Factors That Differentiate the H-bond Strengths of Water Near the Schiff Bases in Bacteriorhodopsin and Anabaena Sensory Rhodopsin

Background: Bacteriorhodopsin (BR) functions as a proton pump, whereas Anabaena sensory rhodopsin (ASR) functions as a photosensor.

Results: pK_a of the conserved Asp residues near the Schiff base significantly differ between BR and ASR.

Conclusion: The O-D stretching frequencies for D₂O are correlated with pK_a(Asp).

Significance: The presence of a strongly H-bonded water results from the proton-pumping activity in BR.

Abstract

Bacteriorhodopsin (BR) functions as a light-driven proton pump, whereas Anabaena sensory rhodopsin (ASR) is believed to function as a photosensor despite the high similarity in their protein sequences. In Fourier transform infrared (FTIR) spectroscopic studies, the lowest O-D stretch for D₂O was observed at ∼2200 cm⁻¹ in BR but was significantly higher in ASR (>2500 cm⁻¹), which was previously attributed to a water molecule near the Schiff base (W402) that is H-bonded to Asp-85 in BR and Asp-75 in ASR. We investigated the factors that differentiate the lowest O-D stretches of W402 in BR and ASR. Quantum mechanical/molecular mechanical calculations reproduced the H-bond geometries of the crystal structures, and the calculated O-D stretching frequencies were corroborated by the FTIR band assignments. The potential energy profiles indicate that the smaller O-D stretching frequency in BR originates from the significantly higher pK_a(Asp-85) in BR relative to the pK_a(Asp-75) in ASR, which were calculated to be 1.5 and −5.1, respectively. The difference is mostly due to the influences of Ala-53, Arg-82, Glu-194–Glu-204, and Asp-212 on pK_a(Asp-85) in BR and the corresponding residues Ser-47, Arg-72, Ser-188-Asp-198, and Pro-206 on pK_a(Asp-75) in ASR. Because these residues participate in proton transfer pathways in BR but not in ASR, the presence of a strongly H-bonded water molecule near the Schiff base ultimately results from the proton-pumping activity in BR.

Introduction

Bacteriorhodopsin (BR)² functions as a light-driven proton pump, and the driving force for its action is provided by photoisomerization of the all-trans retinal chromophore, which is covalently attached to Lys-216 via the Schiff base, to 13-cis. This leads to proton transfer pathways that proceed toward the extracellular side via Asp-96, the Schiff base, Asp-85, and Glu-204 (1–3). Asp-85 is located near the Schiff base and serves as the counterion. The presence of water molecules in this region had been suggested by Fourier transform infrared (FTIR) studies (4, 5), and these were later confirmed as W401, W402, and W406 in the high resolution crystal structure of the BR ground state (PDB code 1C3W; see Fig. 1) (6). The three water molecules form an H-bond network with Asp-85, Arg-82, Asp-212, and the Schiff base. Among the three water molecules, the chemical properties of W402 are of particular interest because the water molecule is located at H-bond distances of 2.63 and 2.87 Å from the carboxyl O atom of Asp-85 and the Schiff base N atom, respectively (see Table 1). From FTIR analysis of various mutants in D₂O, the lowest O-D stretching frequency of 2171 cm⁻¹ in BR was assigned to the O_W402–H…O_Asp-85 bond, implying that this is the strongest H-bond in the network near the Schiff base (7). The band assignment presented in FTIR studies was also supported by previous computational studies by Hayashi et al. (8, 9).

FIGURE 1.

Overview of the BR (PDB code 1C3W) and ASR (PDB code 1XIO) structures. Red spheres indicate O atoms of water molecules. Yellow arrows indicate proton transfer pathways.

TABLE 1.

Comparison of the H-bond geometries near W402 in the BR crystal structures (PDB ID codes 1C3W at 1.55 Å resolution and 2NTU at 1.53 Å resolution) in the QM/MM geometry by Hayashi and Ohmine (8) and in this study (models 1 and 2, Fig. 4)

Distances are in Å, and angles are in degrees. ND, not determined. An underlined number indicates a significant deviation from the distance in the crystal structure. See supplemental S1 for the atomic coordinates of the QM/MM geometries.

Donor…acceptor	1C3W	1C3W; QM/MM Hayashi and Ohminea (model 1)	1C3W; QM/MM (model 1)	1C3W; QM/MM (model 2)	2NTU	2NTU; QM/MM (model 1)	2NTU; QM/MM (model 2)
	Å	Å	Å	Å	Å	Å	Å
W402…Asp-85	2.63	2.58	2.61	2.57	2.52	2.63	2.59
W402…Asp-212	2.85	4.07	2.92	2.99	3.01	2.90	2.95
W401…Asp-85	2.59	2.63	2.51	2.64	2.69	2.59	2.68
(OH without acceptor)b		(W401)	(W401)	(W406)		(W401)	(W406)
W406…W401	2.75	2.77	2.76	3.06	2.69	2.78	3.07
W406…Asp-212	2.75	2.78	2.64	2.64	2.81	2.73	2.65
Lys-216…W402	2.87	2.63	2.78	2.76	2.94	2.80	2.77
Arg82…W406	2.49	2.72	2.74	2.87	2.87	2.73	2.86
Root mean square deviation	ND	ND	0.19	0.18	ND	0.17	0.18
Angle (NLys-216…OW402…OAsp-85) (degrees)	106	ND	108	108	111	111	111

^a See Ref. 8.

^b A water molecule in which one of the two O-H bonds does not possess the H-bond acceptor molecule; W401 in model 1 and W402 in model 2 (Fig. 4).

Water molecule W402 was also found in the crystal structure of Anabaena sensory rhodopsin (ASR) (PDB code 1XIO; see Fig. 1) (10). In contrast to the proton pumping of BR, ASR is believed to function as a photosensor. Despite the high similarity of their protein sequences, some key residues that are functionally important in the proton-pumping event in BR are not conserved in ASR; Asp-96 in BR, which serves as a proton donor to the Schiff base, is replaced with Ser-86 in ASR, and Glu-194 and Glu-204 in BR, which is located in the terminal region of the proton transfer pathway, are replaced with Ser-188 and Asp-198, respectively. Although W402 is present in both BR (6) and ASR (10), Asp-212 in BR is replaced with Pro-206 in ASR (Fig. 1). The absence of W401 and W406 in the corresponding H-bond network of ASR may be associated with the absence of an acidic residue corresponding to Asp-212 in BR. FTIR studies have suggested that the O-D stretch in water molecules was only observed at >2500 cm⁻¹ in ASR (11), which is ∼300 cm⁻¹ higher than the lowest O-D stretching frequency of 2171 cm⁻¹ in BR (7).

The significant difference in the lowest O-D stretching frequency implies that the H-bond properties of W402 are significantly different in BR and ASR. It has been established by FTIR that the lowest O-D stretching frequency of water can be found at less than 2400 cm⁻¹ in a number of proton-pumping rhodopsins (12). The same tendency also holds true for BR and ASR.

In the BR and ASR crystal structures, the angle defined by the Schiff base N, W402 O, and Asp-85 O atoms (N_Lys…O_W402…O_Asp) differs significantly; the angle is 106° in BR (6) and 83° in ASR (10). Thus, it was proposed that the angle difference may differentiate the H-bond strength of the water molecules in BR and ASR (11). On the other hand, it also appears that the O_W402–H…O_Asp angle is more crucial than the N_Lys…O_W402…O_Asp angle to the energetics of the H-bond as it involves a H atom. In general, the O_donor–H…O_acceptor angle strongly depends on the O_donor–O_acceptor distance, i.e. the O_donor–H…O_acceptor angle is more linear when the O_donor–O_acceptor bond is shorter (13).

In addition, the N_Lys…O_W402…O_Asp angle in the crystal structure can be directly altered in response to changes in the O_W402–H…O_Asp distance. Because of the nature of H-bonds (14–17), the O_W402–H…O_Asp-85 distance could also be predominantly determined by the pK_a difference between the H-bond donor (W402) and acceptor (Asp) moieties. In general, a smaller pK_a difference yields a short, symmetrical H-bond (14–17), and the H atom of O_W402–H migrates toward the acceptor moiety, Asp. Unfortunately, the energetics of the O_W402–H…O_Asp bond, particularly the pK_a difference between W402 and the Asp residue, remain unclear. Thus, it is essential to clarify how the O_W402–H…O_Asp distance is energetically determined in the BR and ASR crystal structures.

Here we present calculated O-D stretching frequencies of D₂O near W402 in BR and ASR in the ground state using a large scale quantum mechanical/molecular mechanical (QM/MM) approach. We also present the potential energy profiles of the H-bond between W402 and Asp-85 in BR and between W402 and Asp-75 in ASR and demonstrate that the frequencies and H-bond properties are ultimately determined predominantly by the pK_a values of H-bond acceptors Asp-85 and Asp-75. By calculating pK_a(Asp-85) in BR and pK_a(Asp-75) in ASR through solving the linear Poisson-Boltzmann equation with explicit consideration of the protonation states for all titratable residues, we are finally able to pinpoint the factors that cause a difference of ∼300 cm⁻¹ between the O-D stretching frequencies of W402 in BR and ASR.

EXPERIMENTAL PROCEDURES

As demonstrated in our previous work with photoactive yellow protein (18), we employed the following systematic modeling procedure.

First, we constructed initial molecular models of BR and ASR using their crystal structures and adding hydrogen atoms. Second, to gain better understanding of the electronic structure of W402 and the associated H-bond network, we performed large scale QM/MM calculations for the entire BR and ASR proteins. To gain insight into the QM/MM potential energy profiles of the H-bonds, O_W402–H…O_Asp-85 in BR and O_W402–H…O_Asp-75 in ASR, we calculated pK_a(Asp-85) in BR and pK_a(Asp-75) in ASR by simultaneously titrating all titratable residues in BR and ASR. Technical details of each modeling procedure are summarized below.

Atomic Coordinates and Charges

As a basis for the computations, the crystal structures of BR (PDB codes 1C3W (6) and 2NTU (19)) and ASR (PDB code 1XIO (10)) were used. To generate the initial geometries, the positions of the H atoms were energetically optimized with CHARMM (20) using the CHARMM22 force field. During this procedure, the positions of all non-H atoms were fixed, and the standard charge states of all the titratable groups were maintained (i.e. basic and acidic groups were considered to be protonated and deprotonated, respectively). The Schiff base was considered to be protonated. Atomic partial charges of the amino acids and the Schiff base were adopted from the all-atom CHARMM22 (20) parameter set.

Protonation Pattern and pK_a

The present computation is based on the electrostatic continuum model created by solving the linear Poisson-Boltzmann equation with the MEAD program (21). To facilitate a direct comparison with previous computational results (e.g. see Refs. 22 and 23), identical computational conditions and parameters, such as atomic partial charges and dielectric constants, were used. To obtain absolute pK_a values of target sites (e.g. pK_a(Asp-85) of BR), we calculated the difference in electrostatic energy between the two protonation states, protonated and deprotonated, in a reference model system using a known experimentally measured pK_a value (e.g. 4.0 for Asp (24)). The difference in the pK_a value of the protein relative to the reference system was added to the known reference pK_a value. The experimentally measured pK_a values employed as references were 7.2 for the Schiff base (25, 26), 12.0 for Arg, 4.0 for Asp, 9.5 for Cys, 4.4 for Glu, 10.4 for Lys, 9.6 for Tyr (24), and 7.0 and 6.6 for the Nϵ and Nδ atoms of His, respectively (27–29). All other titratable sites were fully equilibrated to the protonation state of the target site during the titration. The ensemble of the protonation patterns was sampled by a Monte Carlo method with Karlsberg (30). The dielectric constants were set to ϵ_p = 4 inside the protein and ϵ_w = 80 for water. All computations were performed at 300 K, pH 7.0, and an ionic strength of 100 mm. The linear Poisson-Boltzmann equation was solved using a three-step grid-focusing procedure at resolutions of 2.5, 1.0, and 0.3 Å. The Monte Carlo sampling yielded the probabilities (protonated and deprotonated) of the two protonation states of the molecule. The pK_a value was evaluated using the Henderson-Hasselbalch equation. A bias potential was applied to obtain an equal amount of both protonation states (protonated = deprotonated), yielding the pK_a value as the resulting bias potential.

QM/MM Calculations

We employed an electrostatic embedding QM/MM scheme and used the Qsite (Version 5.6, Schrödinger, LLC, New York) program code as performed in previous studies (32). We employed the restricted density functional theory method with B3LYP functional and LACVP**+ basis sets. The geometries were refined using a constrained QM/MM optimization whereby the coordinates of the heavy atoms in the surrounding MM region were fixed to the original x-ray coordinates, whereas those of the H atoms in the MM region were optimized with the OPLS-2005 force field. To investigate the energetics of the entire H-bond network, the QM region was defined as follows (Fig. 2): (a) W402 and the residues/groups in the same H-bond network, i.e. W401, W406, Tyr-57, Arg-82, Asp-85, Ser-85, Thr-89, Tyr-185, Asp-212, Lys-216 (the Schiff base), and retinal for BR and (b) W402 and the residues/groups in the same H-bond network, i.e. Tyr-11, Tyr-51, Asp-75, Trp-76, Thr-79, Ser-47, the backbone of Phe-202, Lys-210 (the Schiff base), and retinal for ASR. Other residues that are not in the H-bond network of W402 (e.g. Trp-86 in BR is not involved in the same H-bond network even though the residue corresponds to Trp-76 in ASR) were approximated by the MM force field. The resulting QM/MM optimized geometries are listed in supplemental Table S1 for BR and Table S2 for ASR. The potential energy profile of the H-bond was obtained in the following manner. First, we optimized the geometry without constraints using QM/MM and used the resulting geometry as the initial geometry for the subsequent steps. Next, we moved the H atom from the H-bond donor atom (O_donor) to the acceptor atom (O_acceptor) by 0.05 Å, optimized the geometry by constraining the O_donor–H and O_acceptor…H distances, and then calculated the energy of the resulting geometry. This procedure was repeated until the H atom reached the O_acceptor atom. After obtaining the stable geometry of the QM fragment, we calculated the O-D stretching frequencies of W401, W402, and W406 in BR and of W402 in ASR. The frequency calculations were performed using a numerical differentiation method at the same level as the QM/MM geometry optimization. The calculated frequencies were scaled using a standard factor of 0.9614 for B3LYP (33).

FIGURE 2.

Residues and groups included in the QM regions for BR (a) and ASR (b). Dotted lines indicate hydrogen bonds. c, geometry of moieties near Lys-210 in ASR before (magenta) and after (yellow) the QM/MM optimization is shown.

Calculation of the O-D Stretching Frequencies of D₂O from an Empirical Equation

The O-D stretching resulting of D₂O were also evaluated on the basis of the resultant QM/MM optimized geometries by using the correlation of stretching frequency with respect to bond distance of the O_donor–D…O_acceptor bond, as previously proposed by Mikenda (34). This correlation was empirically expressed with the O-D stretching frequency, ν_O–D, and the O_donor–O_acceptor bond distance, r_O–O, by

where A = 2.11 × 10⁶, and B = 3.23 for the O_water–D_water…O_acceptor bond (34). Note that these parameters were derived from spectroscopic and x-ray diffraction data for 61 different solid deuterates, e.g. CaSO₄·2D₂O and MnCl₂·6D₂O (34).

To predict the frequencies, Equation 1 requires only the O_donor–O_acceptor distance, which can be preferentially taken from the QM/MM optimized geometry. Remarkably, the O-D stretching frequencies calculated by a QM/MM numerical differentiation method were reasonably well reproduced using Equation 1 solely from the optimized O_donor–O_acceptor distance, particularly for frequencies <2500 cm⁻¹ (see Table 3).

TABLE 3.

Model 2, O–D stretching frequencies (in cm⁻¹) of the water molecules near W402 in BR

The QM/MM results of model 2 on the basis of the 1C3W structure primarily discussed in this study.

Donor…acceptor	FTIR-2a	1C3W		2NTU
		QM/MM (model 2)	Empirical equationb (model 2)	QM/MM (model 2)	Empirical equationb (model 2)
W402…Asp-85	2171(lowest)	2078 (lowest)	2201 (lowest)	2120 (lowest)	2240 (lowest)
W402…Asp-212	2599	2690	2592	2670	2572
W401…Asp-85	2323	2400	2311	2411	2359
W406…(ND)	2690	2678	2727	2689	2727
W406…W401	2636	2658	2620	2656	2622
W406…Asp-212	2292	2312	2303	2302	2328
R2 to FTIRc		0.94c	0.99c	0.96c	0.98c

^a See Ref. 7.

^b See Ref. 34.

^c Relative to FTIR-2.

Notably, Equation 1 indicates that the O-D stretching frequency of 2500 cm⁻¹ corresponds to the O_acceptor–O_donor distance of ∼2.8 Å. The deviation from the values calculated with the QM/MM numerical differentiation method appears to be more pronounced for frequencies >2500 cm⁻¹ (Fig. 3). Because the original parameters were derived from solid hydrates (34), the larger deviation for weak H-bonds at these frequencies may be due to H-bond patterns or mode couplings specific to the protein environments.

FIGURE 3.

The geometric correlation of the O-D stretching frequencies with the O_acceptor…H-O_donor bond of H₂O (correlation with the O_donor–O_acceptor distance). Each open square (□) represents the O_donor–O_acceptor distance of the QM/MM optimized geometries (Tables 1 and 4) and the O-D stretching frequency calculated by a QM/MM numerical differentiation method (Tables 2, 3, and 5). The solid curve was determined using an empirical equation (Equation 1) proposed by Mikenda (34).

RESULTS

H-bond Geometries Near W402 in BR

The O-D stretching frequencies of D₂O predominantly depend on the H-bond geometry, particularly the H-bond donor-acceptor distance (O_donor–O_acceptor) (34). If the assumed H-bond pattern is not correct, the resulting QM/MM optimized geometries significantly deviate from the original crystal structure. Thus, although the QM/MM optimized geometries (especially the heavy atom positions) are not always consistent with the crystal structures, it is necessary to carefully evaluate the H-bond pattern with respect to the atomic coordinates of the crystal structures before frequency calculations.

In BR, two H-bond patterns are geometrically possible for the H-bond network of W402 (Fig. 4). A model previously presented from FTIR studies (7) suggests that an H atom of W401 has no H-bond acceptor (model 1), whereas an alternative model suggests that an H atom of W406 has no H-bond acceptor (model 2). The QM/MM-optimized geometries of the two models appear to represent the actual H-bond patterns of the crystal structures because of the low root mean square deviation (0.17–0.19 Å, Table 1), implying that the two models probably coexist as tautomers in the actual BR protein environment.

FIGURE 4.

H-bond geometries of water molecules near the Schiff base in BR and ASR. Pink arrows in model 1 indicate mode couplings of the two O-D bonds. Red dotted lines indicate “strong H-bonds” reported in FTIR studies (7).

Model 1 resembles the QM/MM geometry reported by Hayashi and Ohmine (8). However, in their QM/MM geometry, the O_W402–O_Asp-212 distance was significantly elongated to 4.07 Å, and the H-bond is absent (8). This varies significantly from the crystal structures and the present QM/MM geometries (O_W402–O_Asp-212 = 2.85–3.01 Å, Tables 2 and 3). In the reported QM/MM geometry, W402 forms a new H-bond with another carboxyl O atom of Asp-212 (Fig. 4). Hence, it is to be considered that the band assignment reported by Hayashi and Ohmine (8) refers to another possible H-bond geometry that differs from the crystal structures.

TABLE 2.

Model 1, O-D stretching frequencies (in cm⁻¹) of the water molecules near W402 in BR

Connecting lines indicate the presence of coupling between the two modes. n.d., not determined; R², correlation coefficient.

^a See Ref. 7.

^b See Ref. 8.

^c See Ref. 34.

The QM/MM geometries imply that model 1 can explain the O_donor–O_acceptor distances in the crystal structure provided by Luecke et al. (PDB code 1C3W) (6) more reasonably than model 2 (Table 1). On the other hand, model 2 appears to explain the O_donor–O_acceptor distances in the crystal structure determined by Lanyi and Schobert (PDB code 2NTU) (19) more reasonably, namely in terms of O_W402–O_Asp-85, O_W402–O_Asp-212, O_W401–O_Asp-85, and O_Arg-82–O_W406 distances (Table 1). However, the small discrepancies could also be due to being artifacts of the crystallization process. Because the resolutions of the two crystal structures are 1.53 and 1.55 Å (which are essentially the same), both of the H-bond tautomers are likely to contribute to the O-D stretching frequencies measured by FTIR. If both tautomers are present, this might corroborate the rapid H/D exchange observed in this region.

Calculated O-D Stretching Frequencies in BR

In FTIR studies (7), the three H-bonds O_W401–H…O_Asp-85, O_W402–H…O_Asp-85, and O_W406–H…O_Asp-212 were classified as being strong H-bonds because of the assignment of low frequencies (Fig. 4).

Model 1

In FTIR studies (7) the lowest frequency of 2171 cm⁻¹ was assigned to O_W402–H…O_Asp-85. However, in model 1, O_W402–H…O_Asp-85 was longer than O_W401–H…O_Asp-85 (Table 1), leading to a larger calculated O-D stretching frequency for O_W402–H…O_Asp-85 than O_W401–H…O_Asp-85 (Table 2), as predicted from the empirical equation by Mikenda (Equation 1) (34).

In model 1, two mode couplings between O_W401–H…O_Asp-85 and O_W401–H…O_Asp-85 and between O_W406–H…O_Asp-212 and O_W406–H…O_W401 were observed. However, the latter coupling is not consistent with the FTIR assignments. In FTIR studies, H-bonds O_W406–H…O_Asp-212 and O_W406–H…O_W401 were assigned as strong (2292 cm⁻¹) and weak (2599 cm⁻¹), respectively (Table 2) (7). Because mode couplings are unlikely to occur between a strong and weak H-bond, model 1 fails to sufficiently explain the FTIR assignments.

Model 2

In contrast to model 1, model 2 can reasonably explain the strong H-bond (2292 cm⁻¹) of O_W406–H…O_Asp-212 and the weak H-bond (2636 cm⁻¹) of O_W406–H…O_W401 that were assigned by FTIR studies (Table 3). Furthermore, the lowest calculated frequency was observed for O_W402–H…O_Asp-85, which is also consistent with FTIR studies (7). Significantly high correlations of the frequencies predicted from Equation 1 suggest that assignment of the O-D stretching frequencies on the basis of model 2 is also justified by the H-bond geometry. Note that no remarkable mode couplings were observed for model 2 in the QM/MM calculations.

In the following section further analysis of BR will be discussed on the basis of model 2 using the crystal structure by Luecke et al. (PDB code 1C3W) (6) unless otherwise specified.

H-bond Geometries near W402 and O-D Stretching Frequencies in ASR

In contrast to BR, the ASR crystal structure only possesses a single water molecule, W402, near the Schiff base (Fig. 4). The H-bond geometry of the crystal structure was reasonably reproduced by the QM/MM geometry optimization (Table 4). In the QM/MM calculations, the lowest O-D stretching frequency of 2376 cm⁻¹ in ASR, which was found for O_W402–H…O_Asp-75 (Table 5), was significantly larger (by ∼300 cm⁻¹) than that for the corresponding O_W402–H…O_Asp-85 in BR (2078 cm⁻¹, Table 3). In FTIR studies, the O-D stretching of D₂O was only observed in the >2500-cm⁻¹ region for ASR (11), which was also ∼300 cm⁻¹ higher than the lowest frequency of 2171 cm⁻¹ in BR (7). Because the main purpose of this study is to understand the reason for the significant discrepancy between the lowest frequencies of D₂O (assumed as W402 (11)) of BR and ASR, these computational results are sufficiently accurate for our purpose to describe the chemical properties of W402.

TABLE 4.

H-bond geometries near W402 in the ASR crystal structure (PDB ID code 1XIO at 2.00 Å resolution) and the QM/MM geometry

ND, not determined. Seesupplemental S1 for the atomic coordinates of the QM/MM geometries. Distances are in Å, and angles are in degrees.

Donor…acceptor	1XIO	QM/MM, deprotonated Asp-75
	Å	Å
W402…Asp-75	2.74	2.71
W402…Tyr-51	2.92	2.97
Trp-76…W402	2.94	2.84
Lys-210…W402	3.02	3.00
Root mean square deviationa	ND	0.17
Angle (NLys-210…OW402…OAsp-75) (degrees)	83	92

^a Excluding Lys-210 due to obvious displacement of the sidechain carbon atoms from the original crystal structure (Fig. 2c).

TABLE 5.

O–D stretching frequencies (in cm⁻¹) of W402 in ASR

Donor…acceptor	FTIRa	Deprotonated Asp-75
		QM/MM	Empirical equation
W402…Asp-75	>2500	2376	2392
W402…Tyr-51	>2500	2573	2583

^a See Ref. 11.

DISCUSSION

Significant Differences between the H-bond Energy Profiles of BR and ASR

To identify the origin of the difference in O-D stretching frequencies of W402, we analyzed the potential energy profiles of H-bonds O_W402–H…O_Asp-85 in BR and O_W402–H…O_Asp-75 in ASR by altering the H atom position along the O_W402–O_Asp-85/75 bond (see Fig. 5). In both H-bonds, the energy minimum was found at the W402 moiety rather than at the Asp moiety, suggesting that in the ground state W402 exists as H₂O in the presence of deprotonated Asp in BR and ASR.

FIGURE 5.

Potential energy profiles along the proton transfer coordinates of O_W402–H…O_Asp-85 in BR (black solid line) and O_W402–H…O_Asp-75 in ASR (blue solid line). ΔE describes the difference in energy relative to the energy minimum. Boxed arrows indicate the shifts in the potential energy curve accompanied by the pK_a(Asp) change from ASR to BR. The marginal energy drop near 1.74 Å in O_W402–H of ASR was due to an alteration of the H-bond pattern; below 1.74 Å, Ser-47 donates an H-bond to the carboxyl O atom of Asp-75, which is simultaneously H-bonded by W402. At 1.74 Å, the hydroxyl H atom is oriented toward another carboxyl O atom of Asp-75 due to the unusual proximity of the H atom of W402 to Asp-75.

On the other hand, the potential energy profiles of the Asp moiety were significantly different for BR and ASR. For an H atom along the O_W402–O_Asp-85/75 bond, the energy of the Asp-85 moiety of BR is significantly lower than that of the Asp-75 moiety of ASR. Thus, it is more favorable for an H atom of O_W402–H to be at the Asp moiety in BR than in ASR, indicating that pK_a(Asp) is significantly higher for BR than ASR (arrow 1 in Fig. 5). As clearly shown in Fig. 5, the lower energy near the Asp-85 moiety in BR leads to greater broadening of the potential-well width near W402 toward Asp-85, which corresponds to an elongation of O_W402–H. Thus, the H atom of O_W402–H migrates toward the H-bond acceptor Asp more in BR than in ASR (arrow 2 in Fig. 5). It should be noted that the longer O_W402–H corresponds to a smaller O-D stretching frequency. It is also obvious from the width of the entire well of the potential energy profiles that the O_W402–O_Asp-85 length in BR (∼2.6 Å, Table 1), which is shorter than the O_W402–O_Asp-75 length in ASR (∼2.7 Å, Table 4), is a result of the smaller pK_a difference between W402 and the Asp residue in BR relative to the corresponding difference in ASR (arrow 3 in Fig. 5). Thus, from the potential energy profiles (Fig. 5), the difference of ∼300 cm⁻¹ in the O-D stretching frequency is mainly due to the difference between the pK_a values for the Asp residues in BR and ASR.

Factors that Differentiate pK_a(Asp) of BR and ASR

By solving the linear Poisson-Boltzmann equation, we calculated pK_a(Asp-85) in BR and pK_a(Asp-75) in ASR to be 1.5 and −5.1, respectively, in the presence of the protonated Schiff base (Table 6). The calculated pK_a(Asp-85) of 1.5 is consistent with an experimentally measured pK_a(Asp-85) of 2.2–2.6 in the presence of the protonated Schiff base (2, 35). Although pK_a(Asp-75) is not known, the significantly low pK_a(Asp-85) in BR relative to pK_a(Asp-75) in ASR is consistent with the QM/MM-calculated potential energy profiles (Fig. 5). The significantly low pK_a(Asp-75) in ASR appears to be consistent with the fact that Asp-75 remains deprotonated throughout the photocycle (36).

TABLE 6.

Contribution of residues to the pK_a shifts of Asp-85 in BR and Asp-75 in ASR (relative to the bulk solvent) in the presence of the protonated Schiff base (in pK_a units)

For clarity, residue pairs that are responsible for differences of <0.7 between pK_a(Asp-85) and pK_a(Asp-75) are not listed except for the Glu-204/Asp-198 pair. Side, side chain; bb, backbone.

BR (1C3W)	pKa(Asp-85) = 1.5			ASR (1XIO)	pKa (Asp-75) = −5.1			Difference total
	Side	bb	Total		Side	bb	Total
Ile-52	−0.1	0.0	−0.1	Trp-46	−0.6	−0.3	−0.9	0.8
Ala-53	0.0	−0.5	−0.5	Ser-47	−2.6	−0.6	−3.2	2.7
Tyr-57	−0.5	−0.4	−0.9	Tyr-51	0.2	−0.4	−0.2	−0.7
Arg-82	−3.9	0.8	−3.1	Arg72	−2.1	0.9	−1.2	−1.9
Glu-194	1.2	−0.1	1.1	Ser-188	0.0	0.0	0.0	1.1
Glu-204	0.0	0.1	0.1	Asp-198	0.6	0.0	0.6	−0.5
Asp-212	6.3	−0.1	6.2	Pro-206			−0.1	6.3
Schiff			−9.3	Schiff			−10.3	1.0

Asp-212 in BR

The most significant difference observed between pK_a(Asp-85) and pK_a(Asp-75) was induced by the substitution of Asp-212 in BR for Pro-206 in ASR (Fig. 6), which is responsible for the pK_a difference of ∼6 (Table 6). Because Asp-212 is an H-bond acceptor for O_W402–H and is thus far the most proximal negative charge to Asp-85 (5.1 Å), Asp-212 upshifts pK_a(Asp-85) by the largest amount among all residues in BR (Table 6).

FIGURE 6.

Residues that contribute to differences between pK_a(Asp) of BR (a) and ASR (b).

It is widely known that both Asp-85 and Asp-212 are ionized in the ground state (2, 37, 38). However, the deprotonated state of Asp-212 is more energetically stable than the deprotonated state of Asp-85 as seen in the lower (pK_a(Asp-212) = −2.0 and pK_a(Asp-85) = 1.5) (Table 7); this is consistent with previous pK_a assignments of <1 for Asp-212 (39, 40) or pK_a(Asp-85) < pK_a(Asp-212) (41, 42). The deprotonated state of Asp-212 was facilitated by H-bond donations from Tyr-57 (decreasing pK_a(Asp-212) by 3.4) and Tyr-185 (decreasing pK_a(Asp-212) by 3.3). In addition, Arg-82 is closer to Asp-212 (3.8 Å) than Asp-85 (6.6 Å) (Fig. 6), stabilizing the ionized state of Asp-212 more effectively than Asp-85 (increasing pK_a by ∼6, Table 7).

TABLE 7.

Key residues that alter the pK_a difference between Asp-85 and Asp-212 in the presence of the protonated Schiff base (in pK_a units)

Side, side chain; bb, backbone; ND = not determined.

BR (1C3W) residue	pKa (Asp-85) = 1.5			pKa (Asp-212) = −2.0
	Side	bb	Total	Side	bb	Total
Met-56	−0.4	−0.8	−1.2	−0.1	−0.3	−0.3
Tyr-57	−0.5	−0.3	−0.9	v3.2	−0.2	−3.4
Ala-81	0.0	0.7	0.8	0.0	0.1	0.1
Arg-82	−3.9	0.8	−3.1	−6.4	0.1	−6.3
Asp-85	ND	ND	ND	6.4	−0.4	6.0
Trp-86	−0.2	−0.6	−0.8	v0.8	−0.3	−1.0
Thr-89	−2.0	−0.9	−2.9	0.1	−0.2	−0.1
Tyr-185	−0.3	−0.1	−0.4	−3.2	−0.1	−3.3
Glu-194	1.2	−0.1	1.2	1.8	−0.1	1.7
Phe-208	0.0	0.2	0.2	0.0	1.8	1.8
Asp-212	6.3	−0.1	6.2	ND	ND	ND
Schiff	−9.5	0.2	−9.3	−9.3	0.6	−8.7

Although the most significant difference between pK_a(Asp-85) and pK_a(Asp-75) induced by the substitution of Asp-212 in BR for Pro-206 in ASR is remarkable, this is not sufficient to entirely explain the difference between pK_a(Asp-85) and pK_a(Asp-75) (see below), which is in agreement with mutational studies of ASR mutant P206D (43).

Ser-47 in ASR

Ser-47 donates an H-bond to Asp-75 in ASR, decreasing pK_a(Asp-75) by 3.2, whereas the corresponding residue in BR is the non-polar Ala-53 (Table 6). This was responsible for a pK_a difference of 2.7 between Asp-85 and Asp-75 (Table 6).

Glu-194–Glu-204 Pair in BR

Another remarkable pK_a(Asp-85) upshift in BR originates from the Glu-194–Glu-204 moiety near the terminal region of the proton transfer pathway (2, 3) (Fig. 6). It is likely that the two acidic residues share a proton. The corresponding acidic residue pair is absent in ASR, thereby contributing slightly to the larger pK_a(Asp) in BR relative to that in ASR (∼0.5 pK_a units, Table 6). In this study Glu-204 is protonated and Glu-194 is deprotonated when Ser-193 donates an H-bond to Glu-194, which is consistent with the previously reported pK_a(Glu-204) of ∼9 (35, 44, 45).

Nevertheless, the geometry of the crystal structure also suggests that protonated Glu-194 and deprotonated Glu-204 are almost equally possible when Ser-193 donates an H-bond to Glu-204, as suggested by Gerwert and co-workers (46). Thus, it would be more plausible to consider the two acidic residues as a pair that possesses ∼1 H⁺ as concluded in previous QM/MM simulations (47). This resembles a pair of acidic residues, Glu-L212 and Asp-L213, in the proton transfer pathway of photosynthetic reaction centers, which has been interpreted as sharing ∼1 H⁺ (48–50). Deprotonated Glu-194 contributes to an increase in pK_a(Asp-85) of ∼1 (Table 6); this, therefore, corroborates the linkage between pK_a(Asp-85) and pK_a(Glu-204) previously reported in mutant BR studies (35) if we consider Glu-194–Glu-204 as a pair of acidic residues sharing ∼1 H⁺.

Orientation of Arg-82 in BR and Arg-72 ASR

Arg-82 was found to contribute to an increase in pK_a(Asp-85) of 3.1 in BR (Table 6), which is in agreement with a similar increase (∼4.5) reported for BR mutants R82Q and R82A (51, 52). Irrespective of the conservation of Arg-82/Arg-72 in BR/ASR, their influence on pK_a(Asp-85) and pK_a(Asp-75) significantly differs by ∼2 because the side chain of Arg-72 is oriented away from the Schiff base and toward the extracellular side in ASR (O_Asp-75–N_Arg-72 = 9.6 Å) (10), which is the opposite of the orientation of Arg-82 in BR (O_Asp-85–N_Arg-82 = 6.6 Å) (6) (Fig. 6). Thus, the differing orientations of the Arg side chains decrease the pK_a(Asp) difference between BR and ASR (Table 7).

In summary, the significantly large pK_a(Asp-85) in BR relative to pK_a(Asp-75) in ASR is largely due to the presence of Asp-212 near Asp-85 in BR and the absence of the corresponding acidic residue in ASR. In addition, the donation of an H-bond from Ser-47 to Asp-75 lowers the pK_a(Asp-75) in ASR by facilitating deprotonation. The corresponding polar residue is absent in BR; this upshifts pK_a(Asp-85) in BR relative to pK_a(Asp-75) in ASR. The Glu-194–Glu-204 pair is located near the terminal region of the proton transfer pathway in BR, whereas only a single acidic residue is located near the corresponding position in ASR. This also upshifts pK_a(Asp-85) in BR relative to pK_a(Asp-75) in ASR.

Conclusions

QM/MM calculations revealed that the heavy atom positions in the BR crystal structures could be explained by one of two models. Model 2 can reasonably explain the strong H-bond (2292 cm⁻¹) of O_W406–H…O_Asp-212 and the weak H-bond (2636 cm⁻¹) of O_W406–H…O_W401, as previously assigned by FTIR (Table 3). In model 2, the lowest calculated frequency was observed at O_W402–H…O_Asp-85, which is also consistent with FTIR studies. Essentially the same frequencies were also predicted solely from the O_donor–O_acceptor distances using Equation 1 proposed by Mikenda (34). This suggests that Equation 1 can be used to support the frequency assignments made from spectroscopic studies. In ASR, the lowest calculated O-D stretching frequency is observed at O_W402–H…O_Asp-75, which is ∼300 cm⁻¹ higher than that of BR (Tables 3 and 6), as previously reported in FTIR studies of BR (7) and ASR (11).

The potential energy profiles of two H-bonds, O_W402–H…O_Asp-85 in BR and O_W402–H…O_Asp-75 in ASR, demonstrate that the significant pK_a difference between Asp-85 in BR and Asp-75 in ASR is ultimately responsible for the difference of ∼300 cm⁻¹ in the O-D stretching frequency of W402 between BR and ASR (Fig. 5). Electrostatic calculations resulted in a pK_a(Asp-85) of 1.5 in BR and a pK_a(Asp-75) of −5.1 in ASR (Table 6). The pK_a difference is mainly due to differences in the influences of Ala-53, Arg-82, Glu-194, and Asp-212 on Asp-85 in BR and the corresponding residues Ser-47, Arg-72, Ser-188, and Pro-206 on Asp-75 in ASR, indicating the presence of long range electrostatic interactions along the proton transfer pathways (31). A strongly H-bonded water near the Schiff base, which results from electrostatic interactions between Asp-85 and these acidic residues in BR, could play a key role in proton transfer.