Functional Dynamics of an Ancient Membrane-Bound Hydrogenase.

The membrane-bound hydrogenase (Mbh) is a redox-driven Na+/H+ transporter that employs the energy from hydrogen gas (H2) production to catalyze proton pumping and Na+/H+ exchange across cytoplasmic membranes of archaea. Despite a recently resolved structure of this ancient energy-transducing enzyme [Yu et al. Cell 2018, 173, 1636-1649], the molecular principles of its redox-driven ion-transport mechanism remain puzzling and of major interest for understanding bioenergetic principles of early cells. Here we use atomistic molecular dynamics (MD) simulations in combination with data clustering methods and quantum chemical calculations to probe principles underlying proton reduction as well as proton and sodium transport in Mbh from the hyperthermophilic archaeon Pyrococcus furiosus. We identify putative Na+ binding sites and proton pathways leading across the membrane and to the NiFe-active center as well as conformational changes that regulate ion uptake. We suggest that Na+ binding and protonation changes at a putative ion-binding site couple to proton transfer across the antiporter-like MbhH subunit by modulating the conformational state of a conserved ion pair at the subunit interface. Our findings illustrate conserved coupling principles within the complex I superfamily and provide functional insight into archaeal energy transduction mechanisms.

Introduction

The membrane-bound hydrogenase (Mbh) is a primordial enzyme that powers energy transduction in the thermophilic archaeon Pyrococcus furiosus.[1,2] Mbh catalyzes ferredoxin (Fd)-driven (Em,7 = −450 mV, redox midpoint potential at pH = 7) hydrogen gas production (Em,7 = −420 mV)[1−4] and employs the small thermodynamic driving force (ΔG ∼ −60 mV) for proton pumping and Na+/H+ exchange.[3,4] The sodium motive force (smf) generated across the archaeal membrane powers subsequent Na+-driven ATP synthesis.[5]

Mbh is a 14-subunit enzyme complex[1,2,6] comprising a hydrophilic domain, responsible for electron transfer and H2 production, and a membrane domain driving Na+/H+ transport, with the recently resolved cryoEM structure revealing key features of its molecular architecture[1] (cf. also ref (7)). Mbh is a predecessor of the modern complex I superfamily,[1,7,11] which catalyzes NADH- or Fd-driven quinone (Q) reduction and couples the much larger (ΔG ∼ −800 to −1200 mV) driving force to proton pumping across the membrane domain. Despite intensive work in recent years,[8−12] the molecular principles of the fascinating energy transduction mechanism employed by this superfamily remain unsolved.

The catalytic cycle of Mbh is initiated by association of the reduced Fd to the positively charged MbhN at the top of the hydrophilic domain, comprising subunits MbhI-MbhL and MbhN (Figure ). The electrons are transferred from Fd via three tetranuclear iron–sulfur (FeS) centers to the NiFe core of MbhL, responsible for the proton reduction.[1] The binuclear NiFe catalytic center is coordinated by four cysteine residues (Cys68L, Cys71L, Cys374L, Cys377L), in addition to two CN— and a CO ligand, and closely related to soluble NiFe-hydrogenases.[1,13] The membrane domain of Mbh comprises eight subunits responsible for ion transport (Figure ). Based on mutagenesis and structural studies of the related Mrp (multiple-resistance and pH adaption) transporters,[1,14,15] the Na+/H+ exchange was suggested to take place in MbhB/C/D/G (MrpF/G/Ant/C in Mrps).[14,15] Together with MbhA/F/G (Figure ), these subunits located on the terminal edge of Mbh establish a bundle of 3 × 4 transmembrane (TM) helices—a symmetric motif common in many antiporters.[8,9,11,14,15]

Figure 1

Structure and function of Mbh. Ferredoxin docks to MbhN (position indicated by black arrow) and transfers electrons via the FeS centers to the NiFe active site (inset), where protons are reduced to hydrogen gas (2H+ + 2e– → H2). Proton transport and Na+/H+ exchange could take place in MbhM/MbhH and MbhA-G, respectively (see the main text). A lipid gap with modeled phosphatidylinositol is shown in the bottom inset.

MbhH is related to the proton pumping NuoN module of complex I (E. coli nomenclature; Nqo14/ND2 in other species), but it is rotated by 180° in the membrane relative to the former (Figure ). This antiporter-like subunit comprises two symmetry-related TM-helix bundles with one broken helix each, buried charged residues, and a conserved ion pair. These features form functional elements for water-mediated proton pumping in complex I[16−24] and could have a similar function in Mbh. MbhH is further clamped by an amphipathic transverse helix of MbhI that could secure tight electrostatic interaction between the subunits.[8]

MbhG and MbhD+E are related to NuoK/J (Nqo11/10, ND4L/ND6)—subunits that are also most likely involved in proton transport in complex I.[8,9,16,18,20−24] However, in contrast to the latter where they reside close to the interface between the hydrophilic and membrane domains, these subunits are located in the middle of the membrane domain in Mbh (Figure ). MbhG/D/E could catalyze Na+/H+ transport, although the molecular principles remain debated. Yu et al.[1] suggested that MbhC is responsible for Na+ transport, while proton transfer was suggested to occur in MbhH and MbhD/G. In contrast, Steiner and Sazanov[15] resolved sodium ions in Mrp subunits homologous to MbhD (N-terminal part of MrpA) and MbhG (MrpC). A sodium pathway was also suggested to reside in MbhA (MrpE), with a putative input channel from the negatively charged side (cytoplasmic side) in MbhA and an output site to the positively charged side (periplasmic side) of the membrane at the MbhG/F/H interface (Figure ). Sodium ions have also been suggested to enter MbhH (MrpD) and exit to the periplasmic side at MbhF/G/H.[14]

The membrane-bound MbhM (related to NuoH/Nqo8/ND1 of complex I) forms an “ankle” region at the interface of the hydrophilic and membrane domain. This region has a special functional role in complex I,[18,23−25] whereas a large, ca. 10 Å wide cleft is observed in the cryoEM structure at the MbhM/MbhH interface (Figure , bottom inset).[1] The structure also revealed conserved loop regions within MbhM (NuoH/Nqo8), MbhI (N-terminal part of NuoA/Nqo7), and MbhL (NuoD/Nqo4), which undergo conformational changes in complex I.[20,23,26,27]

To shed light on the elusive ion transport mechanism of Mbh, we probe here the coupling and dynamics between the redox and proton pumping domains and the sodium/proton exchange domain in the membrane-bound hydrogenase from P. furiosus(1) by combining classical molecular dynamics simulations and data clustering methods, with quantum chemical calculations. Our findings illustrate detailed hydration dynamics, putative sodium and proton binding sites, and possible locations of the proton channels. We also present a mechanistic model that could explain how the coupling between redox-driven proton reduction, proton pumping, and ion transport is achieved.

Results

Global Dynamics and Proton Wires Leading to the Active Site

To gain insight into the functional dynamics of Mbh, we performed 8 μs of MD simulations of the solvated 14-subunit enzyme modeled in a phosphatidylinositol (PI) membrane, which is abundant in P. furiosus,[28] and explored the effects of modeling key residues in different protonation states (Tables S1 and S2). Mbh remains structurally stable during the MD simulations, with an overall root-mean-square deviation (RMSD) of <4 Å relative to the refined cryoEM structure (Figure S1A). The protein shows a high flexibility toward the cytoplasmic side surface of MbhA/B/C and the upper part of the hydrophilic domain (Figure S1B).

The conserved loops connecting the active site and a charged funnel in MbhM are also highly flexible, particularly the long MbhI loop, which we modeled in silico based on the experimentally resolved backbone coordinates. We note that the dynamics projected from the MD simulations are in good overall agreement with B-factors extracted from the cryoEM maps (Figure S1B), supporting that the simulations capture the global dynamics of the protein.

To probe possible H+ and Na+ transfer pathways, we next analyzed buried water networks formed during the MD simulations. The dry cryoEM structure reaches a highly hydrated state within 100 ns of the MD simulations, where a few hundred water molecules form transient interactions within the hydrophilic domain as well as buried parts of the membrane domain (Figure A,B and Figure S2). The water wires connecting the cytoplasmic side with the NiFe site could conduct protons for H2 production. On the basis of clustering analysis (see the Methods section), we observe four possible proton channels (see definition in the Supporting Information), comprising several charged residues and water molecules, leading to Glu21L (Figure A,B, Figure , and Figure S3A), a highly conserved residue within hydrogenases[13] located on the β1−β2 loop next to the NiFe site. The active site Cys374L is further bridged by a water molecule to Glu21L, particularly when His75L is modeled in a protonated state (Figure A). Interestingly, the neighboring Glu20L corresponds to the quinone-coordinating active site histidine (His38Nqo4 in T. thermophilus) in complex I and is conformationally flexible.[1,7−12,18] Water influx occurs at the interface between MbhL and MbhM formed by the lower edge of the β-sheet in MbhL (around 25%; median over all simulations, Figure and Figure S3B). Channels also form along the interface of the β1−β2 and MbhI loops (25%, Figure and Figure S3B) and at the MbhL/MbhJ interface (34%, Figure and Figure S3B). A cleft between MbhM and MbhH (12%) also weakly contributes to the overall water influx into the NiFe-site (Figure and Figure S3B), similar to the E-channel in complex I.[8,18,20-25] These channels, particularly around the β1−β2 loop, comprise functionally important motifs in the canonical complex I,[8,11] thus also supporting their relevance in Mbh.

Figure 2

Global hydration dynamics of Mbh. (A) Conserved residues (see the Methods section) within the complex I superfamily are shown in stick representation and buried water molecules from MD simulations within 6 Å of these residues as red spheres. (B) Around 200–250 water molecules associate with the hydrophilic (MbhI-N, blue) and the membrane domains (MbhA-H, yellow) of Mbh during MD simulations. The plot shows the median over 16 independent simulations (8 μs in total) for each time point, with 25 and 75 percentiles, represented by the shaded areas. Hydration dynamics for individual subunits are reported in Figure S2.

Figure 3

Hydration and protonation dynamics of the Mbh active site. (A) The NiFe cluster is stabilized by conserved contacts to surrounding residues. Glu21L on the β1−β2 loop is bridged to Cys374L by a water molecule in up to 85% of the MD simulation time depending on the state (NIB: OH– ligand; NIC: H– ligand; NIC H75+: NIC with cationic His75L). (B) DFT calculations suggest that water-mediated proton transfer from Glu21L to Cys374L is energetically and kinetically feasible. Only the chemically active atoms are shown. See Figure S4, Table S5, and Movie S1 for further details. (C) Glu21L flips away from the active site upon deprotonation.

Figure 4

Water clusters leading to the cytoplasmic bulk and the cleft between MbhH and MbhM could allow for reprotonation of Glu21L from the bulk. The clusters are present in all simulations, but the number of water molecules varies depending on the protonation states. The figure shows centroid pathways based on cluster analysis (see the Methods section).

Interestingly, when Glu21L is modeled in its protonated state, as predicted by our electrostatic calculations (see the Methods section), the residue forms a water-mediated contact with Cys374L of the NiFe site (Figure A,C). In contrast, upon deprotonation, Glu21L flips away from the NiFe site toward the hydrated channels leading to the cytoplasmic bulk phase at the MbhM/MbhL interface (Figure C and Figure S3).

To probe whether Glu21L could act as a proton donor for Cys374L, we performed quantum chemical density functional theory (DFT) calculations (Figure B), which allowed us to address the reaction energetics along key steps of the NiFe catalytic cycle (see the Methods section, Figure S4). The water-mediated proton transfer from Glu21L to Cys374L has a reaction barrier of around 9 kcal mol–1, and the two states are nearly isoenergetic by ΔE = +0.6 kcal mol–1 in the NiII/FeII state (Figure B, Table S5, and Movie S1).

These findings suggest that the proton uptake is kinetically accessible on physiologically relevant time scales along the conserved Glu21L/Cys374L pathway (Figure B), residues that may also be functionally important in canonical hydrogenases.[29,30,35] Formation of the protonated Cys374L leads to an elongation of the Ni–Cys374L bond (Figure B and Figure S3), which in turn opens up the contact for the proton to the H— bound between the Ni and the Fe. This could lead to the formation of the H2 species in a process, enabled by a subtle (130° rotation of the Cys–H bond (Figure B, Figure S4, Table S5, and Movie S1; cf. also refs (29) and (30)). We note that the conformation of Arg320L, which forms contacts with both the NiFe center and Asp372L, strongly influences the energetics of these reaction steps, and samples both inward and outward conformations in the MD simulations (Figure S9D), while the conserved salt-bridge to Asp372L remains intact. On the basis of the water network analysis and DFT calculations, we suggest that at least one proton could be transferred from the cytoplasmic bulk phase to the NiFe center for the H2 production via Glu21L, Cys374L, and the water channels formed around the β1−β2 loop.

Functional Hydration of the Mbh Membrane Domain

The membrane domain of Mbh undergoes a significant hydration change during the MD simulations, in which around 200 water molecules establish pathways that could enable both proton and Na+ transport across the membrane (Figure A,B and Figure S2). We observe a hydration site from the cytoplasmic side at the large cleft between MbhH and MbhM (Figure S5A,B). These water molecules form hydrogen-bonded arrays toward the central axis of MbhH connecting Lys256H with His346H, His350H, and Lys354H. This pathway leads to the cytoplasmic bulk between MbhH and MbhM via Lys409H (Figure A) and accounts for a significant portion of the water influx toward MbhH. The same site was recently suggested to support Na+ transport in the related Mrps,[14] although the chain of three conserved lysine residues along the pathway is expected to electrostatically disfavor Na+ transport.

Figure 5

Structure and dynamics of MbhH. (A) Left inset: water molecules establish a hydrogen-bonded proton array between the cytoplasmic bulk and the hydrophilic axis, comprising histidine and lysine residues and leading to Lys256H. The figure shows the last snapshot (500 ns) from simulation S3 (Table S1). Right inset: lipid binding sites on the cytoplasmic and periplasmic sides of the membrane showing the 20 most probable binding poses from eight independent simulations, as determined by pyLipid (see Supporting Information, Methods). (B) Water pathways based on clustering analysis of 8 μs MD data (see the Methods section). Three water inlet clusters serve as water conduits to the hydrophilic axis in MbhH: (I) from the cytoplasm via the lipid cleft (contribution 75%), (II) from the cytoplasmic side via the E-channel homologue in MbhM (18.4%), and (III) from the periplasmic side by a water wire between TM1/2 of MbhI and TM2/3 of MbhM via the broken helix TM12 of MbhH (2.7%). (C) Distances between Glu141M and Lys409H across the lipid filled cleft during MD simulations. Protonation of Glu141M increases the distances to Lys409H. See Figure S9 for distance distribution in all simulations. The inset shows the shortest water-mediated pathway between the two residues in the deprotonated state (cyan sticks) and the lack of connectivity in the protonated state (green sticks). (D) Protonation of Glu141M increases the distance to Lys409H (dark blue, in Å) and disrupts the hydrogen-bonding (HB) connectivity, as shown by the probability of the stable HB wire between the two residues (in tan) and the number of HBs along the shortest path (in light blue).

A transient water chain also connects Glu141M with Lys409H across the nonpolar lipid-filled cleft (Figure C,D). However, these residues are ca. 10–12 Å apart when Glu141M is modeled in a deprotonated state and ca. 18 Å upon protonation of the latter, rendering proton exchange between the residues unlikely, especially since no titratable groups stabilize the long hydrogen-bonding wire.

Three to four PI lipids bind to the cleft region (Figure A), sealing the MbhH/MbhM gap from the periplasmic side and the horizontal gap at the MbhH/MbhM interface near the broken helix TM12 (Figure A). This observation is consistent with the blurred density around the region in the cryoEM map.[1]

Our simulations also indicate that binding of at least one additional lipid from the cytoplasmic side is sterically possible, which could block the proton transfer across the MbhM/MbhH interface (Figure S3D).

We observe another water influx site from the periplasmic side that reaches Lys409H between MbhI and MbhH (TM12), with many conserved polar residues lining up along the pathway (Figure A,B). This cluster, which has an overall occupancy of around 5%, is not present in all simulations (Figure S3B), but it could nevertheless be functionally relevant. Although no water molecules enter the hydrophilic axis of MbhH via TM7, we observe a partial pathway leading to the cytoplasmic side around Lys256H, His350H, Lys354H, and water molecules (Figure B). These findings thus indicate that a canonical S-shaped pathway, analogous to those observed in complex I,[20,21,23,24] could also establish a proton pathway in MbhH.

Residues along the lateral proton transfer wire in MbhH strongly interact with the conserved Lys225H/Glu143H ion pair at the interface of MbhH and a putative Na+-binding MbhG/D (see below), but the two regions do not exchange water molecules (Figure A). The conformation of this ion pair could modulate the proton transfer barrier along the lateral pathway in MbhH, and vice versa, similar to what has been observed in complex I (see below).[8,16,19,20,36]

This ion pair also forms strong electrostatic interactions with a charged cluster in MbhG/D, comprising Asp37G, His41G, and Glu69D residues that bind Na+ in Mrp (Figure A, see below).[15] This region, called here the “O-site” (for the occluded Na+ binding site) together with the ion pair, is hydrated from the cytoplasmic side of the membrane during our simulations and could form key elements enabling proton-coupled Na+ transport.

Figure 6

Na+/H+ transport in MbhA-G. (A) Putative sodium binding sites and functional motifs in Mbh based on MD simulations. Insets: the N-site at MbhA/C, here with a bound Na+, is accessible from the cytoplasmic bulk; O-site in MbhG with two modeled Na+ ions; P-site, which spontaneously binds bulk Na+ ions during most MD simulations; and a putative conformational gate at TM2/TM3 of MbhD modulating ion transport between the N- and O-sites. (B) Hydration of the O-site in different protonation and conformational states. Asp37G deprotonation (blue) and conformational changes in the MbhD gate (brown) favor hydration of the O-site. Hydration in the closed-gate conformation (π-bulge form) is low (orange), whereas biasing toward α-helical conformation (gray) increases the hydration levels (see Figure S6). (C) O-site hydration in MD simulations. Influx via MbhD/G and MbhA/F accounts for ca. 75% of the observed water molecules in the region. The MbhA/F pathway is connected from the N-site via the MbhD gate to the O-site. Inlets from pathway cluster analysis are shown as spheres, and areas with volumes of greater than median hydration are depicted as surface.

Hydration Dynamics of the Putative Na+-Binding Site

We find that hydration of the putative O-site occurs via two main pathways, which account for >75% of the water molecules observed in this region. The major channel leads from the cytoplasmic side at the MbhA/MbhF interface and is established around a kink region at Pro88C of the broken helix TM3 of MbhC (Figure A). The water molecules flow in via a cluster of conserved polar residues in MbhB and MbhC (Thr39B/Thr42C and Asp35B/Asn38B/Thr86C) that could form a primary Na+ binding site (Figure A), here called the “N-site” (for negatively charged side Na+-binding site). Interestingly, a similar motif establishes a Na+ binding site in the unrelated light-triggered Na+-pump KR2.[32] In the related Mrps, a Na+ pathway was recently suggested to also involve a Thr/Asn cluster,[15] but leading via His137A toward the proposed N-site.[15] In our simulations, the latter pathway remains sealed from water molecules, whereas in Mrps, two detached TM-helices (in the subunit homologous to MbhA) could open up this channel.

To further probe the principles underlying sodium binding, we placed a Na+ ion around the N-site in MbhB/MbhC. In these simulations, the Na+ rapidly finds a binding pose that is stabilized by Asp35B, Asn87B, Ser68B, Thr86C, and water molecules (Figure A and Figure S5). Mutations of the homologous residues in Mrps block Na+ transport activity, thus further supporting the functional relevance of this site.[14]

Water molecules also enter from the cytoplasmic side between MbhD and MbhG close to Lys23G and Lys28G and lead further to the O-site that could support proton transfer across this region. The MbhD/G pathway is favored in simulations, where Asp37G is modeled in a deprotonated (anionic) state. We note, however, that the high positive charge around this region is unlikely to support Na+ transport (cf. ref (14)), unless protonation changes are involved (cf. also ref (1)).

MbhD Regulates Ion Transport to the Na+/H+ Coupling Site

The conformation of the π-bulge in TM3 of MbhD (residues 71–79) correlates with the overall water influx toward the putative Na+/H+ coupling site in MbhD/G (Figure B). In the hydrated state, the π-bulge moves toward MbhF and opens up a gap between TM2 and TM3 of MbhD (Figure B). This leads to an increase in the hydration level by a factor of 2 (Figure B and Figure S6B), whereas in the dry state, the TM2 and TM3 helices remain in close contact and prevent water diffusion across the site (Figure A).

The conformational state of the TM3 helix of ND3 in the canonical complex I was recently suggested to regulate proton transfer during the active-to-deactive transition and possibly during turnover.[23−27] More specifically, the α-helical form, present in the active state of complex I, favors well-wired proton pathways, whereas the π-bulge, observed in the deactive state, could block proton transfer.[24] Interestingly, the homologous TM3 of MrpA is captured in an α-helical form,[15] suggesting that conformational transitions between a π-bulge and α-helix could also be involved in Mbh.

To further probe how such conformational transitions affect the hydration dynamics in Mbh, we perturbed TM3 of MbhD to form an α-helix during the MD simulations (see Supporting Information Methods). The modeled α-helix remains dynamically stable for 0.5 μs during unrestrained MD simulations (Figure S7) and results in rapid hydration of the sodium cavity from the cytoplasmic side via MbhA (Figure B,C). In stark contrast, in simulations where TM3 forms a π-bulge, the region becomes 90% less hydrated (Figure B). These findings suggest that TM3 of MbhD could act as a gate that controls water and Na+ exchange between the proposed N- and O-binding sites.

Sodium Binding at the P-Site

We observe spontaneous sodium binding in nearly all MD simulations at the periplasmic side surface in a cavity formed by MbhB, MbhC, and MbhF (cf. also ref (1)). The Na+ binds to the conserved Asp59B, Tyr26F, and His75C or, in a few simulations, to the nearby loop of MbhA. Both binding modes remain highly stable throughout the 0.5 μs simulations (Figure S5G), with rapid water exchange with the bulk. However, in contrast to the N-site at the MbhB/C interface, we do not observe pathways leading toward the interior of Mbh or to the O-site, located ca. 18 Å away. However, Na+ transport to/from the P-site could be achieved by conformational switching into an alternate access state,[37] possibly supported by the structurally similar TM motifs also found in other transporter proteins.[38]

Sodium Transport between Binding Sites

To probe the Na+ transport mechanism between the putative N- and O-sites, we placed sodium ions at the respective sites or at the MbhC/MbhD interface. The Na+ ions remain tightly bound at the N- or O-sites on 0.5 μs time scales (Figure A), but at the interface region, the Na+ interacts with TM2 and TM3 of MbhD for ca. 0.5 μs, after which it moves to the O-site to a position that closely resembles the binding mode in Mrp (Figure B and Figure S10).[15] These findings support that the MbhD gate could be involved in Na+ transport in Mbh.

Figure 7

Sodium transport is influenced by the protonation states in the O-site. (A) Dynamics of Na+ ions between the O- and N-sites in unbiased MD simulations from the center of the membrane (z = 0 Å) with Asp37G modeled in deprotonated (blue) and protonated (red) states. The structure depicted in (B) is marked with a yellow circle. (B) Structural snapshot of the sodium motion after 500 ns of classical MD simulation. Intermediate sodium positions during the simulation are marked with dashed circles in the structure and the time trace depicted in panel A. Residues forming contacts with sodium ions during the simulation are listed in black if the ConSurf score is at least eight and otherwise in gray (see Table S4).

We also probed the reverse Na+ transport direction by placing Na+ ions at both sides of His41G, as experimentally resolved for Mrp.[15] When the carboxylates are modeled in their deprotonated states, the Na+ ions remain strongly bound at the O-site (Figures A and 7A), whereas upon protonation of Asp37G and Glu69D, the Na+ moves to the N-site within 50 ns (Figure A). The Na+ diffusion leads to a subtle conformational change in the π-bulge region of MbhD-TM3 (Figure S10). Na+ binding drastically reduces the proton affinity of both Glu69D and Asp37D, suggesting that the Na+/H+ binding events are tightly coupled (Figure S8).

The water network analysis suggests that Asp37G and Glu143H are in direct hydrogen-bonded contact via 2–3 water molecules, which could enable proton transfer between the residues upon Na+ binding. To test this process, we transferred the proton from Asp37G to Glu143H, which leads to rapid dissociation of Lys225H toward Lys256H. These conformational changes lower the pKa of Lys256H and could thus trigger proton transfer in the MbhH subunit (Figure A and Figure S8).

Figure 8

Putative redox-driven Na+/H+ transport and proton pumping in Mbh. (A) Sodium binding to the O-site triggers protonation changes, conformational switching of the Glu143H/Lys225H ion pair and deprotonation of Lys256H. Inset: the predicted protonation fraction of Lys256H for the open ion pair (blue) and closed ion pair conformations (red). (B) Summary of mechanistic model: (1) Na+ uptake to the O-site via MbhB/C/D (N-site) and H+ uptake via MbhB, (2) modulate the conformational state of the buried ion pair in MbhH, and (3) trigger proton transfer across the membrane in MbhH. (4) Reprotonation of Lys256H induces (5) association of the ion pair and (6) ejection of the Na+ and H+ across opposites sides of the membrane. Protonation of the O-site lowers the affinity of the Na+, which could leave via the P-site by conformational changes in the MbhD gate (see text). Protons could exit to the cytoplasmic side via a water cluster observed in MbhG.

Discussion

Mbh was originally suggested to function as a proton pump based on experiments performed in membrane vesicles.[3] However, it is also possible that the enzyme employs the small energy transduced from Fd-driven hydrogen production to drive secondary active sodium/proton exchange,[4,6] an operational mode that is consistent with the sodium-dependent F0F1-ATP synthase in P. furiosus.[33] Sodium functions as the coupling ion in the homologous hydrogenase in Thermococcus onnurineus NA1, supporting a possible similar role also in Mbh.[34]

Although the exact Na+/H+ stoichiometry of Mbh is unknown, the bioenergetic boundary conditions thermodynamically allow Mbh to transfer one ion (Na+ or H+) per 2e– at 120 mV sodium motive force (smf) and thus enable the possible electrogenic function of Mbh (cf. also refs (3 and 35)). In contrast, for the related Mrps, which lack the redox module, mechanisms based on one Na+ exchanged per proton[15] or even sodium transport without H+ exchange were recently suggested.[14]

We propose that the terminal MbhA-MbhG module functions as the sodium translocation domain in Mbh (cf. also refs (1 and 15)), whereas the hydration dynamics in MbhH supports its involvement in proton transfer. Our findings indicate that the conserved Lys225H/Glu143H ion pair of MbhH and the putative Na+/H+ binding O-site establish a coupling element between the proton and sodium transport and proton pumping (Figure A). To this end, the sodium affinity at the O-site is modulated by the protonation state of the Glu69D/His41G/Asp37G cluster, and vice versa. Na+ binding to this site could trigger dissociation of the Lys225H/Glu143H ion pair (Figures and 8A) and, in turn, induce proton transfer from Lys256H (Figure A), either directly across the membrane via His350H/Lys354H and water molecules around TM7 (Figure B) or laterally toward the MbhH/M interface (Figure B). However, because the unresolved lipid cleft poses challenges in the current modeling and the region lacks titratable groups, it remains possible that the MbhH/M interface is not used for proton transfer to the cytoplasmic side. As an alternative, we propose that the protons are taken up via the water-mediated pathway formed along TM12 and transferred via the Lys409H-His346H-His350H chain to reprotonate Lys256H. This is expected to result in reassociation of the Lys225H/Glu143H ion pair and ejection of the proton from the O-site to the cytoplasmic side as well as Na+ ejection to the periplasmic side. The ion release could involve conformational switching into an alternate access state, e.g., at the MbhB/D interface. Conformational changes in MbhD were found to favor Na+ transfer between the N- and O-sites (Figure S10), whereas the proton release could occur via the water cluster observed at MbhD/G.

In this putative model, the transport of Na+ and H+ across the membrane could be triggered by reprotonation of Lys256H in MbhH, thus following overall similar, although simpler, physical principles as proposed for complex I.[8] Although the exact molecular principles of the redox-driven conformational changes in the conserved loop regions and charge arrays in MbhM and MbhI/L remain unclear, we note that the so-called E-tunnel could regulate the proton affinity and accessibility of the terminal lysine in MbhH and be involved in coupling the redox reactions with the charge transport process.

On the basis of the structural similarity to complex I, we note that the functional elements in MbhH are also expected to support proton conduction in the reverse direction, depending on the external conditions, and therefore do not exclude the possibility of its involvement in establishing a secondary proton gradient (Figure B, cf. refs (4 and 6)).

Our study also found evidence for proton pathways leading to the NiFe center, with input sites, around the highly conserved β1−β2 loop interface, leading to Cys374L and the metal bound hydride in the active site via Glu21L. Glu21L undergoes a protonation-state-dependent conformational switching, which could help shuttle protons to the active site. A similar conformational switching of the functionally central Glu242 in cytochrome c oxidase has been suggested to favor kinetic gating and prevent possible back-leaks.[39]

Previous MD simulations on the NiFe-hydrogenase from D. vulgaris revealed three proton pathways leading to Glu34L (equivalent of Glu21MbhL),[40] a residue that is also supported by other experimental[41] and computational[30] studies. More specifically, two pathways were observed in NiFe-hydrogenases that correspond to our channels at the MbhL/I/J interface, whereas another pathway showed overall resemblance to our MbhL/N/J channel.[40] To this end, we note that the membrane subunit MbhM could lead to some structural differences at the MbhL/M interface. A pathway resembling the MbhL/M site was also observed in respiratory complex I (Figure S3),[42] whereas hydration through the E-channel from the cleft region in Mbh has not been described before. Interestingly, His75MbhL has also been suggested to form an alternative pathway to the active site.[21,43] Although we find no significant contribution of the latter channel in the modeled states (Figure S3A), our data provide further support for several of the other pathways previously observed in NiFe-hydrogenases.

We note that oxygen sensitivity in hydrogenases has been linked to a proximal 3Fe4S iron sulfur center as well as to the topology of the hydrophobic gas channels leading to the active site.[41,43,44] Mbh shows an oxygen tolerance with a half-life of around 14 h in oxygen sensitivity assays,[6] despite comprising only 4Fe4S centers. Mutational studies suggest that constriction of the hydrophilic tunnel around the corresponding MbhI/J/L pathway observed here can significantly enhance the oxygen tolerance in NiFeSe-hydrogenases from D. vulgaris.[45] These findings indicate that the MbhI/J/L pathway could also be relevant for the oxygen tolerance of Mbh, whereas explicit diffusion of O2 or H2 along the channels would be necessary to study the effect.

Conclusions

We have presented here functional dynamics of the membrane-bound hydrogenase (Mbh) from Pyrococcus furiosus by using large-scale molecular simulations. In summary, we observed putative water-mediated proton pathways leading along the MbhL/MbhM interface to Glu21L, which could shuttle protons to the NiFe center, responsible for the H2 production. We also observed significant hydration changes in the MbhH subunit during the MD simulations as well as pKa shifts upon conformational changes in a buried ion-pair at the MbhH/MbhG interface—functional elements that could support proton transfer across the archaeal membrane. The simulations also revealed three putative Na+ binding sites that could be responsible for the Na+/H+ transport activity of Mbh. The N-site at the MbhC/B interface is a Thr/Ser-rich region, which, via conformational changes in a transmembrane helix of MbhD, can transfer Na+ ions to the buried carboxylate rich O-site, located at the MbhG/MbhD interface. Conformational switching of an analogous TM-helix was recently suggested to regulate proton transfer also in the canonical complex I[24] (cf. also ref (23)). We further found that the Na+ affinity of the O-site is sensitive to the protonation state of Asp37G, and vice versa. This putative Na+/H+ binding site is electrostatically strongly coupled to the buried ion pair at the MbhH/MbhG interface, which in turn could trigger proton transfer along MbhH. We also found evidence for a putative Na+ binding site at the periplasmic side of the membrane in MbhC/B/F, which upon conformational changes could become accessible to the O-site binding region. Our combined findings provide insight into key conserved coupling principles within the complex I superfamily and detailed functional insight into archaeal energy transduction mechanisms.

Materials and Methods

The cryo-EM structure of Mbh from Pyrococcus furiosus (PDB ID: 6CFW)[1] was embedded in a 1-palmitoyl-2-palmitoleoyl-sn-glycero-3-phosphoinositol (PYPI) membrane by using CHARMM-GUI.[46] We additionally modeled two PYPI lipid molecules in the cleft between MbhH and MbhM, unresolved side chains in MbhI, and missing N/C-terminal residues. The model was embedded in a 200 × 100 × 168 Å3 box comprising TIP3P water molecules and ions to mimic a 250 mM NaCl concentration. MD simulations were performed by using the CHARMM36m force field[47] in combination with force field parameters for the NiFe site for the NIB (FeII/NiIII–OH–) and NIC (FeII/NiIII–H– ligand) states.[48] The MD simulations were performed by using NAMD ver. 2.13[49] with periodic boundary conditions (PBC) and long-range electrostatics modeled with the particle mesh Ewald (PME) approach with a grid size of 1 Å, at 1 bar and 310 K, and with a 2 fs integration time step. The membrane was first equilibrated around the protein, with heavy atoms restrained by a harmonic force constant of 10 kcal mol–1 Å–2. Lipid tails were initially melted with a harmonic restraint on the headgroups. After gradual heating to 310 K, all restraints were removed, followed by creation of the NIB and NIC states and models with different protonation states of the titratable residues. Initial protonation states were assigned based on electrostatic calculations (see the Supporting Information and Table S2). MDAnalysis,[50] Visual Molecular Dynamics,[51] and PyMol[52] were used for analysis and visualization (see the Supporting Information Methods and Table S1 for further simulation details).

Clustering Analysis

Hydration dynamics in the MD simulations were analyzed in all states by aligning the trajectories around subunits MbhA-MbhH, MbhI, and MbhM, or MbhJ-MbhN. Cα atoms of residues in helices and sheets were used to define the scope of the analysis by a convex hull. Water molecules around putative sodium sites were identified within 6 Å of the center of mass of the O-site (E143H, K225H, D37G, H41G, E69D), N-site (D35B), or P-site (D59B). For MbhH, water molecules within 4 Å of K409H, H350H, K354H, or K256H, and for the NiFe site, water molecules within 4 Å of E21L were considered in the analysis. The clustering analysis was performed with Aquaduct ver. 1.0.11[53] using the Barber algorithm (at a cutoff of 1.4 Å), with path trimming for the sodium site analysis. Water analysis was performed on the full data set for all sites in the membrane domain. Water clusters in the active site were analyzed based on simulations S3, S5, S10, and S11 (Table S1). The largest cluster was recursively divided into two subclusters by using the balanced-iterative reducing and clustering using the hierarchies (BIRCH) method for simulations S3 and S10 (for NiFe site) and simulations S1, S3, S4, and S9 (for MbhH). Clusters with <10 members were assigned to the outlier cluster, and conserved residues were identified by using ConSurf.[54] For the visualization of cluster medoid paths, paths with the same input and output cluster were chosen.

DFT Models

Quantum chemical DFT models of the NiFe site were created based on the MD-minimized Mbh model. The DFT model comprised E21L, C68L, I70L, C71L, H75L, T376L, R320L, D372L, C374L, C377L, and N36J, the NiFe core, and three H2O molecules. Amino acids were cut and saturated with hydrogen atoms at the Cα–Cβ bond. The DFT model comprised 126 atoms, which were structure optimized at the B3LYP-D3/ε=4 level[55,56] by using the def2-SVP basis sets for all atoms except Fe and Ni, which were modeled with the def2-TZVP basis sets.[57] The reaction pathway for proton transfer from E21L to C374L and C374L to H– were optimized along minimum-energy pathways, followed by optimization of H3O+ transition state, which showed one imaginary frequency at −924.6 cm–1 (see the Supporting Information Methods). All systems were optimized in the triplet spin state. TURBOMOLE ver. 7.5[58] was used for the DFT calculations (see Supporting Information Methods, Figure S4, and Table S5 for details of all QM calculations).