O to bR transition in bacteriorhodopsin occurs through a proton hole mechanism

Significance

Bacteriorhodopsin (bR) is a molecular ion pump that has been long regarded as a prototypical system that demonstrates the Grotthuss mechanism for long-range proton transport in biology. Using extensive quantum mechanical/molecular mechanical (QM/MM) simulations, we show that the O to ground state transition in bR, in fact, features proton exchange between two distant groups via a proton hole mechanism. The proton hole is stabilized by a highly conserved arginine residue, which also plays the crucial role of modulating the hydration level of the protein cavity via a conformational transition. Therefore, the current study highlights the general importance of protein internal hydration level change and consideration of all possible titration states of water to the mechanism of long-range proton transport in biomolecules.

Abstract

Extensive classical and quantum mechanical/molecular mechanical (QM/MM) molecular dynamics simulations are used to establish the structural features of the O state in bacteriorhodopsin (bR) and its conversion back to the bR ground state. The computed free energy surface is consistent with available experimental data for the kinetics and thermodynamics of the O to bR transition. The simulation results highlight the importance of the proton release group (PRG, consisting of Glu194/204) and the conserved arginine 82 in modulating the hydration level of the protein cavity. In particular, in the O state, deprotonation of the PRG and downward rotation of Arg82 lead to elevated hydration level and a continuous water network that connects the PRG to the protonated Asp85. Proton exchange through this water network is shown by $\sim$0.1-$\mathrm{\mu }$s semiempirical QM/MM free energy simulations to occur through the generation and propagation of a proton hole, which is relayed by Asp212 and stabilized by Arg82. This mechanism provides an explanation for the observation that the D85S mutant of bacteriorhodopsin pumps chloride ions. The electrostatics–hydration coupling mechanism and the involvement of all titration states of water are likely applicable to many biomolecules involved in bioenergetic transduction.

Bacteriorhodopsin (bR) is a biomolecular pump that uses the energy of light to pump protons across the cell membrane (1). Since its first structural characterization more than 45 y ago (2), bR has become a model system for structural biology and bioenergetic transduction mechanisms, due mainly to its relative structural simplicity (Fig. 1A) and experimental advantages (3). The proton pumping mechanism has been studied extensively with diverse spectroscopic methods, which established the various kinetic intermediates and their lifetimes. As shown in Fig. 1B, the photocycle starts with the ground state (referred to as the bR state); absorption of a photon by the chromophore retinal triggers its isomerization from all-trans to the 13-cis,15-anti form, which is followed by a cascade of conformational transitions and elementary proton transfer steps. Upon completion of the photocycle, bR converts the photo energy of ca. 50 kcal/mol into the proton-motive force by releasing one proton to the extracellular side, and the fundamental features of the proton pumping mechanism are clear and undisputed (4). Several key kinetic states have been characterized with crystallography using cryotrapping techniques; there are 153 X-ray crystallographic structures of bR in the protein data bank (PDB) with resolution as high as 1.25 Å (5). Using time-resolved serial femtosecond crystallography with an X-ray free electron laser, Nango et al. (6) have captured 13 structural snapshots of the conformational changes in bR that occur in the nanoseconds to milliseconds after photoactivation, leading to detailed insights into the mechanism of energy transduction.

Fig. 1.

Three-dimensional structure of the bR state and the photocycle at pH $\ge$ 6. (A) Side view of the bR monomer in PDB ID 5B6V (6). Functionally important amino acids are shown and labeled. Key proton transfer steps are labeled 1 through 5. (B) Scheme of the bR photocycle with the important intermediates and their lifetimes. Also see SI Appendix.

Despite this remarkable progress, there is no consensus on the complete reaction mechanism (3, 7, 8). Some atomic-level features have been demonstrated only recently, such as the role of protonated water clusters in long-range proton transfer (PT) (9) or the primary PT being relayed by Thr89 (10), while others still have not been explored. In particular, there is little X-ray structural characterization for the O state; PDB contains four O-like crystal structures, however, all of them are bR mutants. It is assumed that the newly formed all-trans retinal adopts a twisted conformation (11) and that structural rearrangements of the protein helices lead to a more open conformation of the extracellular side. Vibrational changes during the O$\to$bR transition were characterized (12), and a protonation of an unidentified Asp or Glu was noted and subsequently assigned to D212 (13, 14). The detailed mechanism for the transition of O back to the ground (bR) state remains largely unclear, however. During this transition, estimated to occur on a time scale of 0.5 to 5 ms (15–18), D85 has to be deprotonated and the proton release group (PRG, which includes E194 and E204) protonated, as long as the pH of the medium is above 6 (17). The two motifs are separated by as much as 15 Å (Fig. 1A) and thus the transition has to involve a remarkably long-ranged proton exchange. The nature of the PRG itself was revisited in 2017 and “a reevaluation of the last PT steps in bR” was deemed necessary (p. E10909 in ref. 19).

The changes in the protonation pattern are expected to be coupled with an upswing movement of the R82 sidechain toward D85 and D212; the causal relationship between R82 motion and the PTs is not known. For example, for the PTs to proceed, the cavity waters have to form a continuous water chain from D85 to the PRG, which could be hindered by the R82 sidechain located in between. In fact, based on the available crystal structures, it is not obvious that there is enough water in the cavity to facilitate the long-range PT. Indeed, previous quantum mechanical/molecular mechanical (QM/MM) studies based on minimum energy path (MEP) calculations found that a Grotthuss-like PT from D85$\to$E204 through D212 exhibited a very high reaction barrier of 22 kcal/mol and was strongly endothermic by $\sim$10 kcal/mol (20); these results were inconsistent with the experimentally estimated kinetics for the O$\to$bR transition or its exergonic nature established based on the pK_a difference between D85 and the PRG (21, 22). An alternative proposal suggested the R82 sidechain as a simultaneous proton acceptor/donor during the long-range PT (23); a QM/MM-MEP calculation for this mechanism, however, found an even higher barrier of $\sim$36 kcal/mol (24). Moreover, R82 mutants are still able to pump protons although with reduced efficiency (25, 26). Finally, we note that an O-like mutant D85S was shown to pump chloride ions from the extracellular toward the cytoplasmic side (27), similar to the mode of action of halorhodopsin (28); this observation hinted that the O$\to$bR transition may not proceed through the canonical Grotthuss-like mechanism.

In this study, by combining extensive classical and QM/MM molecular dynamics (MD) simulations, we are able to gain insights into the structural features of the O state and its transition back to the bR state. In particular, we show that the hydration level of the protein cavity increases substantially in the O state to enable the long-range proton exchange between D85 and PRG. Moreover, by computing three-dimensional (3D) free energy surfaces with $\sim$0.1-$\mathrm{\mu }$s multiwalker metadynamics, we show that the proton exchange in fact involves the transfer of a proton hole (29, 30), which is stabilized by the intervening R82; the unique simulations were made possible by the use of an efficient, semiempirical QM/MM potential and a collective coordinate for describing PTs without a priori assumption of the transfer mechanism (31). The computed free energy barrier and exergonicity are consistent with experimental estimates, and the mechanism provides an explanation for the Cl⁻ pumping activity of the D85S mutant.

Results

O and O* States Feature Elevated Internal Hydration Levels Relative to the Ground State.

The appearance of an additional C=O stretch band in time-resolved Fourier transform infrared (TR-FTIR) spectra hinted at the transient protonation of D212 during the O$\to$bR reaction (13, 14), constituting an intermediate O*. Starting with the ground-state crystal structure, we modify the protonation states of D85, D212, and the PRG to construct models for the O and O* states; extensive classical MD simulations using Hamiltonian replica exchange (HREX) (SI Appendix) are then used to establish reliable structural models for the O and O* states in an explicit lipid membrane environment.

D85 and D212 are deprotonated in the ground state (Fig. 2A); to help stabilize the negative charges in the protein interior, R82 prefers an upward orientation and is aided by the Schiff base and a cluster of water molecules resolved in crystal structures (6). Moreover, the PRG features an excess proton stored via a strong hydrogen bond between E194 and E204. A small number of water molecules separate the PRG and R82, whose upward orientation prevents the formation of a continuous water wire between the PRG and D85/D212, an essential feature that helps prevent the backflow of the proton from the PRG to D85 (32).

Fig. 2.

(A) Structures of the active site for the states bR, O, and O* obtained from HREX simulations. Green spheres: proton bound to an Asp or Glu; water density is emphasized by pink shading; retinal moiety of K216 is not shown for clarity. (B) Illustration of the collective variables used to analyze HREX simulations and to run metadynamics simulations. (C) Histograms of the swing movement of the sidechain of R82, of the PRG distance, and of the number of water molecules within 10 Å of the center of the protein cavity from the HREX simulations of bR (black), O (blue), and O* (yellow) states. (The point on the line E204–D85, dividing that distance in the ratio of 1:2, is considered as the center of cavity.) Histograms of the swing movement and of the PRG distance were generated from the final 9 ns, while the histograms of water distributions originate from the final 7 ns of HREX simulations.

In the O and O* states, the PRG is deprotonated, while D85 and D212 are protonated, respectively. These changes in the protonation pattern are coupled with several major structural rearrangements compared to the ground state. First, the negatively charged E194 and E204 sidechains become separated (see Fig. 2C for distance histograms for the C$\delta$ atoms of E194 and E204); in particular, E194 rotates away to form a hydrogen-bonding interaction with Y83. Moreover, R82 rotates to a downward orientation to better stabilize the negatively charged PRG through a salt-bridge interaction with E204; the shift in the preferred R82 orientation is supported by distance histograms from unbiased MD simulations (Fig. 2C) as well as explicit umbrella-sampling simulations using the A44N–R82C$\zeta$ distance as the collective variable (SI Appendix). Interaction of the R134 sidechain with E194 was also analyzed because of the previously proposed effect of R134 on the acidity of E194 (33) (SI Appendix); no significant coupling with PT processes was found.

Separation of the E194/E204 sidechains in the O and O* states leads to an elevated hydration level in the protein cavity (Fig. 2A); as shown in Fig. 2C, the number of water molecules in the central part of the cavity increased from $\sim$9 in the ground state to 12 to 13 in the O and O* states. Moreover, the downward rotation of the R82 sidechain leaves enough space in the cavity to form a continuous water network that spans from the PRG to D85/D212, setting the stage for proton exchange between these distant regions.

Competing Proton Transfer Pathways for the O to O* Transition.

Since the O* state was captured in TR-FTIR studies (13, 14), we first analyze the O$\to$O* transition, which involves PT from D85 to D212. As discussed in SI Appendix, upon QM/MM equilibration of the O and O* states, there are minor changes in the configurations of water molecules near D85/D212 compared to those in classical MD simulations. Nevertheless, these water molecules modulate the separation of the D85/D212 sidechains and therefore potentially impact the PT mechanism and energetics. Accordingly, 3D metadynamics simulations are used to probe the transfer mechanism in detail; the collective variables are as follows (SI Appendix): $\zeta$ that describes the progress of PT, the D85C$\gamma$ to D212C$\gamma$ distance, and the average oxygen coordination number $\overline{s_{\mathrm{O}}\left(r_{ij}\right)}$ of the four QM water molecules.

The free energy surface (FES) in two-dimensional representations is plotted with $\zeta$ as one coordinate (horizontal in the plots) and either the distance between D85C$\gamma$ and D212C$\gamma$ (Fig. 3A) or the net charge of the QM waters (computed based on oxygen coordination number $\overline{s_{\mathrm{O}}\left(r_{ij}\right)}$ of the QM waters, Fig. 3B) as the other coordinate (vertical in the plots). Three representative transition-state structures are shown in Fig. 3C.

Fig. 3.

(A) Computed free energy surface for the O$\to$O* transition as a function of the D85C$\gamma$–D212C$\gamma$ distance and the PT reaction coordinate $\zeta$. (B) The same surface as a function of the net charge on the four QM water molecules $q_{\mathrm{n}\mathrm{e}\mathrm{t}}$ and $\zeta$. There are three distinct pathways, which differ in the barrier heights. (C) The geometries of the upper active site found in the transition structures marked by colored dots. Black, minimum free energy path, direct proton transfer; red, proton hole transfer (OH^–); blue, excess proton transfer (H₃O⁺); green spheres, protons bound to an Asp or Glu; orange molecules, charged water species (OH^– or H₃O⁺).

In Fig. 3A, there are two low-energy basins, one for the reactant state at $\zeta =-0.58$ and the other for the product state at $\zeta =0.58$. In terms of energetics, the reactant O is the global minimum on the FES, and the product lies $\mathrm{\Delta }{G}_{\mathrm{O}\to {\mathrm{O}}^{*}}=3.6$ kcal/mol higher. This means the proton is better stabilized on D85 than on D212, and the PT from D85 to D212 and thus the O$\to$O* transition is endergonic. The minimum free energy path connecting both minima leads through a transition state (TS), which lies at $\mathrm{\Delta }{G}_{\mathrm{O}\to {\mathrm{O}}^{*}}^{‡}=7.8$ kcal/mol. A transfer via this path is accompanied by a strong reduction of the D85C$\gamma$ to D212C$\gamma$ distance, so that the approximate TS ($\zeta =0.0$) has the aspartates 0.42 nm apart; as illustrated in Fig. 3C (black dot), this corresponds to a direct proton exchange between D85 and D212 sidechains without the explicit involvement of any water molecule.

In Fig. 3B, plotting the FES along $\zeta$ and the net charge of QM water molecules illustrates two alternative PT pathways, which involve a hydronium (Fig. 3C, blue dot) and hydroxide (Fig. 3C, red dot), respectively, in the TS; the corresponding barriers are $\sim$13.2 and 12.1 kcal/mol, respectively, and thus the direct proton exchange is the dominant mechanism. Apparently, the penalty of rearranging the hydrogen-bonding network is more than compensated for by positioning the proton donor/acceptor groups next to each other to lower the intrinsic PT barrier. These observations highlight the importance of allowing water molecules to fully equilibrate during the PT (34); they also caution against inferring the dominant PT mechanism based solely on the water structure prior to the reaction (35).

O to bR Transition Occurs through a Proton Hole Mechanism.

Since the O$\to$bR transition involves not only proton exchange between D85 and the PRG, but also reorientation of the R82 sidechain as well as separation of E194/E204 in the PRG, 3D metadynamics simulations are required to fully elucidate the underlying mechanism. The three collective variables are as follows (see SI Appendix for details): $\zeta$ that describes the progress of the proton exchange, the A44N–R82C$\zeta$ distance (dR82), and the E194C$\delta$–E204C$\delta$ distance (dPRG). The resulting 3D FES following $\sim$100 ns of QM/MM metadynamics simulations is shown in Fig. 4A. To monitor the mechanism of the proton exchange, we also plot the mean net charge of QM water molecules for snapshots taken from the metadynamics simulations in Fig. 4B; as described in SI Appendix, a net charge was obtained from the number of hydrogen atoms bonded to each oxygen. Also, care was taken to count each QM hydrogen only once to avoid artifacts associated with water molecules explicitly engaged in proton or proton hole exchange with neighboring groups. Representative snapshots are shown in Fig. 4C.

Fig. 4.

(A) The 3D free energy profile of the complete O$\to$bR transition. The relevant states are labeled. Two pathways with similar barrier heights can be identified: O $\to$ O* $\to$1 $\to$ TS_1→2 $\to$ 2 $\to$ 4 $\to$ bR and O $\to$O* $\to$ 1 $\to$ TS_{1→3 → 3} $\to$ 4 $\to$ bR. (B) Net charge of QM water molecules. White, $\overline{q_{\mathrm{n}\mathrm{e}\mathrm{t}}}=0$, neutral water; red, $\overline{q_{\mathrm{n}\mathrm{e}\mathrm{t}}}=-1$, ${\mathrm{O}\mathrm{H}}^{-}$; blue, $\overline{q_{\mathrm{n}\mathrm{e}\mathrm{t}}}=+1$, ${\mathrm{H}}_3{\mathrm{O}}^{+}$. (C) Representative structures corresponding to the states identified with metadynamics. Green, proton bound to an Asp or Glu; orange, charged water species (here, always OH^–).

The local minimum corresponding to the O state has a free energy of $\mathrm{\Delta }G$ = 3.6 kcal/mol relative to the bR ground state (product of the proton exchange reaction studied in this subsection). Its value of $\zeta =-0.82$ corresponds to one of the two oxygen atoms of D85 being protonated. With an A44 to R82 distance of 2.87 nm, R82 is swung down to stabilize the two negative charges of deprotonated E194 and E204, which maintain a long C$\delta$–C$\delta$ distance of 0.62 nm, with a water molecule in between.

The first step toward the bR ground state is the O$\to$O* transition, as described in the previous subsection. In this step, the proton is transferred from D85 to D212 in an endergonic process as the minimum corresponding to the O* state lies 2.5 kcal/mol higher in free energy than the O state. This PT event proceeds over a barrier of $\mathrm{\Delta }{G}^{‡}=5.7$ kcal/mol, either directly or via a proton hole mechanism. The barrier height obtained here is 2.1 kcal/mol lower than that found for the O$\to$O* transition discussed above using a smaller QM region. This modest discrepancy provides an evaluation of sensitivity of our QM/MM free energy simulations to differences in the simulation setup, such as the QM region size and the scaling of QM–MM electrostatics applied or not (SI Appendix).

Subsequently, the R82 sidechain moves slightly upward to an A44 to R82 distance of 2.75 nm while dPRG remains at 0.62 nm. This conformation corresponds to a flat metastable state denoted as 1 with $\zeta =-0.40$, which lies 2.8 kcal/mol higher in energy than O*, to which it is still highly similar in structure. Starting from state 1, the subsequent proton exchanges may follow one of two different pathways, which have comparable energetics according to the computed FES.

Path 1 proceeds via TS_1→2 $\to$ 2 $\to$ 4 $\to$ bR. Here, E204 is protonated by a water molecule resulting in a proton hole, which is stabilized by the positively charged guanidinium group of R82. Afterward, the E194 to E204 distance decreases, suggesting that E204 and E194 form a hydrogen bond. In addition, D85 is reprotonated by D212, and the R82 sidechain moves toward D85 gradually, accompanying the proton hole apparently. In the transition state TS_1→2, a barrier of $\mathrm{\Delta }{G}^{‡}=7.7$ kcal/mol relative to state 1 is passed, before a broad, flat metastable state 2 is reached, leading to a free energy gain of $\mathrm{\Delta }G=-2.8$ kcal/mol relative to TS_1→2. In 2, the water wire is disrupted, and the proton hole is closer to D212 and stabilized by the guanidinium group of R82. Then, a low barrier of $\mathrm{\Delta }{G}^{‡}=1.0$ kcal/mol has to be overcome to move the proton hole farther in the direction of D85, while R82 advances toward D212 simultaneously. Eventually, D85 is deprotonated by the proton hole, and the positively charged sidechain of R82 stabilizes the two negative charges of D85 and D212 at dR82 = 2.33 nm. This results in a large free energy gain of $\mathrm{\Delta }G=-10.6$ kcal/mol in state 4. In 4, E204 is protonated and hydrogen bonded to E194. Finally, after the final PT from E204 to E194 over a barrier of $\mathrm{\Delta }{G}^{‡}$ = 2.9 kcal/mol, the system reaches the global minimum, the bR ground state, which we set to $\mathrm{\Delta }G=0$. In bR, $\zeta =0.82$ with a small E194C$\delta$–E204C$\delta$ distance of 0.41 nm, corresponding to a strong hydrogen bond between E194 and E204.

Path 2 starts with the protonation of E204 from a water molecule as well, and it proceeds through states TS_1→3 $\to$ 3 $\to$ 4 $\to$ bR. D85 is not reprotonated in this pathway, and the distance E194C$\delta$–E204C$\delta$ remains large as the proton hole moves toward the protonated D212. First, the sidechain of R82 swings toward D212, being accompanied by the proton hole. The transition state TS_1→3 poses a barrier of $\mathrm{\Delta }{G}^{‡}=7.8$ kcal/mol relative to state 1. Note that another, minor protonation pattern may occur in this region of the 3D space spanned by the applied reaction coordinates: D85 and E204 protonated (but not D212) with an ${\mathrm{O}\mathrm{H}}^{-}$ proton hole located farther down, closer to the PRG; this observation is detailed in SI Appendix. Once this barrier is passed, D212 is deprotonated by the proton hole, and R82 approaches D212. This results in the local minimum 3, which lies $-8.9$ kcal/mol beneath TS_1→3. Next, the protonated E204 and the negatively charged E194 move closer together, while R82 swings closer to D85 and D212, leading to the local minimum 4 and, finally, the global minimum bR following the final PT from E204 to E194, as described above.

The current simulation setup does not cover the possibility of deprotonation of the R82 sidechain and thus participation of R82 as a PT relay. Although an arginine might deprotonate in principle, a deprotonation of R82 in bR would lead to largely increased energy barriers opposing the long-range PT; a detailed mechanistic analysis is presented in SI Appendix. Such a pathway is therefore unfavorable. Also, with the current simulation setup, E194 is the final acceptor of the excess proton and is also hydrogen bonded to E204. The protonated E194 is favored over the protonated E204, and a proton exchange between them is opposed by a modest barrier of 2.9 kcal/mol. A recent study by Tripathi et al. (36) found the excess proton centered between E204 and E194, resulting in an extremely flat and symmetric free energy minimum along the proton exchange reaction coordinate, similar to our previous QM/MM studies (37, 38). Our current simulations focus on resolving the long-range PT mechanism rather than elucidating the precise molecular features of the PRG, which requires more systematic variations of the QM region, the QM level (e.g., semiempirical vs. full density-functional theory), and the local level of hydration and explicit treatment of nuclear quantum effects.

Notwithstanding, the overall O$\to$bR transition is computed to be an exothermic process with a driving force of $-3.6$ kcal/mol. This value can be compared to the pK_a values of D85 and of the PRG estimated from experimental studies. To this end, Balashov et al. (17) calculated an equilibrium constant of $10^{-2.2}$ for proton exchange between D85 and the PRG in the ground state. Under the assumption that the measured values for the ground state sufficiently resemble the O$\to$bR transition, the estimated $\mathrm{\Delta }{G}_{\mathrm{O}\to \mathrm{b}\mathrm{R}}=-3.0$ kcal/mol. The value of $\mathrm{\Delta }{G}_{\mathrm{O}\to \mathrm{b}\mathrm{R}}=-3.6$ kcal/mol obtained from our simulations agrees well with this estimate.

The involvement of the proton hole during the O$\to$bR transition is clearly illustrated by the mean net charge of QM water molecules shown in Fig. 4B. Naturally, $\overline{q_{\mathrm{n}\mathrm{e}\mathrm{t}}}=0$ in the O state, which means that all water molecules are in their neutral form. The O$\to$O*$\to$1 transition and the O* state itself exhibit $\overline{q_{\mathrm{n}\mathrm{e}\mathrm{t}}}=0$ mostly, corresponding to a direct PT, with a trace of $\overline{q_{\mathrm{n}\mathrm{e}\mathrm{t}}}<0$ that may be attributed to a minor proton hole pathway; these features are consistent with the study of O$\to$O* transition using a smaller QM region discussed in the previous subsection, further supporting the robustness of our computational methodology. The mechanism of the subsequent 1$\to$bR transition is also clear: Both possible pathways passing through TS_1→2 and TS_1→3 proceed via a proton hole with D85/E204 or D212/E204 protonated (Fig. 4C). After the deprotonation of D85 or D212, the proton hole vanishes, and all QM water molecules are present in the charge-neutral form (in the states 3 and 4). The final transition 4$\to$bR proceeds as a direct proton exchange between E194 and E204, as corroborated by the small E204–E194 distance of ca. 0.41 nm in the relevant region of the FES.

Concluding Discussion

Despite more than 45 y of studies, the mechanistic understanding of bacteriorhodopsin remains incomplete due largely to the poor characterization of the O state and its conversion back to the ground bR state. By combining extensive classical and QM/MM simulations, we are able to fill these important voids in the photocycle of this prototypical proton pump.

Multidimensional QM/MM metadynamics enabled us to explicitly probe the mechanism and energetics of the proton exchange process that underlies the O$\to$bR transition. The energetics and kinetics from the free energy simulations are summarized schematically in Fig. 5. The rate-limiting barrier corresponds to a rate constant of $1.8\times 10^3$ ${\mathrm{s}}^{-1}$ using transition-state theory; this calculation of rate relies on a minor compensation of errors, and thus no perfect agreement with reference data can be expected. Still, the resulting time scale of $\sim$0.6 ms for the O$\to$bR transition lies within the experimentally estimated range of 0.5 to 5 ms (15–18). As noted above, the computed exergonicity of $-3.6$ kcal/mol is also consistent with the experimental estimate based on pK_a differences between D85 and the PRG. Therefore, our free energy calculations provide a mechanism that is consistent with experimental kinetic and thermodynamic data.

Fig. 5.

One-dimensional Gibbs free energy profile for the O$\to$bR transition, showing data obtained from the simulation of the complete O$\to$bR process (black curve) and those of the first step, O$\to$O* (blue curve). Free energy values in parentheses are from the simulation of O$\to$O* with a smaller QM region.

The simulations also provided microscopic insights into the transition mechanism and structural motifs that play key roles during the process. We find that the changes of protonation states of buried amino acids (D85, D212, and the PRG) are coupled with structural rearrangements of charged internal groups and water molecules. Specifically, the R82 sidechain swings between two distinct orientations to preferably stabilize the negative charges of D85/D212 and PRG in the bR and O states, respectively. The structural rearrangement of R82 together with that of the PRG also modulates the level of hydration in the protein interior, another feature essential to the long-range proton exchange between D85 and the PRG. In the bR state, the strong hydrogen bond between E194 and E204 ensures a low level of hydration between D85/D212 and PRG, and the upward orientation of R82 prevents the formation of a continuous water wire; together these features constitute a “double-gate security” that prevents the wasteful proton back flow from the PRG to D85 in the bR state. By contrast, in the O state, the level of hydration between D85/D212 and PRG is elevated due to the breakup of the strong hydrogen bond E194–E204 upon proton release; moreover, reorientation of the R82 sidechain allows the formation of a continuous water network between the PRG and D85, again highlighting the regulatory role of R82. The positive charge of R82 is also apparently essential to facilitating the generation and transfer of the proton hole from the PRG to D85/D212.

Many of these mechanistic details are not limited to bacteriorhodopsin. For example, the importance of internal hydration level change coupled with protonation events has been discussed in several proton pumping systems such as cytochrome c oxidase (39–41) and Complex I (42); thus the observations here further support the general relevance of this electrostatics–hydration coupling mechanism. The gating function of the arginine sidechain was previously also observed in voltage-sensitive ion channels by Armstrong et al. (43), who found that plug-like movement of an arginine sidechain controlled the passage of K⁺ ions through the channel.

The observation of a proton hole mechanism in bacteriorhodopsin is somewhat unexpected considering that the system has been regarded as a prototypical proton pump, rather than a hydroxide pump. While our computational methodology has the flexibility of describing the hydronium mechanism as well, the current simulations do not provide an explicit comparison of the hydroxide- and hydronium-mediated pathways; thus the latter cannot be excluded definitely and deserves to be characterized in detail in future studies. On the other hand, our observation provides a plausible explanation that the D85S mutant of bacteriorhodopsin was shown to be a Cl⁻ pump (27) and, additionally, corroborates with the similarities of bacteriorhodopsin and halorhodopsin as pointed out previously (44): Both proteins share a common pumping mechanism that relies on a conserved pattern of electrostatic interactions. Specifically, crucial acidic residues of bacteriorhodopsin are replaced by Cl⁻ binding sites in halorhodopsin. Also, the transition state observed in our current work involves an ${\mathrm{O}\mathrm{H}}^{-}$ anion and protonated D85/D212, and this configuration is analogous to halorhodopsin as well as to the chloride-pumping mutants of bacteriorhodopsin, D85X (X = S,T,N) (45, 46). All of this suggests that the system is poised to stabilize and allow the passage of a negatively charged ion such as hydroxide. Therefore, as discussed by two of us (29, 30), it is important to consider all relevant titration states for water in the analysis of PT reactions, especially in highly heterogeneous environments such as the interior of a protein. Along this line, our discussion of competing PT pathways in the O$\to$O* transition cautions against inferring the dominant PT mechanism based solely on the water structure prior to the reaction (35).

Finally, from the technical point of view, the current study highlights the value of employing a calibrated semiempirical QM method like the density-functional tight-binding to the third order (DFTB3) in QM/MM free energy simulations. The balance of accuracy and efficiency makes it feasible to construct meaningful multidimensional free energy surfaces with $\sim$100 ns of sampling with minimal a priori assumptions regarding the possible reaction mechanism (e.g., Grotthuss vs. proton hole transfer). Therefore, there continues to be pressing need to further develop effective semiempirical QM methods for increasingly complex biological applications (47, 48).

Materials and Methods

The recent structure of bR in the ground state from time-resolved serial femtosecond crystallography (PDB ID 5B6V) (6) was considered as an initial structure; the monomeric bR molecule was embedded in a 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) lipid bilayer surrounded by aqueous solution. Structural models of the protein in each of the states bR, O, and O* were prepared on the basis of HREX extended-sampling classical simulations performed with the CHARMM36 force field (49) using Gromacs (50, 51). The actual proton transfer processes were simulated by means of the hybrid QM/MM framework, using the DFTB3 method (52) with the 3OB parameterization (53) as the QM method, and combined with the multiple-walker metadynamics algorithm (54) to enhance sampling and generate free energy profiles. Two such simulations were performed to investigate the transitions O→O* and O→bR, respectively; each involved a specifically designed QM region and a suitable set of three collective variables to span the relevant configurational space. The QM/MM simulations were performed with Gromacs (55) interfaced with Plumed (56) as well as our in-house implementation of DFTB (57) and DFTB+ (58). Detailed descriptions of simulation setups for classical and QM/MM simulations, additional discussions, benchmark calculations of DFTB3/3OB for proton exchange between carboxylate groups, and additional calculations performed with alternative QM/MM setups are presented in SI Appendix.

Data Availability

All study data are included in this article and/or SI Appendix.