Lifetime imaging of GFP at CoxVIIIa reports respiratory supercomplex assembly in live cells

The assembly of respiratory complexes into macromolecular supercomplexes is currently a hot topic, especially in the context of newly available structural details. However, most work to date has been done with purified detergent-solubilized material and in situ confirmation is absent. We here set out to enable the recording of respiratory supercomplex formation in living cells. Fluorescent sensor proteins were placed at specific positions at cytochrome c oxidase suspected to either be at the surface of a CI1CIII2CIV1 supercomplex or buried within this supercomplex. In contrast to other loci, sensors at subunits CoxVIIIa and CoxVIIc reported a dense protein environment, as detected by significantly shortened fluorescence lifetimes. According to 3D modelling CoxVIIIa and CoxVIIc are buried in the CI1CIII2CIV1 supercomplex. Suppression of supercomplex scaffold proteins HIGD2A and CoxVIIa2l was accompanied by an increase in the lifetime of the CoxVIIIa-sensor in line with release of CIV from supercomplexes. Strikingly, our data provide strong evidence for defined stable supercomplex configuration in situ.

20 cells (n = 2 replicates). Statistics: ***, P=0.001 compared to control (ANOVA one-way). Figure S6: Respiratory efficacy of different HeLa cell lines stably expressing sEcGFP fusion constructs. a, oxygen consumption rate (OCR) determined by an Extracellular Flux Analyzer (® Seahorse XF96e). After recording resting respiration, the following chemicals were added sequentially to the cells: oligomycin (Oligo, 1 µM), to measure the nonphosphorylating OCR, FCCP (2 µM) to achieve maximal OCR, and antimycin A (AA, 0.5 µM) and rotenone (Rot, 0.5 µM) for determination of the extra-mitochondrial OCR. Dashed lines and double arrows explain parameters for calculation (representative for untransfected HeLa cells) b, Calculation of basal, maximal, ATP-linked and proton leak-linked respiration as well as spare respiratory capacity. Mean values ± s.d. (27 wells per cell line, n = 3 independent experiments). Statistics: no significant differences compared to untransfected HeLa cells for all stable cell lines tested (P0.05, ANOVA one-way). Figure S7. Structure of CIV homo-dimer with inward and outward subunits labeled. Top view: subunit I in yellow, CoxVIIIa in pink, subunit CoxVIa in green and subunit CoxVIb in blue (UniProtKB AC: P07471 Blue: Chains H, U, Cytochrome c oxidase polypeptide VIb, UniProtKB AC: P00429). The C-termini of CoxVIIIa subunits in detergent-solubilized, purified dimers are on opposite sides of the homo-dimer, subunits VIa and VIb are probably building the monomer monomer contact side. 2,3

Supplementary Figure S8. Insertion of a longer linker between sEcGFP and CoxVIIIa (CoxVIIIa-Link) results in insensitivity
to changes in SC assembly factors. Figure S9. Time-resolved measurements are insensitive to the fluorescence intensity respectively the expression level of the construct. Average fluorescence lifetimes related to mean intensities of sEcGFP fused to CoxIV (cyan) or CoxVIIIa (pink). For disruption of supercomplexes, CoxVIIIa-sEcGFP cells were transiently transfected (twice) with siRNA directed against supercomplex assembly factor Cox7a2l (black) as described before. One data point represents the mean  per cell (20 cells, 3 independent assays). Figure S10. FLIM of cells transfected with ATeam. ATeam is a fluorescence resonance energy transfer (FRET)-based indicator for ATP, 4 here with Clover as donor and mRuby as acceptor. Fluorescence lifetime images were recorded including cells with donor only (acceptor bleached) and donor and acceptor. Scale bar: 10 µm. Figure S11. FLIM/FRET of cells co-transfected with different OxPhos subunits. a, mEGFP was used as donor and DsRed as acceptor. Complex I (CI) was labeled at subunit 30kD, complex III was labeled at subunit k, complex IV at subunit CoxVb, complex II at subunit B (CII) and ATP synthase at subunit  (CV). b, To check for dimerization of complex V, the following FRET pairs were tested: subunit -mEGFP x subunit -DsRed (CVxCV, first donor, then acceptor construct), subunit b x subunit b, subunit b x subunit . CVb-D-A is a tandem construct of subunit b-eGFP-DsRed where intramolecular FRET can be expected. At least, two independent assays per combination were performed. FRET efficiency was calculated as E = (1-DA/D)*100. Fluorescence lifetime images were recorded in cells 6 days after co-transfection and within one image half, the acceptor was completely bleached to obtain the reference lifetime for the donor in absence of the acceptor.

Supplementary material and methods
In situ pH calibration for determination of the pH dependence of fluorescence lifetime. For pH settings, stable cell lines CoxVIIIa-sEcGFP and mt-sEcGFP were perfused initially for 3-5 min with PBS Determination of mean fluorescence intensity (all mitochondria per cell) was performed using ImageJ software (NIH Image, http://rsb.info.nih.gov/nih image/index.html). For thresholding, the otsu filter was used as a mask for mitochondria and the background was set to NaN.

Oxygen consumption measurements. Oxygen consumption of intact HeLa cells was recorded with
the Seahorse XF e 96 Extracellular Flux Analyzer (Seahorse Biosciences; North Billerica, MA, USA).
30.000 cells were seeded into each well of a 96-well XF cell culture microplate 24 h before the experiment. 60 min before the experiment, cells were washed with XF assay medium made of XF base medium (Minimal DMEM, 0 mM Glucose, 102353-100 from Seahorse Biosciences) with supplements (1 mM pyruvate, 2 mM L-glutamine, and 5.6 mM D-glucose -adjusted to pH 7.4), placed in fresh XF assay medium and incubated at 37°C before loading into the XF e Analyzer.
Supplements were from Roth. After recording resting respiration in the analyzer, the following chemicals from Seahorse Biosciences were added sequentially to the cells: oligomycin (1 µM), to measure the non-phosphorylating OCR, FCCP (2 µM), to achieve maximal OCR, and antimycin A (0.5 µM) and rotenone (0.5 µM), for determination of the extramitochondrial OCR. Three measurements (cycles) were performed for the resting OCR, three after oligomycin addition, three after FCCP and three after antimycin A plus rotenone with a 3-min interval of recording and a 3-min interval of mixing for each measurement.  The correlation of fit ranges from -1 to 1, and was calculated using simulated map with 5Å resolution for the models. Average map volume calculated from density map values averaged over the center positions of all the fit atoms, thus a higher value means the atoms sit in higher density, indicating a better fit. * Density map and full-atom model from the same work. 9 Overall, our model shows high similarity to available experimental models, especially one of the newest models with the highest available to date resolution: model of the "tight" respirasome PDB 5J4Z with resolution 5.8 Å. 9 We used "Orientation of Proteins in Membranes" (OPM) database 14 to obtain protein structures with assigned membrane limits (available at http://opm.phar.umich.edu). We used this data to ensure correct orientation of each protein in membrane with respect to other proteins in the supercomplex model. The OPM service provides spatial arrangements of membrane proteins with respect to the hydrocarbon core of the lipid bilayer, but for visualization we marked borders between the membrane lipids and the solution.
Additionally, we have placed a cytochrome c molecule structures in each of the three known binding sites in the supercomplex: one on the cytochrome oxidase monomer, and two on the cytochrome bc1 complex dimer. The X-ray structure of the cytochrome bc1 complex dimer (PDB 1PP9) 6  cytochrome oxidase the cytochrome c molecule was placed manually in agreement with known particular residue interactions. [15][16][17] For each sequence of labeled COX subunit model structures were generated using structures of GFP (PDB 1GFL) 18  The fraction of models, fitting into the supercomplex structure without clashing with other proteins or the membrane corresponds to the conformational space, available for the linker in a particular location (Table YYY). This number is bigger for mobile labels, which have longer linkers, and are exposed to the solution; and smaller for labels, attached by a short linker and/or surrounded by other proteins, and thus have restricted mobility. To give quantitative estimates of relative mobility of each linker, we also considered the number of "free" residues in each fusion construct, as a number of residues, mainly responsible for the linker and label mobility. Thus, we counted all residues between original subunit and the fluorescent label, not participating in any secondary structure and not restricted by a hydrogen bond or a salt bridge. Finally, we estimated the maximum distance between the sEcGFP label and nearest surface (of membrane or supercomplex protein) achieved with a fully stretched conformation of the linker (Supplementary Table 3) Supplementary Figure S12. Stages of modeling structure of sEcGFP fused with subunits of COX and CIII, shown for subunit IV of COX. a, Using the structure of GFP as a template, 25 models with different conformations of linker are created with MODELLER using build-in molecular dynamics to generate reasonable conformations of the linker. b, All generated models are attached to the structure of the whole protein (here, COX dimer) by superposition of the first 3 residues of the linker with corresponding 3 C-terminal residues of the labeled subunit. c, All model structures of the fused subunit are inspected manually and structures, overlapping with membrane or protein are removed. Thus, only models with reasonable conformations of the linker and realistic possible placement of the sEcGFP are left. COX monomers are shown in different shades of gray, membrane location is indicated by red dots, labeled subunit IV of COX is shown as dark blue cartoon, models of sEcGFP with a linker are shown in cyan cartoon.

Fluorescence lifetimes of sEcGFP decreases with increasing glycerol content of aqueous solutions
To show that the radiation lifetime is a function of the inverse refractive index in square as predicted by Strickler-Berg 23 , the average fluorescence lifetime (amplitude weighted lifetime amp) of sEcGFP was determined in increasing glycerol concentrations. However, measurements in mixtures of more than 90% glycerol were not realizable in an accurate way since homogeneous solubilisation was not possible any more. Nevertheless, it becomes clear that the fluorescence lifetime of purified sEcGFP in 90% -100% glycerol/PBS (w/w) would not be decreased to the lifetime of CoxVIIIa-sEcGFP (1.7 ns) nor could the refractive index be reduced to the value of proteins (n1.6,). 24 As scattering in the solution, which is more dramatic for the amplitude than the intensity weighted lifetime, might reduce the lifetime, we analyzed int as well and compared it to the amp values. We observed that curve progression of these lifetimes was analogical. Thus, determination of amp in solution was reliable.