Origin of the unique topology of the triangular water cluster in Rubrobacter xylanophilus rhodopsin

Abstract

The crystal structure of Rubrobacter xylanophilus rhodopsin (RxR) reveals a triangular cluster of three water molecules (W413, W415, and W419) at the extracellular proton-release site, near Glu187 and Glu197. Using a quantum mechanical/molecular mechanical approach, we identified the structural nature of this unique water cluster. The triangular shape is best reproduced when all three water molecules are neutral H₂O with protonated Glu187 and deprotonated Glu197. Attempts to place H₃O⁺ at any of these water molecules result in spontaneous proton transfer to one of the acidic residues and significant distortion from the crystal structure. The plane defined by the triangular water cluster extends into the guanidinium plane of Arg71, with both aligned along the W413...W419 axis. This extended plane lies nearly perpendicular to a five-membered, ring-like H-bond network involving two carboxyl oxygen atoms from Glu187 and one from Glu197. The resulting bipartite planar architecture, defined by the water triangle, Arg71, and the Glu187/Glu197 network may reflect the exceptional thermal stability in RxR.

Significance

The crystal structure of Rubrobacter xylanophilus rhodopsin (RxR) reveals a compact, triangular water cluster at the extracellular proton-release site, distinct from the canonical architecture of the corresponding water cluster in bacteriorhodopsin. The present study indicates that this triangular geometry arises from three neutral H₂O molecules stabilized by a bipartite planar scaffold formed by the proton-sharing pair Glu187 and Glu197 and protonated Arg71, without involving H₃O⁺. This unique architecture of the water cluster may reflect the exceptional thermal stability of RxR.

Introduction

Among microbial rhodopsins, bacteriorhodopsin (BR) has long served as the prototypical system for investigating light-induced H⁺ transport along the transmembrane. In BR, photoisomerization of retinal triggers a cascade of proton transfer steps via a well-characterized network of titratable amino acids and internal water molecules, resulting in proton release to the extracellular side. The release of the protons toward the bulk surface is facilitated by acidic residues, Glu194 and Glu204, collectively referred to as the “proton release group [1–3]”.

The involvement of waters in the proton-release group of BR has been intensively debated. Bashford and Gerwert were among the first to apply electrostatic calculations to propose a functional role for internal water molecules in the proton-release pathway. In their study, water molecules were manually positioned (“by hand”) based on Fourier transform infrared (FTIR) spectroscopy data, and their effects on the protonation states of surrounding residues were evaluated by solving the linear Poisson–Boltzmann equation [4]. This approach led to the proposal that internal waters are essential for the function of the extracellular proton-release group. Building on this framework, it was later proposed that a stable H-bond network is formed between Glu194, Glu204, and a central water molecule, enabling concerted proton release [5]. While it is true that a stable H-bond network generally facilitates proton transfer more efficiently than a transient, flexible network—since the formation of the H-bond network itself is an activation process [6], this model for BR was later challenged. In particular, Goyal et al. performed quantum mechanical/molecular mechanical (QM/MM) calculations and demonstrated that the water molecules in the Glu194/Glu204 region are dynamically exchanging and not stably positioned, thereby questioning the core assumption of the fixed water model [7]. Their results suggest that a dynamic ensemble of water configurations, rather than a single static structure, governs the proton-release process in BR.

In parallel with this debate on the structural flexibility of the proton-conducting pathway, a mechanistic hypothesis emerged regarding the protonation state of the proton-release group. Garczarek and Gerwert proposed that, during this process, a protonated water species, H₃O⁺ or H₅O₂⁺, forms at a water molecule (W403) situated near Glu194 and Glu204, based on the presence of a broad infrared continuum band in time-resolved FTIR [5]. However, this interpretation was later challenged by several studies. For example, Goyal et al. demonstrated that the same continuum band (1800–2000 cm^–1) can be reproduced by a delocalized proton shared between two glutamates, without invoking H₃O⁺ or H₅O₂⁺ [7].

A key requirement for stabilizing H₃O⁺ in a protein environment is the presence of three H-bond acceptors with nearly identical pK_a values. If one of the acceptors had a significantly higher pK_a, the proton would spontaneously relocate to that site, resulting in a neutral water molecule and a protonated acceptor group [8,9]. In BR, this criterion is not fulfilled. W403 is relatively solvent-exposed, and the H-bonding network lacks both geometric symmetry and electrostatic balance, rendering the formation of a stable H₃O⁺ species energetically unfavorable.

Remarkably, recent structural and functional analyses have revealed microbial rhodopsins with distinct architectures and potentially divergent transport mechanisms. One such example is Rubrobacter xylanophilus rhodopsin (RxR), a phylogenetically distinct and thermally stable proton pump that occupies an evolutionary position between archaeal and eukaryotic rhodopsins. The crystal structure of RxR, resolved at 1.8 Å resolution, reveals an extracellular proton-release site containing two acidic residues, Glu187 and Glu197, along with a cluster of internal water molecules [10] (Figure 1).

Figure 1.

Extracellular proton-release site of RxR identified in the crystal structure (PDB ID: 6KFQ). Dotted lines indicate H-bonds. Red balls represent water molecules forming the triangular water cluster (W413, W415, and W419). Red dotted lines indicate O...O distances between these water molecules. Numerical labels indicate distances (Å).

The extracellular proton-release site in RxR exhibits several distinctive features. Kojima et al. demonstrated that Glu187 and Glu197 function cooperatively as the proton-release group, analogous to the Glu194/Glu204 pair in BR, as mutational analyses revealed that substitution of either Glu187 or Glu197 perturbs the proton release process [11]. Structurally, Glu187 and Glu197 are embedded in a compact, solvent-shielded cavity, bridged by a set of three crystallographically well-resolved water molecules. These water molecules are situated in a predominantly hydrophobic environment, surrounded by residues such as Phe67, Tyr68, Tyr72, and Phe201, further isolating it from bulk solvent and reducing conformational flexibility. This configuration stands in marked contrast to the more solvent-exposed and geometrically diffuse Glu194/Glu204 site in BR.

Notably, the three water molecules in RxR, W413, W415, and W419, form a nearly equilateral triangle, with internal angles of 52°, 59°, and 68° (Figure 1). This triangular geometry is rare among water clusters observed in proteins, which more often form larger, open-ring structures such as pentamers (e.g., the five-membered water cluster observed between two redox-active quinones in photosystem I [12]). For reference, the H–O–H bond angle in an isolated water molecule is 104.5°, which closely matches the internal angle of a regular pentagon (108°), consistent with the frequent observation of five-membered water rings in proteins. This angle is also close to 109.5°, the tetrahedral O...H–O angle formed between two H-bonded water molecules in ice. In contrast, the ~60° internal angles observed in the triangular water cluster of RxR deviate significantly from this optimal H-bond geometry. Such a compact triangular arrangement is incompatible with three simultaneous linear O...H–O alignments, and from a traditional standpoint, may not be explained as accommodating three such linear H-bonds simultaneously.

Here, using a QM/MM approach, we examine the energetics and H-bond topology of the triangular water cluster near the Glu187/Glu197 pair.

Materials and methods

Initial geometry

The atomic coordinates were taken from the X-ray crystal structure of RxR (PDB ID: 6KFQ [10]). During the optimization process for H atom positions using CHARMM [13], all heavy atom positions were kept fixed, and all titratable groups (e.g., acidic and basic groups) were ionized and the retinal Schiff base was protonated. Atomic partial charges for amino acids were obtained from the CHARMM22 parameter set [14]. The topology, parameter, and atomic partial charges for the protonated retinal Schiff base, were adopted from the CHARMM topology/parameter file, toppar_all22_prot_retinol.str, which is equivalent to the updated version, toppar_all36_prot_retinol.str.

Protonation pattern and pK_a

The protonation pattern of the titratable residues was calculated by solving the linear Poisson-Boltzmann equation using the MEAD program [15]. All computations were performed at 300 K, pH 7.0, and an ionic strength of 100 mM.

For pK_a calculations, crystal water molecules were removed, and bulk water was implicitly modeled using a dielectric constant of 80. This treatment is particularly appropriate for titration because the number and quality of crystal water molecules depend strongly on the quality of each crystal structure (e.g. resolution). More importantly, retaining crystal water molecules during titration would artificially constrain their H-atom orientations based on the initial geometry, thereby fixing the H-bond pattern and, consequently, the protonation states of nearby titratable residues. As a result, the system could be biased toward a specific, potentially non-representative protonation state, simply due to the initial (tentative) geometry. In contrast, replacing these crystal water molecules with an implicit solvent model (dielectric constant of 80) provides a spherical solvation environment that statistically represents all possible H-atom orientations of water molecules, allowing for a more flexible and physically realistic determination of protonation patterns and pK_a values. For this reason, unless the H-bond geometry of individual water molecules can be clearly assigned and reliably modeled (e.g., in the case of metal-ligated water molecules), it is standard practice to remove explicit crystal waters for pK_a calculations.

To determine the pK_a values of titratable sites in the protein, the calculated pK_a shift (relative to a reference system) was added to the known reference pK_a value. The experimentally measured pK_a values used as references were 12.0 for Arg, 4.0 for Asp, 9.5 for Cys, 4.4 for Glu, 10.4 for Lys, 9.6 for Tyr [16], and 7.0 and 6.6 for the N_ε and N_δ sites of His, respectively [17–19]. The linear Poisson-Boltzmann equation was solved using a three-step grid-focusing procedure at resolutions of 2.5 Å, 1.0 Å, and 0.3 Å. Protonation patterns were sampled using a Monte Carlo method implemented in Karlsberg [20]. The Monte Carlo sampling provided the probabilities ([protonated] and [deprotonated]) for each titratable site. The pK_a value was evaluated using the Henderson-Hasselbalch equation, by applying a bias potential to equalize the protonated and deprotonated populations ([protonated]=[deprotonated]); the resulting bias potential corresponds to the pK_a value. During the titration of each focusing residue, all other titratable sites were fully equilibrated with respect to the protonation state of the focusing site. The resulting protonation state for each titratable residue was represented explicitly by the presence or absence of H⁺ for subsequent QM/MM calculations.

The electrostatic contribution of each residue to the pK_a value of Arg71 was evaluated by setting the atomic charges of that residue to zero and recalculating the pK_a value; the resulting pK_a shift was defined as the electrostatic contribution of the residue.

QM/MM calculations

Geometry optimization was performed using a QM/MM approach. Restricted density functional theory (DFT) method was employed with the B3LYP functional and LACVP* basis sets, using the QSite [21] program. The QM region included Gly185, Thr186, and Glu187; the side-chains of Arg71, Tyr72, and Glu197; backbone atoms of Glu66, Phe67, Tyr68, and Trp69; and water molecules W413, 415, and 419, which together form the H-bond network surrounding the triangular water cluster. All other protein components and water molecules were treated in the MM region. Atomic coordinates were fully relaxed in the QM region. In the MM region, H atom positions were optimized using the OPLS2005 force field [22], while heavy atom positions were fixed. See the Supporting Information for the atomic coordinates of the QM/MM-optimized structures (Supplementary Code S1).

Although the OPLS2005 force field was used in QSite for the MM region, the atomic charges applied to the MM region were based on those derived from MEAD-based calculations, namely, CHARMM charges for amino acids and restrained electrostatic potential (RESP) charges [23] for cofactors. Ideally, it would be preferable to use the CHARMM force field in QSite as well, but this is not currently supported: QSite mandates the use of the OPLS force field for MM atoms. Likewise, the MEAD framework has been empirically developed and validated using CHARMM charges for both the protonated and deprotonated states of titratable sites. Since QSite treats QM-MM interactions through electrostatic and van der Waals terms, this hybrid scheme may formally affect the electrostatic interaction terms. Nonetheless, previous studies have shown that this approach provides practically sufficient accuracy in describing electrostatic interactions. For instance, strong correlations have been observed between redox potentials calculated using MEAD (e.g., [24]) and HOMO energies calculated using QSite (e.g., [25]), even under such hybrid conditions. Therefore, despite the lack of full consistency between force fields, this approach remains effective when compatibility with MEAD-based calculations is prioritized. In the present study, this hybrid scheme was adopted to leverage the complementary advantages of MEAD and QSite.

Results and Discussion

The triangular water cluster is located near a pair of acidic residues, Glu187 and Glu197. In the crystal structure of RxR, the H-bond pattern of the three water molecules in this cluster is effectively constrained to a single, unambiguous configuration when all three are assumed to be H₂O: W413 donates H-bonds to Trp69 (backbone carbonyl oxygen) and Glu187; W415 donates to Glu66 (backbone carbonyl oxygen) and W419; and W419 donates to W413 and Glu197 (Supplementary Figure S1). This limited H-bond pattern arises primarily because residues with flexible side-chain H-bond orientations (e.g., serine, threonine, and tyrosine) are not involved in the network. Therefore, alternative hydrogen-bond patterns can be predominantly excluded from the subsequent analysis.

The short distance of 2.6 Å between the carboxyl oxygen atoms of these residues, along with the relationship pK_a(Glu187)>pK_a(Glu197) [11], suggest that protonated Glu187 donates an H-bond to deprotonated Glu197. Consistently the calculated pK_a values are 8.2 for Glu187 and 5.8 for Glu197 (Table 1). Under this protonation state, the QM/MM-optimized geometry best reproduces the characteristic triangular shape of the water cluster formed by W413, W415, and W419 when assuming H₂O at all three sites (Table 2, Figure 2a, and Supplementary Figure S2a). The shared proton between Glu187 and Glu197 is predominantly localized on Glu187 rather than Glu197, consistent with pK_a(Glu187)>pK_a(Glu197) reported in previous studies [11].

Table 1.

Calculated pK_a values for titratable residues in RxR

RxR	pKa	(BR)
Asp85	11.8	Asp96
Asp108	9.7	Asp115
Asp74	3.1	Asp85
Asp205	0.0	Asp212
Glu187	8.2	Glu194
Glu197	5.8	Glu204

Corresponding sites in BR are shown for clarity.

Table 2.

Structural comparison of H-bond distance, angle, and total energy for QM/MM-optimized structures relative to the crystal structure

H-bonds	6KFQ	QM/MM (assumed protonation state)
		H2O	(W413: H3O+) → H2O Glu187-H+	(W415: H3O+) → H2O Glu187-H+	(W419: H3O+) → H2O Glu197-H+
Distance
W413...W415	3.17	3.25	3.25	3.31	3.39
W415...W419	2.93	2.89	2.72	2.85	2.62
W419...W413	2.70	2.75	2.66	2.78	2.67
R 2		0.95	0.83	0.86	0.71
Angle
W413...W415...W419	52	53	52	53	51
W415...W419...W413	68	70	74	72	80
W419...W413...W415	59	57	54	55	50
R2		0.95	0.87	0.88	0.79
RMSD in structures		0.24	0.31	0.33	0.38
Total energy		0.0	6.2	7.3	1.2

Distances are in Å, angles in degree (°) and energies in kcal/mol. The original crystal structure corresponds to PDB ID: 6KFQ. The assumed protonation states in the QM/MM initial structures are indicated in parentheses. R²: coefficient of determination; RMSD: root mean square deviation for the entire region; Glu187-H⁺: protonated Glu187; Glu197-H⁺: protonated Glu197.

Figure 2.

Resulting QM/MM-optimized structures of the extracellular proton-release site in RxR. (a) Structure obtained assuming H₂O at all three water molecules in the initial structure. (b–d) Structure obtained assuming H₃O⁺ at W413, W415, and W419, respectively, in the initial structure. Blue arrows indicate the direction of proton transfer along H-bonds. Black balls represent protons in the local H-bond network. Cyan balls represent transferring protons. Dotted lines indicate H-bonds.

In contrast, attempts to model H₃O⁺ formation at any of the three water molecules result in proton transfer to one of the two acidic residues, ultimately yielding H₂O and a protonated glutamate (Figure 2b-d and Supplementary Figure S2b-d). The resulting QM/MM-optimized structures also deviate significantly from the original crystal structure, suggesting that H₃O⁺ is energetically unstable in this environment (Table 2).

A key factor contributing to destabilization of H₃O⁺ is the presence of Arg71. In the present study, the pK_a value of Arg71, calculated by solving the linear Poisson-Boltzmann equation, is 12.8. This high pK_a value reflects substantial stabilization of the protonated form of Arg71 by nearby acidic residues, including Asp205 (+6.6), Glu197 (+4.5), and Asp74 (+3.6) (Table 3). Because protonated Arg71 donates H-bonds to both W413 and W419, it competes with protonated water species at these sites, thereby disfavoring H₃O⁺ formation. Taken together, these results suggest that the triangular water cluster identified in the 1.8-Å crystal structure consists of three H₂O molecules, with Glu187 protonated and Glu197 deprotonated.

Table 3.

Contribution of selected residues to the shift in pK_a(Arg71)

Groups	Contribution to pK a(Arg71)	Distances
Asp205	6.6	4.3
Glu197	4.5	5.4
Asp74	3.6	7.0
Glu187	0.4	6.9
retinal Schiff base	–2.8	8.3

Shifts are given in pK_a units, along with their distances from Arg71 (Å). Note that Glu187 is protonated.

Unlike pentagonal water clusters, where all five water molecules form H-bonds to their adjacent neighbors to generate a fully connected ring stabilized by lateral, side-to-side interactions [12], the present triangular water cluster lacks one such connection. Specifically, the H-bond between W413 and W415 is absent (Figure 2a and Supplementary Figure S2a), which explains why the W413...W415 distance is longer than the other two sides of the triangle (Table 2). Instead of forming an H-bond along this side, W413 orients toward the backbone carbonyl oxygen of Tyr68, opening the W413...W415 side of the triangle. This opening allows the cluster to use W415 as an anchor point, forming an H-bond with the backbone carbonyl oxygen of Glu66, which likely contributes to reduced positional disorder of the cluster (Figure 1).

In addition to their roles in the triangular water cluster, W413 and W419 also accept H-bonds from the guanidinium nitrogen atoms of Arg71, placing the three oxygen atoms of the triangular water cluster and the three nitrogen atoms of Arg71 nearly in the same plane, aligned along the W413...W419 axis (Figure 3a).

Figure 3.

Molecular planes formed by clusters of polar groups at the extracellular proton-release site of RxR. (a) Triangular water cluster (W413, W415, W419) and the guanidinium group of Arg71, forming a coplanar arrangement. The pink line indicates the W413...W419 axis. (b) H-bond network forming a five-membered ring-like plane composed of W413, W419, Glu187, and Glu197. (c) Schematic illustration of the two nearly perpendicular planes. The blue plane includes the triangular water cluster and Arg71, while the red boxed plane includes the five-membered ring involving Glu187 and Glu197. Dotted lines indicate H-bonds.

Furthermore, W413 and W415 participate in an additional H-bond network with two carboxyl oxygen atoms from Glu187 and one from Glu197 (Figure 3b). These five oxygen atoms form a five-membered, ring-like H-bond network whose plane lies approximately perpendicular to that of the triangular water cluster and the Arg71 guanidinium group, again sharing the W413...W419 axis. As a result, the [W413, W415, W419] plane and the [Arg71 guanidinium] plane are approximately coplanar, while the [W413, W415, Glu187, and Glu197] plane is nearly perpendicular to both, forming a bipartite planar arrangement (Figure 3c).

This tightly organized architecture, the bipartite planar arrangement centered on a common W413...W419 axis, may contribute to the extremely high thermal stability of RxR [11,26]. In particular, the triangular water cluster is positioned at the electrostatic interface between the basic guanidinium group of Arg71 and the acidic carboxyl groups of Glu187 and Glu197. Thus, the W413...W419 axis serves as a salt-bridge like scaffold, which likely explains why the three water molecules exhibit relatively low disorder, as reflected in their moderate B-factors (~35), despite being located in a hydrophobic environment surrounded by residues such as Phe67, Tyr68, Tyr72, and Phe201.

While protein dynamics and membrane environments may influence H-bond geometries to some extent, the 1.8 Šcrystal structure provides a reasonable basis to consider that the observed configuration represents a well-defined local energy minimum stabilized under crystallization conditions. In fact, the B-factors of the three water molecules (~35 Ų) are comparable to those of neighboring residues, supporting the view that the observed configuration is not highly dynamic, but rather represents a stable structural arrangement. If molecular dynamics simulations yield configurations that deviate substantially from the crystallographic structure, such deviations may instead reflect conformations that are less relevant to the physiological state.

In addition, the water cluster analyzed here is located in a compact, solvent-shielded cavity surrounded by hydrophobic residues such as Phe67, Tyr68, Tyr72, and Phe201, limiting its conformational flexibility and shielding it from bulk solvent. This environment contrasts with the Glu194/Glu204 region in BR, which is more solvent-exposed. Importantly, even the crystallographically resolved molecules resembling membrane lipids (e.g., the propanoate molecule, MPG) are not in van der Waals contact with the water cluster [10], suggesting that electrostatic or steric perturbation from membrane lipids is minimal. Taken together, the present study provides a justified and reliable starting point for mechanistic insights into the role of internal water molecules in proton release in RxR.

Conclusion

The unique shape of the triangular water cluster formed by W413, W415, and W419 at the extracellular proton-release site of RxR can be explained by a configuration of three neutral H₂O molecules in the presence of protonated Glu187 and deprotonated Glu197, without invoking protonated water species such as H₃O⁺. The absence of an H-bond along the W413...W415 edge gives rise to the unique triangular geometry, characterized by O...O...O angles ranging from 50° to 70° (Table 2).

The plane defined by the triangular water cluster extends into the plane of the Arg71 guanidinium group, with both aligned along the W413...W419 axis. Furthermore, this shared axis also defines a second, approximately perpendicular plane: a five-membered, ring-like H-bond network composed of W413, W415, and the carboxyl oxygen atoms of Glu187 and Glu197. These three orthogonally arranged elements, i) the triangular water cluster, (ii) Arg71, and (iii) the Glu187/Glu197 network, constitute a bipartite planar architecture centered on the W413...W419 axis.

Importantly, this axis lies at the electrostatic interface between a basic residue (Arg71) and acidic residues (Glu187 and Glu197), forming a salt-bridge-like scaffold that stabilizes the water cluster and reduces its positional disorder. This unique configuration is likely a local structural feature consistent with the exceptional thermal stability of RxR [11,26], and may also be associated with a proton-release timing mechanism specific to this rhodopsin [11]. These findings provide new structural insights into the diversity of proton-release site architectures in microbial rhodopsins.

Conflict of interest

The authors declare that they have no conflicts of interest with the contents of this article.

Author contributions

T.N., M.T., K.S., K.K., Y.S., and H.I. performed research; T.N., M.T., and H.I. analyzed data; and H.I. wrote the paper.

Data availability

All data is included in the manuscript and/or supporting information.