Low-barrier hydrogen bond in photoactive yellow protein

Abstract

Low-barrier hydrogen bonds (LBHBs) have been proposed to play roles in protein functions, including enzymatic catalysis and proton transfer. Transient formation of LBHBs is expected to stabilize specific reaction intermediates. However, based on experimental results and theoretical considerations, arguments against the importance of LBHB in proteins have been raised. The discrepancy is caused by the absence of direct identification of the hydrogen atom position. Here, we show by high-resolution neutron crystallography of photoactive yellow protein (PYP) that a LBHB exists in a protein, even in the ground state. We identified ≈87% (819/942) of the hydrogen positions in PYP and demonstrated that the hydrogen bond between the chromophore and E46 is a LBHB. This LBHB stabilizes an isolated electric charge buried in the hydrophobic environment of the protein interior. We propose that in the excited state the fast relaxation of the LBHB into a normal hydrogen bond is the trigger for photo-signal propagation to the protein moiety. These results give insights into the novel roles of LBHBs and the mechanism of the formation of LBHBs.

The idea that the formation of low-barrier hydrogen bonds (LBHBs) plays an essential role in enzyme catalysis was proposed in the early 1990s (1, 2). Although several lines of circumstantial evidence support the existence of LBHBs, negative results have also been published (3–5). This discrepancy is caused by the absence of direct demonstration of LBHBs in proteins. In general, hydrogen bonds in proteins are identified by the distance between a donor and an acceptor within the crystal structure. Because of its abnormally short bond length, a LBHB is accompanied by a quasi-covalent bond feature, whereas an ordinary hydrogen bond can be depicted as an electrostatic interaction between a donor–proton dipole and a dipole (or a monopole) on an acceptor atom (6–8). In LBHBs, the proton is shared by the donor and acceptor atoms, resulting in the distribution of the hydrogen between the two (6). Therefore, to identify a LBHB, it is essential to determine the position of the hydrogen atom and those of the donor and acceptor atoms. Recently, it was shown that a light sensor protein, photoactive yellow protein (PYP), contains 2 short hydrogen bonds (SHBs) adjacent to the reaction center, even in the ground state (9, 10). The hydrogen atoms involved in the SHBs, however, could not be observed either by X-ray crystallography at atomic resolution (9, 11) or neutron crystallography at 2.5-Å resolution (10).

PYP is a putative photoreceptor for negative phototaxis of the purple phototropic bacterium, Halorhodospira halophila (12). The chromophore of PYP, p-coumaric acid (pCA), is buried in a hydrophobic pocket. Absorption of a photon triggers the isomerization of the chromophore and the subsequent thermal reaction cycle (13, 14). The hydrogen-bonding network near the chromophore is modulated during the thermal reaction, resulting in proton transfers within the network that are associated with large conformational changes (15–18). Two SHBs are formed between pCA and E46 and between pCA and Y42. The SHBs are OH···O hydrogen bonds, with the phenolic oxygen of the chromophore located 2.51 Å from the phenolic oxygen of Y42 and 2.58 Å from the carboxylic oxygen of E46 (9). The proton transfer occurs between E46 and the chromophore during the thermal reaction (15, 16). Therefore, the SHBs are essential for the photoreaction and stability of PYP.

In this article, to reveal the properties of the SHBs in PYP, we performed high-resolution neutron crystallographic analysis combined with high-resolution X-ray crystallography. The large crystals, prepared by using the crystallization phase diagram method (19), diffracted neutron and X-ray up to 1.5- and 1.25-Å resolution, respectively, at room temperature. Using the joint method of the X-ray and neutron refinements, clear nuclear density maps of 87% of hydrogen and deuterium atoms in the whole protein were observed. From the nuclear density maps, we succeeded in identifying the deuterium atoms involved in the SHBs and determined that the SHB between pCA and E46 is a LBHB, whereas the SHB between pCA and Y42 is not a LBHB but should be termed as the short ionic hydrogen bond (SIHB). We also revealed that R52, which is believed to be protonated to be a counter ion of pCA, is deprotonated. Finally, based on the observation, we discuss roles of the LBHB in the protein.

Results and Discussion

Observation of the Hydrogen Atoms of PYP.

To identify the positions of hydrogen and deuterium atoms precisely, a joint method of neutron and X-ray crystallography was applied to neutron diffraction data at 1.5-Å resolution and X-ray diffraction data at 1.25-Å resolution, and the positions of 87% of hydrogen and deuterium atoms in PYP were determined (Fig. 1A). Fig. 1B shows the hydrogen-bonding network, including all hydrogen/deuterium atoms around the chromophore. The details of the hydrogen bonds are shown in Fig. 1 C–F, onto which the distributions of the heavy atom electron density and the deuterium/hydrogen atom nuclear density are superimposed. All hydrogen/deuterium atoms responsible for hydrogen bonds are clearly observed. The distances from the phenolic oxygen of the chromophore to the phenolic oxygen of Y42 and to the carboxylic oxygen of E46 are 2.52 and 2.56 Å, respectively (Fig. 1C). These distances are altered by a few 0.01 Å compared with the previous X-ray crystal structure (9). This subtle difference was also observed by the previous neutron study (10), suggesting the effects of the H–D exchange. The hydrogen-bonding network across E46 and Y42 is spread over the molecule and reaches the N-terminal region (Fig. 1D) and the loop connecting β4 and β5 through R52 (Fig. 1E). These regions exhibit substantial conformational changes during the photo-reaction process, ultimately leading to signaling (18, 20). Another important finding is that R52 is deprotonated (Fig. 1E), although R52 is believed to be protonated to serve as the counterion of the negatively-charged chromophore (21).

Fig. 1.

Structure of PYP including hydrogen atom positions. (A) Stereoview of the whole molecule including the determined hydrogen and deuterium atoms. (B) Stereoview of the hydrogen bonding network around the chromophore. (C–F) Stereoviews of the representative hydrogen bonds shown in B. The F_O − F_C difference nuclear density map omitting the hydrogen/deuterium atoms, superimposed on the structural model is shown. The blue mesh (contoured at 5.5 σ) represents the positive nuclear density of the deuterium atoms, and the red mesh (contoured at −5.5 σ) represents the negative nuclear density of the hydrogen atoms. The yellow mesh (contoured at 4.0 σ) shows the 2F_O − F_C electron density map calculated from X-ray crystallographic analysis, which was used for the determination of the heavy atom positions.

Hydrogen Bonds in PYP.

According to the distances between donor and acceptor atoms, PYP contains 115 hydrogen bonds, among which 103 hydrogen (or deuterium) atoms exhibited clear nuclear densities. Fig. 2A shows the correlation of donor-H(D) and H(D)-acceptor bond lengths with the donor–acceptor distances and also the histogram of hydrogen-bond lengths. Among these, 2 SHBs exhibit distances significantly shorter than those of the other hydrogen bonds. Except for the SHB of pCA-E46, the donor-H(D) distances are nearly constant. Average distances are 0.95 Å for O-H(D) and 0.98 Å for N-H(D), and the H(D)-acceptor distances range from 1.8 to 2.5 Å with an average distance of 2.07 Å. Hydrogen atoms involved in these hydrogen bonds were replaced by deuterium atoms except for those located in some of the β-strands. The average distances of the NH and ND bonds were, however, indistinguishable at the present resolution. The nuclear density maps of the deuterium atoms involved in the two SHBs can be clearly observed (Fig. 2 B and C). The interatomic distances involved in these hydrogen bonds are listed in Fig. 2 B and C. In the SHB between pCA and E46, the distances between the phenolic oxygen of pCA and D and between the carboxylic oxygen of E46 and D are 1.37 and 1.21 Å, respectively. Both bond lengths are substantially longer than the average lengths of OH (OD) covalent bonds (0.95 Å), indicating that the deuterium atom is not covalently bound to either of the two oxygen atoms. However, these bond lengths are much shorter than the average H(D)-acceptor bond length (2.07 Å) (see Fig. 2A). These values indicate that the deuterium atoms must be shared by both oxygen atoms. We can conclude that the SHB between the chromophore and E46 is a LBHB. In contrast, the position of the deuterium atom between Y42 and pCA is shifted toward the phenolic oxygen atom of Y42. The interatomic distance between the phenolic oxygen of Y42 and D is 0.96 Å, close to the average covalent OH bond length, 0.95 Å. Therefore, the deuterium atom is covalently bound to the phenolic oxygen atom of Y42. The interatomic distance between the phenolic oxygen of pCA and D is 1.65 Å, which is shorter than the average distance between H(D) and acceptor atom (O), 2.07 Å. The SHB between pCA and Y42 is not a LBHB, but rather a SIHB. Thus, a SHB is not always a LBHB.

Fig. 2.

Hydrogen bonds in PYP. (A) Correlation of donor-H(D) and H(D)-acceptor bond lengths with the donor–acceptor distances and a histogram of hydrogen bond lengths. The open and closed triangles show the donor-H(D) and the H(D)-acceptor distances, respectively. The SHBs are shown as a red circle (pCA-E46) and a blue circle (pCA-Y42). The solid lines represent the calculated donor-H(D) distances, assuming bent angles of 120° and 180°, respectively. (B and C) The nuclear and electron density maps with the structure models of the two SHBs, pCA-E46 (B) and pCA-Y42 (C). The blue mesh represents the positive nuclear density of the deuterium atoms, contoured at 80% of the maximum peak height of the F_O − F_C difference maps omitting each deuterium atom involved in the hydrogen bond; contour levels of 80% of the maximum peak height correspond to 8.26 σ for pCA-E46 and 6.86 σ or pCA-Y42. The yellow mesh (contoured at 4.1 σ) and the red mesh (contoured at −5.3 σ) show the 2F_O − F_C electron density maps of the heavy atoms and the F_O − F_C difference nuclear density map omitting the hydrogen atoms, respectively.

Role of LBHB in PYP.

It has been proposed that pCA takes an anionic form (15, 16). In general, an isolated charge buried in a protein's interior will destabilize the protein structure; to be electrically neutralized, it must be coupled with a counterion. The counterion against pCA's negative charge is thought to be the cationic form of R52 at a distance of 6.34 Å from the phenolic oxygen of pCA (21) (Fig. 3A). However, as shown in Fig. 1E, R52 takes an electrically neutral form in PYP, indicating that the isolated negative charge must be stabilized by another mechanism. One possible explanation can be retrieved by the strong bond strength of LBHB. The bond strength of LBHB is 12–24 kcal/mol (6), which is extremely stronger than that of an ordinary ionic hydrogen bond (a few kcal/mol). Therefore, the energetic gain by the LBHB compensates the energetic disadvantage of the isolated charge buried in the protein interior. We also propose that the isolated negative charge is stabilized partly because of the delocalization of the negative charge within the LBHB-conjugated system (Fig. 3B). This proposal is based on a valence bond representation of LBHB (5, 8, 22). LBHB connecting π-conjugated systems are often observed in homoconjugated betaine complexes (8, 22). Although the stretching frequency of asymmetric CO in a carboxyl group usually appears at ≈1,780 cm⁻¹, the stretching frequency in homoconjugated betaine complexes is shifted toward the lower frequency, ranging from 1,720 to 1,740 cm⁻¹. The asymmetric CO stretching vibration of the carboxyl group of E46 was observed at 1,739 cm⁻¹ (15, 16). The good agreement in the CO stretching frequency supports our proposal in Fig. 3B that the LBHB combines the two π-conjugated systems of E46 and pCA. As a consequence, a negative charge is delocalized over the conjugated system comprised of the two π-conjugate bond groups.

Fig. 3.

The mechanism of the stabilization of the isolated negative charge in the vicinity of the chromophore in a hydrophobic environment of PYP. (A) The neutralization by a counter ion, which is believed so far. We ruled out this mechanism by the finding that R52 is not protonated (Fig. 1E). (B) Strong bond strength of LBHB and charge delocalization caused by the quasi-covalent bond of LBHB stabilizes the energetic disadvantage of the isolated negative charge buried in the protein interior.

The existence of the LBHB between pCA and E46 indicates that the proton affinities of pCA and E46 are close to each other in the protein interior (6, 22), even though in solution the pK_a values of pCA and the carboxyl group of glutamic acid are 8.8 (23) and 4.25, respectively. The pK_a increase of E46 is caused by the transfer of glutamic acid into the hydrophobic environment of the protein interior, and the pK_a decrease of pCA is partly caused by the stabilization of the deprotonated resonance form of pCA, with the aid of the hydrogen bond (NH···O) between the amide proton of C69 and the carbonyl oxygen of pCA (24) (see Fig. 1F). As a consequence, in the dark state, pCA and E46 show similar pK_a values in the protein interior. In the excited state, an instantaneous change in dipole moment brings about a charge translocation from the phenol ring to the ethylene chain of the chromophore (25). This charge translocation leads to alteration of the proton affinity of the phenolic oxygen, resulting in relaxation of the LBHB into an ordinary ionic hydrogen bond; the shared proton is transferred to E46 during the lifetime of the excited state (≈1 ps). Ultra-fast infrared spectroscopy has revealed that the CO stretching mode of the carboxyl group of E46 is altered during formation of the excited state (26). In the ground state, the quasi-covalent bond of the LBHB can be expected to sterically restrain the phenol ring moiety of the chromophore at the specific position. Once the LBHB is disrupted in the excited state, the phenol ring moiety of the chromophore is liberated and promotes the subsequent molecular events, specifically, the fast isomerization of the chromophore.

LBHBs have been proposed to be responsible for enzymatic catalysis, in particular, in the transition state of the catalytic centers of serine proteases (1, 2). However, the transient formation of LBHBs has not been confirmed to our knowledge until this study. Our data clearly show that LBHB can be formed in a protein even in the ground state. We conclude that the pK_a matching between donor and acceptor atoms is the prerequisite for the formation of LBHB, which provides an insight into the transient formation of LBHB. The present observation of a LBHB involved in PYP reveals roles of LBHB in the biophysical aspects of protein structure and function, i.e., stabilization of an isolated charge in the protein interior and mediation of fast proton transfer in the excited state. Recently, spectroscopic study of a GFP variant proposed that fast excited-state proton transfer near the chromophore is facilitated by a LBHB (27), suggesting that the properties of LBHB proposed here are widely used. The extensive experimental and theoretical studies of the LBHB of PYP will establish additional roles for hydrogen bonds and help develop our understanding of the molecular mechanisms operating in a wide range of proteins.

Methods

Crystallization of PYP for Neutron and X-Ray Diffraction Experiments.

Wild-type PYP was overexpressed by using the pET system in Escherichia coli BL21(DE3) (Novagen) and reconstituted with pCA anhydride in 4 M urea buffer (28). The proteins were purified by using DEAE Sepharose CL6B (Amersham Biosciences) column chromatography several times until the optical purity index (absorbance 277 nm/absorbance λ_max) became <0.44. Crystals of PYP were prepared by the hanging-drop vapor diffusion method in conjunction with the microseeding method (a useful means of regulating the number of seeds in a drop) under the supersaturation conditions (19). The crystallization drop solution was adjusted to 24 mg·mL⁻¹ PYP, 2.2 M ammonium sulfate, and 1 M sodium chloride with 20 mM sodium phosphate buffer. The reservoir solution was 2.5 M ammonium sulfate and 1.1–1.2 M sodium chloride. These crystallization buffers were made with 99.9% heavy water (Aldrich). The temperature and pD of these solutions were maintained at 293 K and 9.0, respectively.

X-Ray Data Collection, Processing, and Refinement.

The X-ray diffraction experiments were performed at room temperature by using the BL41XU beam line installed at SPring-8, Hyogo, Japan. The crystal (2.37 × 0.70 × 0.66 mm³) was sealed in a quartz capillary. The diffraction data were recorded with a Quantum 315 CCD detector (ADSC) over a rotation of 180° with an oscillation step of 1.0°. The wavelength was set to 0.45 Å. The camera distance and exposure time for each image were 300 mm and 1 s, respectively. The diffraction data up to 1.25 Å were processed with XDS (29) and SCALA (30). The R_merge of each diffraction image was not altered during the diffraction experiment, suggesting that the radiation damage was negligibly small.

The initial phases were determined by using the molecular replacement program AMoRe (31); the previously reported structural model (Protein Data Bank ID code 2PHY; ref. 21) was used as the initial model. The refinement programs CNS (32) and SHELXL (33) were used to refine the structural model, and Coot (34) was used to build the model. Finally, the anisotropic B factors were refined. During the analysis, several residues were found to occupy more than one conformation; these multiple conformations were subsequently built and refined. The final values of R and R_free were 11.2% and 14.9%, respectively (Table S1). Model quality was assessed by PROCHECK (35) implemented in CCP4. All residues displayed φ and ϕ values in the favored or allowed regions of the Ramachandran plot. The detailed statistics of data collection are shown in Table S1.

Neutron Data Collection and Processing.

The neutron diffraction experiments were carried out at room temperature by using a neutron single crystal diffractometer (BIX-4) installed at the JRR-3 reactor, Japan Atomic Energy Agency (36). The crystal (2.89 × 0.85 × 0.79 mm³) was sealed in a quartz capillary. The neutron wavelength was set to 2.6 Å. A sufficient amount of reservoir solution was added to the bottom of the capillary to prevent the crystal from drying. The data were collected according to the step-scan method, with an oscillation step of 0.3°. The collection time for each frame was 4 h. To improve completeness, after collecting 118 frames, the capillary was rotated by 90° on the plane perpendicular to the incident neutron beam, and an additional 143 frames of the ϕ step-scan were recorded. The net time required to collect the total of 261 frames was 45 days. The dataset from the PYP crystal was processed up to 1.5-Å resolution by using the programs DENZO and SCALEPACK (37), both of which were suitably modified for the processing of neutron data. The detailed statistics of data collection are shown in Table S1.

Refinement of the Hydrogen/Deuterium Atom Positions.

Molecular refinement for the neutron crystallographic analysis was initiated by using the structural model determined by the X-ray crystallographic analysis, with the water molecules removed. The crystals used for the X-ray and neutron diffraction experiments were grown under the same crystallization conditions, suggesting that the observed molecular structures were potentially identical to each other (Table S1). Structure determination was performed by using the refinement programs X-PLOR (38) and CNS (32), in which the topology and parameter files had been specially modified for neutron crystallography. The programs Coot (34) and XtalView (39) were used to build the model. Although the positions of the heavy atoms determined by the X-ray crystallographic analysis were fixed during the refinement procedure, the positions of the hydrogen atoms, deuterium atoms, and water molecules were refined with rigid body, annealing refinement and energy minimization procedures using the neutron diffraction data. The joint method of neutron and X-ray crystallography improved nuclear density map of hydrogen and deuterium atoms, when compared with the ordinary neutron crystallography (see SI Text and Fig. S1) Although 120 water molecules are visible in the X-ray crystallographic analysis, only 73 of the water molecules can be observed in the neutron crystallographic analysis. However, the neutron crystallographic analysis revealed three types of water molecules with different types of rotational freedom (40, 41); the number of molecules of the triangular type, short ellipsoidal stick type, and spherical type are 25, 10, and 38, respectively (Table S1).

Preparation of the Figures.

PyMOL was used in the preparation of the nuclear and electron density maps superimposed on the ball-and-stick model in Figs. 1 and 2 and Fig. S1 (42).