Loss of respiratory complex I subunit NDUFB10 affects complex I assembly and supercomplex formation

2.1 Knockout of NDUFB10 results in loss of complex I and respiratory supercomplexes

Wildtype HAP1 and HAP1 NDUFB10 knock out cell lines were purchased from Horizon Discovery (Horizon Genomics GmbH). A rescue cell line was generated by introducing NDUFB10-MYCFLAG into the AAVS1 (HDR) safe harbor site using CRISPR/CAS9 technology. Crude mitochondria from wildtype HAP1, NDUFB10^KO, and rescued NDUFB10^KO with NDUFB10-MYCFLAG tag were isolated and membrane complexes were solubilized with digitonin under native conditions. The proteins were separated by blue native gel electrophoresis in BNGE (Wittig et al. 2006). To resolve SC and single complexes, immunoblotting was conducted against complex I subunit (NADH:ubiquinone oxidoreductase subunit B10, NDUFB10), complex III subunit (ubiquinol-cytochrome C reductase core protein III, UQCRC2), complex IV (Cytochrome oxidase subunit IV, COXIV) (Figure 1B). In the wild type cell line, plotting against NDUFB10 revealed that most of the subunit was found in supercomplexes of the composition I₁ + III₂ + IV and I₁ + III₂ (Figure 1B). In addition, CIII was found in respirasome complex RC III₂ + IV. CIV assembled in SCs and RCs, but in addition there was a considerable amount of free CIV (Figure 1B). In the NDUFB10^KO line, hardly NDUFB10 could be detected and no supercomplexes with complex I were found, when the membrane was blotted against complex IV (COXIV) and complex III (UQCRC2). Respirasome complexes containing CIII and CIV were still present indicating that respiratory complexes CIII and CIV could assemble in the absence or knockdown of CI. We validated the importance of the accessory subunit NDUFB10 in complex I assembly and SCs formation by rescuing the HAP1 NDUFB10^KO cells via CRISPR/CAS9 mediated insertion of NDUFB10 cDNA into the genomic safe harbor locus site (AAVS1). In the rescue cell line, we observed the full recovery of CI and supercomplexes SC I + III₂ + IV and SC I + III₂ (Figure 1B). Consequently, the P-module-NDUFB10 subunit is crucial for complex I assembly.

In CI-depleted conditions, we found higher levels of RC III₂ + IV as well as free III₂, IV and IV₂, while in rescued cells due to the formation of SC, free CIII₂ and CIV decreased again because CIII and CIV re-assembled into SC I₁ + III₂ + IV_n (Figure 1B). This suggests that loss of CI did not reduce the biogenesis and assembly of CIII and CIV per se. This was further explored by generating expression and protein profiles.

2.2 NDUFB10 is essential for the assembly of the N- and Q-module into full CI complex

Since complex I biogenesis is a progressive process, where modules are sequentially assembled (Guerrero-Castillo et al. 2017), we next asked, how knockout of NDUFB10 would affect the different assembly steps and activity of the modules. For example, in some knockout cell lines for complex I nuclear genes (NDUFS6, NDUFS4) the N module can be assembled but cannot be mounted on the Q/P 830 kDa complex (Kahlhöfer et al. 2021). The NDUFB10 is part of the P_D-a module, which provides part of the assembly platform for the Q- and N-module (Guerrero-Castillo et al. 2017). We determined the activity of the N-module, which is the last module added during complex I biogenesis. Therefore, we tested for NADH:ubiquinone reductase enzyme activity by an in-gel activity assay (IGA) (Figure 1C). While the SC protein band in control cells displayed NADH oxidizing activity, we found no NADH oxidizing function in NDUFB10^KO at the level of SC. However, we found NADH oxidizing activity in a subcomplex (# in Figure 1C), which is likely attributed to a partial N-module/Q-module.

Next, we probed the assembly of the Q-module by staining NDUFS3, a subunit that is attributed to the Q-module. The wild type mitochondrial extract showed the presence of supercomplexes and several subcomplexes of lower molecular weight that were positive for NDUFS3 (Figure 1D, marked as # and *). In the NDUFB10^KO_, no signal in the height of SC assembly was detected, but the subcomplexes detected in the wildtype (#, *) plus an additional subcomoplex with MW<480 were detected (Figure 1D, marked as **). In particular, the subcomplex of MW ∼500 kDa accumulated. Probably, diverse subcomplexes accumulated since they could not be further processed to full complex I respectively supercomplexes, likely due to the lack of the anchoring platform P_D (Guerrero-Castillo et al. 2017). The assembly intermediates were further characterized in Section 3.5.

2.3 Loss of intact complex I and supercomplexes results in perturbation of the respiratory activity and decrease of ΔΨ_m

Next, we asked how the loss of supercomplexes and perturbance of complex I would affect the electron transport in the respiratory chain. We determined the bioenergetic profile of the NDUFB10 deficient cells by monitoring oxygen consumption rates (OCR) with a Seahorse XF⁹⁶ Extracellular Flux Analyzer. Different OXPHOS complex inhibitors were sequentially added to determine basal respiration, ATP synthase-linked respiration, maximal respiratory capacity and non-respiratory oxygen consumption (for details see material and methods). We tested HAP wt cells, NDUFB10^KO and rescued NDUFB10^KO cells. The wt and rescued cells exhibited similar responses to the inhibitors, namely decrease of OCR after oligomycin application, increase of OCR after FCCP application, and shut down of ETC-related respiration after inhibition of rotenone and antimycin addition (Figure 2A). The rescue was not complete, though, since for example full capacity was not reached. Basal respiration (wt: OCR_basal = 1.29 ± 0.25 [SD] pmol min⁻¹ 1000 cells⁻¹ and rescued: OCR_basal = 1.39 ± 0.14 [SD] pmol min⁻¹ 1000 cells⁻¹) and ATP synthesis related respiration (wt: OCR_ATP = 0.87 ± 0.17 [SD] pmol min⁻¹ 1000 cells⁻¹ and rescued: OCR_ATP = 0.98 ± 0.15 [SD] pmol min⁻¹ 1000 cells⁻¹) were comparable in wt and rescued cells (Figure 2A). Interestingly, in both cell lines the uncoupler FCCP restored an OCR slightly lower than the respective basal OCR but did not further stimulate respiration (wt_{spare capacity} = 0.96 ± 0.36%, rescued_{spare capacity} = 0.79 ± 0.48%). This can have several reasons: lack of sufficient substrate to reach maximal OCR, or no spare capacity in wt and rescued cells. Contrary to control and rescued cells, NDUFB10^KO cells showed a significant reduction of mitochondrial respiration and no response to ETC inhibitors and uncouplers (Figure 2A, Supplementary Figure S1).

Figure 2:

Loss of complex I and supercomplex assembly results in pertubation of mitochondrial respiration. (A) Respiration measurement of wild type cells, NDUFB10^KO and NDUFB10^KO rescued. Seahorse Mito Stress test showing oxygen consumption rate (OCR; mean and standard deviation) under high glucose conditions; 1, oligomycin injection (2 µM); 2, FCCP injection (0.5 µM); 3, rotenone and antimycin A injection (0.5 µM); a representative experiment is shown. (B) Mitochondrial membrane potential measurements with TMRE (25 nM) from five independent experiments and 140 cells. Normalization on MTG (200 nM). (C) Relative mitochondrial ATP levels were determined by ATP-Red1™ (5 µM); measurements from three independent experiments and 40 cells. WT, HAP1 wildtype; KO, HAP1 NDUFB10 knockout; two-sided t-test *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001. Charts: The error bars denote the standard error (SE); the charts represent 50 percentiles. The vertical lines in the boxes represent the mean values. Technical replicate #1. Biological replicates: N = 28 for wt, N = 30 for KO and n=13 for rescue.

Reduction of the respiratory activity will likely affect the ΔΨ_m, which is build by the activity of the primary proton pumps CI, CIII and CIV. Missing CI activity reduces the net proton pumping by 8 H⁺/O₂ (N-pathway CI/CIII₂/CIV: 20 H⁺/O_2; CIII₂/CIV: 12 H⁺/O₂).

To validate the loss of function of the ETC, we determined the mitochondrial membrane potential (ΔΨ_m), which is directly linked to OXPHOS activity.

The relative ΔΨ_m was determined with the ΔΨ_m-sensitive dye Tetramethylrhodamine ethyl ester (TMRE). TMRE accumulates in active polarized mitochondria in dependence on the ΔΨ_m and thus its fluorescence intensity is an indicator for ΔΨ_m. The TMRE fluorescence intensity was normalized to MitoTracker™Green (MTG) fluorescence. The experiment revealed a significant reduction of ΔΨ_m in NDUFB10^KO cells (Figure 2B). This data is in accordance with the reduced respiration and thus proton pumping into the intra cristae space in NDUFB10^KO. Eventually, a reduced proton motive force will result in less ATP production. To test this, we determined relative mitochondrial ATP levels with the dye ATP-Red1™. The fluorescence of ATP-Red1 is proportional to mitochondrial ATP levels. To account for mitochondrial mass, the fluorescence was normalized to MTG. As anticipated, the NDUFB10^KO showed significantly reduced ATP levels (Figure 2C).

2.4 Knockout of NDUFB10 does not affect the expression of other ETC complexes

Loss of complex I related to NDUFB10 knockout and its effects on OXPHOS activity could be regulated on the level of gene expression and/or protein level. To check this, relative expression levels of OXPHOS complexes were determined by quantitative PCR. Complex I expression was probed by its subunit NDUFA9, complex II by subunit SDHA, complex III by subunit UQCRC1, complex IV by subunit Cox8a and ATP synthase by subunit ATP5A. Moreover, we tested for the expression of SCAF1, a chaperone involved in supercomplex assembly. All values were normalize on β-tubulin. Compared to HAP1 WT cells, NDUFB10^KO cells showed no significant difference in NDUFA9 mRNA levels, a subunit that resides in the interface of the N- and P-module and is not assigned to a specific complex I module (Stroud et al. 2016) (Figure 3A). This does not necessarily mean that NDUFA9 was also unchanged on the protein level, because likely most of complex I regulation happens at the protein level (Stroud et al. 2016). In contrast, complex II subunit SDHA trended to an increased expression. For CIII and CIV, the mRNA levels of subunits UQCRC1 and Cox8a decreased significantly, while CV expression was apparently not altered in NDUFB0^KO as the expression of ATPFA was unchanged (Figure 3A). Expression profiling also revealed that supercomplex assembly factor 1 (SCAF1) was reduced in NDUFB10^KO cells. SCAF1 is an essential factor in SC formation. In particular the interaction between CIII and CIV is modulated by SCAF1 (also known as COX7A2L) (Cogliati et al. 2016; Maekawa et al. 2020) (Figure 3A). This suggests, that disrupted CI biogenesis might compromise III₍₂₎ + IV assembly/interaction due to loss of SCAF1.

Figure 3:

NDUFB10 effects on the expression of respiratory complexes and ATP synthase. (A) Expression of OXPHOS subunits NDUFA9 (CI), SDHA (CII), UQCRC1 (CIII), Cox8a (CIV), ATP5A (CV) and the supercomplex assembly factor SCAF1. mRNA levels were determined using RT-qPCR analysis, normalized on beta-tubulin. Statistics: N = 3 independent experiments, n = 3 biological samples and four technical replicates each (ΔΔCt). (B) Immunoblot of complex I subunit NDUFS3. WT, wildtype; MW, molecular weight marker. (C) Immunoblot of complex II subunit SDHA. (D) Immunoblot of respiratory complex subunits as indicated. a: lower, b: higher exposure of the same blot. Antibodies against NDUFB8, SDHB, UQCRC2, MTCO1, and ATP5A. (E) Quantification of protein levels normalized on Tom20.

Following, subunits of different OXPHOS complexes were probed on the protein level (Figure 3B–D). We found a significant decrease in the level of complex I subunit NDUFS3, a subunit assigned to the Q-module (Figure 3B and E). Complex II subunit SDHA was not changed in NDUFB10^KO (Figure 3C). Next, we analyzed protein subunits of all OXPHOS complexes: NDUFB8, SDHB, UQCRC2, MTCO1, and ATP5A. Protein levels were normalized on Tom20 levels, that did not change (Figure 3E), (Supplementary Figure S2). SDHA protein was not changed in NDUFB10^KO, although mRNA indicated an increase in expression (Figure 3A and E), but SDHB protein levels were significantly increased. For complex III, we observed a slight increase in the UQCRC2 in NDUFB10^KO (Figure 3E). The mitochondrial encoded complex IV subunit MTCO1 was not affected in NDUFB10^KO. Finally, we assessed complex V subunit α (ATP5A) protein levels. The NDUFB10^KO exhibited the same levels as the WT cells. This was matched by the expression levels of the same subunit, which also displayed no difference between WT and NDUFB10^KO (Figure 3A and E). For complex I, we found a reduction of NDUFS3 and an almost complete loss of subunit NDUFB8, part of the ND5 module, in NDUFB10^KO cells (Figure 3E). From expression and protein analysis data, a tendency towards an increase of complex II can be deduced. The protein data for complex III and complex IV, however, do not indicate a difference between wildtype and NDUFB10^KO. Finally, complex V was also not affected by complex I depletion.

2.5 NDUFB10^KO results in incomplete complex I assembly

To get a more comprehensive understanding of NDUFB10^KO effects on assembly of the different modules of complex I, mass spectrometry analysis was performed. We asked which other subunits of CI were affected by NDUFB10 downregulation. According to quantitative PCR analysis of mRNA, NDUFA9 (Q-module) expression was not changed in NDUFB10^KO cells (Supplementary Figure S3). The expression of mitochondrial encoded subunit ND1, which is part of the P _P -module, in NDUFB10^KO was only 66% of the level in WT, which is in principle a knock down (KD). For several other subunits, expression levels also were lowered, however, this was not significant. In two of the biological replicates of NDUFB10^KO cells no ND1 was detected. It has to mentioned, though, that CI mt-DNA encoded subunits are always very hard to find due their high hydrophobicity. ND1 together with TIMMDC1 provides the anchoring platform for the Q-module in the membrane at an early step during CI biogenesis (Fang et al. 2021; Guerrero-Castillo et al. 2017). Starting from this building block, CI continues to be assembled from both the N-, Q- and P-modules as more N, Q, and P subunits are added (Guerrero-Castillo et al. 2017). The P _D -module, which contains NDUFB10, is added several steps later. Therefore, the loss or downregulation of ND1 is critical for the assembly of CI. Although we could not assess ND1 levels, subunits NDUFA3 and NDUFA13, part of the ND1 (P_P-_a) module were not decreased in NDUFB10^KO, while subunits NDUFA2 and NDUFA7, assigned to the N-module, were downregulated (Figure 4B). In addition, the abundance of NDUFA10, part of ND2-(P_P-b)module, was lower in NDUFB10^KO. From this data it appears that the assembly of the N-module and the P_P-_b-module, is affected in NDUFB10^KO, while the ND1 (P_P-_a) module is not concerned. The subunits ND1, NDUFA3, NDUFA8 and NDUFA13 are part of an intermediate Q/P_P-_a complex (Guerrero-Castillo et al. 2017) and indeed, we found no changes in Q-subunit NDUFS2, NDUFS3, NDUFS7 and NDUFS8 levels in NDUFB10^KO (Figure 4B and C). In contrast, the abundance of N-(NDUFS1, NDUFS4, NDUFS6, NDUFV1, NDUFV2) and ND2/P_P-b-(NDUFC1, NDUFC2, NDUFS5) module were less abundant in NDUFB10^KO cells. Also, when looking at the subunits of the P_D-b and P_D-a modules, it is obvious, that the abundance of these sub-modules is decreased in NDUFB10^KO cells (Figure 4D).

Figure 4:

Changes in the abundance of complex I subunits in NDUFB10^KO. (A) Volcano plot showing log2 changes in NDUFB10^KO cells in comparison to the wildtype. (B) Abundance of CI subunits of the N-, Q- and ND1/ND2 modules. (C) Abundance of N-(NDUFS1, NDUFS4, NDUFS6, NDUFV1, NDUFV2), ND2/P_P-b-(NDUFC1, NDUFC2, NDUFS5), and Q-(NDUFS2, NDUFS3, NDUFS7, NDUFS8) module subunits in WT and NDUFB10^KO cells. (D) Abundance of subunits of the P_D-a (NDUFB1, NDUFB5, NDUFB6, NDUFB10, NDUFB11) and P_D-b modules (NDFUB7, NDUFB8, NDUFB9, NDUFB3, NDUFB2).

Together, the CI-proteome data suggest that in NDUFB10^KO cells no fully assembled CI but only subcomplexes containing Q-module subunits, ND1 and partial ND2/P_P-b subunits are built, while N-, P_D-b and P_D-a module subunits are reduced. The immune-staining of NDUFS3, a Q-module subunit, was mildly decreased and NDUFB8, a P_D-b subunit was strongly reduced. These observations are in line with the MS-data. Together, the data suggests that the assembly of CI is halted at an early stage. Likely, this is just the last step before ND1 and TIMMDC1 anchor the Q-subunits to the inner mitochondrial membrane (Guerrero-Castillo et al. 2017). Thus, neither the N-module nor the P_D-arm will be assembled.

2.6 Loss of CI and respiratory supercomplexes affects mitochondrial morphology

Complex I and SC are mainly found in lamellar cristae sheets (Davies et al. 2011, 2018). Mitochondria are highly dynamic organelles whose architectural variations are closely linked to function (Hackenbrock 1968). Therefore, one can expect an impact of loss of CI and respiration on mitochondrial architecture. Transmission electron micrographs show tubular and more roundish mitochondria. Cristae were rather sparse (Figure 5A). It has to be mentioned, however, that cells had to be kept in high glucose conditions (25 mM Glucose) because NDUFB10^KO cells were exclusively relying on glycolysis and were not able to survive with low glucose supply (5.6 mM glucose). For accurate quantification of subtle changes in mitochondrial morphological parameters, the analyze particles function in ImageJ software was used. After analyzing about 200 mitochondria for each condition (WT and KO cell lines), we found no difference in mitochondrial perimeter, while the circularity of mitochondria decreased in NDUFB10^KO conditions (Figure 5B). Statistical analysis further revealed a significant decrease in mitochondrial width but no change in mitochondrial length for NDUFB10^KO mitochondria compared to WT (Figure 5C). Consequently, the calculated aspect ratio (AR; length-to-width ratio) (Krebiehl et al. 2010) was significantly increased in NDUFB10^KO, without an effect on the mitochondrial area. The cristae architecture was also affected in NDUFB10^KO conditions (Figure 5D). The number and length of cristae per mitochondrion decreased, while the cristae width increased, resulting in an altered aspect ratio for cristae architecture. In sum, we found moderate changes of mitochondrial morphology and ultrastructure in NDUFB10^KO cells, whereby cristae widening was the most prominent outcome.

Figure 5:

Mitochondrial morphology changes in NDUFB10^KO cells. (A) Transmission electron micrographs of mitochondria in wild-type and NDUFB10^KO cells. Images are representative of 190 (WT), respectively 197 (KO) imaged mitochondria. (B) Quantification of mitochondrial shape parameters. Mitochondrial perimeter and circularity. (C) Mitochondrial width, length and resulting aspect ratio (AR), which is the ratio between the minor and major axes of the ellipse equivalent to the object – that represents mitochondrial elongation and reflects the “width-to-length ratio”. Also, mitochondrial area is shown. (D) Characterization of cristae. Two-sided t-test, *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001, ****p ≤ 10⁻¹⁰; error bars indicate standard deviation HGlc, high glucose culture.

4.1 Cell culture

HAP1 cells have a near-haploid genotype, thereby facilitating genetic manipulation. In this project, HAP1 parental wildtype (WT) and NDUFB10 knockout cells were purchased from Horizon Discovery (Horizon Genomics GmbH) and used to generate different isogenic cell lines. Under normal culture conditions, HAP cell lines were cultured in Iscove’s Modified Dulbecco’s Medium (IMDM) (Gibco, #12440; containing 25 mM glucose) supplemented with 10% FBS at 37 °C and 5% CO₂. When cells reached 70–75% confluency, cells were detached with Trypsin/EDTA for 1 min at 37 °C and the desired number of cells was passaged into a cell culture flask or imaging dish.

4.2 Generation of rescued HAP1 NDUFB10^KO cell lines

We rescued HAP1 NDUFB10 knockout cells via CRISPR-Cas9 targeted NDUFB10 insertion into the AAVS1 Safe Harbor locus (GSHs). Two plasmids were transfected into the cells, the pAAVS1-EF1a-Puro-DNR (AAVS1-donor plasmid), containing the labeled NDUFB10 sequence with MYC-FLAG (small Tag) and pCas-Guide-AAVS1 plasmid, containing the CAS9 and gRNA sequences. For transfection, Turbofectin 8.0 reagent was mixed 4:1 to total plasmid DNA concentration, whereas AAVS1-donorplasmid and the pCas-Guide-AAVS1 plasmid were prepared in a 1:3 ratio. After 48 h, cells were selected via puromycin resistance (1 μg/mL). Cells were cultivated in the puromycin-containing medium until single cell clones could be observed. Single cells were picked using serial dilution and expanded in 96-well plates. When confluent, cells were transferred to a larger growth surface and frozen for long-term storage.

4.3 Transmission electron microscopy (TEM)

For high-resolution imaging of mitochondrial ultrastructure, HAP1 cells were seeded on coverslips that were previously coated with RGD-grafted poly-L-lysin-graft-(polyethylene glycol) (PLL-PEG-RGD). When the desired confluency was reached, cells were fixed with 2.5% glutaraldehyde in 0.1 M sodium cacodylate for 20 min on ice. Then cells were washed with 0.1 M sodium cacodylate buffer containing 3 μM calcium chloride. Next, a solution containing 1% OsO₄, 0.8% potassium ferrocyanide, 3 μM calcium chloride in 0.1 M sodium cacodylate was prepared and added to the cells. After 30 min incubation, cells were washed with double distilled water and stained with 2% uranyl acetate. Next, cells were incubated in EtOH solutions, increasing the concentration gradually (20%, 50%, 70%, 90%, 100%). A mixture of 50% EtOH and 50% Durcupan ACM resin (Sigma Aldrich, #44611) was prepared and added to the cells. Subsequently, cells were incubated 2 × 3 h in 100% Durcupan, which was polymerized for at least 48 h at 60 °C in an oven. For imaging, thin sections of 70 nm were prepared and recorded with a Zeiss EM902 operated at 80 kV (Zeiss) and 300 kV.

4.4 MitoTracker Green FM staining

MitoTracker Green (MTG) is a membrane potential-independent mitochondrial tracker with excitation/emission maxima λ = 490/λ = 516 nm. However, a minimum amount of membrane potential is needed to allow the entrance of the dye. In this project, we used 200 nm MTG (Invitrogen, #M7514) to stain mitochondrial mass, whereby the DMSO concentration was below 1%. Cells were incubated for 30 min at 37 °C and 5% CO₂ before medium exchange and subsequent imaging.

4.5 Membrane potential measurements

Tetramethylrhodamine ethyl ester (TMRE) is a cell-permeable, cationic, fluorescent dye with excitation/emission maxima λ = 549/λ = 575 nm. It is used to determine mitochondrial membrane potential ΔΨ_m since its accumulation in mitochondria is ΔΨ_m dependent_. Cells were stained with 25 nM TMRE (BioMol #ABD-22220) for 30 min. Simultaneously, cells were stained with 200 nM MitoTracker™Green (MTG) as a reference for mitochondrial mass. After 30 min staining, cells were washed twice with PBS solution and imaging medium without phenol red was added.

4.6 Determination of ATP

The BioTracker™ ATP-Red dye (Sigma Aldrich, #SCT045) allows for live-cell imaging for mitochondrial targeting ATP molecules. The dye is not fluorescent until a negatively charged ATP is present that breaks the covalent bonds between boron and ribose, causing ring-opening and fluorescence. The red fluorescent dye has excitation/emission maxima of λ = 510 nm/λ = 570 nm. Cells were incubated with 5 μM ATP-red for 15 min at 37 °C. Cells were simultaneously stained with MTG as mitochondrial mass reference. After staining, cells were washed twice with PBS solution and fresh imaging medium was added.

4.7 Blue native PAGE (BNGE)

For BNGE, cells were collected from two confluent T175 flasks and processed as described previously (Salewskij et al. 2020). Briefly, harvested cells were lysed mechanically using the cell homogenizer. Then mitochondria were enriched via differential centrifugation. Digitonin ∼50% (TLC) (Sigma-Aldrich, #D5628) was used for mitochondrial solubilization in a ratio of 6 g/g as described in (Wittig et al. 2006). 50–100 μg protein were loaded per pocket and separated via a vertical native gel 3–12% SERVAGel™ (Serva, #43251.01). NativeMark™ unstained was used as a protein standard (Thermo Fisher, #LC0725) for molecular weight estimation. However, it should be noted that in gradient gels, native membrane proteins do not move at the same speed as a free protein standard, particularly in the more increased molecular weight range. All buffers were prepared as described in (Wittig et al. 2006).

4.8 In-gel activity assay

The native gel was incubated for 45 min in 20 mL complex I substrate solution until violets bands were clear indicating active complex I. The reaction was stopped by denaturing the native complexes with 10% acetic acid solution. Then the gel was washed with water and documented. Complex I substrate solution is prepared as described in (Jha et al. 2016).

4.9 Immunoblotting

PVDF membranes were blocked with 10% non-fat dry milk in TBS-T (200 mM Tris, 1.37 M NaCl and 0.1% Tween 20). Super Signal West Pico by Thermo Scientific (#34080) was used as a chemiluminescent substrate to visualize protein bands. Immunoblotting was conducted against complex I subunits NDUFB10 and NDUFS3 and NDUFB8, complex II subunit SDHA, complex III subunit UQCRC2, complex IV subunit Cox4L1. Tom20 was used as endogenous control. The secondary antibody was anti-mouse DyLight™ 800 from Rockland (#610-145-002) or anti-rabbit (#611-645-122). Antibodies are listed in Table 1. The signal was detected using the ChemiDOC™MP Infrared Imaging System (BioRad^®). The mean density of the detected bands was determined using the analysis tool measure of ImageJ^®.

4.10 Respiration measurements

4.11 Quantitative polymerase chain reaction

Total RNA of HAP1 cell lines was obtained with the Monarch^® Total RNA Miniprep Kit (NEB #T2010), and cDNA was transcribed with the RevertAid First Strand cDNA Synthesis Kit (Thermo Scientific #K1621). Quantitative PCR (qPCR) was performed on the StepOnePlus™ Real-Time PCR System (Applied Biosystems). The PCR reaction was prepared from PowerUP™ SYBR™ Green Master Mix (Applied Biosystems # A25741), 30 ng cDNA, and 0.1 nM of each forward and reverse primer per sample. The PCR protocol started with an initial denaturation step of 10 min, followed by 40 cycles of 95 °C for 15 s and 60 °C for 60 s. Primers were purchased from Eurofins Genomics. The list of primers used is found in Table 2.

4.12 Mass spectrometry