 μl/ml emetine (for pulse assays) or 100 μl/ml anisomycin (for pulse-chase assays) to inhibit cytoplasmic protein synthesis as described (52). Subsequently, whole cell extracts were prepared by solubilization in RIPA buffer (1% NP-40, 0.1% SDS, 50 mm Tris–HCl pH 7.4, 150 mm NaCl, 0.5 sodium deoxycholate, 1 mm EDTA) with 1 mm PMSF and 1× mammalian protease inhibitor cocktail (Sigma). In yeast, mitochondrial gene products were labeled with [35S]-methionine (7 mCi/mmol, Perkin Elmer) in whole cells at 30°C in the presence of 0.2 mg/ml cycloheximide to inhibit cytoplasmic protein synthesis (53). Equal amounts of total cellular protein (human or yeast) were loaded in each lane and separated by SDS–PAGE on a 17.5% polyacrylamide gel, transferred to a nitrocellulose membrane and exposed to Kodak X-OMAT X-ray film.
For immunoprecipitation of COX20-FLAG, 600 μg of mitochondria were extracted in 400 µl of 20 mm Tris pH 7.5, 50 mm KCl, 0.4% lauryl maltoside, 1 mm PMSF and 1× protease inhibitor cocktail. The lysate was centrifuged at 22 000g, 30 min at 4°C. The extract (E) was incubated for 4 h at 4°C with 30 µl beads conjugated with a FLAG antibody (anti-DYKDDDDK beads, Clontech) or SCO1 antibody conjugated to protein A beads (Invitrogen). The unbound material (Un) was removed and the beads were washed three times in 400 µl of 20 mm Tris pH 7.5 50 mm KCl buffer. The beads were then supplemented with 60 µl of Laemmli buffer to release the IP material. Representative amounts of all fractions were boiled, and loaded on an SDS–PAGE gel in the following proportions; E (1×), Un (1×), Wash (1×) and IP (10×).
For co-immunoprecipitation studies of COX20-FLAG with newly synthesized mitochondrial translation products, subconfluent cells were radiolabeled with [35S]methionine for 30 min in the presence of emetine to inhibit cytoplasmic protein synthesis. Cells were then washed with PBS and trypsinized. 7.5 × 106 cells were incubated 10 min with 2 mg/ml of digitonin (Sigma), spun at 10 000 g for 5 min at 4°C. Digitonized cell pellets were washed with PBS, and solubilized in 380 µl of 20 mm Tris pH 7.5, 50 mm KCl, 0.4% lauryl maltoside, 1 mm PMSF and 1× protease inhibitor cocktail. Finally, the resultant extracts were used for COX20-FLAG immunoprecipitation using 20 µl of anti-FLAG-conjugated agarose beads, as described earlier.
Determination of cellular respiration and mitochondrial enzyme activities
Potassium cyanide (KCN)-sensitive endogenous cell respiration was measured polarographically in cells from a 70% confluent flask. Growth media was changed 3 h prior to trypsinizing the cells. Cells were resuspended in DMEM + 10 mm HEPES pH 7.2 buffer, and oxygen consumption was measured using a Clark oxygen electrode in a micro water-jacketed, magnetically stirred cell set at 37°C (Hansatech Instruments Limited, Norfolk, UK). The reaction was inhibited with KCN (700 μM) to obtain the baseline rate, as described (54). All values were normalized to the total cell number.
The activities of the OXPHOS complexes [CI + CIII or NADH-cytochrome c reductase (NCCR) and CIV] and the TCA enzyme citrate synthase were determined spectrophotometrically in freeze-thawed cells as described previously (54). In-gel complex I activity was determined by incubating a BN-PAGE in 2 mm Tris–HCl pH 7.4, 0.1 mg/ml NADH (Sigma) with 2.5 mg/ml iodotetranitrozolium (Sigma) (55).
The abundance of OXPHOS complexes was analyzed by BN-PAGE using a linear 3–12% gradient gel. 2.5 × 106 cells were incubated in 200 μl of PBS and 70 μl of 8 mg/ml digitonin (Sigma) at 4°C 10 min. Cells were spun at 10 000g for 5 min at 4°C and washed two times in PBS. Pellets were resuspended in 100 μl of 1.5 m amminocaproic acid, 50 mm Bis–Tris pH 7.0 and 1 μl of 20% lauryl maltoside (for isolated monomeric complexes analysis) or 10 μl of 10% digitonin (for supercomplexes analysis). After spinning at 22 000g for 30 min at 4°C, the clarified extracts were supplemented with 10 μl of 750 mm aminocaproic acid, 50 mm Bis–Tris pH 7.0, 0.5 mm EDTA and 5% serva blue G. For each sample, extracts from ∼2.5 to 5 × 105 cells were loaded in the gel.
Sucrose gradient analysis
The native mass of COX20 was estimated by following its sedimentation properties in linear sucrose gradients. Mitochondria (1 mg) were extracted in 300 µl of 20 mm Tris pH 7.5, 50 mm KCl, 0.4% lauryl maltoside, 1 mm PMSF and 1× protease inhibitor cocktail. The lysate was spun at 22 000 g for 30 min at 4°C, and the clarified extract was loaded onto a 5 ml 7–20% sucrose gradient containing 20 mm Tris pH 7.5, 50 mm KCl, 0.1% lauryl maltoside and 1 mm PMSF buffer and centrifuged at 28 000 g for 14 h at 4°C. The gradients were fractionated into 13 fractions, TCA precipitated and the pellets resuspended in Laemmli buffer and analyzed by immunoblotting. For each round of ultracentrifugation, a control gradient was loaded with two calibration standards in extraction buffer; 2 mg of hemoglobin and 20 µl of bovine heart lactate dehydrogenase (1000 units/ml, Sigma).
Saccharomyces cerevisiae strains and methods
All yeast strains were in the W303 background (a ade2-1 his3-1,15 leu2-3,112 trp1-1 ura3-1). A W303Δcox20 strain was obtained from Dr A. Tzagoloff (20). The growth media compositions of yeast strains have been described elsewhere (56). Yeast mitochondrial isolation and in vivo protein synthesis were performed as reported (53,57). The ScCox2 and porin antibodies were obtained from Molecular Probes and Abcam, respectively.
All experiments were done at least in triplicate. All data are presented as means ± SD of absolute values or the percentage of control. Values were analyzed for statistical significance by Student's t-test. P < 0.05 was considered significant.
This research was supported by an American Heart Association (AHA) postdoctoral fellowship (to M.B.), and grants-in-aid of research from the NIH (GM071775-06, A.Ba), MDA (A.Ba) and CIHR (S.C.L.). S.C.L. is a New Investigator of the CIHR.
We thank Dr Iliana Soto for critical reading of the manuscript. We thank Dr Tzagoloff for providing reagents.
Conflict of Interest statement. None declared.