Quinone Catalysis Modulates Proton Transfer Reactions in the Membrane Domain of Respiratory Complex I

Abstract

Complex I is a redox-driven proton pump that drives electron transport chains and powers oxidative phosphorylation across all domains of life. Yet, despite recently resolved structures from multiple organisms, it still remains unclear how the redox reactions in Complex I trigger proton pumping up to 200 Å away from the active site. Here, we show that the proton-coupled electron transfer reactions during quinone reduction drive long-range conformational changes of conserved loops and trans-membrane (TM) helices in the membrane domain of Complex I from Yarrowia lipolytica. We find that the conformational switching triggers a π → α transition in a TM helix (TM3^ND6) and establishes a proton pathway between the quinone chamber and the antiporter-like subunits, responsible for proton pumping. Our large-scale (>20 μs) atomistic molecular dynamics (MD) simulations in combination with quantum/classical (QM/MM) free energy calculations show that the helix transition controls the barrier for proton transfer reactions by wetting transitions and electrostatic effects. The conformational switching is enabled by re-arrangements of ion pairs that propagate from the quinone binding site to the membrane domain via an extended network of conserved residues. We find that these redox-driven changes create a conserved coupling network within the Complex I superfamily, with point mutations leading to drastic activity changes and mitochondrial disorders. On a general level, our findings illustrate how catalysis controls large-scale protein conformational changes and enables ion transport across biological membranes.

Introduction

The respiratory Complex I (NADH/ubiquinone oxidoreductase) is a large (0.5–1 MDa) redox-driven proton pump that transduces the free energy from NADH-driven quinone (Q) reduction into the pumping of protons across mitochondrial or bacterial membranes. This proton-coupled electron transfer (PCET) process is highly efficient and fully reversible,¹⁻³ extending across a remarkable distance of nearly 300 Å (Figure 1). The reduced quinol pool shuttles the electrons via Complex III (cytochrome bc₁) to Complex IV (cytochrome c oxidase),⁴ while the generated proton motive force (pmf) powers active transport and synthesis of ATP by F_oF₁-ATP synthase.^4,5 During hypoxia, Complex I operates in reverse mode, catalyzing a pmf-driven quinol (QH₂) oxidation and reverse electron transfer (RET),⁶ which can lead to the formation of reactive oxygen species and various pathophysiological conditions.⁷ As mutations in Complex I are linked to nearly half of all human mitochondrial disorders,^8,9 understanding its mechanistic principles is also of significant biomedical interest.

Figure 1.

Structure and function of respiratory Complex I. (A) NADH-driven electron transfer in the soluble arm of Complex I reduces quinone (Q) to quinol (QH₂), triggering proton pumping across the membrane domain. The figure shows Y. lipolytica Complex I (based on simulation Y12, Table S1). (B) Primary site for Q reduction in NDUFS2 (site 1) and the membrane-bound quinol docking site between ND1 and NDUFS7 (site 2). (C) Rotation of TM3 of ND6 (TM3^ND6) between the π-bulge and an α-helical conformation in the mammalian Complex I (based on simulations M9 and M11, Table S1). (D) Superposition of TM3^ND6 from different Complex I isoforms (mammalian, plant, Y. lipolytica, E. coli, D. melanogaster, and C. thermophilum), showing different conformations of the helix (α/π) and highlighting the hydrophobic residue (Met, Ile, or Val) that blocks the proton transfer pathway between Asp^ND3 and Glu^ND4L (shown for the mouse Complex I structure) in the π-bulge conformation (see also Figure S10 for multiple-sequence alignments).

The L-shaped Complex I comprises up to 45 subunits, including 14 conserved core subunits that are responsible for the coupled electron and proton transfer reactions.¹⁰⁻¹² The hydrophilic arm of Complex I supports a 100 Å electron transfer (ET) process, enabled by a chain of 8–9 iron-sulfur (FeS) centers that bridges the NADH oxidation site with the site for Q reduction in NDUFS2 (Figure 1A).¹³⁻¹⁵ The stepwise reduction of the Q to quinol leads to local proton transfer from Tyr144^NDUFS2 and His95^NDUFS2 (Figure 1B, Yarrowia lipolytica residue/mammalian subunit numbering).¹⁶⁻¹⁸ Re-arrangements of the charge distribution within the primary Q site favors the dissociation of the formed quinol to a membrane-bound Q-binding site (Figure 1B),^16,18 originally identified based on free energy simulations¹⁹ and supported by several experimental and computational studies.²⁰⁻²⁵ The Q reduction and motion of the quinol from a low redox potential site (E_m ∼ −320 mV) near the FeS chain toward higher potential sites within the membrane domain (E_m ∼ +90 mV) could provide the main energy transduction event that drives the proton pump¹ (but cf. also refs (18,19,26−29)).

The proton pumping activity takes place in the 200 Å long membrane domain of Complex I that comprises the three antiporter-like subunits ND2, ND4, and ND5, as well as the ND1/ND3/ND4L/ND6-module, which is adjacent to the membrane-bound Q-binding site (Figure 1A). The membrane domain harbors a central charged/hydrophilic axis,²⁹⁻⁴⁴ which enables the proton pumping activity via wetting transitions around symmetry-related sites³² (but cf. also refs (24) and⁴⁴ for different interpretations). Despite major structural, biochemical, and computational advances in the last decades, the long-range proton pumping mechanism in Complex I has remained a major challenge. In particular, the molecular principles establishing the action-at-a-distance effect between the electron transfer in the hydrophilic arm of Complex I and the proton pumping in the membrane domain remain highly debated, including several alternative models and different interpretations of the data.^{1,24,29,30,44−48}

Molecular dynamics (MD) simulations together with mutagenesis experiments suggest that each antiporter-like subunit of Complex I pumps one proton.³²⁻⁴² In the electrical wave-propagation mechanism (the “electrical cradle”-model),¹ the proton pumping is achieved by stepwise conformational changes of conserved ion pairs, which modulate the proton transfer barrier and enable proton transfer toward the subsequent subunit interface. The “protonation signal” propagates in the “forward direction” from the Q-binding site toward ND5, whereas proton release to the positively charged side (P-side) of the membrane takes place by a backward propagation of the electrical wave by subsequent proton ejection, proton uptake from the negatively charged side (N-side) of the membrane, and “closing” the ion pairs.¹ This model is consistent with the reversible pumping machinery and mutagenesis experiments,³⁵⁻⁴² including experiments suggesting that removal of ND5 and ND4 leads to approximately half of the pumping activity,³⁹ although proton output channels to the P-side of the membrane have not yet been experimentally resolved in all subunits^{29,30,32,36,43,44} (but cf. also ref (46)). To explain the lack of water molecules leading to the P-side in the cryoEM structures, the “electrical wave/cradle”-mechanism was recently challenged by a model in which the terminal ND5 subunit pumps all protons, either as a response of a “forward wave”²⁴ or by the transfer of protons in the ND2 → ND5 direction, while the ND3/ND4L/ND6-module was suggested to transfer protons toward the Q-binding site.^30,44 Other variations of the “ND5-only” pumping model have also recently been proposed, e.g., a model where ND2, ND4, and ND5 take up two protons,⁴⁷ or an alternative model where the conserved ion pairs are responsible for the proton uptake and/or release, whereas pumping is achieved across all antiporter-like subunits.⁴⁸ It also remains unclear whether the proton transfer can take place between the antiporter-like subunit interface and the conserved ion pair elements as these reactions may involve rather high kinetic barriers.³⁶

Several water molecules have been experimentally revealed by recent cryoEM structures^{24,43,44,46,49} and validating in part key results of previous MD simulations.^29,32,35,36 However, while water molecules are indeed central for the long-range proton transfer reactions, they do not define the directionality or energetics of the reactions alone^50,51 (cf. also ref (52)), e.g., in aquaporin.⁵³⁻⁵⁵ Thus, despite the significant structural, computational, and biophysical advances in the field, the energy transduction mechanism of Complex I still remains unsolved and highly puzzling.

In addition to redox-driven proton pumping during turnover, the mammalian Complex I can also undergo a transition from the active (A) (“turnover”) state into a so-called deactive (D) or dormant state,⁵⁶ which provides a protective form of the enzyme against the pmf-driven RET.⁵⁷ CryoEM studies suggest that the deactivation involves conformational changes around the Q site, switching of an α-helical segment of TM3 in subunit ND6 (TM3^ND6) to a π-bulge (Figure 1C,D), as well as other changes that result in a partial loss of the cryoEM density, e.g., in ND3.^9,10,58,59 Moreover, recent structural²⁴ and computational⁶⁰ studies suggest that the deactive form of the mammalian Complex I could block the formation of a proton wire across the ND6 region. These conformational changes were suggested⁶⁰ to disrupt the coupling between the proton and electron transfer reactions in Complex I, preventing the ΔpH-powered RET reactions. Although it still remains unclear whether the deactive state is a part of the catalytic cycle linked to proton pumping (cf. refs (45),⁴⁹ and⁶¹), it was also suggested⁶⁰ that the conformational switching around the TM3^ND6-region is involved in all Complex I isoforms and could enable proton transfer across this constricted site (cf. also ref (24)). In this regard, most Complex I isoforms have been structurally resolved with the TM3^ND6 helix in the π-bulge form, while the α-helix form was first observed in the active state of the mammalian isoforms (see Figure 1D,^{10,11,20,24,59}) and recently also in other species.^{24,44,62−65} Although the common functional motifs are expected to be conserved, the coupling between the conformational changes in TM3^ND6, surrounding structural elements, and loop regions differs among the Complex I isoforms. It was recently suggested²⁴ that the bending angle between the hydrophilic and membrane domains strictly correlates with the conformational changes in the TM3^ND6 region in the ovine Complex I (but cf. ref (45)). Although several Complex I variants indeed undergo bending and twisting motions, which have been linked, e.g., to the quinone motion along the substrate tunnel,⁶⁶ this correlation does not seem to be identical in all species, including in Escherichia coli Complex I,⁴⁴ or in the recent structures of Drosophila melanogaster Complex I,^62,63 where no significant differences in the bending angle of the hydrophilic domain relative to that of the membrane domain are observed between different states. Thus, despite the likely involvement of the conformational switching at ND6,^{24,44,45,59,60,62−65} its causal link to the Q catalysis remains poorly understood.

Here, we study the coupling between the Q catalysis and conformational changes linked to proton translocation in the membrane domain of Complex I by combining >20 μs of atomistic MD simulations with quantum/classical (QM/MM) free energy calculations based on high-resolution cryoEM data of Complex I from the aerobic yeast Y. lipolytica(43) (lacking an A/D transition), where no α/π-transition has been experimentally observed that we compare with simulations of the mouse²⁰ and ovine isoforms²⁴ (with A/D states and an experimentally characterized α/π transition). Our multiscale simulations provide a powerful approach to probe the relationship between molecular structures in different intermediate states, their dynamics, and chemical reactivity—and thus highly complementary information to structural and biochemical experiments that guide the elucidation of the intricate energy-transduction mechanism of Complex I.

Results

Conformational Switching of TM3^ND6 Induces Formation of Proton Wires

In order to probe the coupling between the catalysis and proton translocation in Complex I, we performed atomistic MD simulations based on a high-resolution cryoEM structure of the eukaryotic Complex I from Y. lipolytica(43) that we embedded in a lipid membrane-water-ion environment and compared these to MD simulations of the mouse and ovine Complex I (Figure S1A–C) modeled in different states (Table S1).

Our MD simulations of Y. lipolytica Complex I suggest that ND1 becomes highly hydrated from the N-side of the membrane, establishing a hydrated contact from the Q chamber toward ND6. The local hydration levels equilibrated on ca. 200 ns timescales (Figures 2C–G and S2), whereas the membrane–protein interface remained tightly sealed from leak pathways. TM3^ND6 retains the experimentally resolved π-bulge conformation and blocks the connectivity between Asp67^ND3 and Glu30^ND4L/Glu66^ND4L, particularly at Ile67^ND6, which disrupts the water wire due to steric clashes (Figures 2A–G and S2A,D). The TM3^ND6 helix dihedral (Phe55:Cα-Cβ-Phe71:Cβ-Cα) is around φ ∼ 20–40° (Figure S2I), which compares well with the geometry of the helix in different experimentally resolved structures of both the bacterial and eukaryotic Complex I (Figure S2J). The hydration state of the pathway leading from the Q-chamber to ND6 is similar in both the ovine and mouse simulations (Figure S2B–D), with the blocked π-bulge resembling the deactive MD simulations of the mouse Complex I.

Figure 2.

Local conformational changes and hydration dynamics around ND6 induced by the π → α transition of TM3^ND6. (A) Rotation by the π → α transition of TM3^ND6 in MD simulations of Complex I from Y. lipolytica. (B) The dihedral angle of TM3^ND6 (Phe55:Cα-Cβ-Phe71:Cβ-Cα) shows distinct stable conformations during MD simulations (see Figure S2I for MD data in various states). Rotation of the TM3^ND6-helix in independent targeted simulations is shown as dotted (purple and pink) lines (see Extended Methods). (C) Mapping the proton pathway in the open proton channel, with symbols indicating probe positions of the hydration level (shown in panel F). (D) The π-form of TM3^ND6 leads to a closed proton channel with a broken contact between Asp67^ND3 and Glu30^ND4L. (E) The α-form of TM3^ND6 leads to an open proton channel with a well-connected hydrogen-bonded network between Asp67^ND3 and Glu30^ND4L. (F) Hydration level along the proton channel in the open (blue) and closed (red) channel conformations, averaged over the 0.5 μs MD simulations. (G) Hydration level between Asp67^ND3 and Glu30^ND4L in the open and closed-channel states. The average number of water molecules was calculated from the respective MD state by excluding the first 50 ns (see Supporting Information Methods). The lower and upper error bars show the 1st and 3rd quartiles, respectively. (H) Snapshot of the transition state for the proton transfer reaction across the TM3^ND6 region. (I) QM/MM free energy profiles for the proton transfer across the TM3^ND6 region in the open (blue) and closed (red) channel conformations, with statistical (WHAM) errors (see Extended Methods for further error estimation).

We next used density functional theory (DFT)-based QM/MM free energy calculations (Figures 2H,I and S3) to probe the energetics of the proton transfer reactions based on the MD-sampled conformations. To this end, we chose a structure with partial hydration around Asp67^ND3 (Figure S2E) that allowed us to model a sterically strained water chain between Asp67^ND3 and Glu30^ND4L (Figure S3C), although we note that this region remains mostly dry during the MD trajectory (Figures 2G and S2E).

We find that the proton transfer barrier in this “closed channel”-structure is significant, ∼15 kcal mol^–1 (Figure 2I), and comparable to a rate in the millisecond timescale based on transition state theory, TST, using a standard pre-exponential factor (κ k_BT/h ∼ 6 ps^–1, with the transmission coefficient, κ, set to 1), which could effectively block proton transfer on physiologically relevant timescales. We note that these estimates are likely to provide a lower boundary of the free energy as only a subpopulation of the closed channel states partially hydrates around TM3^ND6, in contrast to the open channel conformation where the region becomes highly hydrated (Figures 2G and S2E). Basis set effects further increase the barrier by 2–3 kcal mol^–1 (∼18 kcal mol^–1/0.8 s^–1, Figure S3D), while accounting for re-crossing events could also affect the predicted rates.⁶⁷ We expect that the effects of friction will further decrease the reaction rates, although we note that these affect low-barrier processes much stronger than high-barrier reactions, where the exponential factor dominates.⁶⁸ Moreover, the enforced proton transfer across this sterically strained region induces partial opening of the channel (Figure S3H), whereas restraining the conformation of Ile67^ND6 in the closed channel state increases the free energy barrier (Figures 2I and S3J), indirectly supporting the coupling between the proton transfer and conformational changes at this region.

To further probe how the conformation of ND6 affects the proton transfer reaction across this constriction site in ND6, we next induced a conformational change of TM3^ND6 into the α-helical form in the Y. lipolytica Complex I (Figure 2A,B, see Materials and Methods, and Figure S2H,I, Extended Methods), with the bias applied only locally on TM3 and TM2 of ND6. After the initial relaxation, TM3^ND6 remains in the α-helical conformation during the subsequent unrestrained MD simulations (Figures 2B and S2I) and establishes a water wire across the TM3^ND6 region with ca. 3–4 water molecules connecting Asp67^ND3 to Glu30^ND4L, which remains stable during the simulated timescales (Figure S2E).

Interestingly, we find that the hydration state of this region is sensitive to Q binding as well as the protonation state of the E-channel (Figure S2A–G), suggesting that the hydration level could modulate the proton transfer rate across the ND6 region (cf. also ref (44)). The sampled state resembles the conformational dynamics and hydration level observed not only in the active forms of the mouse and ovine Complex I (Figures 2A,B and S2A–F) but also in other species, such as the recently determined structures of Chaetomium thermophilum Complex I,⁶⁴ with an experimentally characterized π-to-α transition and with several water molecules around the coupling region, consistent with our MD data (Figure S2G). As noted above, Y. lypolytica Complex I has not been experimentally resolved in the α-helical state of the TM3^ND6 region,^43,47,49 thus preventing a quantitative comparison of the sampled dynamics and the cryoEM structures.

We find that the general large-scale global motion of the Y. lypolytica Complex I shows dominant twisting and bending modes that are sensitive to both the TM3^ND6 conformation and substrate binding (Movie S5, Figure S16, cf. also ref (66)). These modes are similar but not identical to those of the mammalian Complex I (Figure S16) and could arise from the lack of the NDUFA10 subunits at the interface of the hydrophilic and membrane domains.⁶¹ The experimental B-factors further support the conformational dynamics inferred from our MD simulations (Figure S15).

The TM3^ND6 rotation re-positions ND3 and ND4L, particularly Asp67^ND3 that forms a hydrogen-bonded wire to Glu30^ND4L (Figures 2C and S2A), which further connects to Glu66^ND4L. These carboxylates are adjacent to the conserved Glu131^ND2-Lys211^ND2 ion pair, the conformation of which could modulate the proton transfer reactions in ND2.^1,36,60 The α-helical form of TM3^ND6 leads to a drastic decrease of the proton transfer barrier to around 10 kcal mol^–1 (Figures 2H,I and S3, Movie S3), comparable to a microsecond rate constant (according to TST) and consistent with the overall turnover timescale of Complex I. The proton transfer reaction takes place via a Grotthuss-type mechanism, although we also observe partial diffusion of the proton-transfer intermediates and the water chain along the reaction coordinate (Figure S3K,L) that could further decrease the reaction barriers between adjacent sites. These observations suggest that the conformational switching of TM3^ND6 provides a kinetic gate that controls the proton transfer barrier across the gating region toward the interface of ND2. The proton transfer across this region is thus both kinetically and thermodynamically feasible in both forward and reverse directions in the α-helical conformation. This is indeed required not only for the electrical wave (“cradle”) model but also in the alternative “ND5-only” models that would require forward transfer²⁴ or reverse transfer^44,47 across the same region.

Coupling between Conformational Switching and the Active Site for Q Reduction

Our MD simulations suggest that the local conformational changes around ND6 propagate to the Q-binding site, which is located around 60 Å away from the TM3^ND6 helix (Figure 3, Movie S1). This process takes place by re-arrangements of ion pairs, particularly between ND6/ND4L and ND3/NDUFS2 (Figure 3B–D). We observe in the MD simulations that the TM3^ND6 rotation re-positions ion pairs in the TM1-2 loop of ND3, including Glu39^ND3 that breaks its contact with Arg99 of NDUFS2 (Figure 3B,D). These re-arrangements lead to a contraction of the β1-β2 loop of NDUFS2 and induce dissociation of the His95^NDUFS2—Asp196^NDUFS2 ion pair (Figure 3C). His95^NDUFS2 functions as the proton donor during the Q reduction to quinol, with deprotonation of the residue also triggering the dissociation of the His95^NDUFS2—Asp196^NDUFS2 contact (Figures S4–S6, see below, cf. also refs (16) and¹⁸). These conformational changes could favor the diffusion of the formed quinol to its membrane-bound binding site (site 2, Figures 4A and S4, see below). The “outward”-flipped Glu39^ND3 establishes a new contact with Arg112^NDUFS7 of the PSST loop, which in turn re-positions Arg108^NDUFS7 and breaks the ion pair with Glu210^ND1 (Figure 3C–E). We observe overall similar conformational switching networks also in mouse and ovine Complex I (Figures S5 and S6), supporting that the conformational crosstalk is conserved within the Complex I superfamily (see Discussion). The TM1-2 loop of ND3 is unresolved in several experimental structures,^{9,10,24,44,59} but we note that Arg99^NDUFS2 faces toward Glu39^ND3 in some structures (Figure S12B). These findings partially support the proposed ion paired contacts between the residues observed in our MD simulations, but also show that the region is dynamically highly flexible.

Figure 3.

Long-range conformational switching network in Complex I. (A) Overview of the conformational switching network in Y. lipolytica Complex I, with dashed lines showing key electrostatic interactions. Conserved residues are shown in bold black font and partially conserved or non-conserved residues in gray. See also Figure S10 for multiple sequence alignment and Figures S12–S14 for comparison of interactions in different cryoEM structures and MD trajectories. (B) Rotation of TM3^ND6 induces shifts of ion pairs connecting ND4L, ND6, and NDUFS2 (Asp79^ND6-Arg81^NDUFS2 and Arg81^ND4L-Glu67^NDUFS2) in simulations of the open/closed channel. (C,E) The effect of the shifted ion pairs propagates to NDUFS2, leading to conformational changes of the β1-β2 loop. Conformational changes at NDUFS2 lead to the opening of the His95^NDUFS2-Asp196^NDUFS2 contact and changes in ion pairs in the second Q-binding site (Arg108^NDUFS7-Glu210^ND1). (D) The TM3^ND6 rotation induces a conformational change on TM1-2 of ND3, resulting in ion pair contacts between Glu39^ND3 and NDUFS2 (Arg99^NDUFS2) or NDUFS7 (Arg112^NDUFS7). (E) Dissociation of the Arg108^NDUFS7- Glu210^ND1 ion pair leads to conformational changes in the PSST loop of NDUFS7. In panels (B–E), colored structures correspond to an open channel state, whereas gray structures correspond to a closed channel state.

Figure 4.

Proton pathways leading from the membrane-bound quinol site (site 2) toward TM3^ND6 and the energetics of the proton transfer reaction. (A) The quinol in site 2 forms a cation−π interaction with Arg108^NDUFS7 and a hydrogen bond with Glu210^ND1, which is connected via a hydrogen-bonded water wire to Asp67^ND3. The figure shows the conformation of sidechains from simulations with QH₂ (in cyan) and QH^– (purple, see Table S1). (B) QM/MM simulations of proton transfer between QH₂ and Glu210^ND1. The proton transfer leads to a further protonation cascade in the E-channel via Glu206^ND1 → Glu196^ND1 (see Movie S2). (C) Free energy profile for the proton transfer between QH₂ and Glu210^ND1. See Figures S3 and S14 and Movie S2 for further details.

Quinol Binding and Dynamics Trigger a Protonation Cascade

Interestingly, the conformational changes within the charged network propagate from the primary Q-binding site also to the membrane modules in ND1 and NDUFS7 (Figures 3C–E and S8), favoring the diffusion of the quinol to the membrane-bound Q-binding site (site 2, Figure 4A). Our simulations suggest that the quinol is energetically somewhat more stabilized in the open channel conformations relative to the closed channel form (Figure S8C, see also ref (44)), although we note that the quinol forms overall similar interactions in both states. At this membrane-bound site, the quinol headgroup forms a π-stacking interaction with Arg108^NDUSF7 and a hydrogen bond with Glu210^ND1 (Figure 4A), interactions that strongly tune the pK_as of the E-channel residues (Figure S8A). The region is dynamically flexible, with several alternative conformations (or unresolved loops) observed in experimental structures (Figures S7, S12, and S14) and consistent with the dynamics extracted from the MD simulations.

To probe the chemical reactivity of the quinol at this membrane site, we performed further QM/MM free energy calculations. Our data suggest that QH₂ could transfer its proton to Glu210^ND1 in a reaction that is thermodynamically and kinetically feasible (Figure 4B,C). Interestingly, protonation of Glu210^ND1 leads to further proton transfer along Glu206^ND1 and Glu231^ND1 to Glu196^ND1/Glu147^ND1 (Figure S3O, Movie S2), whereas the formed QH^– species is stabilized by either one of the surrounding arginine residues (Arg108^NDUFS7 and Arg112^NDUFS7). The dynamic flexibility is supported by the high variability in the positions of Arg108 and Arg112 in different structures (Figures S12, S14). From Glu196^ND1/Glu147^ND1, we observe a continuous hydrogen-bonded wiring via Tyr146^ND1 to Asp67^ND3 toward the interface of ND2 when TM3^ND6 is in its α-helical conformation (see above). The proton wire from the quinol to the E-channel is similar in mouse and ovine Complex I, although some of the key residues are structurally shifted due to the different conformation of the TM5-6 loop of ND1 (Figures S7J, M, S10). We find that an electric field forms between the positively charged Arg cluster of PSST and the negatively charged carboxylate residues of the E-channel. Electric fields can induce hydration effects by orientating water dipoles and decreasing the energy barrier of charge transfer processes.²⁹ It should, however, also be noted that residues, which remain in strong salt bridges are unlikely to undergo protonation changes unless the ion pair undergoes a conformational change. In this regard, Glu210^ND1 (open channel state), Glu196^ND1, Glu147^ND1, Asp67^ND3, Glu30^ND4L, and Glu66^ND4L (Figures 4A, S2A) could participate in the proton transfer process, whereas the carboxylates of the TM5-6 loop of ND1 can both form as well as break the contacts with the Arg residues of the PSST loop, depending on the conformation of both ND1 and the PSST loops (Figure S4). The electric field effects could thus provide the driving force favoring both the hydration and proton transfer toward ND2 (Figure S8D,E). Similar electric field-modulated proton transfer reactions have recently been found in several other bioenergetically relevant enzymes (cf. refs (29,52,69) and references therein).

The interactions of the quinol within the substrate binding channel are central to positioning functional elements involved in the proton transfer along the E-channel. We find that the interactions between the Q/QH₂ and the substrate tunnel affect the conformation of TM1^ND3 that could position Asp67^ND3via the long TM1-2 loop (Figure S9). These interactions could be important for the proton transfer across the constricted site at ND6 (Figures 2 and S4–S6) but also affect the conformation of the TM3^ND6 itself (Figure S9). The interaction of the quinol with the surrounding subunits affects the conformation of Tyr146^ND1 (Figures 4A and S4E, S5E, and S6E), which bridges the water wire from the membrane-bound Q site toward Asp67^ND3 (cf. also refs (24,36,44 and60)). This leads to an increased hydration level between Asp67^ND3 and Glu30^ND4L/Glu66^ND4L, with similar trends also found in our MD simulations of the mouse and ovine Complex I (Figures 4 and S2, S4–S6).

The quinol diffusion to its membrane-bound site leads to conformational changes in the ion pairs between the arginines (Arg108, Arg112) of the PSST loop and carboxylates (Glu210 and Glu206) of ND1 (Figures 3C,E and S4). These changes result in rather large pK_a shifts (Figure S8A) that could favor the formation of the QH^– species and proton transfer along the E-channel.

Discussion

We propose here a conformational switching network in Complex I that links the PCET reactions responsible for quinone reduction to conformational changes that control the proton transfer from the E-channel toward the antiporter-like subunit ND2. We suggest that these conformational changes are enabled by electrostatic effects, which induce the switching of ion pairs along helices and flexible loops. The process could involve both enthalpic and entropic effects that thermodynamically drive the conformational changes in the loops/helices by the motion of the Q/QH₂ substrate between its distinct binding sites. The entropic contributions are indirectly supported by the strong temperature dependence of the A/D transition,⁵⁷ and show a resemblance to the described Q-induced conformational switching during turnover in Y. lipolytica Complex I (cf. also refs (44,45,61 and66)).

Several of the identified residues within the network are highly conserved (Figure S10), with several mutations leading to a drastic decrease in Complex I activity (Table S5 and Figure S11A).^37,40−42 Mutation of Glu210^ND1 leads to an almost complete loss of the Q oxidoreductase activity, consistent with the role of the residue as a transient proton acceptor in the membrane-bound Q-site (Figure 4A, Table S5).⁴⁰ Moreover, the surrounding carboxylates,^17,37 which are involved in channeling the proton along the E-channel, as well as the arginines of the PSST loop that tune the pK_a of the quinol—have strong inhibitory effects upon substitutions (Table S5 and Figure S11A).⁷⁰ Interestingly, we also identify several mitochondrial disease mutations⁷¹⁻⁷⁴ (Table S6 and Figure S11B) in the vicinity of these coupling sites, further supporting their functional relevance.

Our data suggests that the TM1-2 loop of ND3, harboring Glu39^ND3, plays an important role in mediating the conformational switching between the Q site and the proton pathways. In this regard, we observe that Glu39^ND3 forms key contacts both with residues involved in the Q-binding site (in NDUFS2) and the membrane domain (ND1/NDUFS7) via the ion pairs to Arg99^NDUFS2 and Arg112^NDUF7. Mutations of these contact points show a strong effect on the Q reduction activity (<10–30% of WT) (Table S5).^17,37,70

These observations are also consistent with biochemical data,⁷⁶ suggesting that restraining the TM1-2 loop of ND3 blocks proton pumping. Glu39^ND3 and the TM1-2 loop of ND3 are dynamically flexible, and the carboxylate can also interact with other residues, such as Lys130^ND1 (Figure S12). We find that Cys40^ND3, which is located next to Glu39^ND3, changes its solvent accessibility during the switching process (Figures 3A,D and S4–S7) that could rationalize the sensitivity of the residue toward chemical (N-ethylmaleimide)-labeling during deactivation of the mammalian enzyme.⁵⁷ The Y. lipolytica Complex I lacks the supernumerary subunit NDUFA10 at the interface of the hydrophilic and membrane domains that could be important in locking the enzyme in a stable deactive form, possibly via interactions with NDUFA5.⁶¹ We note that the suggested sites, including the TM1-2 loop of ND3, are also conserved in distant members of the Complex I superfamily, such as the archaeal membrane-bound hydrogenases (Figure S7).

In addition to specific Q headgroup interactions, the long isoprenoid tail of the Q substrate establishes a significant non-polar interaction surface with the nearly 40 Å long substrate tunnel that could position key elements along the proton wire, e.g., via interactions with ND3, ND1, NDUFS7, and NDUFS2 (Figures S8, S9). Fedor et al.(75) indeed observed that the non-polar contacts formed between the Q and the substrate tunnel are important for tuning both the catalytic activity (k_cat) and efficiency (k_cat/K_m), with bovine Complex I achieving maximum k_cat and k_cat/K_m with a long-tailed Q₁₀ that fills the entire substrate cavity.⁷⁵

The conformational dynamics linked to the TM3^ND6 rotation is indirectly supported by cryoEM data^24,58 from different isoforms of Complex I (Figure S7). Interestingly, recent cryoEM data of E. coli Complex I⁴⁴ support that the bacterial isoforms could also employ conformational changes around the Q site and TM3^ND6, similar to those previously suggested for the mammalian enzyme²⁴ (cf. also ref (60)). However, the mechanistic suggestions and directionality of the proton transfer reactions⁴⁴ drastically differ from those suggested here, with protons moving from the E-channel to the Q site, and all pumped protons leaving ND5. In this regard, the correlation of the proposed “open/closed”-bending angle between the hydrophilic and membrane domains and the conformation of the TM3^ND6 region²⁴ show different trends between the mammalian,²⁴E. coli,⁴⁴D. melanogaster Complex I,^62,63 and other species^64,65 (see also Figure S16).

Putative Mechanistic Model

Based on our combined findings, we propose a mechanistic model for the coupling between quinone reduction and proton transfer reactions along the E-channel (Figure 5). In terms of the catalytic cycle, our results suggest that the reduction of Q to QH₂ triggers proton transfer from Tyr144 and His95 of NDUFS2 (Figure 3C). This PCET process breaks the His95-Asp196 contact, which favors the diffusion of the quinol toward the second Q-binding site (Figure 5, site 2, cf. also refs (1,16,18 and19)). The motion of the quinol could involve an enthalpic (ΔΔH) modulation of the substrate binding affinity (Figure S8C), which transduces the driving force not only into conformational changes within the membrane domain by entropic (TΔΔS) effects but possibly also lead to accumulation of molecular strain at central sites.⁷⁸ The broken His95-Asp196 ion pair shifts the position of the β1-β2 loop of NDUFS2, an effect that is mediated by TM3 of ND6 (Figure 5B: I-III, Movies S1 and S4). The conformation of the β1-β2 loop locally also affects the conformation of the TM5-6 loop of ND1, where contacts between Glu210^ND1/Glu206^ND1 and Arg112^NDUFS7/Arg108^NDUFS7 are broken. These conformational changes may stabilize the quinol in its membrane-binding site but also lead to an increase in the pK_a of the nearby carboxylates (Glu210^ND1/Glu206^ND1), which favors deprotonation of the QH₂ to QH^– (cf. see also refs (1,73) and Movie S2).

Figure 5.

Putative Q-reduction triggered proton pumping mechanism in Complex I. Red circles represent carboxylates (Glu, Asp), blue circles represent Lys, Arg, and (protonated) His residues, whereas N- and P-side protons are labeled in red and black, respectively. (A) Left: Q reduction to quinol triggers conformational changes in ND3, ND1, ND4L, and ND6 (indicated by blue lines), opening up the proton channel across TM3^ND6 that favors the quinol dissociation to the second membrane-bound site. Here, the quinol is transiently deprotonated, which triggers a protonation cascade toward the interface of ND2. The accumulated charge induces the subsequent opening of the ion pair and a lateral proton transfer step that propagates in the forward wave via ND2 and ND4 to ND5 (blue arrow). Right: proton ejection, re-protonation, and closing of the ion pairs lead to subsequent protons release across the membrane in the backward wave (red arrow). The proton release across the ND4L/ND2 interface triggers pK_a shifts via the conformational switching network in the E-channel and at the primary Q-binding site. Proton uptake favors quinol release and re-protonation of the Q-binding site. (B) Schematic closeup of the conformational switching network, coupling the Q sites with the proton pathway, and steps involved during the forward/backward wave in the electrical cradle-mechanism. (I) Q enters the substrate cavity, leading to conformational changes in the surroundings; (II) Q binding to site 1 stabilizes the β1-β2 NDUFS2 loop, followed by 2e^–/2H⁺ PCET to form QH₂. (III) QH₂ formation triggers conformational changes that allows the quinol to move to the membrane-bound site. The difference in entropic (conformational dynamics) and enthalpic (affinity change) contributions from the PCET and subsequent quinol motion drive the proton pump and initiate the proton transfer between QH₂ (forming QH^–) and the E-channel, enabled by the π-to-α transition of TM3^ND6. Proton transfer along the E-channel initiates propagation of the forward electrical wave (blue arrow). (IV) Upon the reverse wave (red arrow), conformational changes and re-protonation of the active site and QH^– release QH₂ and restore the binding site for a new cycle.

These processes trigger proton transfer along the E-channel, most likely via an already existing proton within the E-channel due to the high pK_a of this cluster (Figure 5B: III, Figure S8A,B). The proton transfer toward the gating point around TM3^ND6 is mediated by an electric field that hydrates the E-channel and re-positions key residues establishing the proton wire (e.g., Tyr146^ND1 and Asp67^ND3) (Figures 4A,B and S8D,E). We suggest that these orientated electric fields between the positively charged arginine cluster (at QH^–) and the carboxylates of the E-channel provide a downhill gradient for the proton transfer process toward ND6 (Figure S8D,E). The thermodynamic driving force for the proton transfer arises from the complete reduction of quinone to quinol (Q → QH₂) that in turn triggers conformational and protonation changes by affinity modulations between the different sub-sites.

The shifted β1-β2 loop of NDUFS2 also has more distant effects in contact points around ND6: this motion affects Arg81^NDUFS2, favoring an “outward” position of Asp79^ND6, which could re-position the loop next to TM3^ND6 (Figures 3B and S4). This further pulls the hydrogen bond between Asp79^ND6 and Tyr78^ND4L, breaking the contact between Arg81^ND4L and Glu67^NDUFS2 (Figure 3B). These interactions are also mediated by ion pairs in our MD simulations of the mouse and ovine Complex I (Figures S5 and S6), although with variations in exact residues due to the different length of the N-terminal loop of NDUFS2 in different species (Figure S10, S13).

The TM3^ND6 rotation correlates with the hydration of the gating region in our MD simulations and enables the proton transfer toward the interface of ND2. Protonation of the terminal carboxylates (Glu30/66^ND4L) triggers the opening of the ion pair at ND2, which in turn enables proton transfer toward the ND4 interface (Figure 5B). This induces ion pair dissociation and proton transfer toward the ND4/ND5 interface and dissociation of the ion pair(s) in ND5, mediating the proton transfer across the P-side of the membrane. Re-protonation of a key lysine residue on the central axis¹ and closing of the ion pair in ND5 destabilize and eject the ND4-proton across the membrane, which leads to proton uptake and closing of the ion pair in ND4, with subsequent ejection of the “ND2-proton” across the membrane. Re-protonation of the central axis lysine¹ and closing of the ion pair in ND2 destabilize similarly the proton at the ND4L interface, which is then ejected across the membrane. Now, the increased electrostatic repulsion between Glu30/66^ND4L and Asp67^ND3 could mediate, e.g., via the TM1-2 loop of ND3, a conformational change that increases not only the pK_a of the E-channel residues but also the initial proton donating residues of NDUFS2, which are connected to this network via the extended ND3/NDUFS2 loop (Figure 5). The increased pK_a leads to proton uptake from the N-side of the membrane to the E-channel cluster, which could donate back the “borrowed” proton from Glu210^ND1 to QH^– and trigger the release of QH₂ to the membrane (Figure 5B: IV). The catalytic residues (His95 and Tyr144) in the Q site then re-protonate upon these pK_a shifts, e.g., via channels previously identified in molecular simulations⁶⁶ and/or in cryoEM structures,^43,77 initiating a new turnover cycle.

Our proposed mechanism presents a model for the coupling between the quinone reduction and the proton transfer reactions in the E-channel. Although the general long-range energy transduction mechanism of Complex I still remains unsolved, our findings provide an important understanding of how the catalysis controls large-scale protein conformational changes, enabling the proton transfer reactions.

Conclusions

We have studied here the primary events linked to the redox-driven proton pumping in Complex I. Our multi-scale simulations based on experimental high-resolution cryoEM structures from different species identify a conformational switching network that mediates redox-induced conformational changes in the active site to the membrane domain of Complex I. This switching network operates by electric field effects that are mediated via the conformational switching of ion pairs. The unique loop structures and helices are highly conserved within the Complex I superfamily and suggested to operate by similar mechanistic principles. The unique ND3 subunit has a central role in mediating the cross-talk between the two Q-sites and the proton pathways through long-range effects. Our simulations suggest that the protonation and conformational changes coupled to the Q oxidoreduction drive the rotation of the transmembrane helix TM3 of ND6 between a π-bulge and α-helical form, which enables proton transfer from ND1 toward the antiporter-like ND2 subunit, and initiate the proton pumping across the membrane via the electrical cradle process.¹ On a general level, our findings suggest that catalysis and substrate binding control large-scale protein conformational changes by tightly controlled networks of ion pairs, with physically similar functional principles possibly operating also in several other proteins.^{29,52,69,79,80}

Materials and Methods

Atomistic MD simulations of Y. lipolytica (PDB ID: 6YJ4(43)), ovine (PDB ID: 6ZKC(24)), and mouse (PDB ID: 6ZTQ(20,60)) Complex I were performed in a POPC/POPE/cardiolipin (40%, 40%, and 20%) membrane, with the system solvated with TIP3P water and 150 mM NaCl. The simulations were performed in different protonation, conformational, and ligand states at T = 310 K and p = 1 bar using a 2 fs integration timestep and treating long-range electrostatics with the particle mesh Ewald approach. The systems, comprising 1–1.3 million atoms, were treated with the CHARMM36 force field⁸¹ in combination with in-house parameters for the co-factors^18,32 (Table S2). Base protonation states were determined with electrostatic calculations combined with Monte Carlo sampling.^82,83 The MD simulations (around 20 μs) were carried out using NAMD v. 2.14 and 3.0.⁸⁴ VMD,⁸⁵ PyMol,⁸⁶ and UCSF Chimera⁸⁷ were used for analysis and visualization. QM/MM free energy simulations were carried out on QM/MM systems around TM3^ND6 and the membrane-bound Q site, with the QM-MM partitioning (see Figure S1F,G) established with the link-atoms approach, introduced at the Cα-Cβ bond. QM/MM umbrella sampling was carried out with a harmonic force constant of k = 100 kcal mol^–1 Å^–2 introduced on the proton transfer reaction coordinate (center of excess charge or linear combination of bond forming/breaking distances, see Figure S3 and Extended methods). The QM/MM MD simulations were performed at T = 310 with a 0.5–1 fs integration timestep and with the MM part modeled as in the classical MD simulations. Free energies were computed using the weighted histogram analysis method.⁸⁸ The QM/MM simulations were carried out using CHARMM/TURBOMOLE⁸⁹⁻⁹¹ and FermiONs++.⁹² See Supporting Information and Extended Methods for a detailed description of the methods.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.3c03086.

• Extended methods, MD simulation data, DFT models, and comparison of MD states with cryoEM structures (PDF)

• Movies of the conformational coupling (MPG)

• Proton transfer reactions (MPG)

• Proton transfer across TM3^ND6 region in the open and closed channel conformations (MOV)

• PCA (MOV)

• Global PCA (MPG)

Author Contributions

All authors have given approval to the final version of the manuscript.

The authors declare no competing financial interest.