Structural basis of the redox switches in the NAD+-reducing soluble [NiFe]-hydrogenase

See allHide authors and affiliations

Science  01 Sep 2017:
Vol. 357, Issue 6354, pp. 928-932
DOI: 10.1126/science.aan4497

How a hydrogenase protects its active site

Hydrogen-metabolizing organisms use an [NiFe]-hydrogenase to catalyze hydrogen oxidation. One type of [NiFe]-hydrogenase, the NAD+-reducing soluble [NiFe]-hydrogenase (SH), couples reduction of NAD+ to the oxidation of hydrogen. Shomura et al. solved the structure of SH from an H2-oxidizing bacterium in both the air-oxidized and the active reduced state. In the reduced state, the NiFe catalytic center in SH has the same ligand coordination as in other [NiFe]-hydrogenases. However, the air-oxidized active site has an unusual coordination geometry that would prevent O2 from accessing the site and so may protect against irreversible oxidation.

Science, this issue p. 928


NAD+ (oxidized form of NAD:nicotinamide adenine dinucleotide)–reducing soluble [NiFe]-hydrogenase (SH) is phylogenetically related to NADH (reduced form of NAD+):quinone oxidoreductase (complex I), but the geometrical arrangements of the subunits and Fe–S clusters are unclear. Here, we describe the crystal structures of SH in the oxidized and reduced states. The cluster arrangement is similar to that of complex I, but the subunits orientation is not, which supports the hypothesis that subunits evolved as prebuilt modules. The oxidized active site includes a six-coordinate Ni, which is unprecedented for hydrogenases, whose coordination geometry would prevent O2 from approaching. In the reduced state showing the normal active site structure without a physiological electron acceptor, the flavin mononucleotide cofactor is dissociated, which may be caused by the oxidation state change of nearby Fe–S clusters and may suppress production of reactive oxygen species.

Hydrogenases play central roles in the hydrogen metabolism of microorganisms by catalyzing H2-oxidation and H+-reduction (1). [NiFe]-hydrogenases enclose a Ni-Fe(CN)2CO complex at the active site in the large subunit, and the small subunit harbors one to three Fe–S clusters for electron transfer (2). The minimum hydrogenase catalytic unit consisting of the large and small subunits generally accompany other functional units, such as quinone reductase, proton pump, and diaphorase. The [NiFe]-hydrogenases have been classified into five groups on the basis of their function and structure. NAD+ (oxidized form of NAD:nicotinamide adenine dinucleotide)–reducing soluble [NiFe]-hydrogenases (SH), a representative of group III [NiFe]-hydrogenases, regulate redox potential by reversibly catalyzing electron transfer from NADH (reduced form of NAD+) to two protons to produce H2 (3, 4). A minimum unit of SH comprises four subunits, HoxFUYH, and all the subunits show amino acid sequence similarities with soluble subunits of NADH:quinone oxidoreductase (complex I), which generates a proton-motive force across the membrane by coupling oxidation of NADH and reduction of quinone (5). The HoxFU subcomplex in SH is responsible for the conversion between NAD+ and NADH as diaphorase, whereas the HoxHY subcomplex functions as a hydrogenase.

A draft genome sequence analysis of Hydrogenophilus thermoluteolus TH-1, a thermophilic H2-oxidizing bacterium (6), identified four structural genes, hoxFUYH, for SH. Although photosynthetic bacteria and certain species of H2-oxidizing bacteria additionally harbor hoxE and hoxI genes (79), respectively, no other structural gene has been found in the hox operon except for those of maturation factors. SH from H. thermoluteolus (Ht-SH) was isolated from aerobically cultured cells, purified, and crystallized under aerobic conditions, and its structure in the air-oxidized state was determined by means of the single-wavelength anomalous dispersion method and refined to 2.58-Å resolution (table S1). The overall structure (Fig. 1A) shows a shape-complementary interaction between the diaphorase HoxFU subcomplex (~90 kDa) and hydrogenase HoxHY subcomplex (~73 kDa), with an interface area of 2084 Å2 sharing 4.4% of the total surface of the heterotetramer. The interface comprises three intersubunit contacts—HoxU–Y, HoxF–Y, and HoxU–H—with areas of 1071, 239, and 774 Å2, respectively. No interaction between HoxF and HoxH is observed. The residues involved in the intersubcomplex interaction are moderately conserved (fig. S1). As predicted from the amino acid sequence analyses (5), the four subunits show structural similarities to the subunits of complex I. HoxF (residues 165 to 438), HoxU, HoxY, and HoxH are respectively superposed on Nqo1, Nqo3, Nqo6, and Nqo4, the subunits of complex I from Thermus thermophilus HB8 (Tt-complex I) (10, 11), with root mean square deviations (RMSDs) of 1.6, 2.1, 2.6, and 2.3 Å for 256, 174, 83, and 228 Cα atoms. In addition, the N-terminal part of HoxF (9 to 127) is superposed on Nqo2 with RMSD of 1.8 Å for 115 Cα residues, revealing that HoxF has evolved as a fusion protein of Nqo1 and Nqo2 in contrast to the other evolutionarily related diaphorases (12, 13) that consist of separate subunits corresponding to Nqo1 and Nqo2. A flavin mononucleotide (FMN) molecule is found in the same position and orientation as in Tt-complex I (Fig. 1).

Fig. 1 Structure comparison of Ht-SH and the soluble part of Tt-complex I.

(A and B) Overall structure and electron transfer pathway of Ht-SH. (C and D) Overall structure and electron transfer pathway of the soluble part of Tt-complex I (PDB ID: 4HEA). In (A) and (C), homologous subunits are represented by the same color. The numbers shown in (B) and (D) indicate the center-to-center distances between redox centers, and the numbers in parentheses indicate the edge-to-edge distances in angstroms. The redox centers in red characters in (D) are not found in Ht-SH.

In total, Ht-SH is equipped with a minimum set of Fe–S clusters for electron transfer between diaphorase and hydrogenase active sites compared with Tt-complex I, which includes two off-line Fe–S clusters (N1a and N7) and an additional subunit harboring two in-line Fe–S clusters (N6a and N6b) (supplementary text). The orientation between the HoxFU and HoxHY subcomplexes is completely different from that between the Nqo1–3 and Nqu6·4 subcomplexes (Fig. 1 and fig. S3). Despite the difference in the binding partners of HoxU/Nqo3 between Ht-SH and Tt-complex I, however, the distance between the U3 and Y1 clusters of Ht-SH and that between the N5 and N6a clusters of Tt-complex I are comparable, as are the distances between other homologous clusters (Fig. 1). These findings support the hypothesis that the subcomplexes have independently evolved and are assembled as prebuilt modules into each energy metabolism machinery (14).

The Ni–Fe active site of Ht-SH in the air-oxidized state shows unprecedented coordination geometry; the carboxy group of Glu32 binds Ni as a bidentate ligand (Fig. 2A). The model-building and the structure-refinement procedures of the active site were based on previously reported crystal structures of inactive model compounds (15). The geometry restraints used in the structure refinements were similar to the bond distances of the model compounds, except for the distance of 2.86 Å between Ni and Fe, which is comparable with that of the group I [NiFe]-hydrogenases in the oxidized states rather than the model compounds (fig. S3). Four cysteinyl thiolates coordinate the Ni–Fe cluster, as is the case in the group I [NiFe]-hydrogenases. One is the terminal ligand for Ni, and three are bridging ligands between Ni and Fe; in contrast, two are terminal, and two are bridging ligands in the group I [NiFe]-hydrogenases. The thiolate group of Cys462 in the third-bridging ligand position implies that the enzyme is in an inactive state because hydride has been shown to occupy this position in the catalytic intermediate (16). Both Ni and Fe in Ht-SH show octahedral coordination, whereas Ni shows square-pyramidal coordination in the group I [NiFe]-hydrogenases. The structure is consistent with a previous x-ray absorption spectroscopic study of the oxidized SH from Ralstonia eutropha H16 (Re-SH), in which the octahedral coordination of Ni and the presence of O-donor ligands were proposed (17).

Fig. 2 Structure comparison of the Ni–Fe active site in the air-oxidized and H2-reduced states.

(A) The air-oxidized state. (B) The H2-reduced state. Meshes represent the simulated annealing omit map contoured at 3.5σ, where nonprotein atoms and side chains of cysteines and glutamic acid were omitted in the simulated annealing refinement. (C) EPR spectra of air-oxidized (top) and H2-reduced (bottom) enzymes recorded at 77 K. Signals at g = 1.94 and 2.03 are attributed to reduced [2Fe–2S].

The air-oxidized Ht-SH showed electron paramagnetic resonance (EPR) signals (gx = 2.25 and 2.26, gy = 2.13, and gz = 2.04) attributable to Ni(III) (Fig. 2C). This result is in contrast to most previous EPR studies of SH from other species, in which no EPR signals of Ni have been observed under various oxidized conditions (1822). The g values in the present study do not match those identified for the oxidized states of group I [NiFe]-hydrogenases (23, 24), implying that the distinct configuration of the Ni–Fe active site of the air-oxidized Ht-SH is not an artifact caused by the crystallization or x-ray exposure during the diffraction data collection. In contrast to air-oxidized Ht-SH, the EPR spectrum of the H2-reduced enzyme activated by the addition of H2 and incubation at 50°C for 1 hour showed weak Ni(III)–attributable signals (gx = 2.21 and gy = 2.14; the gz signal was overlapped with other signals), with g values close to those of the Ni–C state of Re-SH and group I [NiFe]-hydrogenases (Fig. 2C), supporting the idea that Ht-SH adopts a cluster configuration similar to that of group I [NiFe]-hydrogenases in the active state and that the two states are reversible. In addition, the EPR spectrum of the H2-reduced enzyme changed upon light exposure in a similar way to that of other [NiFe]-hydrogenases (fig. S4) (24). To prepare the crystals in the active state, the air-oxidized crystals were soaked in the buffer containing 1 mM oxidized BV and saturated H2 at 0.1 MPa for 24 hours at 293 K. The reduction of BV was confirmed by the blue color around the crystals. The Ni–Fe active site of Ht-SH in the H2-reduced state shows the same coordination geometry as those of the group I [NiFe]-hydrogenases in the H2-reduced state (Fig. 2B). The carboxy group of Glu32 shows no coordination to Ni, and the thiolate group of Cys462 is the terminal ligand to Ni rather than the bridging ligand, as observed in the air-oxidized state. The distance between Ni and Fe is shortened to 2.5 Å, which is also in agreement with the previously reported values of the group I [NiFe]-hydrogenases in the active states (25).

Aside from the Ni–Fe active site, the structure comparison between the air-oxidized and H2-reduced states revealed two major differences: the displacement of the C-terminal region of HoxF (555 to 591), and the dissociation of FMN upon H2 reduction. In the air-oxidized state, the C-terminal region of HoxF extends from the interface between HoxF and HoxY to the NADH-binding site in HoxF (Fig. 3, A and B). The structure comparison with Tt-complex I in the NADH-bound form (26) reveals that the residues involved in the binding to NADH are well conserved, implying that Ht-SH binds NADH in a similar way to Tt-complex I (Fig. 3, B and C) and the C-terminal region of HoxF occupies the NADH binding site in the air-oxidized state. The interaction between Tt-complex I and NADH consists of π-stackings and a few hydrogen bonds. Similar hydrogen bonds are observed in Ht-SH in the air-oxidized state, in which Glu585 and Gln586 form hydrogen bonds with the hydroxy groups of FMN and the carboxy group of Glu319 of HoxF, respectively. It remains unclear whether the conformation is an artifact of crystallization or represents a physiological state, but the amino acid sequence alignment shows that the corresponding region of HoxF is not conserved among species (fig. S1), suggesting that the conformation of the C terminus of HoxF observed in the air-oxidized state is specific to Ht-SH. In contrast, in the H2-reduced state the C terminus of HoxF is wedged between HoxF and HoxY, resulting in the rigid-body displacements of HoxF and HoxYH, with HoxU remaining fixed (Fig. 3A). No major conformational difference in each subunit between the reduced and oxidized states has been observed, except for the interface between HoxY and HoxH, which is related to the difference in the coordination geometry of the Ni–Fe active site. In particular, Glu32, Glu56, Arg58, Ser464, and Thr467 are cooperatively displaced owing to the ion interaction between Glu56 and Arg58, steric hindrance between Glu56 and Thr467, and hydrogen bonds among these residues (Fig. 4A). Ser464 and Thr467 move together because they are in the same α-helix (463 to 467).

Fig. 3 Comparison of the FMN- and NADH-binding sites of Ht-SH and Tt-complex I.

(A) Superposition of the overall structures of Ht-SH in the air-oxidized and H2-reduced states. Only HoxU was used for the superposition. The C-terminal regions of HoxF are represented by models in different colors (air-oxidized, yellow; H2-reduced, blue). (B) Ht-SH in the air-oxidized state. (C) Ht-SH in the H2-reduced state. (D) Tt-complex I in the NADH-bound form (PDB ID: 3IAM). FMN and NADH are represented by green and yellow stick models, respectively. Another FMN molecule around HoxY that was previously proposed (28) was not found. Key residues involved in the NADH-binding in Tt-complex I, and spatially corresponding residues in Ht-SH are labeled and shown as brown stick models.

Fig. 4 Possible structural determinant for the coordination of Glu32 to Ni in Ht-SH.

(A) An enlarged view around Glu32 in Ht-SH. The air-oxidized state is shown in the same colors as in Fig. 1, and the H2-reduced state is shown in white. The gray dotted lines represent hydrogen bonds. For clarity, the Ni–Fe active site and Y1 cluster are shown only for the air-oxidized state. (B) Sequence alignments of HoxH and large subunits. The topmost four sequences are those of SH, and the bottom seven are large subunits of the group I [NiFe]-hydrogenases. From the top, the sequences shown are HoxH of SH from H. thermoluteolus TH-1 (GI: 675402517), R. eutropha H16 (GI: 32527094), Synechococcus elongatus PCC7942 (GI: 81301364), and Allochromatium vinosum (GI: 288941585); large subunits of [NiFeSe]-hydrogenases from Desulfovibrio vulgaris Hildenborough (GI: 46580327) and Desulfomicrobium baculatum (GI: 256829745); standard [NiFe]-hydrogenases from Desulfovibrio gigas (GI: 130103) and D. vulgaris Miyazaki F (GI: 218885374); and membrane-bound [NiFe]-hydrogenases from E. coli (GI: 15800894), Hydrogenovibrio marinus H110 (GI: 329564792), and R. eutropha H16 (GI: 38637670).

We assume that the conformational change is triggered by reduction and oxidation of nearby clusters. When activated, the enzyme prefers the position of Arg58 observed in the H2-reduced state because the charge of the Y1 cluster is decreased. However, when oxidized, the Ni–Fe cluster is stabilized by the coordination of Glu32. Glu32 in the hydrogen-bond network of Glu56–Ser464–Glu32 would be protonated in the H2-reduced state. The finding that the addition of H2 and oxidized BV activated and reduced the enzyme suggests that the two conformations are in equilibrium even in the crystals, allowing H2 to attack the Ni–Fe active site. Comparing the amino acid sequences of SH and the group I [NiFe]-hydrogenases reveals that Glu56 and Ser464 are specific to SH, with one exception (Fig. 4B), although it remains unclear whether the residues are a prerequisite for the redox-dependent conformational change observed in Ht-SH. Considering the low structure similarity of the HoxH/large subunit between SH and the group I [NiFe]-hydrogenases, as shown by the RMSD of ~2 Å for ~350 Cα residues, it is also assumed that the conformational change is derived from local conformational differences or deletions that give SH higher flexibility around the Ni–Fe active site compared with the group I [NiFe]-hydrogenases.

The dissociation of FMN upon reduction of the enzyme has also been reported for complex I from Escherichia coli, where the dissociation occurred only when nonphysiological electron acceptors, such as potassium ferricyanide and hexammine ruthenium, were used (27). It has been proposed that the dramatic decrease in the affinity of FMN for Tt-complex I in the reduced state is attributed to the additional negative charge of the reduced N1a cluster. If this is the case with Ht-SH, the affinity of FMN for Ht-SH would be decreased when the H2/BV (in the oxidized form) ratio is high and BV in the reduced form is abundant, similar to conditions under which the H2-reduced crystals of Ht-SH were prepared. In contrast, under normal physiological conditions, in which the NADH/H2 ratio is high and oxidation of NADH is coupled to the reduction of H+, FMN would not dissociate because the Fe–S clusters are not fully reduced during the catalytic cycle. As proposed in previous research, the dissociation of FMN suppresses the production of reactive oxygen species (ROS) because most ROS are produced at the FMN in complex I (27). Although it is uncertain whether the F1 cluster that is closest to FMN is a determinant of the dissociation of FMN, it is reasonable to speculate that the redox-dependent dissociation of FMN would suppress ROS production in highly reducing environments in the cells. In this context, the C-terminal region of HoxF in the air-oxidized state may be physiologically relevant to preventing ROS formation at the reduced FMN.

The crystal structure analysis of Ht-SH revealed two major differences between the oxidized and reduced states, which may be functionally important and attributable to the change in the oxidation state of nearby Fe–S clusters. Under aerobic conditions, the Ni–Fe active site reversibly adopts an inactive form, most likely to prevent O2 from approaching the active site, which would cause irreversible inactivation during the activation process. The reduction of the Y1 cluster triggers the conformational change of the Ni–Fe active site to the active form. When H2 is abundant and less NAD+ is available for the electron acceptor in the cells, all the Fe–S clusters should be continuously in the reduced state, causing dissociation of FMN. This would limit the risk of ROS production caused by reduced FMN.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S4

Table S1

References (2938)

References and Notes

  1. Acknowledgments: This work was supported by Japan Science and Technology Agency CREST grant JPMJCR12M4, Japan (Y.H. and S.H.), Japan Society for the Promotion of Science [Category B, grant JP25291038 (Y.H.); Challenging Exploratory Research, grant JP24657077 (Y.H.); Scientific Research on Innovative Areas, grant JP15H00945 (S.H.); Specially Promoted Research, grant 26000008 (S.O.); and Young Scientists B, grant JP16K17936 (H.T.)], and by research grants from The Mitsubishi Foundation (Y.H.) and ENEOS Hydrogen Trust Fund (Y.H.). We thank M. Habukawa, K. Hataguchi, and K. Matsumoto for their support with enzyme preparation. The synchrotron radiation experiments were performed at the BL38B1 (proposals 21013 and 25476), BL41XU (proposal 22996), BL44XU (proposals 21817, 23301, 24798, and 26479), and BL32XU [Platform Project for Supporting in Drug Discovery and Life Science Research (Platform for Drug Discovery, Informatics, and Structural Life Science) from the Japan Agency for Medical Research and Development (AMED) (project 1243)]. The charge-coupled device detector MX225-HE (Rayonix) at BL44XU was financially supported by Academia Sinica and the National Synchrotron Radiation Research Center (Taiwan, ROC). Coordinates and structure factors for the x-ray crystal structures of SH in the air-oxidized and H2-reduced states have been deposited in the Protein Data Bank (PDB) with accession codes of 5XF9 and 5XFA, respectively.
View Abstract

Stay Connected to Science


Navigate This Article