Protons serve as substrate or product in many chemical and biological reactions. In proteins, protons often travel 10 Å or more from the surface to reach an active site. Proton gradients across the membranes of bacteria, mitochondria and chloroplasts contribute to the electrochemical gradients, ∆Ψ, used to store cellular energy (Mitchell, 1961; Rich, 2008; Nicholls, 2010; Gunner et al., 2013). The proton gradient can be generated by vectorial electron transfer, where reactants are oxidized and reduced on different sides of the membrane. Here the electrons cross the membrane, but the protons only move to or from the separated redox sites. In contrast, proton pumps transfer protons through the transmembrane proteins, requiring mechanisms to avoid downhill proton transfer.

Regardless of the mechanism a protein uses, it takes energy to build a proton gradient. The input energy is light in photosynthetic proteins (Cardona et al., 2012; Ge and Gunner, 2016; Cardona and Rutherford, 2019), redox reactions in the electron transfer chain (Kaila and Hummer, 2011), ATP hydrolysis (Vasanthakumar and Rubinstein, 2020) or the dissipation of the gradient of another ion (Fowler et al., 2015; Brandt, 2019). The protons flow down the electrochemical gradient then fuel processes such as ATP synthesis in F₁/F₀ ATPase (Walker et al., 1991; Futai et al., 2012; Yanagisawa and Frasch, 2017) and the active transport of ions and metabolites (Accardi and Picollo, 2010; Gunner et al., 2013).

To build the gradient, protons are transferred from the more negative, N-side of the membrane, where they are at lower concentration (higher pH) to the positive, P-side where they are at higher concentration (lower pH). The P-side is in the periplasm of bacteria, the outer membrane space of mitochondria and in the lumen on the inside of the chloroplast thylakoid membrane. The N-side is toward the bacterial cytoplasm, the mitochondrial matrix and the chloroplast stroma. The electrochemical gradient, ∆Ψ, is made up of the gradient of protons (the ∆pH) but also has contributions from other ions, adding to a voltage change, ∆V, across the membrane (Decoursey, 2003). The ∆Ψ across a given membrane determines the energy needed to push a proton uphill in the protein pumps described here or the energy liberated when protons run from P- to N-side as used for ATP synthesis.

Although we refer to “protons”, H⁺ does not travel alone. Rather it is associated with a water (hydronium, H₃O⁺) or two water molecules as a Zundel cation (H₅O₂ ⁺) or as a larger, Eigen complex (H₉O₄ ⁺) (Agmon, 1995; Wraight, 2006; Farahvash and Stuchebrukhov, 2018). In proteins, the proton can also be bound to redox cofactors, to acidic or basic residues or trapped as a stabilized hydronium (Xu and Voth, 2006; Freier et al., 2011; Ikeda et al., 2017).

Protons move through a chain of oriented molecules by a Grotthuss proton transfer mechanism (Agmon, 1995; Cukierman, 2006; de Grotthuss, 2006; DeCoursey and Hosler, 2014). An active group in the middle of the chain is: 1) a hydrogen bond donor to the next group in the direction of proton transfer and 2) has a lone pair of electrons that is a hydrogen bond acceptor from the neighbor toward the proton input side. In the Grotthuss mechanism no proton moves more than one bond, as the proton acceptor takes ownership from the neighboring proton donor. However, overall the coupled transfers lead to a proton rapidly leaving the input side and appearing at the end of the chain. There are many reviews of the chemistry of proton transfer reactions as well as of proton transfer reactions in proteins (Hammes-Schiffer, 2001; Pomès and Roux, 2002; Blomberg and Siegbahn, 2006; Swanson et al., 2007; Knight and Voth, 2012; Ishikita and Saito, 2014; Miyake and Rolandi, 2015; Wikström et al., 2015; Sakashita et al., 2020).

Two requirements create barriers for Grotthuss proton transfers. First, the chain of hydrogen bonds between proton donors and acceptors needs to be pre-organized. Then, once the proton has transferred, the hydrogen bonds are arranged to return the proton back to the origin, not to move another proton in the same direction. The hydrogen bonded chain needs to fully reorient for the next proton to transfer, so overall proton flux is limited by this slow “hop and turn” process (Nagle and Morowitz, 1978).

Vectorial proton coupled electron transfer. Proteins such as PSII, cytochromes bc₁ and b₆f use vectorial electron transfer reactions where oxidation and reduction reactions are spatially separated to add to the proton gradient. Thus, oxidation occurs on the P-side, where protons are released because the loss of an electron lowers the oxidized product pK_a below the pH of the nearby compartment. Reduction occurs on the N-side, where reduction shifts the product pK_a to be higher than the compartment pH (Rich, 2008; Nicholls, 2010; Gunner et al., 2013; Gunner and Koder, 2017). Within the protein, a sequence of electron tunneling reactions pass the electrons 30 Å or more between the terminal electron donor and acceptor (Gray and Winkler, 2003; Moser et al., 2006). The interior electron transfer reactions are not coupled to gain or loss of protons. Thus, a proton gradient is generated without moving protons through the membrane by a redox loop mechanism as suggested by Mitchell (Mitchell, 1977). The intra-membrane, middle of these proteins are mostly non-polar side chains with few associated water molecules, so discourage proton transfer. However, as will be seen in the discussion of PSII and bRCs, the sites of final, proton coupled oxidation or reduction can be 10 Å or more from the surface, requiring long-range proton transfer to move the protons to the active site.

Proton pumps. The proton pumps include the well-studied, light-driven Bacteriorhodopsin (Balashov, 2000; Luecke, 2000), Complex I (Mathiesen and Hägerhäll, 2002; Hirst, 2013; Sazanov, 2014) and the heme copper oxidase (HuCuOx) family (Kaila et al., 2010; Lee and Ädelroth, 2013). To ensure protons move in the correct direction pumps require three elements. These are proton transfer paths, as found in vectorial electron transfer proteins. However, pumps need Proton Loading Sites, PLS, placed along the proton transfer path, and gates. PLS transiently change their proton affinity to load a proton when the gate is open to the N-side and releases it when it is open to the P-side. Pathway gating and proton loading must be synchronized to guard against energy dissipating proton transfer from P- to N-side.

This review will compare and contrast the residues that make up the proton transfer elements in three light activated proteins: Bacteriorhodopsin (bR) and the photosynthetic proteins, Photosystem II (PSII) and the purple non-sulfur photosynthetic bacterial reactions centers (bRCs) and in the proton pumps Complex I and Cytochrome c oxidase, which are the first and last protein in the erobic electron transfer chain.

Green Fluorescent Protein (GFP). GFP provides a simple example of the role of side chains in and around the proton transfer path (Figure 1). It also shows how fast protons can transfer via a pre-organized Grotthuss competent chain (Brejc et al., 1997; Zimmer, 2009; Donati et al., 2018). GFP is well studied as it has revolutionized cell biology. When introduced into a genome it is co-expressed with a specific protein of interest and its characteristic florescence allows the targeted protein to be localized within a living cell. The chromophore in GFP is a photoacid that absorbs light in the near UV and emits in the green (Zhou and Han, 2018). The large Stokes shift results from the ground state absorption and excited state emission occurring from molecules with different charges. Thus, in the ground state the chromophore is the PLS, while Glu 222 is the proton acceptor when the chromophore proton affinity is diminished by excitation. In this system the path for rapid proton release must be ready to carry the proton away prior to relaxation of the chromophore (Chattoraj et al., 1996).

The proton is transferred through bound water and Ser 205 to Glu 222. A nearby His 148, Thr 203 and Asn 146 are on the outskirts of the proton transfer wire (Stoner-Ma et al., 2005; Di Donato et al., 2011). These provide a polar residue fence to pre-orient the hydrogen bond network. As the protein is light activated experiments can synchronize the protein for kinetic measurements to follow changes in the hydrogen bonding network. The proton arrives on the Glu in less than 10 ps (Donati et al., 2018). Vibrational spectroscopy shows that there are rapid changes in hydrogen bond orientation that precede proton transfer, presumably to fine tune the hydrogen bond connections for Grotthuss proton transfer. The kinetic transients are distinguished as rearrangements, which do not have a kinetic isotope effect, and proton transfers, which do (Di Donato et al., 2011).

Gramicidin (gA). The gA channel is a proton and cation conducting channel that has been used to study proton transfer through a linear water wire, with no intervening side chains. gA is made of two short ß-strand peptides, capped on N- and C-terminal ends. The N-termini meet in the center of the membrane. In the ß-helix the side chains are to the outside, with sufficient space to hold a linear chain of ≈8 water molecules in the middle of the helix. gA is an antibiotic, allowing protons and other cations to cross, depolarizing the cell’s electrochemical gradient (Kelkar and Chattopadhyay, 2007; Li et al., 2016). The simplicity of this system has made it ideal for the experimental (Dorman and Jordan, 2004; Ryu et al., 2015; Paulino et al., 2020) and computational (Roux, 2002; Allen et al., 2004; Till et al., 2008; Lazaridis and Hummer, 2017; Zhang et al., 2020) studies of the Grotthuss proton transfer mechanism.

The water molecules in gA form hydrogen bonds with the two neighboring water molecules and with the amides of the surrounding peptides. The balance of the water-water and water-amide interactions determines the stability of the water wire, the stability of an excess proton within the wire and the barrier for flipping the wire orientation once a proton has translocated to transport another proton via the hop and turn mechanism. The rate determining step for transfer can be initially orienting the water molecules or flipping the oriented water dipoles to the correct direction (Pomès and Roux, 2002; Agmon et al., 2016; Bozdaganyan et al., 2019).

Recent computer simulations showed a rather substantial sensitivity of the water wire orientation to the force field and simulation method (Zhang et al., 2020). In molecular dynamics simulations, with a classical force field the water molecules are fully aligned within the channel with rare flips from one orientation to the other. In contrast, MD with a the Drude polarizable force field shows more disorganized water molecules. Monte Carlo sampling with a Continuum Electrostatic force field also show relatively disorganized water chain. Thus, the balance of the forces that determine the orientation of the water molecules are such that different simulation conditions induce different behavior.

Experiments have also supported a range of structures for water in the channel. The experimentally derived rate of proton translocation through the channel under a transmembrane voltage gradient appears to be diffusion limited (≈2 × 10⁹ s⁻¹) even at pH 0 (Cukierman, 2000; Decoursey, 2003). The reorientation of the water chain is likely to be the rate determining step in Grotthuss proton transfer (Pomès and Roux, 1998). In the MD simulations with a classical force field the water chain flips its direction at ≈4 × 10⁸ s⁻¹ and this process is faster with the Drude force field. Thus, the simulations are in general agreement with the measured rate of proton transfer. However, recent solid-state NMR studies show a well-organized water-wire with flip rates on the millisecond time scale, which would suggest very slow turnover for proton transfers. The NMR studies point to hydrogen bonds between water molecules and amides near the first and last turns of each ß-helix leading to this stability. Thus, despite the simplicity of its structure, gA remains a protein where our understanding of the channel water structure and the mechanism of proton transport remains incomplete.

bR is the simplest and best studied proton pump (Balashov, 2000; Baudry et al., 2001; Gunner et al., 2006; Lanyi, 2006; Lórenz-Fonfría et al., 2008; Clemens et al., 2011). The bacteriorhodopsin family uses retinal not chlorophyll based photoactivation to fuel proton or ion pumping (Figure 2). Absorption of a 568 nm photon initiates a reaction cycle that removes a proton from the cell interior (N-side) and releases one to the outside (P-side) adding to the proton gradient. The overall reaction is simply: H N - side +   +  hv  →  H P - side +

The proton transfers are driven by the light induced conformational changes of the retinal from all trans to 13-cis, which is coupled to small changes in the helical packing in the protein core. There are three PLS in bR that illustrate different mechanisms to control PLS protonation and the connection to the proton transfer path.

Characterization of proton transfer intermediates. In light activated proteins such as GFP, bR, PSII and bRCs experiments can follow the sequential reaction mechanism since the flash of light used to start the reaction synchronizes the population. Time resolved visible and IR spectroscopy, combined with site directed mutations allow assignment of kinetic features to individual residues. bR intermediates were identified with the proton resting on acidic amino acids making up three well separated PLSs, each with metastable intermediates with different ionization states (Balashov, 2000; Lórenz-Fonfría and Kandori, 2009; Lórenz-Fonfría et al., 2011; Lórenz-Fonfría and Heberle, 2014). These intermediates clearly demonstrate the role of transient resting places for protons. Changes occur in times ranging from the picosecond transitions that trap the photon’s energy in the isomerized retinal to the milliseconds required to complete the full photocycle.

In bR, it is possible to crystallize protein trapped in different intermediates by a combination of mutation and temperature changes (Edmonds and Luecke, 2004; Hirai et al., 2009; Wickstrand et al., 2015). Simulations using these structures have shown that the calculated equilibrium proton distribution changes between intermediates as expected (Bashford and Gerwert, 1992; Spassov et al., 2001; Onufriev et al., 2003; Song et al., 2003; Song and Gunner, 2014). More recent time resolved crystal structures have been obtained using X-ray free electron lasers (XFEL) (Nogly et al., 2018; Wickstrand et al., 2019). These structures show many of the motions of water molecules and side chains and helices seen in earlier trapped structures. However, as the XFEL structures are not in deeply trapped intermediates, the dynamic structures provide additional information. However, as the transitions between photocycle intermediates are not all well separated in time, the XFEL structures each contain a mixture of states.

PSII (Umena et al., 2011) is a multi-subunit protein pigment complex present in the thylakoid membrane of plants and cyanobacteria (Cox et al., 2013; Vinyard and Brudvig, 2017; Pantazis, 2018). The reaction is initiated by chlorophyll excitation with a 680 nm photon. Water is the primary electron donor and plastoquinone, PQ, is the final electron acceptor (Figure 3) (McEvoy and Brudvig, 2006). The overall reaction is: 2 H 2 O  +   4 h v   +   2 PQ  +   4  H stroma +   →  O 2   +   2 PQH 2   +   4  H lumen +

The Oxygen Evolving Complex (OEC), an inorganic Mn₄CaO₅ cluster, catalyzes water oxidation following four sequential oxidations of the cluster (through five S-states) releasing O₂ (Suga et al., 2019). The OEC is ≈20 Å from the surface, requiring paths for water entry and O₂ and proton release. The region around the OEC is filled with water molecules that separate into three discrete water-filled channels moving to the lumen (Figures 4A,B) (Vassiliev et al., 2012; Vogt et al., 2015). On the electron acceptor side, Q_B in PSII is quite close to the stroma requiring only a short proton transfer path to bring in protons (Saito et al., 2013).

The reaction center, bRC, of the purple non-sulfur bacteria is the first membrane protein whose structure was solved at atomic resolution (Deisenhofer et al., 1985). RCs are light activated proteins so as with GFP, bR and PSII, time resolved measurements allow individual steps in the series of electron and/or proton transfer reactions to be monitored by time-resolved spectroscopy, showing the individual steps in the reactions (Okamura et al., 2000; Wraight, 2006).

bRCs and PSII are Type II reaction centers where a fully reduced, QH₂ is the final product. The quinol dissociates into the membrane to serve as the substrate of the b₆f complex in oxygenic photosynthesis and the bc₁ complex in bacteria (Cardona et al., 2012; Cardona and Rutherford, 2019). The D1 and D2 subunits of PSII are related to the L and M subunits in bRCs (Raymond and Blankenship, 2004). The bacterial systems use a photon, in the range of 860–960 nm. Thus, they do not have enough energy to carry out the PSII reaction, which uses a 680 nm photon to fuel the uphill transfer of electrons from water to quinone (Heathcote et al., 2002). The primary electron donor in bRCs is periplasmic (P-side) cytochrome c. The redox reactions of cytochrome c are not coupled to proton binding/release. The overall reaction is: 2 cyt c P - side 2 +   +   2 h v   +  UQ  +   2 H N - side +   →   2 cyt c P - side 3 +   +   2 UQH 2

In contrast to PSII, which has a very short distance to the N-side, bRCs have an H subunit, capping the N-side of the protein, requiring a much longer path for the protons to reach the Q_B site (Figure 6).

Quinones as a model redox coupled proton transfer reactant. Electrons move one at a time between cofactors in proteins generating free radical intermediates. While some cofactors such as chlorophylls, hemes, iron sulfur clusters and Tyr are stable one electron redox cofactors, unpaired electrons are often sources of toxic reactive oxygen intermediates (Weisz et al., 2017). Quinones function as single electron donors/acceptors within proteins, but accumulate two electrons and protons (Paddock et al., 2003; Müh et al., 2012). These lipid soluble cofactors thus transport electrons from many proteins including PSII, bRCs and complex I described here to the bc₁ complex in mitochondria and bacteria or b₆f complexes in chloroplasts as electrons move down the electron transfer chains.

As quinones cycle between oxidized quinone (Q), semiquinone (Q^•-) and fully reduced quinol (QH₂) their proton affinity changes. The quinone pK_a is <0, it is <5 for the semiquinone (Zhu and Gunner, 2005; Gunner et al., 2008; Hasegawa et al., 2017). Hence, the anionic semiquinone is often the stable state within the protein. However, the generation of Q^•- can trigger proton binding to nearby residues (Graige et al., 1996; Abresch et al., 1998; Alexov and Gunner, 1999; Paddock et al., 2003). The pK_as for the fully reduced quinol is >10 so the second Q_B reduction is coupled to binding two protons to the cofactor. The difference in the hydrogen bonding pattern for the two quinone carbonyls and two quinol hydroxyls promote quinone dissociation in PSII (Shevela et al., 2012; Saito et al., 2013). As will be described below, in Complex I quinone reduction leads to a large conformational change that trigger proton pumping through distant pathways (Gupta et al., 2020; Gutiérrez-Fernández et al., 2020; Kampjut and Sazanov, 2020).

The quinone electrochemistry is modified within the protein, but the underlying proton affinity of each redox state in the isolated compound strongly influences the order in which protons are bound. The sequence of reactions in bRCs (Graige et al., 1996) and likely in PSII is: 1) Q_B is first reduced to the anionic semiquinone. The negative potential causes protons to be bound to nearby amino acids (Okamura and Feher, 1992; Wraight, 2006; Gunner et al., 2008); 2) uphill protonation of the semiquinone is the rate-determining step preceding the second reduction; 3) reduction is followed by binding a second proton and release of the quinol.

Proton transfer pathways to Q _B in bRCs. The proton transfer paths to Q_B have been well studied (Abresch et al., 1998; Okamura et al., 2000). These are different than those described above near the OEC, as residues play a much larger role in the network. However, they are similar in that there is a tangled complex of proton transfer paths. There is a large number of acidic and basic residues buried in the protein near Q_B that influence the electrochemistry of the quinone and provide paths for proton transfer (Figure 7) (Sebban et al., 1995; Lancaster et al., 1996; Abresch et al., 1998; Alexov and Gunner, 1999; Rabenstein et al., 2000). FTIR difference spectra obtained on Q_A and Q_B reduction shows broad features characteristic of a polarized, interconnected hydrogen bonded network of water molecules and amino acids around the two quinones (Breton and Nabedryk, 1998). L-Asp 210 and Asp 213 may share a proton in the ground state, serving as a PLS. Protonating one of the acids removes a negative charge, stabilizing the semiquinone Q_B ^•- and keeping a proton available for passage to the quinone itself (Lancaster et al., 1996; Alexov and Gunner, 1999; Ishikita et al., 2003). The mutants L-Asp213Asn and L-Ser223Ala slow the rate of this reaction, with the mutation of L-Asp 213 having a bigger impact (Paddock et al., 1994; Paddock et al., 1995). In the absence of L-Asp 213, H-Glu 173 may provide an alternative location for the proton in this extended PLS (Paddock et al., 2003). Thus, this web of acidic residues combines the functions of a cluster PLS and complex proton transfer path.

L-Glu 212, which is a protonated PLS in the ground state, provides the second proton to Q_B (Wraight, 2004). The pK_a of the Glu is ≈10, trapping a proton near the quinone in the ground state ready when needed (Kleinfeld et al., 1984; Okamura and Feher, 1992). The mutant L-Glu212Gln does not affect the delivery of the first proton, supporting this site being neutral in the presence of Q_B ^- . However, the transfer of the second proton is totally blocked, indicating L-Glu 212 is a unique single site PLS (Paddock et al., 1989; Shinkarev et al., 1993; P. H.; McPherson et al., 1994; Okamura et al., 2000; Wraight, 2004).

Three possible paths. The complex web of acidic and other polar residues near Q_B leads to the question of what is the route for proton transfer from the N-side surface to the quinone binding site. The crystal structures reveal three likely paths (Figure 7) (Abresch et al., 1998). The longest path, P1, is ≈20 Å long. It enters the protein near H-Asp 224 or M-Asp 240 and passes to L-Glu 212, which provides the second proton to Q_B. P2, also ≈20 Å long, starts near M-Tyr 3 and moves via H-Glu 173 to L-Asp 213, which donates the first proton to Q_B. P3 is the shortest path, with only ≈7 Å between L-Asp 213 and the surface M-Asp 17 with one water molecule in the middle (Abresch et al., 1998).

Surface PLS as a proton collection site. RCs also have a well characterized external cluster near the entrance to P3 made up of H-Asp 124, H-His 126, H-His 128. The cluster is a proposed proton collection site (Utschig et al., 1998; Paddock et al., 1999; Okamura et al., 2000) (Figure 7). Zn²⁺ or Cd²⁺ bind here and slow proton transfer to Q_B. Clusters of protonatable groups near the surface of proton transfer paths are found in other proteins. A similar proton accumulation site is found in the D-channel (Cai et al., 2018). The broad channel in PSII exits to a cluster of surface acidic residues that can trap the released proton (Bondar and Dau, 2012; Kaur et al., 2021).

P3 appears to carry protons in wild-type bRCs. Mutation of L-Asp 210 and M-Asp 17, have a larger impact when Zn²⁺ or Cd²⁺ are present, showing an additive effect of multiple changes to this pathway. However even with P3 blocked, protons still enter to Q_B, indicating that other routes can serve as pathways, but with slower transfer rates (Okamura et al., 2000). Thus, in the tangled potential proton transfer network, multiple paths are possible, but some are preferred.

Complex I is the first and the largest protein in the erobic respiratory electron transfer chain of bacteria and mitochondria (Brandt et al., 2003; Hirst, 2013; Sazanov, 2015; Kaila, 2018). Within the protein electrons are transferred from NADH to a flavin and through a series of Iron Sulfur (FeS) complexes to a quinone (Verkhovskaya et al., 2008; Efremov et al., 2010; Hirst, 2013; Zickermann et al., 2015). The overall reaction is: NADH  +  H +   +  Q  +   4 H N + → NAD +   + QH 2   +   4 H P +  

The quinone, Q, is often ubiquinone but can be menaquinone in bacteria such as Thermus thermophilus, the source of the protein for the first complete crystal structure (Baradaran et al., 2013). The transfer of the two electrons, which occurs in the N-side peripheral arm, leads to the pumping of four protons from the N- to P-side of the membrane embedded portion of the protein (Figure 8).

Complex I combines elements of two disparate proteins. The long, peripheral arm is likely derived from a soluble hydrogenase and three of the proton channels are related to Mrp Na-H antiporters (denoted antiporter channels), however, the fourth, E-channel, is unique to Complex I and its close relatives (Efremov and Sazanov, 2012; Brandt, 2019). Complex I is remarkable, as the tightly coupled electron transfer and proton pumping elements are separated by as much as 300 Å from the NADH binding site, at the end of the peripheral arm, to the distal proton pumping subunit (Baradaran et al., 2013; Kaila, 2018). Quinone binding and reduction lead to a rotation of the soluble arm that connects the redox reactions to proton pumping in some way (Gupta et al., 2020; Gutiérrez-Fernández et al., 2020; Kampjut and Sazanov, 2020).

Antiporter: simple pathway. Complex I provides examples of both simple and complex proton transfer pathways. There are four proton paths, three through the antiporter subunits and one through the E-channel (Hirst, 2013; Ripple et al., 2013; Sazanov, 2015; Di Luca et al., 2017; Haapanen and Sharma, 2018; Saura and Kaila, 2019). The crystal structures show likely, linear paths through each antiporter subunit (Efremov and Sazanov, 2011; Baradaran et al., 2013; Zickermann et al., 2015) which have chain of well conserved acidic and basic residues in the center running parallel to the membrane (Figures 8A,B) (Fearnley and Walker, 1992; Torres-Bacete et al., 2007; Efremov and Sazanov, 2012). Recognizable water chains leading to the N- and P-sides are seen in computational studies (Kaila et al., 2014; Di Luca et al., 2017; Haapanen and Sharma, 2017; Röpke et al., 2020). Moving along each pathway from the N-side is a Glu/Lys pair then a central Lys followed by either a Lys or Glu (Baradaran et al., 2013; Kaila, 2018). Their protonation states change as the proton is handed from one ionizable residue to the next. Thus, the antiporter channels are simple linear proton transfer paths.

With a linear proton transfer path, it is often possible to identify a unique gating element. Simulations have been carried out to investigate the behavior of Complex I with different protonation states for these residues (Kaila et al., 2014; Di Luca et al., 2017; Haapanen and Sharma, 2017). Increasing the net charge in the interior leads to water molecules being brought into the protein in MD trajectories and they are expelled when the charges are neutralized (Kaila et al., 2014; Hummer and Wikström, 2016; Di Luca et al., 2017). These hydration/dehydration changes will gate proton transfer through the channels, similar to that described above for Asp96 in bR. They can be validated by seeing waters in different locations in structures trapped in different intermediates or by interpretation of IR spectroscopy (Lórenz-Fonfría et al., 2008). However, a buried charge will attract water in MD simulations so it is important that the residue protonation states be correctly assigned in the simulation (Hummer and Wikström, 2016).

Comparison of the structures of Complex I from different organisms shows conservation of the P-side proton release paths in the antiporter subunits. However, on the N-side MD studies (Kaila et al., 2014; Di Luca et al., 2017; Haapanen and Sharma, 2017) found a pathway similar to one identified in the crystal structure of Y. lipolytica (Zickermann et al., 2015) Complex I but different from the one proposed from the T. thermophilus (Sazanov, 2015) crystal structure. Thus, it is not known if the exit path is conserved. Similar changes in pathways through evolution are also found comparing A- and B-type CcO as will be described below.

E-channel: A complex proton transfer path. In contrast to the linear proton transfer path seen through the three antiporter subunits, the fourth proton travels through a path directly under the periplasmic arm denoted the E-channel. This region has a web of water molecules and polar and protonatable residues characteristic of a complex proton transfer path with several PLS clusters (Di Luca et al., 2017; Saura and Kaila, 2019; Gutiérrez-Fernández et al., 2020; Khaniya et al., 2020). There are several competing proposals for the proton transfer path through the E-channel. It has been suggested to use subunits Nqo10 and Nqo11 (Efremov and Sazanov, 2011; Zickermann et al., 2015) or subunit Nqo8 (Baradaran et al., 2013). Various computational studies also provide different answers (Kaila et al., 2014; Di Luca et al., 2017; Haapanen and Sharma, 2017). This uncertainty about the route is characteristic of complex proton transfer paths. There is a growing consensus that residues in subunits Nqo7, 8 10 and 11 are important for E-channel function (Figure 8A). Network analysis, which can accommodate complexity, has proposed a complete path through subunit Nqo4 and Nqo8 at the N-side entry, moving through subunit Nqo8 and Nqo7 in the center, and exiting through subunit Nqo10 and Nqo11 (Khaniya et al., 2020).

Role of quinone in Complex I as a gate. Complex I is able to couple the energy releasing redox reactions in the peripheral arm to the energy requiring proton pumping through four, distant well separated pumping sites (Baradaran et al., 2013; Kaila, 2018). The quinone binding site in complex I is ≈25–30 Å above the membrane surface, which is different from its location in any other quinone dependent membrane protein (Baradaran et al., 2013; Zickermann et al., 2015). Quinone binding leads to the rotation and tilt of the peripheral arm (Gutiérrez-Fernández et al., 2020). MD simulations (Gamiz-Hernandez et al., 2017; Warnau et al., 2018; Gupta et al., 2020) and Monte Carlo sampling (Khaniya et al., 2020) find changes in the connectivity of the hydrogen bond network that depend on the presence and redox state of the quinone. Movement of subunit Nqo4 and Nqo8 leads to changes in the E-channel hydrogen bond network that propagate into the first antiporter channel by a distortion near Nqo10 (Tyr 59) (Gutiérrez-Fernández et al., 2020). The shifts in hydrogen bond network and residue protonation initiated by the quinone reactions thus yield changes in the interaction between the key Glu/Lys residues in the very distant antiporter channels (Efremov and Sazanov, 2011; Kampjut and Sazanov, 2020). This web of long-range communication is not needed in smaller proteins such as CcO, described below, where the change in electrostatic potential due to the redox reactions can directly modify PLS proton affinity coupled to proton pumping.

MD trajectories show quinone binding influences the E-channel by enriching the number of hydrogen bonds near the N-side, which are proposed to open the channel for proton uptake (Gupta et al., 2020; Khaniya et al., 2020). However, when QH₂ is bound the charge of conserved residues change leading to modification of the water wires in the proton transfer channels (Gamiz-Hernandez et al., 2017; Kaila, 2018). The E-channel central region has extended clusters of protonatable residues including Nqo7 (Asp 72), 8 (Glu 130), 8 (Glu 163), 8 (Glu 213) (Khaniya et al., 2020) [the nomenclature uses residue numbering from the T. thermophilus complex I in the form as Nqo subunit (residues)].

Beyond the central cluster of polar residues there is a hydrophobic barrier that blocks the proton transfer to the P-side in the E-channel. Thus, while there have been several studies of the proton entry, connections are rarely drawn from the center to the P-side (Baradaran et al., 2013; Kaila et al., 2014; Zickermann et al., 2015; Di Luca et al., 2017; Haapanen and Sharma, 2017). Network analysis of the hydrogen bonds made in MD trajectories suggests several paths that rely on transient wetting events (Khaniya et al., 2020). One lies near Nqo7 (Tyr 7) and 8 (Tyr 124). Another possibility is from Nqo8 (Glu 130) to 10 (Tyr 59) (Gutiérrez-Fernández et al., 2020; Steiner and Sazanov, 2020). However, these proposed bridging residues are not well conserved. Thus, what permits the proton to cross the hydrophobic barrier, and whether it conserved through Complex I evolution, is still an open question. A similar hydrophobic barrier is also seen in the voltage-sensing domain (VSD) of voltage-gated ion channels. Here mutations of the hydrophobic residues make the system leaky, showing the importance of non-polar residues to block uncontrolled proton transfers (Banh et al., 2019).

Residue conservation. The conservation and sensitivity of residues to mutation can provide evidence that there is a unique pathway for protons. The multisequence alignment of the residues in the linear paths through the three antiporter subunits (Baradaran et al., 2013; Di Luca et al., 2017; Haapanen and Sharma, 2017; Gutiérrez-Fernández et al., 2020) were compared with that found for the central cluster in the complex E-channel pathway (cluster 4 residues) (Khaniya et al., 2020). The WebLogo (Crooks, 2004) provides a graphical comparison of the results (Figures 8C,D). The residues along the linear antiporter paths are highly conserved. In contrast, the E-channel central cluster shows much weaker conservation, suggesting that, while the cluster as a whole must function, individual residues may not be uniquely important.

Mutation The sensitivity to mutation may also distinguish linear from complex proton transfer paths. Thus, mutation of residues along a linear path should severely impair activity. In contrast, a complex path may be less sensitive as there are multiple routes for the proton, though as shown above for bRCs, not all need be equally favorable. Many of the residues in the well-defined antiporter channels have been subjected to site-directed mutations. Mutation of the residues shown in (Figure 8A) severely reduce quinone oxidoreductase activity that is tightly coupled to proton transfer since there is no alternative paths for proton transport (Torres-Bacete et al., 2007; Euro et al., 2008; Michel et al., 2011). However, the E-channel is more complex and there is less consensus about the path. Mutations of proposed E-channel residues often modify but do not kill activity (Taylor et al., 2002; Yang et al., 2009).

CcO is a proton pump belonging to the heme-copper oxidase superfamily (Kaila et al., 2010; Liang et al., 2017; Kaur et al., 2019). The energy for proton pumping comes from electrons from cytochrome c (Cyt c) reducing O₂ to water in the binuclear (Heme & Cu) center (BNC), located in the protein center (Figure 9). The BNC is reduced stepwise, one at a time to store four electrons. O₂ reduction takes place in one step in the fully reduced BNC (Kaila et al., 2010; Cai et al., 2020). O₂ production likewise takes place in one step in the fully oxidized OEC of PSII. This mechanism protects against the release of toxic reactive oxygen intermediates. The overall reaction is: 4 cyt c P 3 +   +  O 2   +   ( 4 +  m )  H N +   →   4 cyt c P 2 +   +   2 H 2 O  +   ( 4 +  m )  H P +

Four electrons come from the cytochrome c on the P-side, and four protons from the N-side to the BNC for chemistry. m is the number of protons pumped across the membrane. There are several related classes of CcO, denoted A, B and C. The A-type CcO is found from bacteria to mammals, while B- and C-type are found in bacteria that live at low O₂ levels. In the A-type CcO, m = 4. B- and C- type CcOs differ in the types of heme used and in the number of protons pumped/electron, with m generally less than four (Lee et al., 2012). CcO provides examples of simple and complex proton transfer paths and simple and cluster PLS as well as a gate generated by hydration/dehydration changes.

Proton transfer paths through CcO. In the A-type CcO two linear water filled channels (D- and K-channels) are seen. Each has an essential ionizable residue at the entry on the N-side and at the end near the BNC, but none within the channels. The D-channel has Asp 132 at the entry and the essential, isolated Glu 286 as the PLS (Wikström et al., 2000; Brändén et al., 2001). The K channel has Lys 362 near the BNC and B-Glu 101 near the entry (Ma et al., 1999; Brändén et al., 2002). The buried Glu and Lys are both isolated in hydrophobic parts of the protein. In the ground states, their proton affinity has shifted so both are neutral (Rb. sphaeroides CcO numbering used here).

The D-channel carries six of the eight protons in the A-type CcO reaction cycle, while the K-channel carries two. At the center of CcO is heme a and heme a₃ and Cu_B of the active site BNC. As the retinal does in bR, the large cofactors may help to block proton transfer through the protein. The protons from the K channel exit into the BNC to be added to the reduced product water, while the D channel exits between the two hemes. The importance of these linear pathways were demonstrated by mutation of the residues at the beginning and end of the channels leading to loss of activity (Jünemann et al., 1997; Qian et al., 1997; Mills and Ferguson-Miller, 2002). Oddly, the D-channel is missing in B- and C-type CcOs, with only a K-channel remaining (Lee et al., 2012). Thus, as suggested in complex I, proton transfer paths may shift through evolution. In the B- and C- type CcOs it remains unclear how the pumped protons move around the active site to be delivered to the P-side.

Complex proton transfer pathways. The P-side of all CcOs has a tangled cluster of strongly interacting polar and protonatable residues that do not provide an obvious single exit path, although linear paths have been suggested (Popović and Stuchebrukhov, 2005; Björck et al., 2019). The hydrogen bond network on the P-side of A- and B-type CcO, was analyzed using Monte Carlo sampling and network analysis (Cai et al., 2018; Cai et al., 2020). Calculations were initiated with experimental crystal structures as well as with snapshots from MD trajectories carried out in different redox states of the hemes and protonation states of key residues. This analysis recognized a linear proton transfer path through the D-channel in the A-type CcO. A very large cluster of interconnected residues was identified as the P-side PLS (Figure 9). This cluster exits through several paths to a region near the cytochrome c binding site. Thus, there is an exit region not a unique exit for protons in this complex path.

Single residue PLS in CcO. Glu 286 is located at the top of the D-channel of A-type CcO (Kaila et al., 2010). It is isolated from other protonatable residues so forms a simple PLS. It plays an essential role, releasing a proton to the BNC for chemistry and to a P-side PLS cluster for pumping. The X-ray crystal structures show it is in a dry region and all simulation techniques give it a high proton affinity as there is nothing in the structure to stabilize an anionic residue (Hummer and Wikström, 2016). Measurements found a pK_a of ≈10 for turnover that is assigned to Glu286 (Namslauer et al., 2003). However, MD simulations showed protonation of a propionic acid in the P-side PLS breaks a hydrogen bond and opens a cavity which then fills with water (Goyal et al., 2013; Son et al., 2017).

The hydration of the water cavity near Glu 286 also serves as a gate for the proton transfer pathway. In the crystal structures and in protein equilibrated in MD trajectories without the water cavity there is no exit found from the D-channel to the P-side PLS. Thus, the closed cavity blocks the backflow of protons (Cai et al., 2018). However, when the cavity is hydrated Glu 286 becomes well connected to the extensive PLS cluster on the P-side. Changes in hydration also help control the proton transfer from Glu 286 to the BNC (Wikström et al., 2003) Thus, hydration tunes the proton affinity of an isolated residue and serves as a gate in the proton transfer pathway, a pattern described for Asp 96 in bR and for the antiport subunits in complex I.

The role of tautomer shifts in a complex PLS. A proton pump must regulate the thermodynamics of PLS loading and unloading, and then change proton affinity when the reaction progresses. This requires tuning the free energy difference between the loaded and unloaded states at the pH of interest as well as the shift in this value as the protein goes through the reaction cycle (See fuller description in Supplementary Material S1). The PLS must remain in the appropriate loaded/unloaded state until the reaction has progressed and the accessibility of N- and P-side is modified by the gates opening/closing (Kim et al., 2007; Kim et al., 2009; Kim and Hummer, 2012; Stuchebrukhov, 2018; Stuchebrukhov, 2019). The challenge is to find the changes that can trap, hold and then release the proton. An analysis of the proton distribution in the PLS cluster in B-type CcO provides some insight into the atomic details of one mechanism.

On the P-side of the B-type CcO, an extended cluster of six residues was found to behave as a PLS (Figure 10) (Cai et al., 2020). The unloaded PLS has one proton bound (net charge -4) while the loaded cluster has two protons. The protein surrounding the PLS provides sufficient long-range positive potential to stabilize the cluster negative charge. There are six tautomers with one proton and 12 with two protons distributed over the six residues (Eq. 1). By investigating the proton affinity of different tautomers in snapshots derived from MD trajectories it was found that a shift in the hydrogen bond pattern changes the tautomer selected. This resulted in dynamic states where a loaded and unloaded state are close in energy so changes at the active site lead to proton binding or release. However, the BNC is ≈15 Å from the PLS and the change in the BNC do not shift the PLS proton affinity enough to fully load a proton to the PLS cluster. This behavior is seen Supplementary Figures S2A,B when there is only a small shift in proton affinity in a group with a pK_a near the pH (Supplementary Figure S2C, middle titration). The problem of incomplete loading/unloading appears to be solved by moving the bound protons and rearranging the hydrogen bonding pattern, which is described as a tautomer trap. This leads to the PLS being trapped loaded or unloaded because the cluster proton affinity is either too high or too low (Supplementary Figure S2C, right-most titration) or unloaded (Supplementary Figure S2C, left-most titration).

The proton shift in the CcO complex PLS shows how a tautomer trap can solve the problem that a PLS, which is sensitive to changes in the protein, may not strongly trap the proton. Thus, in active structures, the change in proton affinity due to the BNC will lead to some changes in the PLS protonation state. But this shift in free energy is insufficient to reliably move the PLS between being fully loaded and unloaded (Supplementary Figure S2). Then a tautomer shift moves the PLS from the dynamic configuration (where the proton was bound or lost) to the stable, fully loaded or unloaded locked configuration. This mechanism may be similar to activation and inactivation process in voltage gated ion channels, where conformational changes occur when the channels are an active open state, then block the channel and transit to inactive state (Aldrich, 2001). A tautomer trap is only available to a PLS cluster, it is not possible in a single residue PLS.

The structure of multiple proton transfer paths in several proteins that add to the transmembrane electrochemical gradient show a range of motifs. Thus, they can be simple linear paths as found in the D- and K- channels of CcO and the antiporter subunits of complex I. They can also be complex paths as seen on the P-side of CcO and the E-channel of complex I. Paths can be filled with water so that a proton never needs to use a side chain as found around the OEC in PSII or to be handed through a mixture of side chains and water molecules as found in GFP and in bRCs. With simple, single site PLS, as at the exit from the D-channel and the N-side of bR, changes far from the PLS trigger water influx that leads to proton release and production of a water chain to ferry protons. A similar mechanism is used in the antiporter subunits of complex I. In contrast, in the PLS cluster on the proton release side of bR and CcO, small, local rearrangements of a cluster of strongly interacting residues leads to large changes in proton affinity to cause the PLS to load and unload.

Thus, the framework that proteins will have proton transfer paths, Proton Loading Sites (PLS) and gates allows the analysis of each of these proteins. However, the motifs vary in the residues that make up the needed elements and in their complexity. Each proton pump reviewed here use structures with different complexity for different parts of the proton transfer paths. The advantages of different motifs remain to be determined.