Protein-induced membrane strain drives supercomplex formation
Figures
Structure and function of Complex I, Complex III2, and the SCI/III2.
(A) The structure of the OXPHOS complexes with lipid headgroups from atomistic molecular dynamics (MD) simulations. (B, C) Overview of the structure and function of redox-driven proton pumping in CI and the Q-cycle of CIII2. Molecular models of the (D) SCI/III2, (E) CI, and (F) CIII2 used in the atomistic MD simulations (water and ions were omitted for visual clarity).
SC formation affects the concentration of lipids and the membrane thickness.
(A–C) OXPHOS complexes viewed from the N-side of the membrane. Membrane shift induced by (D) the SC, and (E, F) the isolated CI and CIII2. The membrane shift relative to the average membrane plane is colored by the shift in z-position viewed from the N-side. (G) Local membrane thickness around SC (top), CI and CIII2 (bottom). (H) Local cardiolipin concentration. (I) Local concentration of Q (Q + QH2). (J) Cryo-electron microscopic (cryo-EM) difference map. (K) Membrane shift determined from the cryo-EM map. (L) Membrane thickness determined from cryo-EM map. (M) Density of Q/QH2 (headgroup) from coarse-grained molecular dynamics (cgMD) simulations. Insets: Protein–Q interactions from atomistic molecular dynamics (MD). (N) Density of cardiolipin (headgroup) from cgMD simulations. Insets: Protein–CDL interactions from atomistic MD.
Conformational dynamics of the OXPHOS proteins and the SC.
(A) Subunits comprising the SC interface. Inset: Interaction near the DED-loop. (B) Normal modes derived from essential dynamics analysis for the SC (left), CI (middle), and CIII2 (right). See Videos 1–5 for normal modes of the SC, CI, and CIII2. (C) Decomposition of interaction energies within the SC. (D) Distribution of the opening/closing mode (left) and the twisting mode (right) for the isolated CI (in orange) and the CI within the SC (in blue). (E) Differences in the dynamics of CIII2 for minimum and maximum values of Mode 1. (F) Distribution of the CI domain angle in the SC (in blue) and CI (in orange). (G, J) Conformational changes in the Rieske subunit between SC (in blue) and CIII2 (in green) affect (G) the QH2 binding in the proximal Qo site and (J) the Rieske FeS-heme c1 distance in the distal monomer. (H) Distribution of the dihedral angle in TM3 of ND6. (I) Overview of the SC showing the locations of ND6 and Rieske subunits, as well as the proximal and distal protomers of CIII2.
Lattice model of SC formation and crowding effects in the IMM.
(A) Possible specific and non-specific interactions between CI (blue/white) and CIII2 (red) and their respective energies in the lattice model. (B) SC population as a function of the specific interaction energy (Especific) and molecular strain (Estrain). (C) Specific and non-specific assemblies as a function of the protein–lipid ratio with varying specific interaction energies (Especific). (D) Representative protein arrangement in crowded IMMs. (E) Average edge-to-edge distance distributions and nearest neighbor distance (red line) for specific protein–Q/protein contacts. (F) Nearest neighbor distance between Q and CI as a function of the Q/QH2 ratio. The dashed line indicates the 7 nm distance between active sites in the SC.
Proposed thermodynamic and functional effects of SCs.
(A) The individual OXPHOS complexes, CI and CIII2, induce prominent molecular strain in the surrounding lipid membrane (dark green area) due to a hydrophobic mismatch between the protein and the membrane. Condensation of locally strained membrane patches entropically drives the SC formation and leads to an overall reduction of the membrane strain (light green area around the SC), favoring the accumulation of cardiolipin around the SC. (B) During normal respiratory conditions, each OXPHOS complex is surrounded by multiple (ca. 6) quinol/quinone molecules that can act as substrates for the proteins. At limiting QH2 concentration, the quinol diffusion between CI and CIII2 becomes rate-limiting and leads to a kinetic advantage of the SC (see also Figure 4F).
Analysis of membrane properties.
Mismatch of the hydrophobic region of CI and CIII2 relative to the lipid membrane. (A) The hydrophobic region in CI (left) and CIII2 (right), with hydrophobic amino acids in gold, and hydrophilic amino acids in cyan. The maps were created using ChimeraX. (B) Ion distribution as a function of z position in a protein-free slab of simulation S1. (C) Dielectric constant (at z = 0.5 nm from the membrane–water interface) around the SC (left) and CI (right), with the SC leading to larger membrane surface with a perturbed ε. (D) Distribution of the membrane thickness in a POPC/POPE/CDL/QH2 membrane and a CDL membrane. (E) Dielectric profile (ε) computed perpendicular to the membrane plane for a POPC/POPE and a pure CDL membrane. The Stern layer, characterized by the local increase in the ε, extends ca. 1.5 nm from the membrane plane and is affected by the lipid composition. (F) Position of sampled dielectric profiles shown in (G–L).
Concentration of lipids and quinones, and analysis of membrane thickness in atomistic molecular dynamics (aMD) simulations.
(A) Local membrane thickness in the SCI/CIII2 (top), CI (middle), and CIII2 (bottom). (B) Local cardiolipin concentration. (C) Local QH2 concentration. Difference in the membrane thickness around the SC relative to CI (left) or relative to CIII2 (right) from (D) aMD and (E) coarse-grained molecular dynamics (cgMD). (F–H) Error bars for the local CDL and Q/QH2 enrichment (data presented in Figure 2H, I) were estimated by averaging over the membrane in the vicinity of the protein (within 1.5 nm distance) divided into six domains (see panel F). The average concentrations for each domain and the calculated errors of (G) CDL and (H) Q/QH2.
Quinone dynamics in the membrane.
(A) Flip-flop motion of the Q headgroup in the membrane from atomistic molecular dynamics (aMD) simulations. Lifetime spent in upper/lower leaflets indicated by bars above. (B) Lifetime analysis of the Q and QH2 headgroup, with the half-life (t1/2) for Q/QH2 spent on a given membrane leaflet during consecutive steps during the simulations. (C) Representative aMD structure of QH2 headgroup interaction with water and lipid headgroups at the membrane interface. (D–F) Average diffusion of Q/QH2 around the SC/IIII2, CI, and CIII2 from coarse-grained molecular dynamics (cgMD) simulations. (G–I) Concentration gradient of Q/QH2 around the SCI/III2, CI, and CIII2 from cgMD simulations. CI – in blue, CIII2 – in beige. The Q species show a directed rotational diffusion around the OXPHOS complexes that could arise from conservation laws due to the Q depletion/increase in unique regions.
Shifts in the membrane leaflet from coarse-grained molecular dynamics (cgMD) simulations.
Shifts in the membrane leaflet viewed from the N-side of the (A) SCI/III2, (B) CI, and (C) CIII2. Shifts in the membrane leaflet viewed from the P-side of the membrane for (D) CIII2 and (E) SCI/III2. The location of the Qo site of the proximal CIII protomer is indicated by a red star. (F) The CIII2 angle was measured between the z-axis of the simulation box and the principal axis of inertia of the protein. (G) Time evolution of the membrane angle averaged over atomistic molecular dynamics (aMD) simulations of the SCI/III2 and CIII2. (H) Projection of the principal axis onto the xy-plane. (I) Visualization of the membrane distortion effect based on trajectory-averaged membrane position.
Analysis of membrane-induced distortion effects.
(A) Relative strain effect relative to a lipid membrane from atomistic MD simulations of the SCI/III2, CI, and CIII2, suggesting reduction of the membrane strain (blue patches) in the SC surroundings. The figure shows the non-bonded energies relative to the average non-bonded energies from membrane simulations (simulation M4, Appendix 1—table 2). (B) The lipid strain contribution for different lipids calculated from non-bonded interaction energies of the lipids relative to the average lipid interaction in an IMM membrane model (simulation M4). The figure shows the relative strain contribution for nearby lipids r < 2 Å, in color from panel (C), and lipids >5 Å from the OXPHOS proteins. (C) Selection of lipids (<2 Å) interacting with the OXPHOS proteins. (D) Potential of mean force (PMF) of membrane thickness derived from thickness distributions from coarse-grained molecular dynamics (cgMD) simulations of a membrane, the SCI/III2, CI, and CIII2. (E) Membrane thickness as a function of CDL concentration from cgMD simulations. (F) ΔGthick of the SC as a function of membrane thickness based on cgMD simulations. (G) Membrane curvature around the SCI/III2 (left), CI (middle), and CIII2 (right) from atomistic simulations. (H) Squared membrane curvature obtained from cgMD simulations, within a 20-nm radius around the center of the system. These maps correspond to the curvature field used in the calculation of the bending deformation energy term (Gcurv).
Analysis of lipid end-to-end distance from atomistic molecular dynamics (aMD) simulations of (A) SC, (B) CI, and (C) CIII2.
Local lipid concentrations from coarse-grained molecular dynamics (cgMD) simulations.
Concentration of (A) POPC, (B) POPE, (C) CDL, (D) QH2, and (E) Q from cgMD simulations of the SCI/III2 (left), CI (middle), and CIII (right).
Surface area per lipid and local distortion of membrane leaflet.
(A) The area per lipid on the P-side of the membrane. (B) The local leaflet shift in the P-side of the membrane. (C) The area per lipid on the N-side of the membrane. (D) The local leaflet shift in the N-side of the membrane.
Diffusion of Q molecules in coarse-grained molecular dynamics (cgMD) simulations.
Map of Q diffusion for (A) SCI/III2, (B) CI, and (C) CIII2.
Relative movement of lipids in coarse-grained molecular dynamics (cgMD) simulations.
Map of average movement of (A) CDL and (B) QH2 for SCI/III2 (left), CI (middle), and CIII2 (right). The movement is normalized relative to the local movement of POPC and POPE.
Correlated motion within the SC and allosteric network analysis.
(A, B) The Pearson correlation coefficients from Cα positions during atomistic molecular dynamics (aMD) simulations for (A) within the SCI/III2. (B) Correlation within the individual CI and CIII2. Core subunits are indicated by labels. Differences between A and B are indicated by yellow boxes. (C, D) Allosteric network analysis (see Materials and methods). (C) Path length difference of SCI/III2 network for the CI QH2 bound state and CI apo state. (D) Difference in average path length between the CI QH2 bound state and CI apo state.
Interaction area between CI and CIII2 from atomistic molecular dynamics (aMD) simulations.
(A) Structure of the SCI/III2 interface. Inset: Specific interactions at the CI and CIII2 interface. (B) Distribution of the pair distances within the SCI/III2. (C) Distribution of interface area between different subunits. (D) Distribution of the total interface area for simulations in the apo and QH2-bound states of CI. (E) Total interface area as a function of simulation time. (F) Multiple sequence alignment of UQCRC1, with the carboxylate motif forming the SC contact highlighted in red. (G) Standard deviation of the inter-subunit distance matrix for the CI apo state (upper triangle) and CI QH2 bound state (lower triangle). (H) Distance distribution between C-terminus of NDUFB7 and CIII2. No specific contacts form between these regions during the MD simulations, supporting that substitution of this region resulted in unaltered SCs.
Analysis of the TM3ND6 dihedral angle in CI.
(A) Position of the ND6 subunit within CI. Average conformation of ND6 from the last 100 ns – (B) the side view and (C) the top view of ND6. (D) Distribution of φ for SCI/III2 and CI in the apo and QH2-bound states of CI. (E) Position of the NDUFS2 subunit of CI. (F) Average conformation (over last 100 ns) of the β1–β2 loop of NDUFS2 from MD simulations for SCI/III2 (top) and CI (bottom). (G, H) Distance distribution between His59 and Asp160 of NDUFS2.
Interaction of cardiolipin and quinone/quinol with the SCI/III2.
(A) Overview of the CDL interaction sites shown in (C–H). (B) Overview of the Q/QH2 interaction sites shown in (I–L).
Videos
Normal modes of the SC.
Normal modes of the isolated CI.
Normal modes of the isolated CIII2.
Comparison of the normal modes between the SC and the isolated CI.
Comparison of the normal modes between the SC and the isolated CIII2.
Tables
Estimated membrane deformation energies associated with individual complexes and their supercomplex (SC).
Energies are reported as mean ± standard deviation from bootstrap analysis. The deformation energies are reported for a membrane patch with an area of ca. 1000 nm2.
| (kcal mol–1) | (kcal mol–1) | |
|---|---|---|
| CI | 123.6 ± 2.8 | 23.9 ± 0.6 |
| CIII2 | 59.6 ± 4.4 | 28.0 ± 1.4 |
| SC | 103.6 ± 1.6 | 49.1 ± 0.9 |
| SC – (CI + CIII2) | –79.2 ± 5.2 | –2.8 ± 2.0 |
List of atomistic MD and coarse-grained molecular dynamics (cgMD) simulations.
| cgMD | |||||
|---|---|---|---|---|---|
| Simulation name | Protein model | CI ligand state | CIII proximal ligand state | CIII distal ligand state | Simulation time (µs) |
| A1 | SCI/III2 | apo | apo | apo | 2 × 0.5 |
| A2 | apo | Q in Qi, QH2 in Qo | apo | 2 × 0.5 | |
| A3 | apo | apo | Q in Qi, QH2 in Qo | 2 × 0.5 | |
| A4 | QH2 | apo | apo | 2 × 0.5 | |
| A5 | QH2 | Q in Qi, QH2 in Qo | apo | 2 × 0.5 | |
| A6 | QH2 | apo | Q in Qi, QH2 in Qo | 2 × 0.5 | |
| A7 | CI | QH2 | n/a | n/a | 2 × 0.5 |
| A8 | CI | apo | n/a | n/a | 2 × 0.5 |
| A9 | CIII2 | n/a | apo | apo | 2 × 0.5 |
| A10 | n/a | Q in Qi, QH2 in Qo | apo | 2 × 0.5 | |
| A11 | n/a | apo | Q in Qi, QH2 in Qo | 2 × 0.5 | |
| Total | 11 μs | ||||
| cgMD | |||||
| Simulation name | Protein model | CI ligand state | CIII proximal ligand state | CIII distal ligand state | Simulation time (µs) |
| C1 | SCI/III2 | QH2 | Q in Qi, QH2 in Qo | Q in Qi, QH2 in Qo | 75 + 50 |
| C2 | CI | QH2 | n/a | n/a | 50 + 40 |
| C3 | CIII2 | n/a | Q in Qi, QH2 in Qo | Q in Qi, QH2 in Qo | 50 + 50 |
| Total | 315 μs |
Simulation details of membrane models.
| aMD | ||||
|---|---|---|---|---|
| Simulation name | System size (Å3) | Lipid composition | Lipid/Q/QH2 molecules | Simulation time (µs) |
| M1 | 78 × 78 × 82 | POPC/POPE (1:1) | 204 | 2 × 0.2 |
| M2 | 78 × 78 × 82 | CDL | 102 | 2 × 0.2 |
| M3 | 187 × 187 × 182 | POPC/POPE/CDL/Q (38:38:19:5) | 1000 | 2 × 0.35 |
| M4 | 187 × 187 × 182 | POPC/POPE/CDL/QH2 (38:38:19:5) | 1000 | 2 × 0.35 |
| cgMD | ||||
| Simulation name | System | Lipid bead distance (nm) | Simulation time (µs) | |
| cgM1/2/3 | SCI/III2 | 0.44/0.50/0.53 | 5/5/5 | |
| cgM4/5/6 | CI | 0.44/0.50/0.53 | 5/5/5 | |
| cgM7/8/9 | CIII2 | 0.44/0.50/0.53 | 5/5/5 | |
| cgM10–13 | Membrane | 0.44/0.47/0.50/0.53 | 5/23/5/5 |
Non-standard protonation states used in atomistic molecular dynamics (aMD) simulations.
*The protonation states of H59NDUFS2 and Y108NDUFS2 were modeled in their neutral (His0)/deprotonated (TyrO−) states with QH2 in CI.
| Subunit | Residues |
|---|---|
| ND1 | E192, E206, H247(ε), H287(ε) |
| ND2 | H25(ε), K46, H48(ε), H112(ε), H186(ε), H232(ε), K263 |
| ND4 | H82(ε), H213(ε/δ), H220(ε), Lys283, H293(ε), H319(ε), H338(ε/δ), H419(ε), H422(ε) |
| ND5 | H27(ε), H56(ε), H109(ε), K119, H230(ε), H248(ε), H323(ε), H348(ε), K392, H484(ε), H509(ε), H605(ε) |
| NDUFS2 | H55(ε), H59(ε/δ)*, Y108*, D104, H150(ε), H157(ε), H190(ε), H200(ε), D292, E343, H348(ε) |
| ND3 | D66, E68, E105 |
| NDUFS7 | D68, E154 |
| NDUFS3 | H19(ε), H53(ε), H145(ε) |
| NDUFV2 | H9(ε), H42(ε), H99(ε) |
| NDUFV1 | H29(ε), D98, H113(ε), H116(ε), H261(ε/δ), H283(ε), H356(ε/δ), H437(ε) |
| NDUFS1 | H43(ε), D232, H255(ε), H293(ε/δ), D324, E347, H401(ε), H421(ε), H437(ε), H494(ε), H549(ε) |
| NDUFS8 | H65(ε/δ), K81, H144(ε/δ) |
| ND6 | E100 |
| ND4L | H52(ε/δ) |
| 18 kDa | H29(ε) |
| 9 kDa | H43(ε), H44(ε), H75(ε) |
| B8 | H21(ε) |
| B12 | E29 |
| B17 | H67(ε), H74(ε), H83(ε), H89(ε), H127(ε/δ) |
| B18 | H3(ε), H60(ε), H81(ε/δ), H84(ε), Asp87, Glu90, H91(ε) |
| B22 | H11(ε), H25(ε), H32(ε), H50(ε), H72(ε), H75(ε), H107(ε) |
| AGGG | H6(ε), H42 (ε/δ), H50(ε) |
| ASHI | H66(ε), H78(ε), H104(ε), H155(ε) |
| ESSS | H45(ε) |
| MNLL | H10(ε), H13(ε) |
Estimation of volume changes from coarse-grained molecular dynamics (cgMD) simulations.
The volume change of the membrane with embedded OXPHOS proteins. The SCI/III2 formation leads to the membrane strain relative to the individual CI and CIII2.
| System | Membrane volume (nm3) | ΔV (nm3) | ||
|---|---|---|---|---|
| Membrane | 4160 | ± | 1.0 | |
| CI | 3946 | ± | 0.8 | 214 |
| CIII2 | 4093 | ± | 0.7 | 68 |
| SCI/III2 | 3865 | ± | 0.7 | 296 |
| ΔV | 14 nm3 | |||
Estimation of protein copy numbers and effective surface area in the IMM.
The data is based on Schlame, 2021, see also Petrache et al., 2000; Morgenstern et al., 2021; Fedor and Hirst, 2018. The crowding model, with a square length of 163 nm, describes the flat membrane regions of the IMM, thus excluding the area of ATP synthase and 10 lipids bound to the c-ring, located at the cristae (cf. Petrache et al., 2000).
| Component | Copy number | Surface area/molecule (nm2) | Total surface area (nm2) |
|---|---|---|---|
| CI | 13 | 190 | 2470 |
| CII | 15 | 16 | 240 |
| CIII2 | 18 | 110 | 1980 |
| CIV | 64 | 64 | 4096 |
| ATP synthase | 21 | 200 | 4200 |
| ATP/ADP carrier | 160 | 10 | 1600 |
| Lipids | 48,300 | 0.7 | 16,905/leaflet |
| of which Q/QH2 1% | 483 | ||
| 291 proteins 48,300 membrane | 14,586 nm2 protein 16,905 nm2 membrane Total: 31,491 nm2 |
Energy terms in the lattice model.
The protein–protein interaction is described by specific interactions term (Especific <0 kBT) and non-specific interactions (Enon-specific >0). The membrane–protein interaction determines the strain energy of the membrane (Estrain), based on the number of neighboring ‘lipid’ occupied grids that are in contact with proteins (Figure 4A). The interaction between the lipids was indirectly accounted for by the background energy of the model. The proteins can occupy four unique orientations on a grid ([North, East, South, West]). The table summarizes the unique energies linked to the respective microstates.
| State | CI and CIII2 position | d(CI – CIII2) | Energy |
|---|---|---|---|
| 1 | Neighbors with specific interaction | 1 | Especific + 10 Estrain |
| 2 | Neighbors with non-specific interaction | 1 | Enon-specific + 10 Estrain |
| 3 | Diagonal neighbors | 12 Estrain | |
| 4 | Separated by one lattice position in either cardinal direction | 2 | 13 Estrain |
| 5 | Separated by one lattice position in cardinal direction | 14 Estrain | |
| 6 | Diagonal neighbors separated by one lattice position | 15 Estrain | |
| 7 | No interaction | 16 Estrain |
Monte Carlo (MC) simulations of the lattice model.
The conformational landscape was sampled by MC using 107 MC iterations with 100 replicas. Temperature effects were modeled by varying β, and the effect of different protein-to-lipid ratios by increasing the grid area. The following simulations were performed (energy units are given in kB × 310 K).
| Simulation | Temperature (K) | Grid size, N | Especific | Enon-specific | Estrain |
|---|---|---|---|---|---|
| G1–G5 | 300, 305, …, 320 | 4 | −1 | 1 | 0.1 |
| G6–G10 | 300, 305, …, 320 | 4 | −1 | 1 | 0.5 |
| G11–G15 | 300, 305, …, 320 | 4 | −1 | 1 | 1 |
| G16–G20 | 300, 305, …, 320 | 4 | −1 | 1 | 2 |