Mammalian CcO (cytochrome c oxidase) is a hetero-oligomeric protein complex composed of 13 structural subunits encoded by both the mitochondrial and nuclear genomes. To study the role of nuclear-encoded CcO subunits in the assembly and function of the human complex, we used stable RNA interference of COX4, COX5A and COX6A1, as well as expression of epitope-tagged Cox6a, Cox7a and Cox7b, in HEK (human embryonic kidney)-293 cells. Knockdown of Cox4, Cox5a and Cox6a resulted in reduced CcO activity, diminished affinity of the residual enzyme for oxygen, decreased holoCcO and CcO dimer levels, increased accumulation of CcO subcomplexes and gave rise to an altered pattern of respiratory supercomplexes. An analysis of the patterns of CcO subcomplexes found in both knockdown and overexpressing cells identified a novel CcO assembly intermediate, identified the entry points of three late-assembled subunits and demonstrated directly the essential character as well as the interdependence of the assembly of Cox4 and Cox5a. The ectopic expression of the heart/muscle-specific isoform of the Cox6 subunit (COX6A2) resulted in restoration of both CcO holoenzyme and activity in COX6A1-knockdown cells. This was in sharp contrast with the unaltered levels of COX6A2 mRNA in these cells, suggesting the existence of a fixed expression programme. The normal amount and function of respiratory complex I in all of our CcO-deficient knockdown cell lines suggest that, unlike non-human CcO-deficient models, even relatively small amounts of CcO can maintain the normal biogenesis of this respiratory complex in cultured human cells.
- cytochrome c oxidase
- protein assembly
- respiratory supercomplex
- RNA interference (RNAi)
Cytochrome c oxidase (CcO), the terminal enzyme complex of the mitochondrial electron-transport chain couples the electron transfer from reduced cytochrome c to molecular oxygen with vectorial proton translocation across the inner membrane. The mammalian CcO complex is composed of 13 different polypeptide subunits, which are encoded by both the nuclear and mitochondrial genomes. Mitochondrially encoded Cox1 and Cox2 form the redox site involved in electron transfer. Electrons enter the CcO complex at the binuclear copper site (CuA) in the Cox2 subunit, which also mediates electrostatic binding of cytochrome c . From CuA, electrons pass to other metal centres in the Cox1 subunit, first to haem a and then to a heterobimetallic haem a3/CuB centre . Together with Cox3, mitochondrially encoded subunits constitute the evolutionarily conserved structural core of the enzyme. The remaining ten subunits, which are encoded by nuclear DNA, are associated with the surface of the complex core. These small polypeptides are required for the assembly and stability of the holoenzyme and are thought to function in the regulation of its activity [3–5]. Tissue-specific isoforms of subunits Cox4, Cox6a, Cox6b, Cox7a and Cox8 have been identified in mammals [6,7]. Most CcO subunits have one or more transmembrane domains, with the exception of Cox5a and Cox5b, which are located at the matrix side, and Cox6b, which is associated with the surface of the complex in the intermembrane space .
Subunit Cox4 is the largest nuclear-encoded subunit of the complex. It was shown to be involved in the allosteric inhibition of CcO activity by ATP, which binds to the matrix portion of the subunit . Isoforms 1 and 2 of Cox4 are encoded by two separate genes and are likely to differ with respect to ATP-induced inhibition of CcO activity . In mammalian cells, the first step of CcO assembly is the membrane integration of Cox1, followed by the association of the Cox4–Cox5a heterodimer . Subunit Cox5a binds indirectly to subunit Cox1 via the matrix domain of subunit Cox4 and the extramembrane segment of Cox6c. Subunit Cox6a is involved in the stabilization of the dimeric state of CcO and may contribute to the formation of an interaction site for cytochrome c . Liver-type subunit Cox6a (Cox6a1/Cox6aL) is found in all non-muscle tissues, whereas heart/muscle-type subunit Cox6a (Cox6a2/Cox6aH), which is encoded by a different gene, is expressed only in striated muscles . Subunit Cox6a was shown to associate with the complex, together with subunits Cox7a or Cox7b, at a very late stage of CcO assembly .
In the present study, we generated HEK (human embryonic kidney)-293 cell lines with stably [shRNA (short hairpin RNA)] down-regulated levels of CcO subunits Cox4, Cox5a and Cox6a. We analysed the steady-state levels of the CcO holoenzyme and the presence and composition of CcO subcomplexes. We have demonstrated directly that depletion of each of the selected subunits results in decreased levels of CcO with a parallel accumulation of CcO subcomplexes and unassembled subunits. These changes were accompanied by a reduction in the activity of CcO and a decreased affinity of CcO for oxygen. The subunit composition of CcO subcomplexes in Cox6a-depleted cells, along with analyses of the ectopic expression of subunits Cox7a2 and Cox7b in these cells, provided novel insights into the late stages of human CcO assembly. Moreover, CcO deficiency in COX6A1-KD (knockdown) cells was complemented by ectopic expression of the Cox6a2 isoform. We have demonstrated further that CcO deficiency in COX5A- and COX6A-KD cells affects the formation of respiratory supercomplexes containing the dimeric form of CcO.
The nucleotide sequences of 33 different candidate miR-30based shRNAs [shRNAmirs (microRNA-adapted shRNAs)] targeted to COX4I1, COX5A and COX6A1 mRNAs were designed with shRNA Retriever (http://katahdin.cshl.org/siRNA/RNAi.cgi?type=shRNA), synthesized and cloned into the pCMV-GIN-ZEO plasmid (Open Biosystems) as described previously . A pCMV-GIN-ZEO derivative that expresses negative control (non-silencing) shRNAmir was obtained from Open Biosystems. The coding sequences of COX7A2 (GenBank® accession number BC100852; IMAGE ID: 40002220) and COX7B (GenBank® accession number BC018386; IMAGE ID: 3861730) were amplified from the respective full-length cDNA clones (ImaGenes), fused to the C-terminal FLAG epitope and cloned (EcoRI/NotI) into the modified pmaxFP-Red-N plasmid (Amaxa). The fidelity of all constructs was confirmed by automated DNA sequencing. Plasmids pReceiver-M02 (EX-C0224) and pReceiver-M13 (EX-C0224) (GeneCopoeia) were used to express Cox6a2 and Cox6a2–FLAG respectively.
Cell culture and transfection
HEK-293 cells (CRL-1573) were obtained from A.T.C.C. (Manassas, VA, U.S.A.) and grown at 37 °C in a 5% (v/v) CO2 atmosphere in high-glucose Dulbecco's modified Eagle's medium (PAA) supplemented with 10% (v/v) fetal bovine serum Gold (PAA). Cell lines stably expressing shRNAmir were prepared using the Nucleofector™ device (Amaxa) essentially as described previously . The transient expression of selected CcO subunits in HEK-293 cells was accomplished using the Express-In Transfection Reagent (Open Biosystems).
Reverse transcription and qRT-PCR (quantitative real-time PCR)
Total RNA was isolated from HEK-293 cells using TriReagent solution (MRC). First-strand cDNA was synthesized from 4 μg of total RNA with the use of Superscript III reverse transcriptase (Invitrogen) and Oligo-dT primers (Promega). Pre-amplification and relative quantification was performed according to the manufacturer's instructions (Applied Biosystems). Ten pre-amplification cycles were run with 12.5 μl of cDNA and a 0.05× pooled mixture of eight TaqMan Gene Expression Assays [Hs00971639_m1, COX4I1; Hs00261747_m1, COX4I2; Hs01924685_g1, COX6A1; Hs00193226_g1, COX6A2; Hs00427620_m1, TBP (TATA-box-binding protein); Hs00173304_m1, PPARGC1A (peroxisome-proliferator-activated receptor γ co-activator 1α); Hs00188166_m1, SDHA (succinate dehydrogenase complex subunit A; Hs01082775_m1, TFAM (transcription factor A, mitochondrial)]. Relative quantification was performed in duplicates twice on the 7300 Real-Time PCR System (Applied Biosystems). The transcript levels of all mRNAs were normalized to the level of TBP mRNA. Because the amplification efficiency of the reference and target genes was the same, the comparative ΔΔCt method was used for relative quantification.
Protein sample preparation and signal acquisition for SDS, BN (blue native) and BN-SDS/PAGE immunoblot analysis were performed essentially as described in [11,14]. The immunoblots were developed with SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Scientific). For analysis of respiratory supercomplexes by BN-SDS/PAGE and BN/BN-PAGE, isolated mitochondria were extracted using digitonin (detergent/protein ratio of 6). Primary detection of BN-SDS/PAGE and BN/BN-PAGE immunoblots was performed with mouse monoclonal antibodies (Mitosciences) raised against the complex I subunit NDUFB6 [NADH dehydrogenase (ubiquinone) 1β subcomplex 6], complex III subunit Core1 and complex IV subunit Cox1. The first dimension gel strips for BN/BN-PAGE immunodetection were soaked in cathode buffer containing 0.1% DDM (n-dodecyl-β-D-maltoside) for 15 min and then in cathode buffer containing 0.02% DDM for another 15 min. The second dimension of BN-PAGE separation was performed in the presence of 0.02% DDM as described in .
The activities of respiratory chain complexes were measured spectrophotometrically with a UV-2401PC instrument (Shimadzu). Rotenone-sensitive complex I activity (NADH:ubiquinone oxidoreductase) was measured by incubating 45 μg of mitochondrial protein in 1 ml of assay medium (50 mM Tris/HCl, pH 8.1, 2.5 mg/ml BSA, 50 μM decylubiquinone, 0.3 mM KCN and 0.1 mM NADH without or with 3 μM rotenone). The decrease in absorbance at 340 nm due to the NADH oxidation was followed. Complex II activity (succinate:2,6-dichloroindophenol oxidoreductase) was measured by incubating 20 μg of mitochondrial extract in 1 ml of assay medium (10 mM potassium phosphate, pH 7.8, 2 mM EDTA, 1 mg/ml BSA, 0.3 mM KCN, 10 mM succinate, 3 μM rotenone, 0.2 mM ATP, 80 μM 2,6-dichloroindophenol, 1 μM antimycin and 50 μM decylubiquinone). The decrease in absorbance at 600 nm due to the reduction of 2,6-dichloroindophenol was monitored. CcO activity was measured by incubating 15–18 μg of mitochondrial protein in 1 ml of assay medium (40 mM potassium phosphate, pH 7.0, 1 mg/ml BSA, 25 μM reduced cytochrome c and 2.5 mM DDM) and the oxidation of cytochrome c (II) at 550 nm was monitored. All assays were performed at 37 °C. The total protein was determined using the method of Lowry .
High-resolution respirometry and oxygen kinetics
Oxygen consumption measurements were performed as described previously , except that the cells were permeabilized with 50–75 μg/ml digitonin and 0.5 μM FCCP (carbonyl cyanide p-trifluoromethoxyphenylhydrazone) was used to uncouple and maximally stimulate respiration. The P50 (partial pressure of oxygen at the half-maximal respiration rate) value was measured in the presence of 0.5 μM FCCP and 10 mM succinate essentially as described in . All measurements were performed independently three to six times for each cell line.
Cox4i1 and Cox6a1 isoforms account for the majority of Cox4 and Cox6a transcripts expressed in HEK-293 cells under normal conditions
Because two tissue-specific isoforms of subunits Cox4 and Cox6a were described in humans, we measured the levels of COX4I1, COX4I2, COX6A1 and COX6A2 mRNAs in HEK-293 cells. To quantify both isoforms using qRT-PCR, pre-amplification of the cDNA was necessary. The transcript levels of all mRNAs were normalized to the level of TBP mRNA. The relative normalized level of COX4I1 mRNA was found to be ~2×105 times higher than COX4I2, whereas the relative normalized level of COX6A1 mRNA was ~5×106 times higher than that of COX6A2 (Figure 1A). This finding suggests that the vast majority of Cox4 and Cox6a expressed and assembled within the CcO complex in HEK-293 cells are represented by Cox4i1 and Cox6a1 tissue isoforms.
Subunits Cox4, Cox5a and Cox6a are depleted in HEK-293 cells by the stable expression of shRNAmir
Owing to the considerably long half-life of the CcO complex  and the relatively large amount of material required for subsequent analyses, a stable vector-based RNAi (RNA interference) approach was utilized to down-regulate the expression of Cox4, Cox5a and Cox6a in HEK-293 cells. We designed and constructed 13, nine and 11 pCMV-GIN-ZEO derivatives encoding miR-30-based shRNAs (shRNAmirs) targeting human COX4I1, COX5A and COX6A1 mRNAs respectively. These plasmids were used to generate stable KD in HEK-293 cells. Six out of the 33 produced cell lines, which were found to have substantially lower levels of the target proteins, were selected for further analyses (Figure 1B).
The residual levels of the target subunits in mitochondrial preparations were quantified using denaturing Western blots. Serial dilutions of protein isolated from the mitochondria of the control line expressing non-silencing shRNA were loaded on the same gels as the KD samples so that the steady-state levels of the respective polypeptides in KD cells could be assessed as a percentage of the control values (see Supplementary Figure S1 at http://www.BiochemJ.org/bj/428/bj4280363add.htm). The residual amount of the Cox4 polypeptide was found to be 45% in COX4I1 (sh1) KD and 55% in COX4I1 (sh14) KD mitochondria. In the COX5A (sh5) KD and COX5A (sh7) KD samples, the amount of Cox5a polypeptide was decreased to 35 and 20% respectively. The Cox6a1 polypeptide was decreased to 25 and 35% in COX6A1 (sh11) KD and COX6A1 (sh12) KD mitochondria respectively.
Relative normalized mRNA levels (2−ΔΔCt×100) of the major COX4I1 and minor COX4I2 isoform were 33 and 24% in the COX4I1 (sh1) KD sample, and 18 and 77% in the COX4I1 (sh14) KD sample respectively. The relative normalized mRNA levels of the major COX6A1 and minor COX6A2 isoform accounted for 13 and 21% in the COX6A1 (sh11) KD sample, and 12 and 105% in the COX6A1 (sh12) KD sample respectively.
Steady-state levels of a subset of respiratory chain subunits as well as several other mitochondrial proteins are not affected in COX4-, COX5A- and COX6A-KD cells
The mitochondrial preparations used to quantify target subunits were utilized for SDS/PAGE immunodetection to determine the steady-state levels of selected CcO subunits, representative respiratory chain subunits and several other mitochondrial proteins (see Supplementary Figure S1).
Depletion of subunit Cox4 led to a reduction in the level of Cox5a to 45% in COX4I1 (sh1) KD and 55% in COX4I1 (sh14) KD samples compared with the control. The levels of Cox1 and Cox2 were both 80 and 90% of the control values in COX4I1 (sh1) and (sh14) KD samples respectively. Depletion of subunit Cox5a resulted in a decrease in the levels of Cox4 and Cox2 to 35% in COX5A (sh5) KD and 20% in COX5A (sh7) KD, whereas the level of Cox1 remained unchanged. Down-regulation of subunit Cox6a caused a reduction in the levels of subunits Cox1, Cox2, Cox4 and Cox5a to 80% of the control values. The steady-state levels of PDH-E2 (pyruvate dehydrogenase subunit E2), porin, cytochrome c, the 70-kDa flavoprotein subunit of SDHA, the ATP synthase subunit F1-β (ATPase F1-β), complex III core protein 2 and the complex I subunit NDUFA9 [NADH dehydrogenase (ubiquinone) 1α subcomplex 9] were not changed compared with the control.
Depletion of subunits Cox4, Cox5a and Cox6a decreases the quantity of CcO holoenzyme and increases an accumulation of CcO subcomplexes
To study the impact of the depletion of subunits Cox4, Cox5a and Cox6a on CcO holoenzyme levels, we performed 5–15% BN-PAGE Western blot analyses. The same mitochondrial fractions used for SDS/PAGE Western blots were solubilized in DDM and analysed along with serial dilutions of control samples (Figure 2A). The amount of CcO holoenzyme was diminished to 65% of the control value in COX4I1 (sh1) KD and to 75% of the control value in COX4I1 (sh14) KD samples. The COX5A (sh7) and (sh5) KD samples showed a decrease in the amount of the CcO holoenzyme to 20 and 60% of the control values respectively. In COX6A1 (sh11) KD and COX6A1 (sh12) KD samples, CcO holoenzyme was reduced to 25 and 30% of the control values respectively.
To investigate the assembly pattern of CcO, we performed BN-PAGE as well as two-dimensional BN-SDS/PAGE, in which we utilized 8–16% polyacrylamide gradients followed by Western blot analysis of the selected subunits. The BN anti-Cox1 immunoblots revealed the presence of several distinct bands in KD mitochondria, which were denoted a–e (Figure 2B). The holoenzyme band a was, in fact, found to be composed of two bands (a1 and a2) . COX6A1-KD mitochondria showed a reduction in the levels of both bands a1 and a2 to 20% of the control values (Figures 2B and 2C). In contrast, the decrease in band a1 in COX4- and COX5A-KD samples, which was comparable with that for COX6A1-KD mitochondria, was accompanied by a more profound reduction in the amount of band a2 (Figures 2B and 2C). In addition to band a, anti-Cox1 immunoblotting detected three faster migrating bands (b, x and c) in all samples with approximate molecular masses of 180, 155 and 110 kDa respectively. Band b, which was barely discernible between control, COX4-KD and COX5A-KD mitochondria, was increased in COX6A-KD samples (Figure 2B). This band probably represents the previously identified assembly intermediate S3 (Figure 2D) . Band x, which was almost undetectable in controls and slightly increased in COX6A-KD mitochondria, was markedly increased in the COX5A (sh7) KD sample. Despite its high apparent molecular mass, this band was detectable only with the anti-Cox1 antibody. Band c, which was detected in all samples, was increased in COX4- and COX5A-KD mitochondria. This band was detected with antibodies against Cox1, Cox4 and Cox5a in COX6A1-KD samples, but only with the anti-Cox1 antibody in COX4- and COX5A-KD samples (Figure 2D). Finally, bands d and e, with approximate molecular masses of 100 and 85 kDa respectively, were found only in COX4- and COX5A-KD samples. These bands were increased in the severely CcO-deficient COX5A (sh7) KD sample and were detected exclusively with the anti-Cox1 antibody.
Whereas CcO activity is significantly decreased, activities of respiratory chain complexes I, II and III are not affected in COX5A- and COX6A1-KD cells
To assess residual CcO activity in mitochondrial preparations, we determined CcO activity and normalized it to the activity of SQR (succinate:coenzyme Q10 reductase, complex II) (Figure 1B). The same mitochondrial preparations used in the electrophoretic analyses were utilized in these assays. The relative CcO/SQR ratio values obtained for mitochondria from the COX4I1 (sh1) (97%) and COX4I1 (sh14) (103%) KD cells were comparable with that obtained for control cells. In COX5A (sh5) KD and COX5A (sh7) KD mitochondria, the relative normalized CcO/SQR ratio accounted for 72 and 57% of the control values respectively. The mitochondrial preparations from COX6A1 (sh11) KD and COX6A1 (sh12) KD cells revealed a decrease in the CcO/SQR ratio to 51 and 73% of the control values respectively. The activities of the remaining respiratory chain complexes were found to be unchanged in all of the KD samples investigated (results not shown).
Isolated CcO deficiency due to COX5A and COX6A1 KD affects the organization of respiratory supercomplexes in HEK-293 cells
To study the impact of COX5A and COX6A1 KD on the composition of respiratory chain supercomplexes, mitochondrial fractions were solubilized with digitonin (digitonin/protein ratio of 6) and analysed by BN-SDS/PAGE and BN/BN-PAGE Western blotting. Both approaches showed that COX5A and COX6A1 KD significantly decreased the levels of the dimeric form of the CcO holoenzyme (Figure 3). Furthermore, BN/BN-PAGE Western blots revealed markedly reduced levels of supercomplexes III2IV2 and I1III2IV2, which apparently contained the dimeric form of the CcO holoenzyme (Figure 3B). In contrast, the amount of the major mammalian supercomplex I1III2IV1 was normal or slightly increased (Figure 3B). The appearance of the faint spot corresponding to undissociated IV2 below the spot corresponding to III2 of the III–IV-containing supercomplex substantiates the presence of the III2IV2, rather than the III2IV1 supercomplex in the samples (Figure 3B, long exposure). The two spots migrating above complex I probably represent undissociated supercomplexes (e.g. I1III2IV1 and I1IV1) .
In all samples, including the controls, BN/BN-PAGE immunoblots revealed additional spots that migrated at the level of supercomplexes in the first dimension, but had apparent molecular masses lower than that of the CcO holoenzyme (Figure 3B). These spots were most apparent in COX5A-KD mitochondria, which exhibited the most severe CcO holoenzyme defect and the highest accumulation of incomplete CcO assemblies (Figure 2B). Indeed, these spots might represent CcO assembly intermediates already in the form of respiratory supercomplexes.
Normal normoxic respiration is accompanied by increased P50 values in COX5A- and COX6A1-KD cells
To further characterize mitochondrial respiratory function in COX5A (sh7) KD and COX6A1 (sh11) KD cells, we performed high-resolution respirometry using the Oroboros Oxygraph. The only difference observed in this assay was diminished oxygen consumption (~80% of control) after the addition of ascorbate/TMPD (N,N,N′,N′-tetramethyl-p-phenylenediamine) substrates in COX5A (sh7) KD cells (Figure 4A). Therefore we analysed the respiratory response of the cells to low oxygen. The digitonin-permeabilized cells were treated with FCCP to uncouple and maximally stimulate respiration (state 3u) and were then treated with succinate. The oxygen kinetics were quantified by determining the P50. The average P50 value was 0.072±0.02 (S.D.) kPa (n=6) in control (non-silencing) cells, 0.163±0.01 kPa (n=5) in COX5A (sh7) KD cells and 0.105±0.01 kPa (n=3) in COX6A1 (sh11) KD samples (Figure 4B). Thus the P50 value was elevated more than 2-fold in COX5A (sh7) KD and ~1.5-fold in COX6A1 (sh11) KD cells.
Overexpression of the Cox6a2 isoform complements the quantitative CcO holoenzyme defect of COX6A1-KD cells
Despite the substantial decrease in CcO holoenzyme levels, the levels of heart/muscle-specific COX6A2 mRNA remained low in Cox6a1-depleted CcO-deficient cells (Figure 1A). Therefore we tested whether the Cox6a2 isoform can substitute for subunit Cox6a1 during CcO assembly in HEK-293 cells. To obtain optimal accumulation of the subunit polypeptide, the cells were transfected twice consecutively, leading to transgene expression for a total of 4 days. Both untagged and FLAG-tagged forms were used to transfect wild-type and COX6A1-KD cells. This analysis was simplified by the fact that both COX6A1 shRNAmirs used to prepare the KD cells targeted the 3′-UTR (untranslated region) of COX6A1 mRNA, which was not present in our Cox6a2 expression constructs.
The ectopic expression of the Cox6a2 isoform in COX6A1 (sh11) KD cells led to the restoration of wild-type CcO levels. The total pool of CcO in these cells consisted of the Cox6a2-containing enzyme as well as residual Cox6a1-containing molecules (~25%) (Figure 5B). Therefore both untagged (Figures 5A and 5B) and FLAG-tagged (Figures 5A and 5C) Cox6a2 polypeptides could substitute for the Cox6a1 isoform during CcO assembly in HEK-293 cells. From these experiments, it was also apparent that ectopic expression of Cox6a2 in wild-type HEK-293 cells increased the amount of CcO holoenzyme (band a1) as well as the level of band a2. In all transfected samples, including controls, an increase in the amount of the assembly intermediate S3 (band b) was observed, probably as a result of the transfection treatment (Figures 5A and 5B).
The consequences of increased CcO content after ectopic expression of the Cox6a2 isoform in COX6A1-KD cells were analysed by spectrophotometric measurement of CcO activity. CcO activity in Cox6a2–FLAG expressing cells, which was normalized to the activity of the control enzyme (SQR), did not change significantly compared with either untreated wild-type cells or wild-type cells transfected with the control vector (1.26–1.29). In COX6A1-KD cells, ectopic expression of the Cox6a2–FLAG isoform led to a ~1.5-fold increase in CcO/SQR activity ratio compared with transfection with the control vector (CcO/SQR ratio was 0.75 and 0.49 respectively) (Figure 5D).
Subunit Cox7a2 enters the CcO assembly after Cox7b, but before Cox6a2
To elucidate the entry points of subunits Cox7a2 and Cox7b with respect to the formation of the assembly intermediate S3, which shows increased accumulation in COX6A1-KD cells, we transfected COX6A1-KD cells with expression constructs containing Cox7a2- and Cox7b-coding sequences fused to a C-terminal FLAG epitope. To obtain optimal accumulation of the subunit polypeptides, the cells were transfected twice consecutively, leading to transgene expression for a total of 4 days. Subsequently, isolated mitochondria were resolved by BN (7–10% acrylamide gradient)-PAGE and subjected to Western blot analysis using antibodies specific to Cox1 and the FLAG epitope. No changes in the amount of the CcO holoenzyme or S3 intermediate were observed in COX6A1-KD mitochondria upon expression of these polypeptides (Figure 5C).
In contrast with the anti-Cox1 immunoblot, which detected the S3 intermediate in all of the analysed samples (Figure 5C, long exposure of band b), the FLAG-specific antibody cross-reacted with this assembly intermediate only in mitochondria from cells expressing Cox7b–FLAG. The fact that Cox1-specific, but not FLAG-specific, antibodies detected S3 in the samples expressing Cox7a2–FLAG indicated that, unlike Cox7b, Cox7a2 is incorporated after the formation of the assembly intermediate S3 (Figure 5C).
Mitochondrial preparations from cells transfected with the Cox6a2–FLAG construct were also analysed for the presence of expressed subunits in the late assembly forms of CcO. Despite the high cross-reactivity of both bands a1 and a2 with the anti-Cox1 antibody in these samples, the FLAG-specific antibody detected band a2 only in Cox7a2–FLAG- and Cox7b–FLAG-expressing samples (Figure 5C). Furthermore, when endogenous Cox6a1 was probed, only a1 was detected in samples, whereas both bands a1 and a2 were confirmed with antibodies specific to Cox1, Cox2 or Cox6c (Figures 5A and 5B). These results indicate that subunit Cox6a is incorporated after the formation of band a2, probably as the last structural subunit incorporated into the human CcO complex.
CcO is a crucial cellular enzyme with a central role in oxidative metabolism [2,20]. The significant interest in the biogenesis of CcO stems from its clinical importance. Defective CcO biogenesis is frequently related to severe mitochondrial diseases that often involve tissues with a high energy demand [2,21,22].
It was demonstrated previously in various non-human eukaryotic models that KD of Cox4 or Cox5a affects CcO holoenzyme content [23–26]. Deficiency of subunit Cox6a was described to cause severe CcO deficiency in Drosophila  and mice . However, the yeast Cox6a homologue is dispensable for assembly of the yeast CcO complex . In the present study, we have demonstrated that the impact of stable down-regulation of Cox4i1, Cox5a and Cox6a1 in cultured human cells on the content of CcO corresponds to similar studies in higher eukaryotes. Furthermore, our results provide detailed KD-specific patterns of CcO assembly intermediates and indicate several novel aspects of the sequential incorporation of subunits during the assembly process, including identification of a novel assembly intermediate in human CcO assembly.
Overall, the accumulation of CcO subcomplexes seen in our KD cells reflected the proposed entry of individual subunits into the assembly. The effects of COX6A KD on CcO subcomplex pattern was manifested mainly by the marked increase in the assembly intermediate S3. Unlike COX4- and COX5A-KD mitochondria, in which the reduction in the holoenzyme band a1 was accompanied by a more pronounced decrease in band a2, Cox6a-deficient mitochondria retained significantly higher levels of the latter species. These results indicate that band a2 represents another rate-limiting step in human CcO assembly. Both Cox6a and Cox6b subunits, which are thought to be responsible for dimerization of CcO, are the first to dissociate from the bovine complex under various destabilizing conditions [30,31]. Thus it appears plausible that the elevated levels of the assembly intermediate S3 in COX6A-KD cells could stem from compromised binding of Cox6b, which might be contingent upon the concomitant assembly of Cox6a. KD of both early-assembled subunits Cox4 and Cox5a resulted in accumulation of four subcomplexes consisting merely of subunit Cox1. The absence of Cox1–Cox4/Cox5a heterodimers confirmed further the interdependence of the assembly of Cox4 and Cox5a [11,32]. Together with the lack of accumulation of higher-molecular-mass intermediates, our findings suggest that the assembly of the Cox4–Cox5a heterodimer with Cox1 is necessary for the subsequent association of Cox2, and thus for the rest of the assembly to proceed. The Cox1-containing subcomplex x of ~155 kDa, which apparently does not contain any of the CcO subunits that we tested, was found to be particularly increased in COX5A-KD mitochondria. The composition and migration of this subcomplex suggests that it might represent an off-path complex that is not relevant to the normal route of assembly. However, the other samples, including controls, contained small amounts of this subcomplex as well. A similar situation in terms of discrepancy between subunit composition and the apparent molecular mass was observed in the case of the 110 kDa subcomplex c from COX4- and COX5A-KD mitochondria. These results strongly suggest that individual CcO subunits as well as CcO subcomplexes associate during the assembly with several non-subunit proteins, as reported for yeast Cox1 . The expression of FLAG-tagged versions of Cox7a2, Cox7b and Cox6a2 in both wild-type and Cox6a1-deficient backgrounds allowed us to elucidate for the first time the very late events in human CcO assembly. We identified the particular entry points of these three subunits and demonstrated the significance of the CcO holoenzyme bands a1 and a2. According to our results, band a1 probably represents the 13-subunit CcO holoenzyme (S4). In contrast, band a2 lacks subunit Cox6a, which appears to be added as the last assembled structural subunit. Band a2 (S4*) is formed by the addition of Cox7a2 and probably Cox6b1  to the assembly intermediate S3. Finally, subunit Cox7b was found to join the assembling complex during or at the end of the formation of the assembly intermediate S3 (Scheme 1).
The amount of CcO holoenzyme in COX5A (sh7) KD and both COX6A1-KD samples corresponded to the residual content of the targeted subunits. On the basis of our assertion that the Cox6a subunit completes the assembly of the CcO holoenzyme, the amount of fully assembled CcO seems to be strictly contingent upon the amount of the residual Cox6a. Furthermore, transient overexpression of subunit Cox6a in wild-type HEK-293 cells increased the content of CcO holoenzyme. Lower levels of S3 intermediate and a shift in the a1/a2 ratio towards a1 in COX5A (sh7) KD samples suggest changes in the kinetics of CcO assembly in terms of the increased utilization of intermediates subsequent to the integration of Cox5a, thereby increasing the level of complete CcO complex to the level of Cox5a. A similar reduction in the levels of RNAi-target subunits in COX5A (sh5) KD and COX6A1 (sh12) KD cells resulted in a milder reduction of CcO holoenzyme levels in Cox5a-deficient cells. This observation suggests the existence of different threshold values for the availability of various CcO subunits during enzyme assembly. The diminished content of CcO holoenzyme in our KD cells resulted in a milder reduction in specific CcO activity. Indeed, the relatively mild Cox4 reduction in COX4I1-KD cells resulted in virtually normal CcO activity. These observations are in line with the well-established high excess capacity of CcO [20,34].
Diaz et al.  demonstrated that the total loss of CcO leads to complex I deficiency in a mouse COX10-knockout model. Furthermore, RNAi KD of Cox4 in murine cultured cells resulted in decreased stability and activity of complex I . A similar effect of COX4- and COX5A-KD on complex I was described in Caenorhabditis elegans, where the content of complex I was normal, but its enzymatic activity was reduced significantly . In contrast, the CcO deficiency in our COX4-, COX5A- and COX6A-KD cells did not lead to any significant changes in the amount and/or activity of other complexes of oxidative phosphorylation system, including complex I. Similarly, normal levels and activity of complex I were also found in murine COX5B-KD cells with severe CcO deficiency . Further work is needed to establish the potential role of fully assembled CcO in complex I biogenesis.
High-resolution respirometry in our CcO-deficient cellular model showed significantly increased P50 values. This may indicate either a change in mitochondrial capacity or decreased affinity for oxygen (=1/P50) of the residual CcO . The lack of a significant decrease in maximal (state 3u) normoxic respiration in our KD cells argues against the possibility that P50 is increased due to lower mitochondrial capacity. The second possibility, which involves a decrease in oxygen affinity in active states when the demand for oxygen is highest, is paradoxical from a functional point of view . However, if the increase in P50 is accompanied by a greater increase in maximum flux (Jmax), the Jmax/P50 ratio still increases and indicates elevated apparent catalytic efficiency with activation by ADP . When plotted as a function of the specific oxygen flux normalized to protein content, the P50 values in both our KD cell lines increased without a corresponding increase in oxygen flux (Figure 4C). Therefore the increase in P50 in our KD cells is most likely to be due to decreased affinity of CcO for oxygen, i.e. an increased apparent Km for oxygen. Marked CcO deficiency with almost unchanged normoxic cellular respiratory rates accompanied by increased P50 values were also found in fibroblasts from patients with Leigh syndrome owing to Surf1 deficiency . In addition to reduced holoenzyme levels, Surf1-deficient fibroblasts were characterized by increased accumulation of CcO subcomplexes. It was hypothesized that, under conditions of high oxygen pressure, the accumulated subcomplexes could restore near normal respiration, and contribute to the increased P50 value [17,37]. However, it was repeatedly demonstrated that subcomplexes found in Surf1-deficient fibroblasts are devoid of the catalytic core subunit Cox2 [11,38]. Similar subcomplexes, i.e. devoid of Cox2, also accumulate in COX5A-KD mitochondria. Cox2 is known to mediate electrostatic binding of cytochrome c and represents the initial entry site for electrons. The subcomplexes in COX6A1-KD cells are represented mainly by incomplete CcO assemblies that lack only few peripheral nuclear-encoded subunits. Exposure of isolated monomeric bovine CcO to hydrostatic pressure results in a mixture of 13-, 11- and nine-subunit CcO complexes . Analysis of separated forms showed that both the 13-subunit complex as well as the 11-subunit form devoid of subunits Cox6a and Cox6b retained 85–90% of electron transport activity of the untreated enzyme. The nine-subunit form that lacks two additional subunits (Cox3 and Cox7a) had only 40–45% activity of intact enzyme. The authors were not able to determine the loss of which of the two subunits led to such marked reduction in enzyme activity . Nevertheless, the results of other studies suggest that the loss of Cox3 is likely to be responsible for the majority of CcO inactivation observed [39,40]. Both of the major intermediates found in our Cox6a-deficient cells (a2 and S3) are thus likely to be capable of electron transfer, unlike the subcomplexes from COX5A-KD cells. Indeed, the capacity of CcO in COX5A-KD cells as measured by oxygen consumption after ascorbate/TMPD was significantly decreased, and the P50 value was significantly increased when compared with both control and COX6A1-KD samples. Therefore it appears likely that the presence of the high-molecular-mass CcO subcomplexes in COX6A-KD cells might contribute to cellular respiration at low oxygen levels and lead to higher apparent CcO excess capacity. In contrast, uncoupled respiration did not change at normoxic levels of oxygen after natural substrates, and spectrophotometric measurement indicated a decrease of specific CcO activity to the same extent in both cell lines.
Nonetheless, the decreased oxygen affinity of CcO in both KD cell lines appears paradoxical, as there apparently seems to be no rational reason for the decreased catalytic competence of the residual CcO enzyme. Furthermore, the substantial reduction in CcO content was accompanied by less diminished specific activity of CcO in COX5A- and COX6A1-KD cells, suggesting compensatory catalytic stimulation of the residual CcO. Taken together, we assume that decreased oxygen affinity of CcO in our KD cells results from quantitative changes in the CcO pool which, when reduced, undergoes normal flux at high oxygen pressure, but is significantly more sensitive to decreased oxygen levels. Such an assumption is in line with the oxygen dependence of CcO flux control demonstrated previously . In this study, a lower experimental concentration of oxygen was related to a steeper initial slope of the cyanide titration curve for maximal succinate oxidation rate. Moreover, a lower concentration of CcO inhibitor was necessary to cause complete inhibition of respiration .
Both COX5A- and COX6A-KD mitochondria showed, in addition to a reduction in the CcO monomer, an even more pronounced reduction of the dimeric form of CcO. Indeed, the pattern of respiratory supercomplexes showed that species containing the dimeric form of the CcO complex (III2IV2 and I1III2IV2) are significantly reduced in KD cells. In contrast, supercomplexes containing monomeric CcO were unaltered or even slightly increased. A similar effect was shown recently to stem from a CcO defect induced by KD of the human CcO-specific copper metallochaperone Cox17 . The majority of mammalian CcO is thought to exist in a dimeric form within the inner mitochondrial membrane . Indeed, it appears that dimerization of CcO plays a crucial structural and functional role, conferring maximal structural stability on the complex .
In yeast, it was shown that complex IV associates with complex III already in the form of incomplete subcomplexes, and that some of the late assembled subunits are probably added directly to the III/IV supercomplex . An example of such a subunit is Cox13, the yeast homologue of human Cox6a . Indeed, BN/BN-PAGE immunoblots for both control and down-regulated cells revealed additional spots that migrate at the level of supercomplexes in the first dimension, but with an apparent molecular mass lower than that of the CcO holoenzyme. Although we cannot, in principle, exclude the removal of some of the peripheral subunits because of detergent treatment, these spots might represent incomplete CcO assemblies that are already present in supercomplexes. Further analyses are required to confirm this hypothesis.
Despite remarkable conservation of secondary structure in Cox6a isoforms, mature Cox6a1 and Cox6a2 subunits (non-muscle and heart/muscle isoform) were found to share lower intra-species (approx. 60%) than inter-species (80–88%) amino acid sequence identity in humans, rats and cows, suggesting that the COX6A1 and COX6A2 genes arose before mammalian radiation . The tissue-specific pattern of these isoforms is established during tissue differentiation [47–49]. A switch from COX6A1 to COX6A2 isoforms was described to occur during mammalian postnatal development in skeletal muscle and heart as well as during differentiation of myogenic cells in vitro, and is assumed to be essential for normal function of tissues with high aerobic metabolic demands [48,49]. In the heart of mice lacking the COX6A2 gene, the content of the CcO holoenzyme was virtually equal to the level of Cox6a1 isoform in wild-type control, which accounted for 20% of the total Cox6a content. Thus the content of CcO comprising Cox6a1 appeared to be unchanged; in other words, without apparent compensation by induction of the expression of subunit Cox6a1 in the knockout heart . HEK-293 cells used in this study are derived from embryonic kidney tissue, which is specific by almost exclusive (both prenatal and postnatal) expression of the COX6A1 isoform [47–49]. Indeed, the vast majority of the COX6A transcripts in HEK-293 cells is represented by the COX6A1 isoform. The COX6A1 RNAi constructs were designed to target the 3′-UTR of the COX6A1 transcript, which is lacking in the COX6A2 isoform. Indeed, similar to the results from COX6A2-knockout heart, the decrease in the major isoform in our COX6A1-KD cells was not accompanied by up-regulation of the minor isoform (COX6A2). However, ectopic expression of Cox6a2 in COX6A1-KD cells tends to complement the CcO defect. It therefore appears that the mechanisms responsible for maintenance of tissue-specific patterns of COX6A isoforms cannot react in response to actual cellular state. In contrast, it is also possible that the pronounced biochemical defect still did not provoke sufficient physiological impairment, which would otherwise trigger the expression of the minor subunit.
In conclusion, our results indicate that, whereas nuclear-encoded CcO subunits Cox4 and Cox5a are required for the assembly of the functional CcO complex, the Cox6a subunit is required for the overall stability of the holoenzyme. Consequently, the heterogeneous CcO population of Cox6a-deficient cells exhibits higher residual respiration at low oxygen levels than the various CcO forms found in COX5A-KD cells. The fact that the ectopic expression of heart/muscle-specific isoform of Cox6a can complement the CcO defect in COX6A1-KD cells is in sharp contrast with unaltered levels of this isoform in our CcO-deficient model, and suggests the existence of a fixed differentiation programme regarding human Cox6a isoforms. The description of a novel assembly intermediate at the very last step of CcO assembly suggests additional regulatory level of the process. The normal amount and function of complex I in all of our CcO-deficient cell lines suggest that even relatively small residual amounts of CcO can maintain normal biogenesis of this respiratory complex in human cells.
Daniela Fornuskova designed the study, prepared the KD cell lines, performed Western blot analyses and wrote the paper. Lukas Stiburek helped to interpret the data and write the paper. Laszlo Wenchich performed high-resolution respirometry. Kamila Vinsova quantified mRNA levels. Hana Hansikova measured specific enzyme activities. Jiri Zeman was involved in experimental planning, data analysis and writing the final paper.
This work was supported by grants from the Grant Agency of Charles University [grant number GAUK 1/2006/R], the Grant Agency of Czech Republic [grant number GACR 303/07/0781], the Internal Grant Agency of the Ministry of Health of the Czech Republic [grant number IGA MZ NS 10581/3] and the Center of Applied Genomics [grant number CAG 1M0520], and by institutional project [grant number MSM 0021620806].
Abbreviations: BN, blue native; CcO, cytochrome c oxidase; DDM, n-dodecyl-β-D-maltoside; FCCP, carbonyl cyanide p-trifluoromethoxyphenylhydrazone; HEK, human embryonic kidney; KD, knockdown; NDUFB6, NADH dehydrogenase (ubiquinone) 1β subcomplex 6; P50, partial pressure of oxygen at half-maximal respiration rate; qRT-PCR, quantitative real-time PCR; RNAi, RNA interference; SDHA, succinate dehydrogenase complex subunit A; shRNA, short hairpin RNA; shRNAmir, microRNA-adapted shRNA; SQR, succinate:coenzyme Q10 reductase, complex II; TBP, TATA-box-binding protein; TMPD, N,N,N′,N′-tetramethyl-p-phenylenediamine; UTR, untranslated region
- © The Authors Journal compilation © 2010 Biochemical Society