Tissue- and species-specific differences in cytochrome c oxidase assembly induced by SURF1 defects.

Mitochondrial protein SURF1 is a specific assembly factor of cytochrome c oxidase (COX), but its function is poorly understood. SURF1 gene mutations cause a severe COX deficiency manifesting as the Leigh syndrome in humans, whereas in mice SURF1(-/-) knockout leads only to a mild COX defect. We used SURF1(-/-) mouse model for detailed analysis of disturbed COX assembly and COX ability to incorporate into respiratory supercomplexes (SCs) in different tissues and fibroblasts. Furthermore, we compared fibroblasts from SURF1(-/-) mouse and SURF1 patients to reveal interspecies differences in kinetics of COX biogenesis using 2D electrophoresis, immunodetection, arrest of mitochondrial proteosynthesis and pulse-chase metabolic labeling. The crucial differences observed are an accumulation of abundant COX1 assembly intermediates, low content of COX monomer and preferential recruitment of COX into I-III2-IVn SCs in SURF1 patient fibroblasts, whereas SURF1(-/-) mouse fibroblasts were characterized by low content of COX1 assembly intermediates and milder decrease in COX monomer, which appeared more stable. This pattern was even less pronounced in SURF1(-/-) mouse liver and brain. Both the control and SURF1(-/-) mice revealed only negligible formation of the I-III2-IVn SCs and marked tissue differences in the contents of COX dimer and III2-IV SCs, also less noticeable in liver and brain than in heart and muscle. Our studies support the view that COX assembly is much more dependent on SURF1 in humans than in mice. We also demonstrate markedly lower ability of mouse COX to form I-III2-IVn supercomplexes, pointing to tissue-specific and species-specific differences in COX biogenesis.

Introduction

Mammalian oxidative phosphorylation system (OXPHOS) consists of five multisubunit protein complexes and two mobile electron carriers — ubiquinone and cytochrome c. Electron transporting complexes I–IV (cI–cIV) form the respiratory chain (RC), where transfer of electrons from reducing equivalents to molecular oxygen leads to proton pumping across the inner mitochondrial membrane (IMM), resulting in mitochondrial proton gradient formation. ATP synthase, complex V (cV), then uses electrochemical potential of the proton gradient as a driving force for ATP synthesis. Organization of RC complexes in the IMM appears to be rather dynamic and individual RC complexes coexists with respiratory supercomplexes (SCs) composed of cI, cIII and cIV [2]. As proposed by the “plasticity model” of the RC organization, SCs differ in various tissues and cell types and their composition could be regulated according to actual energetics demands and substrate availabilities [1].

Complex IV - cytochrome c oxidase (COX, cIV), the terminal enzyme of the RC transfers electrons from reduced cytochrome c to oxygen molecule embedded in its structure. In mammals, COX can be detected as a monomer, dimer or as a part of several SCs. COX is formed by 14 different subunits. Three largest subunits COX1, COX2 and COX3 are coded for by mitochondrial DNA (mtDNA) and represent the catalytic core of the enzyme. Ten subunits (COX4, COX5a, COX5b, COX6c, COX7b, COX7c, COX8, COX7a, COX6b, COX6a) encoded by nuclear genes are involved in COX regulation, assembly, stability and dimerization [20], [26], [58]. Recently, the NDUFA4, formerly described as complex I subunit was recognized as the 14th nuclear encoded subunit of COX [6]. NDUFA4 is loosely attached to the assembled COX complex and appears to be essential for enzyme biogenesis [48]. COX molecules also contain several metal cofactors in two copper sites (CuA, CuB) and two heme moieties (heme a and a). Mammalian COX assembly pathway proceeds via four/five step-by-step assembly intermediates S1–S2–S3–S4*–S4, where S4 represents a fully assembled COX monomer [20]. COX biosynthesis and assembly of individual subunits is a highly regulated process, depending on many ancillary/assembly proteins. They are essential for different steps of COX biogenesis, from regulation of expression of catalytic core subunits (LRPPRC, TACO1, hCOA3, COX14) [15], [64], [65], [68], through copper metabolism and insertion (COX17, SCO1, SCO2, COX11, COX19, COA6, COX20) [10], [22], [31], [32], [33], [34], [43], heme a biosynthesis and insertion (COX10, COX15, FDX2) [3], [4], [51], to membrane insertion and processing of catalytic core subunits (OXA1l, COX18) [60]. A few other COX assembly proteins have been identified; they participate in early (SURF1, COA5) [24], [60] or intermediate stages (PET100) [36] of COX biogenesis, but their precise function is as of yet unknown.

SURF1 is a 30 kDa hydrophobic protein localized in the IMM, encoded by a SURF1 nuclear gene, which is part of a highly conserved housekeeping gene cluster, the surfeit locus [19], [60], [62], [69]. Up to now, SURF1 is supposed to be involved in a formation of S2 assembly intermediate, most likely in association of COX2 subunit with COX1–COX4–COX5a subassembly [59], [66]. However, its function might be more redundant, because studies on yeast homolog Shy1 indicate, that Shy1/SURF1 might play a role in heme a transfer/insertion into COX1 subunit [7], [57]. Tissue-dependent copper deficiency was found in patients harboring SURF1 gene mutations, which points to a possible function of SURF1 in maintaining of proper cellular copper homeostasis [58]. Moreover, a recently identified MITRAC12 protein [38] was found to interact with SURF1 and COX1 in a mitochondrial translation regulation assembly intermediate of COX1, which further extends possible roles of the SURF1 in COX biogenesis.

Mutations in the human SURF1 gene result in a severe reduction of fully assembled, active COX and accumulation of COX assembly intermediates in patients' cells and tissues. SURF1 mutations manifest usually several months after birth as a fatal neurodegenerative mitochondrial disorder, Leigh syndrome (LS) [52], [70]. In recent studies 74 known SURF1 gene mutations have been summarized and linked to LS and atypical LS [35], but without genotype–phenotype correlation [5], [12], [46], [47], [61], [63]. To better understand SURF1 function, SURF1 knockout mouse (SURF1−/− mouse) model was generated [18]. SURF1−/− mice were smaller at birth, had mild reduction in motor skills at adult age and COX activity was found to be mildly reduced in all tissues examined. Interestingly, SURF1−/− mice showed prolonged lifespan compared to wild-type littermates that was later assigned to enhanced insulin sensitivity and increased mitochondrial biogenesis [17], [49]. Animals were also protected from neuronal damage induced by kainic acid accompanied by reduced mitochondrial uptake of calcium ions [18]. In addition, recent study revealed that loss of SURF1 initiates mitochondrial stress response pathways, including mitochondrial biogenesis, the UPRMT and Nrf2 activation [49].

In the present study, we used the SURF1−/− mouse model for detailed analysis of disturbed COX assembly and COX ability to incorporate into respiratory SCs in different tissues and fibroblasts. Furthermore, we examined SURF1−/− mouse fibroblasts in comparison with human fibroblasts of patients with SURF1 mutations to reveal interspecies differences in kinetics of COX biogenesis pathway, from assembly intermediates to SCs. We show an accumulation of abundant COX1 assembly intermediates and preferential recruitment of COX into I–III2–IVn SCs in SURF1 patient fibroblasts, whereas SURF1−/− mouse fibroblasts were characterized by much milder decrease in COX monomer, which was also more stable. Interestingly, murine COX, both in the wild type and in SURF1 knockout showed only limited preference towards the formation of SCs.

Material and methods

Experimental material

For experiments different tissues were obtained from 3-month-old SURF1−/− knockout B6D2F1 mice [18], generated by the insertion of a loxP sequence in exon 7 of the mouse SURF1 gene, leading to an aberrant, prematurely truncated and highly unstable protein, and from control wild type SURF1+/+ mice. Immortalized skin fibroblasts from control and SURF1−/− mouse [18] were cultured at 37 °C in 5% atmosphere of CO2 in a DMEM medium supplemented by 10% fetal bovine serum, 20 mM HEPES (pH 7.5) and geneticin (50 μg/ml). The same conditions were used for cultivation of human patients' skin fibroblasts lacking the SURF1 protein due to 845 del CT mutations of SURF1 gene [44] and from controls, except that geneticin was replaced with penicillin (10 μg/ml) and streptomycin (10 μg/ml). The project was approved by the ethics committees of Institute of Physiology, CAS. Informed consent was obtained from the parents of the patients according to the Declaration of Helsinki of the World Medical Association.

Isolation of mitochondria

Muscle (hind leg) was minced in a K medium (150 mM KCl, 2 mM EDTA, 50 mM Tris, pH 7.4) supplemented with protease inhibitor cocktail (1:500, PIC from Sigma) and homogenized by ultra turrax IKA (2 × for 15 s, level 4) and glass-teflon homogenizer (600 rpm, 5 strokes). 5% (w/v) homogenate was centrifuged 10 min at 600 g and postnuclear supernatant was centrifuged 10 min at 10,000 g. Pelleted mitochondria were washed once (10,000 g, 10 min) and resuspended in K medium.

Liver mitochondria were isolated from 10% homogenate prepared in STE medium (250 mM sucrose, 10 mM Tris, 2 mM EDTA, pH 7.2) supplemented with PIC (1:500) using glass-teflon homogenizer (600 rpm, 7 strokes). Postnuclear (800 g, 10 min) supernatant filtered through a gauze was centrifuged for 15 min at 5200 g, pelleted mitochondria were washed twice (13,000 g, 10 min) in STE with PIC and then resuspended in STE medium.

Heart mitochondria were isolated essentially as liver mitochondria, except that postnuclear supernatant was centrifuged for 10 min at 13,000 g.

Fibroblast mitochondria were isolated according to Bentlage et al. [9] with slight modifications. Cells harvested using 0.05% trypsin and 0.02% EDTA were sedimented (600 g, 5 min) and washed twice in phosphate-buffered saline (PBS — 140 mM NaCl, 5 mM KCl, 8 mM Na2HPO4, 1.5 mM KH2PO4, pH 7.2). Weighed cell pellet was suspended in ten times (w/v) the amount of 10 mM Tris-buffer with PIC (1:500) and homogenized by teflon-glass homogenizer (8 strokes, 600 rpm). Immediately afterwards 1/5 volume of 1.5 M sucrose was added. Homogenate was centrifuged at 600 g, 10 min and mitochondria containing supernatant was kept on ice. Pellet was suspended in original volume of SEKTP (250 mM sucrose, 40 mM KCl, 20 mM Tris, 2 mM EGTA, pH 7.6, PIC 1:500), rehomogenized (5 strokes, 800 rpm) and centrifuged at 600 g, 10 min. The supernatants were pooled and centrifuged 10,000 g, 10 min. Sedimented mitochondria were washed with SEKTP (10,000 g, 10 min) and suspended in SEKTP.

All isolations were performed at 4 °C and mitochondria were stored at − 80 °C. Protein concentration was measured according to [11].

Protein analysis by Blue Native PAGE (BNE) and BNE/SDS PAGE

Mitochondrial pellets were suspended in MB2 buffer (1.5 M ε-aminocapronic acid, 150 mM Bis-tris, 0.5 mM EDTA, pH 7.0), solubilized with digitonin (8 g/g protein) for 15 min on ice and centrifuged for 20 min at 20,000 g, 4 °C. Samples for BNE were prepared from supernatants by adding 1/20 volume of 5% SBG dye (Serva Blue G 250) in 750 mM ε-aminocapronic acid and 1/10 volume of 50% (v/v) glycerol.

Frozen cell pellets were resuspended in sucrose buffer (83 mM sucrose, 6.6 mM imidazole/HCl, PIC 1:500, pH 7.0) [67] and sonicated for 10 s to obtain 10% (w/v) suspension. Cell membranes were sedimented for 30 min at 100,000 g, 4 °C, solubilized with digitonin (4 g/g protein) in an imidazole buffer (2 mM ε-aminocapronic acid, 1 mM EDTA, 50 mM NaCl, 50 mM imidazole, pH 7.0) for 10 min and centrifuged at 30,000 g, 20 min at 4 °C [67]. For BNE analysis supernatants were mixed with 1/10 volume of 50% (v/v) glycerol and with 5% SBG dye in 750 mM ε-aminocapronic acid at a dye/digitonin ratio 1:8 (w/w).

Solubilized mitochondria were analyzed by Bis-tris BNE [56] on 5–12% polyacrylamide gradient gels, cell membranes were analyzed by imidazole BNE [67] on 5–16% polyacrylamide gradient gels using the Mini-Protean apparatus (BioRad).

For two-dimensional separation by BNE/SDS PAGE, the stripes of BNE gel were incubated in 1% SDS and 1% 2-mercaptoethanol for 1 h and then subjected to SDS PAGE on a 10% slab gel [55].

Western blot analysis

Proteins were transferred from the gels to PVDF membranes (Immobilon-P, Millipore) using semidry electroblotting. The membranes were blocked with 5% (w/v) non-fat dried milk in TBS (150 mM NaCl, 10 mM Tris, pH 7.5) for 1 h and incubated 2 h or overnight at 4 °C with primary antibodies diluted in TBS with 0.1% Tween-20. Monoclonal primary antibodies to the following enzymes of OXPHOS were used: SDHA (ab14715, Abcam), CORE1 (ab110252, Abcam), NDUFB6 (ab110244, Abcam), NDUFS3 (ab110246, Abcam), COX1 (ab14705, Abcam). The detection of the signals was performed with the secondary Alexa Fluor 680-labeled antibody (Life Technologies) using the Odyssey fluorescence scanner (LI-COR). Quantification of detected signals from BNE/SDS PAGE was carried out in Aida Image Analyzer program, version 3.21.

Spectrophotometric assays

COX, cII (succinate:cytochrome c oxidoreductase (SCCR)) and citrate synthase (CS) activities were measured as previously [44], cI (NADH:cytochrome c oxidoreductase (NCCR)) activity was measured as in [50].

Doxycycline treatment of the cells

Experiment was performed as described in [41]. Briefly, fibroblasts (grown to 70% confluence in DMEM medium) were treated with 15 μg/ml doxycycline (DOX) for 7 days and then washed 3 times with PBS to withdraw DOX. Subsequently, the cells were collected at different time points (0, 6, 16, 24, 48, 72 and 96 h) after DOX removal. Weighed pellets of cells were stored at − 80 °C for analysis of solubilized cell membranes by 2D BNE/SDS-PAGE. Two independent experiments of DOX inhibition in human and mouse fibroblasts were performed. COX1 antibody signals from 2D Western blots were quantified using Aida Image Analyzer v. 3.21 (Raytest). The relative distribution of signal between individual COX forms (assembly intermediates, monomer, dimer, supercomplexes) was determined from 2D blots. The total COX1 signal for each given time point was calculated from 1D SDS PAGE (see Fig. 4 E–H in the Data in Brief appendix) and normalized to SDHA signal. COX1 signal was then divided by the relative quantities obtained in the first step. The resulting datasets from each experiment were resampled to [0, 100] interval and averaged values from two experiments plotted as comparative 2D maps.

Fig. 4

Pulse-chase metabolic labeling of mitochondrially synthesized proteins in SURF1-deficient mouse and human fibroblasts.

(A, B) Human control and SURF1 patient fibroblasts, (C, D) SURF1+/+ and SURF1−/− mouse fibroblasts. Mitochondrial translation products of mouse and human fibroblasts were labeled with [35S] methionine + cysteine for 2 h in the presence of cycloheximide. After indicated time of chase (0.5 h, 6 h, 16 h, 24 h) with cold methionine and cysteine, cell membranes were isolated, solubilized by digitonin (4 g/g protein) and analyzed by BNE/SDS PAGE. Radioactivity was detected in stained dried gels. On the right side of each gel, individual mtDNA coded subunits are marked and mtDNA coded COX subunits COX1, COX2, COX3 are highlighted in frames. COX assembly intermediates (AI), COX monomer (M) are marked by dotted lines; COX SCs (SC).

Metabolic pulse-chase labeling of mtDNA encoded proteins

Proteins encoded by mtDNA were labeled using 35S-Protein Labeling Mix (Met + Cys; Perkin Elmer NEG072) by procedure described in [37]. Briefly, cells were incubated for 16 h with chloramphenicol (40 μg/ml), washed twice in PBS, and after 15 min incubation in DMEM medium without methionine and cysteine (DMEM-Met-Cys) and 15 min incubation in DMEM-Met-Cys with cycloheximide (CHX; 0.1 mg/ml), 35S-Protein Labeling Mix (350 μCi/150 mm dish) was added. Cells were incubated for 2 h, then 250 μM cold Met and Cys was added and after 15 min cells were washed with PBS + 250 μM cold Met and Cys and finally with PBS. Cells were grown in standard DMEM medium supplemented with 5% (v/v) fetal bovine serum and harvested at different times (0.5 h, 6 h, 16 h, 24 h). Pellets of labeled cells were mixed with the same w/w of unlabeled cells, cell membranes were isolated, solubilized with digitonin and analyzed by 2D BNE/SDS PAGE. Gels were stained in a Coomassie R 250 dye, dried and radioactivity was detected using Pharos FX™ Plus Molecular Imager (Bio-Rad). COX1 radioactive signals from 2D gels (Fig. 4 A, B, C, D) were quantified using Aida Image Analyzer v. 3.21 and relative quantities of individual COX forms (assembly intermediates, monomer, supercomplexes) were used to divide the respective COX1 signal for given chase-time from 1D SDS PAGE, normalized to overall radioactive signal in each time point. The resulting datasets from each experiment were resampled to [0, 100] interval and plotted as comparative 2D maps.

Results

Decreased COX activities in SURF1−/− mouse tissues and fibroblasts

Analysis of RC activities in isolated mitochondria from SURF1−/− mouse tissues and in fibroblasts whole cell lysates showed that COX activities related to activity of citrate synthase (CS) were decreased to 37–62% of control (heart 55%, liver 37%, brain 50%, muscle 48%, fibroblasts 62%) (Table 1), as previously described in [18]. Activities of other RC enzymes were not significantly changed in SURF1−/− mouse tissues/fibroblasts (not shown). The activity of CS was increased (22.7%) in SURF1−/− liver mitochondria but not in other tissues. This may suggest some compensatory upregulation of mitochondrial biogenesis, as observed previously in heart and skeletal muscle [49], or in SURF1 patient cells [28]. In general, the COX defect in SURF1−/− mouse tissues and fibroblasts is less pronounced than in SURF1 patients' fibroblasts and tissues, where COX activity was decreased to 10–30% of control values [44], [59]. These results are in agreement with differences in phenotype severity between SURF1−/− mice and SURF1 patients: while patients suffer from fatal Leigh syndrome, SURF1−/− mice show increased lifespan without considerable mitochondrial dysfunction [18], [49].

Table 1

Changes in COX and CS activities in SURF1−/− mouse.

	SURF1+/+			SURF1−/−			SURF1−/−/SURF1+/+
COX	CS	COX/CS	COX	CS	COX/CS	COX	COX/CS
Heart	3385.9 ± 120.8	2047.9 ± 136.7	1.7	2228.5 ± 190.7	2433.7 ± 218.2	0.9	65.8⁎⁎	55.4⁎⁎
Muscle	2486.7 ± 39.5	451.5 ± 50.2	5.5	1110.7 ± 121.3	420.5 ± 27.8	2.6	44.7⁎⁎	47.9⁎⁎
Liver	1495.2 ± 63.4	253.7 ± 11.5⁎	5.9	675.9 ± 23.0	311.3 ± 14.2	2.2	45.2⁎⁎	36.8⁎⁎
Brain	1527.2 ± 97.1	627.9 ± 28.2	2.4	744.5 ± 17.7	609.9 ± 24.1	1.2	48.7⁎⁎	50.2⁎⁎
Fibroblasts	459.3 ± 33.6	333.7 ± 15.3	1.4	248.3 ± 49.3	280.3 ± 28.4	0.9	54.1⁎⁎	62.3⁎

COX and CS enzyme activities (nmol/min/mg protein) were measured spectrophotometrically in SURF1+/+ and SURF1−/− mouse tissues and fibroblasts, ratio between SURF1−/− and SURF1+/+ values was expressed in %. The values are mean ± S.E. (n = 5–7).

p < 0.05.

p < 0.01.

Decreased content of assembled COX complexes and accumulation of COX assembly intermediates in SURF1−/− mouse tissues and fibroblasts

We performed 2D BNE/SDS PAGE analysis in combination with Western blot to detect and quantify various COX forms from assembly intermediates to supercomplexes in examined SURF1+/+ and SURF1−/− mouse tissues and fibroblasts (Fig. 1, Fig. 2; Fig. 3 in the Data in Brief appendix). Generally, decreased COX activities in various SURF1−/− mouse tissues and fibroblasts corresponded to decreased total COX content on Western blots. The amount of fully assembled forms of COX (monomer, dimer and COX-containing SCs) was downregulated in SURF1−/− mouse. In heart, liver and brain (Fig. 1 A, C, D; Fig. 3 in the Data in Brief appendix) it was in good agreement with measured COX activity (heart 53%, liver 39%, brain 64%), whereas in muscle it was somewhat higher (82%) and in fibroblasts lower (30%) (Fig. 1 B, E), than expected from activity measurements, possibly due to semiquantitative character of WB immunodetection (antibody reactivity in different tissues, large differences in amounts of different COX forms).

Fig. 1

Two-dimensional electrophoretic analysis of different COX forms present in SURF1−/− mouse tissues and fibroblasts and SURF1 patient fibroblasts.

Respiratory complexes and supercomplexes were solubilized using 8 g digitonin/g protein of isolated mitochondria, separated by BNE in the first dimension and SDS PAGE in the second dimension and detected by Western blots using specific antibodies to COX1 (cIV), CORE1 (cIII), NDUFB6 (cI) and NDUFS3 (cI). For analysis, heart (A), muscle (B), liver (C), brain (D) and fibroblast (E) of SURF1+/+ and SURF1−/− mice as well as fibroblasts (F) of human control and SURF1 patient were used. COX1 assembly intermediates (COX1 AI), COX monomer (M), COX dimer (D), III2–IV SC (III2–IV), I–III2 SC (I–III2), I–III2–IVn SCs (I–III2–IVn), complex III dimer (cIII2), complex I (cI).

Fig. 2

Distribution profiles of the COX1 signal in different COX forms resolved by BNE/SDS PAGE analysis.

COX1 signals from two-dimensional electrophoretic analysis in Fig. 1 were expressed as quantitative distribution profiles. SURF1+/+ and SURF1−/− mice heart (A), muscle (B), liver (C), brain (D) and fibroblasts (E) and human control and SURF1 patient fibroblasts (F). Individual COX forms are indicated: COX1 assembly intermediates (COX1 AI), COX monomer (M), COX dimer (D), III2–IV SC (III2–IV), I–III2–IVn SCs (I–III2–IVn).

Fig. 3

Analysis of COX assembly in SURF1-deficient mouse and human fibroblasts following the release of doxycycline-arrested mitochondrial protein translation.

BNE/SDS PAGE representative Western blot analysis using antibody to COX1 subunit performed in doxycycline treated (A) human control and (B) SURF1 patient fibroblasts and (C) SURF1+/+ and (D) SURF1−/− mouse fibroblasts. For BNE analysis, cell membranes were isolated and solubilized by digitonin (4 g dig/g protein). COX1 assembly intermediates (AI), COX monomer (M), COX supercomplexes (SC) are marked. Control cells without DOX (C-DOX), times t0–t96 represent time points in hours after DOX removal, when the cells were harvested. Two independent DOX experiments for each cell line were performed to generate 2D maps showing distribution of COX1 forms along the DOX experiments at time points t0–t96 h (0 h–96 h) in (E) human control and (F) SURF1 patient fibroblasts and (G) SURF1+/+ and (H) SURF1−/− mouse fibroblasts. Relative quantities of individual COX forms (assembly intermediates, monomer, dimer, supercomplexes) were used to divide the respective COX1 signal for given time point from 1D SDS PAGE (see Fig. 4 E–H in the Data in Brief appendix), normalized to SDHA signal. The resulting datasets from each experiment representing individual COX forms in human and mice cells were rescaled (minimum = 0, maximum = 100) and averaged to plot in comparative 2D maps. COX assembly intermediates (AI), COX monomer (M), COX supercomplexes (SC).

Relative distribution of individual COX forms varied considerably between the studied control mouse tissues. Monomer represented the dominant form in all tissues, with relative amount ranging from 50% in heart to as much as 85–95% of total COX in brain and liver. Significant content of COX dimers (18–27%) and III2–IV SC (13–15%) was detected in heart and muscle (Fig. 1 A, B; Fig. 2 A, B; Fig. 3 in the Data in Brief appendix), whereas in liver and brain we found only negligible amount of these COX forms (Fig. 1 C, D; Fig. 2 C, D; Fig. 3 in the Data in Brief appendix). Weak signals of COX were also detected above 1 MDa. These SCs were larger than I–III2 SC (detected by strong cI and cIII signals) and therefore presumably represent the I–III2–IVn SCs.

In SURF1−/− mouse tissues, the amount of assembled COX forms was lower compared to SURF1+/+, but the COX monomer was still the dominant form. COX dimer still represented 10–20% of total COX in heart and muscle but it almost disappeared in liver and brain (Fig. 1 A–D; Fig. 2 A–D; Fig. 3 in the Data in Brief appendix). III2–IV SC was preserved in small amount in all tissues, and weak COX signals of higher I–III2–IVn SCs were detected only in heart and muscle. In contrast to controls, we detected increased content of COX1 assembly intermediates (AI) in muscle and heart, representing approximately 10% of total COX signal. In liver and brain, the accumulation of AI was negligible (Fig. 1 A–D; Fig. 2 A–D).

In SURF1+/+ mouse fibroblasts, the COX monomer represented more than 90% of total COX signal, the remainder being comprised of small contribution of COX dimer, III2–IV and I–III2–IVn SCs (Fig. 1 E; Fig. 3 in the Data in Brief appendix). In SURF1−/− mouse fibroblasts, the COX defect was more accentuated than in other mouse tissues — we detected reduced signal of COX monomer, negligible content of I–III2–IVn SCs and markedly accumulated COX assembly intermediates, which represented 30% of total COX signal (Fig. 1 E; Fig. 3 in the Data in Brief appendix).

Other RC complexes (cI and cIII) were not affected by the COX defect in SURF1−/− mouse tissues and fibroblasts. As expected, the content of COX-containing III2–IV SC was reduced (Fig. 1 A–E; Fig. 1, Fig. 2 in the Data in Brief appendix).

SURF1 patient fibroblasts preserve large I–III2–IVn supercomplexes

Given the marked differences in COX activities between SURF1 patient and SURF1−/− mouse fibroblasts and respective controls, we also performed 2D BNE/SDS PAGE analysis on human control and SURF1 patient fibroblasts to obtain interspecies comparison of the COX assembly defect consequences. In human control fibroblasts, COX was mainly found as a monomer and I–III2–IVn SCs (Fig. 1 F). There was also higher amount of COX dimer and III2–IV SC in comparison to SURF1+/+ mouse fibroblasts (Fig. 1 E, F; Fig. 3 in the Data in Brief appendix). In SURF1 patient fibroblasts, two dominant COX forms were detected: the majority of fully assembled COX was detected in the I–III2–IVn SCs and as the large amount of COX1 assembly intermediates. In contrast the signal of COX monomer represented less than 10% of total COX, a pattern significantly different to SURF1−/− mouse fibroblasts (Fig. 1 E, F; Fig. 3 in the Data in Brief appendix). Taken together, we show that the COX defect due to lack of SURF1 caused by SURF1 gene mutations/knockout exerts both tissue and species specificity.

COX supercomplexes assembly kinetics

Our present experiments indicate significant differences in COX association into III2–IV and I–III2–IVn SCs between human and mouse, which can be observed both in control and SURF1-deficient fibroblasts. To explore the dynamics of COX incorporation into supercomplexes, we transiently treated cells with doxycycline (DOX), a reversible inhibitor of mitochondrial translation, to deplete the cells of mtDNA encoded OXPHOS subunits [41]. Human and mouse control and SURF1-deficient fibroblasts were cultured for 7 days in the presence of DOX and, after DOX removal, cells were collected at different time points (t0, t6, t16, t24, t48, t72, t96 h) to follow the assembly of newly synthesized mitochondrial-encoded subunits into RC complexes and subsequently into associated supercomplexes.

First we checked the amount of remaining COX in cell homogenates after DOX treatment. Using SDS PAGE and Western blot analysis (Fig. 4 in the Data in Brief appendix), we clearly showed lower levels of COX1 subunit in comparison with controls (DOX untreated cells) in all cell lines studied. COX1 antibody signals normalized to signals of SDHA were decreased to 30% and 34% in control and SURF1 patient fibroblasts, and to 20% and 23% in SURF1+/+ and SURF1−/− mouse fibroblasts, respectively. DOX treatment caused also slight decrease of cIII and cV levels, yet the greatest decrease was, unsurprisingly, observed in the content of cI, with its 7 mitochondrially encoded subunits.

Isolated membranes from fibroblasts were subsequently solubilized with digitonin (4 g/g protein) and analyzed by 2D BNE/SDS PAGE in combination with Western blot. In both human fibroblast lines, antibody detection of COX1 subunit signals revealed decrease in all COX forms at time point t0 (Fig. 3 A, B, E, F). Apparently, human cells preserved the reported stoichiometric distribution of COX (COX1) among COX assembly intermediates (AI), COX monomer, dimer and COX SCs (III2–IV, I–III2–VIn) throughout the doxycycline treatment. The balance between COX AI and monomer was also maintained during the time course after DOX withdrawal. Specifically, in human control cells (Fig. 3 A, E), the longer the cells were cultured after DOX removal, the more COX monomer and SCs were synthesized and at time point t96 h COX1 signal reached the original steady state (DOX untreated cells). At all time points, the monomer was still the prevalent COX form. In SURF1 patients' fibroblasts (Fig. 3 B, F) at t0 (after DOX removal), COX monomer was the main COX assembled form as its level decreased less than the content of COX associated into I–III2–IVn SCs. At t0 and all successive time points, signals of COX1 assembly intermediates represented the dominant form of COX and COX monomer further decreased. Compared to control cells, COX1 signal in SURF1 patient did not reach the steady state levels observed in cells without DOX treatment even at t96 h.

The shift in balance between COX monomer and AI towards the AI was also clearly observable in SURF1−/− mouse fibroblasts and was maintained throughout the time course as in SURF1 patients' cells. However, in other aspects, mouse and human models differed. First, the incorporation of newly synthesized COX subunits into higher SCs was delayed in mouse fibroblasts and only started to appear at t16–24 h (Fig. 3 C, D, G, H). Also, the COX SCs represented only minor portion of COX, which was just barely detectable by COX1 antibody in mouse cells without DOX treatment both on SURF1+/+ and SURF1−/− background. The COX monomer was the dominant COX form in all mouse cells; it increased significantly in SURF1+/+ cells from t0 to t96 h, while in SURF1−/− cells AI represented up to 50% of COX1 signal, gradually accumulating at the respective time points.

The interspecies differences between mouse and human fibroblasts are clearly visualized in 2D maps of COX distribution into individual forms (Fig. 3 E–H). In human fibroblasts (Fig. 3 E, F), COX incorporation into SCs was much more prevalent than in mouse cells (Fig. 3 G, H), control human cells maintained stable proportion between COX monomer and SCs, whereas SURF1 patient cells preferentially incorporated the assembled COX into SCs rather than into COX monomer. In SURF1+/+ and SURF1−/− mouse cells (Fig. 3 G, H), COX SCs amount was negligible when compared to COX monomer. DOX experiment thus clearly showed COX assembly defect due to SURF1 gene mutation/knock out. SURF1 patients' cells accumulated more COX assembly intermediates and most of assembled COX incorporated into SCs, whereas SURF1−/− mouse cells accumulated less COX AI and were characterized by more stable COX monomer compared to SURF1 patient.

Pulse-chase labeling of mitochondrially encoded COX subunits

As a follow up on the doxycycline experiments aimed at comparison of control and SURF1 defective human and mouse fibroblasts, we analyzed the assembly kinetics of COX directly by pulse-chase 35S labeling of mitochondrial translation products. Fibroblast cell lines pretreated with chloramphenicol (16 h/overnight) were washed and pulsed with 35S labeled Met-Cys in the presence of cycloheximide (CHX). Cells were then chased with cold Met-Cys without CHX at time points 0.5 h, 6 h, 16 h, 24 h to follow the time course of newly synthesized COX1, COX2, and COX3 subunits incorporation into assembly intermediates, monomer and SCs of COX. Isolated cell membranes were analyzed by 2D BNE/SDS PAGE after solubilization with digitonin (4 g/g protein).

In control human fibroblasts we observed the major portion of COX1 incorporated in assembly intermediates, whereas a small part was already present in the COX monomer and I–III2–IVn SCs at chase time 0.5 h (Fig. 4 A; Fig. 5 A). COX2 and COX3 subunits followed the same pattern but the migration distance and thus the identity of their respective assembly intermediates differed from COX1 (Fig. 4 A). At chase time 16 h, COX AI signal considerably decreased, accompanied by the increase of remaining forms, i.e. COX monomer, COX dimer and COX SCs. This pattern was consistent for signals from all three mtDNA encoded COX subunits. At the 24 h chase, COX monomer became the dominant form, and the COX1 assembly intermediates almost disappeared. In the case of SURF1 patient cells (Fig. 4 B; Fig. 5 B), COX1 assembly intermediates clearly dominated at chase 0.5 h. Small portion of COX monomer and SCs were also present, but they are only distinguishable from the signals of COX2 and COX3 subunits (Fig. 4 B). At chase 16 h, COX1 assembly intermediates still represented the dominant COX form, with small portion of COX1 shifting to the monomer as well as into COX SCs. At the end of the chase periods at t24 h, the signals of COX subunits weakened in patient cells, but the amount of COX1 assembly intermediates still prevailed over the fully assembled COX forms.

Fig. 5

2D maps of pulse-chase metabolic labeling of mitochondrially synthesized COX1 subunit.

(A) Human control and (B) SURF1 patient fibroblasts, (C) SURF1+/+ and (D) SURF1−/− mouse fibroblasts. 2D maps show biogenesis of COX1 subunit along the pulse-chase experiment at chase times t0.5, t6, t16 and t24 h (0.5 h–24 h). Relative quantities of individual COX forms (assembly intermediates, monomer, supercomplexes) were used to divide the respective COX1 signal for given time chase from 1D SDS PAGE, normalized to overall radioactive signal in each time chase. The resulting datasets from each experiment representing individual COX forms in human and mouse cells were rescaled (minimum = 0, maximum = 100) and plotted in comparative 2D maps. COX assembly intermediates (AI), COX monomer (M), COX supercomplexes (SC).

When we analyzed the COX assembly kinetics in SURF1+/+ mouse fibroblasts (Fig. 4 C; Fig. 5 C), we observed high amount of COX1 assembly intermediates, but also considerable signal of fully assembled new COX monomer and beginning of SCs formation at chase time 0.5 h. At longer chase periods COX1 assembly intermediates almost disappeared. We were able to detect COX monomer together with COX dimer, and COX-containing SCs, similar to the observation in DOX experiment using COX1 antibody from time point t24 h onwards. In SURF1−/− mouse fibroblasts at time points 0.5 h and 6 h, the formation of COX1 AI was prevailing over the signal of monomer (Fig. 4 D; Fig. 5 D), similar to human SURF1 patient cells. However, at later time points COX1 assembly intermediates rapidly disappeared, whereas newly synthesized COX monomer appeared stable, and only its content was clearly decreased when compared to SURF1+/+ mouse cells. As in SURF1+/+ mouse cells, large I–III2–IVn SCs were clearly detected.

In both SURF1−/− and SURF1 patient fibroblasts the amount of newly synthesized COX1 subunit during the pulse (t0.5 h) reaches approximately 70–80% of respective control levels. While an increase in COX1 translation could result in the apparent increase in the amount of COX monomer in SURF1−/− cells, this does not seem to be the case in our model.

Taken together, also when followed by pulse-chase, the COX assembly kinetics in mouse SURF1+/+ and SURF1−/− fibroblasts showed significant differences from the respective human cells, particularly regarding COX assembly intermediates depletion/accumulation and stability of COX monomer. As can be seen from 2D maps in Fig. 5 A–D, control human cells became depleted of COX AI at later time point (chase t24 h) than SURF1+/+ mouse cells (chase t16 h), which indicates possible slower COX1 biogenesis in humans. Also, SURF1−/− mouse fibroblasts coped better with SURF1 protein absence; they accumulated less COX AI and were able to synthesize more stable COX monomer when compared to SURF1 patient cells, where the SURF1 seems to be crucial for effective COX1 biogenesis and its incorporation to COX monomer.

The assembly kinetics of mtDNA subunits for other OXPHOS complexes, e.g. cytochrome b of cIII or ATP6 and 8 of cV, was comparable in human control and patient cells as well as in SURF1+/+ and SURF1−/− mouse cells (Fig. 4 A–D), as can be expected for isolated COX defect.

Discussion

In the present study we analyzed the COX biogenesis from assembly intermediates (AI) to supercomplexes (SC) in tissues and fibroblasts from SURF1−/− mouse and SURF1 patient fibroblasts to address the observed tissue and interspecies differences in phenotype severity. To characterize changes at steady state conditions as well as the dynamics of COX de novo synthesis, we combined enzyme activity measurements, 2D PAGE/immunodetection on doxycycline arrested cells and 35S-labeling of mtDNA encoded proteins.

Numerous studies indicate that SURF1 protein (SURF1) promotes early stages of COX biogenesis, from COX1 translation regulation to its association with other COX subunits into COX AI [20], [42], [58], [62]. Despite the absence of SURF1, 10–30% of control amount of active COX is assembled in SURF1 patients' fibroblasts [44]. Thus, while improving its efficacy, SURF1 is not absolutely essential in the COX assembly process [52]. Decreased levels of COX holoenzyme and reduction of COX activity were found to be accompanied by decreased levels of COX subunits [44], [69] and accumulation of specific COX AIs. In SURF1 patients, COX subcomplexes were detected in primary fibroblasts, skeletal muscle and heart, while their amount was very low in liver and brain [59], [60], [62]. Likewise, we have also observed tissue variance in COX assembly process in SURF1−/− mouse tissues on Western blots, where we detected markedly variable amount of accumulated COX1-containing subcomplexes (Fig. 1 A–D). However, the severity of COX assembly defect in tissues from SURF1−/− mouse was not as pronounced as in SURF1−/− mouse fibroblasts possibly because the assembly defect better manifests in proliferating cell cultures than in largely post-mitotic tissues studied (Fig. 1 E, F).

COX subassemblies and formation of monomer

Observed COX1-containing subcomplexes can either be COX degradation products or more likely represent true AIs, as they are predominantly labeled already at the shortest chase times (t0.5 h, Fig. 4 A–D). Furthermore, AIs labeling gradually decreased during the chase, in contrast to the increase in the COX monomer signal. However, AI dynamics differed between human and murine models of SURF1 deficiency. In SURF1 patient fibroblasts, COX1-subassemblies persisted throughout the chase without the progression into COX monomer. Contrariwise, SURF1−/− mouse fibroblasts had higher content of COX monomer from the beginning and its levels remained stable from t6 h onwards, while COX1 AIs gradually disappeared between t16–24 h, arguing for considerably faster turnover of COX1 AIs in mouse cells.

This is in line with our doxycycline (DOX) treatment experiments. COX1-subassemblies detected in lesser amount also in both types of control cells most likely reflect significant limiting step of COX assembly, which proceeds at slower pace e.g. because of incorporation of catalytic components of the enzyme (hemes, CuB). Fittingly, in bacteria [13], [23] and yeast the SURF1 homologs were linked with heme a incorporation into COX1. Shy1 — yeast homolog of SURF1 — forms early assembly intermediate with COX1 [39]. In this intermediate, both heme a cofactor sites are most likely formed in a stepwise process — heme a in a transition to the Shy1-containing complex and heme a3 within Shy1 complex. CuB site was also suggested to be formed at this stage, since Shy1 was found to transiently interact with CuB metallochaperone COX11. Thus, formation of heterobimetallic CuB-heme a3 site should occur in the Shy1 complex [27]. In addition, Shy1 was found also in association with COX15, heme a synthase, most likely cooperating on heme a transfer and insertion into early COX1 assembly intermediates [7]. While SURF1 can likely function analogously in the mammalian cells, this has yet to be experimentally proved. As in case of mammalian SURF1, Shy1 seems not to be absolutely required for COX1 maturation steps, because yeast strains lacking Shy1 have still residual fully assembled and active COX.

Interestingly, the assembly process seems to proceed faster in mouse than in human fibroblasts. Thus, at chase time point 16 h, all COX1 in mouse cells was assembled in holoenzyme while in human fibroblasts a significant portion of the subunit was still present in assembly intermediates. This could be due to a higher reservoir of COX1 AIs in control human cells and, along with faster decomposition of stalled AIs, may partially explain the milder phenotype of SURF1 absence in mouse. Faster assembly and rapid recycling of unincorporated subunits in mouse system would result in more frequent assembly “attempts” producing higher steady state COX content than in human patient cells.

On the contrary, translational activation of COX1 specifically in mouse cells may yield analogous outcome. In this respect, the phenotype of yeast Shy1 mutant was partially restored by the suppressor MSS51 through increased translation of COX1 [8]. Mss51p is primarily specific translation factor for COX1 mRNA, that acts on the 5′ untranslated region (UTR) of COX1 mRNA to promote translation initiation [27]. However, mammalian mitochondrial mRNAs do not have significant 5′ UTRs [40] and this may be the reason why functional homolog of translation activator Mss51p has not been found yet. On the other hand, human homolog (LRPPRC) of the yeast translational activator Pet309 has been reported and, in addition, translational activator of COX1 TACO1 is necessary for efficient translation of COX1 [64]. These proteins, or other yet unidentified factor(s), may effectively act as species-specific suppressors of COX defects in SURF1−/− mouse cells. Larger pool of translated COX1 accessible for finishing of COX1 maturation without SURF1 may then lead to synthesis of higher amount of COX and thus be the cause of the milder COX defect in SURF1−/− mouse cells. However, our data do not support such possibility. As already mentioned the decrease in the amount of newly synthesized COX1 subunit was approximately equal in both mouse and human fibroblasts with SURF1 defect without any indication of compensatory upregulation in mouse cells.

Impaired COX biogenesis due to SURF1 deficiency is characterized by accumulation of the S2 intermediate containing COX1–COX4–COX5a subunits [20], [42]. S2 represents an important rate limiting step and can usually be detected even in control cells/tissues. Recently, MITRAC complexes (mitochondrial translation regulation assembly intermediates of COX) were described as part of COX biogenesis pathway [38]. Their suggested function lies in regulation of COX1 translation by coordinating the interaction of COX1 with specific assembly proteins before entering the further steps of COX assembly. This regulatory cycle should be actually considered as an alternative mechanism of COX1 biogenesis in mammals to that in yeast. SURF1 was also found to be part of MITRAC complexes together with hCOA3 (MITRAC12) [15] and other COX assembly proteins (COX15, COX16, C12orf62). However, it is possible that the S2 subassemblies merely co-migrate with MITRAC complexes on native gels. The presence of several types of complexes and/or their dynamics in recruitment and release of various assembly factors has been reflected in changes of migration patterns of COX1-containing assemblies in 2D gels. We propose that absence of SURF1 does not exclude formation of MITRAC complexes and their accumulation with S2 subassemblies. Indeed, we consistently detect complexes that may represent such associations in COX defective cells/tissues characterized by impaired efficiency of COX1 biogenesis and its delayed interaction with other COX subunits.

COX incorporation into supercomplexes

COX holoenzyme is known to interact with other RC complexes and form supercomplexes (SCs) [21], [25], [41], [54]. However, in all mouse tissues examined in this study, COX monomer represented the dominant COX form. We also detected the presence of COX dimer and III2–IV SC and these forms were decreased or even disappeared in SURF1−/− mouse. The signal of large I–III2–IVn SCs was quite weak in SURF1+/+ mouse heart and muscle despite the fact that these SCs are usually observed in digitonin solubilizates [53], [67]. Interestingly, several recent studies reported the absence of COX SCs (III2–IV and I–III2–IVn) in liver and fibroblasts from C57Bl/6 J mouse strain, which served also as the parental strain for formation of SURF1−/− mouse [16], [29]. This was associated with the presence of short isoform of COX7a2l (SCAFI) subunit in this strain, which may preclude SC formation.

The relative absence of SCs in mouse samples contrasted with our observations in human fibroblasts, where we found strong signal of I–III2–IVn SCs. Therefore, we used DOX to analyze species-specific differences of COX ability to interact into SCs in control and SURF1 deficient human/mouse fibroblasts. In control human fibroblasts the COX1 signal increased equally in all detectable COX forms (AI, M, SCs), indicating a stable balance in distribution of COX forms. Due to decreased amount/stability of COX monomer, the SURF1-deficient patient fibroblasts preferably incorporated COX into I–III2–IVn SCs for its stabilization [28], [30]. On the other hand, in mouse fibroblasts assembly kinetics of COX proceeded mostly just to the level of COX monomer. COX SCs formation was delayed, and although their content even exceeded the levels in untreated cells, these higher forms of COX represented just a small portion of the total COX quantity. This is the most important difference between SURF1−/− mouse and SURF1 patient fibroblasts we uncovered. The preference towards SCs incorporation is not unique to SURF1 human patients. Recent study on cybrid clones carrying the heteroplasmic cytochrome b m.15579 A > G mutation demonstrated that its deleterious effects were attenuated when cIII was assembled into I–III2–IVn SCs [14].

There is only limited amount of information available about COX assembly kinetics after DOX mediated inhibition. The most comprehensive study was performed in control cybrids (143B cells) and DOX caused strong decrease of overall COX signal and disappearance of COX SCs. Gradual incorporation of the COX into SCs was observed at 6 h after DOX removal and COX1 appeared in SCs even later [41]. In our human fibroblasts, about 30% of various COX assemblies remained after DOX treatment and also COX SCs were partially preserved. Similarly, mouse fibroblasts started to synthesize COX SCs a few hours after DOX removal, but in the end favored COX monomer as a main functional COX form. Control human and mouse fibroblasts resembled cybrid cells [41], [45] as the amount of COX totally recovered 96 h after DOX treatment. Milder COX defect in SURF1−/− mouse fibroblasts also allowed the cells to reach the steady state after 96 h period, but SURF1 patient cells were delayed in COX recovery, again pointing towards slower and more affected COX biogenesis.

In conclusion, our present study shows that COX in tissues or cells from the SURF1+/+ and SURF1−/− mice exerts lower preference to be incorporated into SCs. On the other hand, I–III2–IVn SCs represented an important functional forms of COX in human cells, and even more so in SURF1 patient fibroblasts. Another reason why the defect caused by SURF1 absence in mice is less dramatic when compared to SURF1 patients is that COX monomer seems to be more stable in SURF1−/− mice and at the same time the turnover of accumulated COX assembly intermediates is faster.

Transparency document

Transparency document