Revisiting the concept of peptide bond planarity in an iron-sulfur protein by neutron structure analysis.

The planarity of the peptide bond is important for the stability and structure formation of proteins. However, substantial distortion of peptide bonds has been reported in several high-resolution structures and computational analyses. To investigate the peptide bond planarity, including hydrogen atoms, we report a 1.2-Å resolution neutron structure of the oxidized form of high-potential iron-sulfur protein. This high-resolution neutron structure shows that the nucleus positions of the amide protons deviate from the peptide plane and shift toward the acceptors. The planarity of the H─N─C═O plane depends strongly on the pyramidalization of the nitrogen atom. Moreover, the orientation of the amide proton of Cys75 is different in the reduced and oxidized states, possibly because of the electron storage capacity of the iron-sulfur cluster.

INTRODUCTION

Peptide bonds are amide bonds that are generated by the dehydration reaction between α–amino acids in proteins. The C─N bond has a partial double-bond character because of its resonance stabilization with a separation of charge between the nitrogen and oxygen atoms (), and thus the bond is energetically most stable when the six atoms that constitute the peptide group are in the same plane. The partial double-bond character is important for resistance to hydrolysis () and forming secondary structures (–). The amide hydrogen and oxygen atoms in the peptide bond often act as donors and acceptors in hydrogen bonds, respectively. Hydrogen bonds between nitrogen and oxygen atoms are typically 2.5 to 3.5 Å, and the energy of an N─H…O═C hydrogen bond contributes about 1 to 10 kcal/mol to protein stability (–). The strength of hydrogen bonds in proteins depends on bond length, bond angle, polarity, pressure, and temperature, among other parameters (–). Despite the partial double-bond character of peptide bonds, several high-resolution x-ray structures and computational analyses have shown the distortion of the ω angle, which describes the Cα─C─N─Cα plane, where i is the residue number (–). The peptide bond is not perfectly planar in these structures, and the distortion depends on the φ angle, which describes the C─N─Cα─C plane, and the ψ angle, which describes the N─Cα─C─N plane (, , ). The deviations of the position of the amide protons from the peptide plane are not well understood because of the lack of structural data for hydrogen atoms, and only a few neutron structures and nuclear magnetic resonance (NMR) analyses have revealed the deviations of these protons (–). Neutron crystallography is a powerful method for directly detecting the position of hydrogen atoms, because the magnitudes of the coherent scattering lengths of protium and deuterium atoms are comparable to that of carbon (). Moreover, high-resolution neutron structures can determine the precise locations of hydrogen nuclei, which are different from the center of electron density because of polarization.

High-potential iron-sulfur protein (HiPIP) is responsible for electron transfer from cytochrome bc1 complex to light-harvesting–reaction center complexes (LH1-RC) in bacterial photosynthesis (). HiPIP has a cubane-like Fe4S4 cluster at its center, with a redox potential of above +300 mV (). This high potential arises from the hydrogen bonds and hydrophobic environment around the iron-sulfur cluster (–). The Fe4S4 cluster is divided into two Fe2S2 clusters; subcluster 1 (Fe1-S4-Fe2-S3) and subcluster 2 (Fe3-S1-Fe4-S2). In HiPIP, subcluster 1 is largely responsible for electron storage (). We have previously reported the x-ray crystal structures of reduced (~0.48-Å resolution) and oxidized (~0.80-Å resolution) HiPIP from a thermophilic purple bacterium, Thermochromatium tepidum (–). The ultrahigh-resolution x-ray structure at 0.48-Å resolution in the reduced state revealed large distortions of the peptide bonds around the iron-sulfur cluster (). Moreover, some electron densities of amide protons deviated from the peptide plane. Although the electron density of hydrogen atoms is detected in the high-resolution x-ray structure of HiPIP, the neutron structure is important to determine the precise locations of hydrogen nuclei to elucidate the effect of hydrogen bonds on the redox potential and reaction mechanism.

RESULTS AND DISCUSSION

X-ray/neutron joint refinement of oxidized HiPIP

We collected x-ray and neutron diffraction data of oxidized HiPIP at 0.66- and 1.2-Å resolution, respectively, from the same crystal at a cryogenic temperature of 100 K (fig. S1 and table S1). Initially, we performed the x-ray structure refinement and subsequently determined the coordinates of hydrogen atoms using neutron diffraction data (table S2). The final Rwork and Rfree are 7.41 and 8.15% for the x-ray data and 15.4 and 16.8% for the neutron data, respectively. There are no major structural differences compared with the 0.48-Å resolution structure in the reduced state (fig. S2A). These structures can be superimposed with a root mean square deviation for all the nonhydrogen polypeptide atoms of 0.62 Å. The ω angles are also identical between the two states, including the highly distorted peptide bond around the iron-sulfur cluster (fig. S2B). The CN bond lengths have a negative correlation with the CO bond lengths (fig. S2C and table S3), as observed in other high-resolution structures (). These bond lengths indicate the degree of the double-bond character of the peptide bond. In the electron and neutron scattering length (or nuclear) densities, each of the atoms, including hydrogen atoms, can be observed separately (Fig. 1A). The nuclear densities of D and Dγ1 of Thr13 around the protein surface are observed to be positive because of hydrogen (protium)–deuterium (H/D) exchange. In contrast, the nuclear density of H of Val71, which is buried in a hydrophobic core, is negative. The protonation states of ionizable residues correspond to the differences in the CO bond lengths (fig. S3). The centers of the electron and nuclear densities of hydrogen atoms are different because of the polarization (Fig. 1, B and C). The neutron diffraction data of oxidized HiPIP are high quality enough to determine the precise locations of the hydrogen nuclei.

Fig. 1.

Electron and nuclear density maps of oxidized HiPIP.

(A) Electron and nuclear density maps around the peptide bond between Pro12 and Thr13 and between Asn70 and Val71. The 2Fobs − Fcalc electron density map is shown in gray at a contour level of +4σ. The Fobs − Fcalc hydrogen omit nuclear density map is shown in green and orange at a contour level of +5σ and −5σ, respectively. (B) Difference between the center of electron density and nuclear density of the indole ring of Trp78. The Fobs − Fcalc hydrogen omit electron density map and Fobs − Fcalc hydrogen omit nuclear density map are shown in yellow (+4σ) and pink (−7.5σ), respectively. (C) Close-up view of Nε1─Hε1 and Cη2─Hη2 of Trp78. The Fobs − Fcalc hydrogen omit electron density map and Fobs − Fcalc hydrogen omit nuclear density map are shown in yellow (+4.5σ, +4σ, and +3.5σ) and pink (−8σ, −7.5σ, and −7σ), respectively.

Distortion of the amide proton

The bond lengths and angles for hydrogen atoms are generally constrained in the x-ray and neutron structural refinement as a riding model. However, the nuclear densities of the amide protons of oxidized HiPIP deviate from the peptide plane (Fig. 2A). Therefore, the coordinates of amide protons whose occupancies are greater than or equal to 0.60 of deuterium or 0.75 of protium, the nuclear density of which was clearly observed, were refined without geometry restraints. In addition, the deuterium atoms of side chains in a single conformation were also refined. The coordinates of 64 amide protons and 36 deuterium atoms on the side chains were refined without restraints. The data-to-parameter ratio of the neutron diffraction data was sufficient to refine the coordinates of the amide protons. The refined hydrogens fit well into the omit map (Fig. 2A). The hydrogen bond patterns of the refined amide protons are shown in tables S4 and S5. A hydrogen bond usually consists of a single acceptor and donor pair. However, a hydrogen bond pattern can sometimes contain more than one donor or acceptor and is called as a bifurcated hydrogen bond. To understand the effect of the bifurcated hydrogen bonds, we divided the amides into those with a clear single acceptor (table S4) and those with two acceptors near the amide proton (table S5). In single-acceptor hydrogen bonds, most of the donor-hydrogen-acceptor refined angles (∠DH…A) are greater than those of the riding model (Fig. 2B and table S4). The positions of the nuclei of the amide protons shift toward the acceptor atoms (Fig. 2A). In bifurcated hydrogen bonds, some amide protons (Ala14, Ile15, Ala32, Gly35, Trp58, and Gly60) point away from the nearest acceptor atoms (Fig. 2C and table S5). These hydrogen atoms are attracted from two directions by the electrostatic force, and the N─H bond is shorter than the ideal value of 1.02 Å obtained from neutron scattering or NMR studies (, ). The N─H bond length is critically affected by the strength of the hydrogen bond, which depends mainly on the donor-acceptor distance and angle (). The bonding energy of the bifurcated hydrogen bonds is slightly smaller than that of the single-acceptor hydrogen bond (). Therefore, it is likely that the length of the N─H bond in the bifurcated hydrogen bonds is short. Only the amide proton of Asn70 points away from the nearest acceptor atom, although there are no other acceptor atoms nearby. Sγ of Cys43 is 4.62 Å from the amide proton of Asn70 (fig. S4), which may affect the position of the amide proton of Asn70. The refined ∠DH…A of three of the four cysteine residues covalently bound to the iron-sulfur cluster (Cys43, Cys61, and Cys75) deviate substantially from the riding model (table S3). This deviation is correlated with the distortion of the ω angles caused by the Fe─Sγ bonds between the iron-sulfur cluster and the cysteine residues (fig. S2B). Although most of the N─H bond lengths are around the ideal value of 1.02 Å, some of the N─H bond lengths deviate from the ideal value (Fig. 2D). The hydrogen bonds between the amide protons and Sγ of the cysteine residues that are bound to the iron-sulfur cluster (Leu63 and Ser77) change the N─H bond length substantially (table S4). Pyramidalization at the peptide nitrogen atom is observed because of the attraction of hydrogen atoms by acceptor atoms and the distortion of the ω angle (fig. S5A). The degree of nitrogen pyramidalization (θN), which is the dihedral angle between the C─N─H and C─N─Cα planes, does not show a clear correlation with the C─N length [correlation coefficient (r) = 0.23] (fig. S5B). This result indicates that the change in the double-bond character of the peptide bond is not decisive for the nitrogen pyramidalization. In general, the planarity of the peptide bond is discussed by referring to the ω angles due to the insufficient structural data for the amide protons. In this study, we focus on the planarity of the H─N─C═O planes in the peptide bond because we refined the amide protons without geometry restraints. The dihedral angle of H─N─C═O is defined as the ω′ angle. The distributions of the ω angle and ω′ angle have σ of 6.1° and 5.7°, respectively (Fig. 3, A and B). The degree of carbon pyramidalization (θC) ranges from about −5° to 5° from the plane, whereas θN ranges from −20° to 25° (Fig. 3C). θC is much smaller because the oxygen atoms must rotate simultaneously with the Cα─C─N plane due to the sp2 hybridization of the C atom. There is a correlation between θN and the ω′ angle (r = 0.78) (Fig. 3D). In contrast, there is no clear correlation between θC and the ω′ angle (Fig. 3E). The planarity of the H─N─C═O plane depends strongly on the pyramidalization of the nitrogen atom. In other words, the attraction of amide protons by acceptor atoms determines the planarity of the H─N─C═O plane (Fig. 3F).

Fig. 2.

Distortion of the peptide bond.

(A) Deviations of hydrogen atoms from the peptide plane. The Fobs − Fcalc hydrogen omit nuclear density map is shown in magenta. The amide hydrogen atoms of riding models are shown in green. The distance between amide hydrogen and oxygen atoms are shown in black. The angles of NH…O for experimentally determined models and riding models are shown in magenta and green, respectively. (B) Histogram of the frequency of differences in donor-hydrogen acceptor angles in single-acceptor hydrogen bonds between the experimentally refined model and the riding model. (C) Deviations of hydrogen atoms from the acceptor atoms in bifurcated hydrogen bonds. (D) Histogram of the frequency of N─H bond lengths for amide protons.

Fig. 3.

Position of hydrogen and oxygen atoms in the peptide bond.

(A) Histogram of the frequency of the deviation from the plane of the ω angle. The peptide bonds including the refined amide protons are counted. (B) Histogram of the frequency of the deviation from the plane of the dihedral angle of H─N─C═O (ω′ angle). (C) Scatter plot of θC against θN. (D) Scatter plot of θN against ω′ angle. Linear regression is shown as a solid line. (E) Scatter plot of θC against the ω′ angle. (F) Side-view schematic of the peptide bond. The attraction of the amide proton by the acceptor atom shifts the H─N─C═O plane.

In arginine residues, the guanidino group can deviate from the ideal conformation (). In HiPIP, two arginine residues (Arg28 and Arg33) are present on the protein surface. In Arg28, the Nε─Cζ, Cζ─Nη1, and Cζ─Nη2 bond lengths are similar (table S6). In contrast, in Arg33, the Nε─Cζ, and Cζ─Nη1 bond lengths are shorter than the Cζ─Nη2 bond length. In addition, the Cδ atom of Arg33 deviates from the guanidinium plane, whereas the Cδ atom of Arg28 does not (fig. S6, A and B). Thus, the planarity of the guanidino group is affected by the surrounding conditions. In addition, nitrogen pyramidalization of the Nη1 and Nη2 atoms is observed (fig. S6, C and D). The hydrogen atoms in the guanidino group also shift toward the acceptor atoms. Nitrogen pyramidalization in the guanidino group has already been reported in the neutron structure of photoactive yellow protein (); thus, pyramidalization in the arginine residues may be common in proteins.

Amide protons around the iron-sulfur cluster that is involved in electron transfer

The distances and angles including hydrogen atoms around the iron-sulfur cluster are shown in table S7. In the oxidized state, the distances between the sulfur atoms and hydrogen bond donors are slightly longer. This elongation indicates that all sulfur atoms undergo a change in electronic state caused by oxidation and follows the same trend as in our previous study of the x-ray structures at 0.8-Å resolution (). In the redox protein, the hydrogen atoms that are coordinated to sulfur atoms play a major role in redox potential (, , ). The four amide protons interact with the sulfur atoms of the cysteine residues covalently bound to the Fe atoms (Fig. 4A). The two amide protons of Phe48 and Thr79 interact with the sulfur of Cys46, which is bound to subcluster 1, and the amide protons of Leu66 and Ser77 interact with the sulfur atoms of Cys61 and Cys75, respectively, which are bound to subcluster 2. The NH bond lengths of 0.97 Å of Phe48 and 1.02 Å of Thr79 are shorter than or equal to the ideal value of 1.02 Å (, ), whereas the NH bond lengths of 1.14 Å of Leu63 and 1.15 Å of Ser77 are substantially longer. Because the hydrogen bonds coordinating to the sulfur atoms are important in regulating redox potential (, ), different hydrogen bond patterns may be an important factor in the difference between subclusters.

Fig. 4.

Iron-sulfur cluster including hydrogen atoms.

(A) Hydrogen bond with Sγ of cysteine residues covalently bound to Fe atoms. N─H bond lengths and H…O distances are shown in red and black, respectively. (B) Position of the amide proton of Cys75 in the oxidized state. The distance between the amide proton of Cys75 and S3 atom is shown in black. The deviations from the C─N─Cα plane of the amide proton of Cys75 and oxygen of Trp74 are shown in blue and red, respectively. (C) Positions of the amide proton of Cys75 and oxygen of Trp74 in the reduced state [Protein Data Bank identifier (PDB ID) 5D8V]. (D) Structure of the HiPIP–LH1-RC complex around the contact surface between HiPIP and RC (PDB ID 7C52). The structural changes of the amide proton of Cys75 and the oxygen of Trp74 between the oxidized and reduced states are indicated by arrows.

In the HiPIP iron-sulfur cluster, subcluster 1, which consists of Fe1, Fe2, S3, and S4, is largely responsible for electron storage (). The position of the amide proton of Cys75 near the S3 atom is different in the reduced and oxidized states (fig. S7, A and B). This difference was also observed in our previous study (). In the oxidized state, the amide proton and oxygen atom of Trp74-Cys75 deviate by 18.8° and 11.7° from the C─N─Cα plane, respectively (Fig. 4B). The amide proton of Cys75 is not oriented toward the S3 atom and shifts toward the indole ring of Trp74 because of the peptide bond distortion. The distance and angle between the amide proton of Cys75 and the center of the five-membered ring of Trp74 are consistent with the NH-π interaction (fig. S7C). This distance and angle are reasonable for obtaining the energetic stability due to the NH-π interaction (). In contrast, in the reduced state, the amide proton of Cys75 is oriented toward the S3 atom (Fig. 4C). In the reduced state, the iron-sulfur cluster has one more electron than in the oxidized state; therefore, the amide proton of Cys75 is attracted by the S3 atom in subcluster 1. The structure of the HiPIP–LH1-RC complex has been reported (). Heme-1 of LH1-RC is distorted by the interaction of Leu63 of HiPIP (Fig. 4D). Heme-1 receives electrons directly although its redox potential is lower than that of HiPIP (). Because heme-1 is distorted by Leu63 and buried in a hydrophobic environment by complex formation, the redox potential increases, and the energy gap with HiPIP decreases (). Phe64, which is the next residue of Leu63 that interacts directly with heme-1, interacts with Val6 and Tyr19 via its side chains (Fig. 4D). Consequently, the position of the amide proton of Asn20 changes, affecting the direction of the oxygen atom of Trp74. The propagation of the structural change by the complex formation of HiPIP and LH1-RC probably leads to the orientation change of the amide proton of Cys75 that interacts with the S3 atom of the iron-sulfur cluster (Fig. 4D). The change in the electron storage capacity of the S3 atom located on the distal side of the binding surface will reduce the redox potential of HiPIP and facilitate the electron transfer from HiPIP to LH1-RC.

MATERIALS AND METHODS

Preparing the large crystal of oxidized HiPIP

HiPIP was extracted from T. tepidum and purified as reported previously (, , ). The purified HiPIP was oxidized with 10 mM potassium ferricyanide and concentrated to 18 mg/ml in 20 mM tris-HCl (pH 8.0). The crystals were obtained in a solution that was a 1:1 mixture of protein solution and reservoir solution (1.8 M ammonium sulfate and 0.1 M Na citrate buffer, pH 4.5) by the vapor diffusion method at 20°C. Large crystals for the neutron diffraction study were obtained by macroseeding using the vapor diffusion method at 20°C. The buffer of the purified sample was exchanged for the deuterium buffer (20 mM tris-DCl in D2O) by ultrafiltration. The single crystal (about 0.5 mm by 0.1 mm by 0.1 mm) was transferred to a solution consisting of a 500-μl aliquot of the purified protein solution (4.5 mg/ml) with 1.7 M deuterated ammonium sulfate (D8) and 0.1 M Na citrate buffer (pD 4.5) in D2O. The reservoir solution was a 1-ml aliquot of 1.7 M deuterated ammonium sulfate (D8) and 0.1 M Na citrate buffer (pD 4.5) in D2O. The crystals with typical dimensions of 2.0 mm by 1.0 mm by 0.5 mm were obtained within 3 to 6 months.

Neutron and x-ray diffraction data collection

The crystal was flash-cooled in a nitrogen gas stream with a cryobuffer that consisted of reservoir solution with 20% (v/v) deuterated glycerol (D8). The time-of-flight neutron diffraction data were collected at 100 K using a neutron diffractometer (iBIX) at BL03 of the Materials and Life Science Experimental Facility at Japan Proton Accelerator Research Complex (J-PARC) (Tokai, Japan) (, ). A total of 28 datasets was collected using wavelengths of 2.00 to 4.40 Å. The exposure time for each dataset was 8.5 hours at a beam power of 200 kW. The neutron diffraction data were processed using the program STARGazer (, ). The −3σ cutoff was applied to remove the strongly negative intensities. The same crystal was also used for the cryogenic x-ray diffraction data collection using an Q315r detector [Area Detector Systems Corporation (ADSC), Poway, CA] at BL5A of the Photon Factory (Tsukuba, Japan). The x-ray diffraction data were processed and scaled using the program HKL2000 (). Two datasets for high- and low-resolution regions were merged into a complete dataset.

Structure refinement

The structure of HiPIP was solved by the molecular replacement method using the oxidized HiPIP [Protein Data Bank identifier (PDB ID) 5WQQ] as a search model with the program MOLREP (). The output structures were manually improved with the program Coot (). The initial refinement of x-ray data was performed by the program Phenix () at 1.2-Å resolution. Then, the refinements with anisotropic atomic displacement parameters were performed using the program SHELXL () at 0.66-Å resolution. The coordinates and occupancies were refined with geometry restraints using the standard values in SHELXL. After the high-resolution x-ray data refinement, the neutron structure refinement was performed using the program Phenix. The refinement parameters of heavy atoms (C, N, O, S, and Fe) were fixed. The coordinates of the amide protons that had occupancies more than or equal to 0.60 that of deuterium or 0.75 that of protium and the deuterium atoms of the side chains in a single conformation were refined without geometry restraints. All figures for the molecular models were prepared using the program PyMOL (The PyMOL Molecular Graphics System, version 2.4.0a0; Schrödinger LLC).