Crystal Structure of Formate Dehydrogenase H: Catalysis Involving Mo, Molybdopterin, Selenocysteine, and an Fe4S4 Cluster

See allHide authors and affiliations

Science  28 Feb 1997:
Vol. 275, Issue 5304, pp. 1305-1308
DOI: 10.1126/science.275.5304.1305


Formate dehydrogenase H from Escherichia coli contains selenocysteine (SeCys), molybdenum, two molybdopterin guanine dinucleotide (MGD) cofactors, and an Fe4S4 cluster at the active site and catalyzes the two-electron oxidation of formate to carbon dioxide. The crystal structures of the oxidized [Mo(VI), Fe4S4(ox)] form of formate dehydrogenase H (with and without bound inhibitor) and the reduced [Mo(IV), Fe4S4(red)] form have been determined, revealing a four-domain αβ structure with the molybdenum directly coordinated to selenium and both MGD cofactors. These structures suggest a reaction mechanism that directly involves SeCys140 and His141 in proton abstraction and the molybdenum, molybdopterin, Lys44, and the Fe4S4 cluster in electron transfer.

Formate dehydrogenase H (FDHH), a 79-kD polypeptide that oxidizes formate to carbon dioxide with the release of a proton and two electrons, is a component of the anaerobic formate hydrogen lyase complex of E. coli (1). Essential to its catalytic activity are an Fe4S4 cluster, a Mo atom that is coordinated by two MGD cofactors, and a SeCys residue (24). With the recent determination of the crystal structures of three other molybdopterin (MPT)-containing enzymes (58), a functional role for the Mo-MPT cofactor has begun to emerge. However, the precise role of the active site selenium in this type of selenoenzyme and its interaction with Mo-MPT cofactors and the Fe4S4 cluster remains to be elucidated (9).

The structure of E. coli FDHH, as solved by multiple isomorphous replacement (MIR) and multiwavelength anomalous dispersion (MAD) methods (Table 1), consists of four αβ domains (Fig. 1). The first domain (residues 1 to 60, 448 to 476, and 499 to 540), comprising two small antiparallel β sheets and four helices, coordinates the Fe4S4 cluster just below the protein surface. The MGD-binding domains II (residues 61 to 135, 336 to 447, and 477 to 498) and III (residues 136 to 335) are each αβα sandwiches with overall topologies that closely resemble the classical dinucleotide-binding fold (10). A marked twofold pseudosymmetry is observed relating the central portions of domains II and III. Despite their low sequence homology (<20% identity), the two domains can be superimposed to a root-mean-square (rms) deviation of 1.2 Å for 56 α carbons. SeCys140, an essential ligand to Mo, is located in a short loop at the NH2-terminus of domain III. The COOH-terminal domain (residues 541 to 715) consists of a six-stranded mixed β barrel and five helices.

Table 1.

Data collection, phasing, and refinement statistics for FDHH structure determination. Purification, crystallization, and cryofreezing of reduced FDHH crystals [Mo(IV), Fe4S4(red)] were performed in a nitrogen atmosphere at <1 ppm of oxygen as described (20). Crystals belong to the tetragonal space group P41212 with cell dimensions of a = b = 146.3 Å and c = 82.3 Å containing one monomer in the asymmetric unit. Crystals of FDHH in the [Mo(VI), Fe4S4(ox)] state (10) were obtained by serially washing crystals in a formate-free solution and then soaking crystals in 10 mM benzyl viologen for 30 min before freezing. Crystals of nitrite-inhibited FDHH were obtained by adding 30 mM sodium nitrite to the benzyl viologen solution during oxidation. Diffraction data were processed with DENZO and SCALEPACK (21) or R-AXIS software (22) and scaled with CCP4 programs (23). MIR phases from five derivatives [K2PtCl4, Sm(OAc)3, AuCN, TMLA, and Pb(OAc)2] together with MAD and anomalous scattering phases (AuCN and TMLA derivatives, respectively) were refined using MLPHARE (23), combined with SIGMAA (23) and subsequently improved through solvent flattening and histogram matching using the program DM (23). The resulting electron density maps were readily interpretable. Model building and refinement were carried out with the programs O (24) and X-PLOR 3.1 (25). The refinement process used all data for which ∣F∣ > 2σF. Values of II for the reduced, oxidized, and NO2-bound data sets were 26.1, 22.1, and 19.5, respectively. During the refinement, ligand bonds to Mo were only loosely restrained (1.0 kcal/Å) and the position of the selenium in reduced FDHH was fixed.

View this table:
View this table:
Fig. 1.

Topology and fold of FDHH. (A) Stereo view of the overall fold of FDHH. Domains I, II, III, and IV are shown in blue, green, yellow, and red, respectively. The MGD cofactors are shown as ball-and-stick models with MGD801 on the left and MGD802 on the right. Mo is shown as a magenta ball in the center, and the Fe4S4 cluster is shown as a ball-and-stick model in domain I. [Prepared with MOLSCRIPT and RASTER3D (26).] (B) Secondary structure schematic of FDHH. Domains I to IV are color-coded as in (A); α helices are numbered from α1 to α26, 310 helices are numbered from h1 to h4, and β strands are numbered from β1 to β23. The location of SeCys140 is denoted by an asterisk; the location of the Fe4S4 cluster is denoted by a solid blue circle.

Similar to the active sites observed in aldehyde ferredoxin oxidoreductase and dimethyl sulfoxide (DMSO) reductase, the active site Mo of FDHH is coordinated by two tightly bound MGD cofactors, each containing a tricyclic ring system with a pyran ring fused to the pterin (6, 7). The Mo di(MGD) of FDHH is ligated within the interfaces of all four domains through an extensive network of hydrogen bonds, salt bridges, and van der Waals interactions, most of which involve domains II, III, and IV (Fig. 2). Domain II exclusively coordinates MGD801 while domain III coordinates MGD802. Domain IV forms a cap over the bound pterin cofactors as it straddles domains II and III. Of the 35 residues that coordinate the Mo di(MGD) cofactor through hydrogen bonds, 23 are well conserved among the known MGD-containing formate dehydrogenases (11); the remaining 12 residues interact primarily through main chain hydrogen bonds (Fig. 2).

Fig. 2.

Schematic representation of the hydrogen-bonding network coordinating MGD801 and MGD802. Residues are color-coded according to sequence conservation among the known sequences of MGD-containing FDH enzymes (11): magenta, invariant; blue, well conserved; and green, not conserved. The eight water molecules involved in hydrogen-bonding interactions are designated by orange circles. Hydrogen bonds present in the oxidized state, the reduced state, or both states are represented by blue, red, or black dashed lines, respectively. The binding of both MGD cofactors resembles the dinucleotide binding observed in the classical dinucleotide-binding proteins (10).

In both the reduced Mo(IV) and oxidized Mo(VI) structures, Mo is ligated to the four cis-dithiolene sulfurs of the MGD cofactors and the selenium of SeCys140. The coordination geometry of Mo in formate-reduced FDHH is closely approximated by a square pyramid in which the sulfur atoms provide the four equatorial ligands and the selenium provides an axial ligand (Fig. 3) (12). The Mo atom is ∼0.4 Å above the equatorial plane in the direction of the selenium. Upon oxidation of Mo(IV) to Mo(VI), however, the pterin portion of MGD801 is rotated 27° away from the equatorial plane and 16° along its own long axis. In contrast, little movement is observed in MGD802 or in the guanine nucleotide portion of MGD801. The fifth ligand, the selenium atom of SeCys140, moves 0.9 Å closer to S12 of MGD802, resulting in a slightly longer bond to Mo (2.7 Å). The four sulfur ligand distances to Mo also increase slightly (to between 2.3 and 2.6 Å) in the oxidized state (12). Fobs - Fcalc electron density maps revealed the presence of a sixth ligand in the Mo(VI) state, giving the Mo a trigonal prismatic coordination geometry (Fig. 3). This ligand was modeled as a hydroxyl group that refined to a distance of 2.2 Å from Mo with a B factor of 16.8 Å2. The position of the SeCys140 β carbon in the oxidized state also enables a new water molecule, H2O64, to interact with the amide of His141 and the NH1 of Arg333, whose side chain moves 3.9 Å toward H2O64. Redox-induced conformational changes result in minor domain movements and minor alterations in the hydrogen-bonding network coordinating MGD801 (Fig. 2).

Fig. 3.

(left). Superposition of the Mo center of FDHH in the reduced [Mo(IV)] state (red) and oxidized [Mo(VI)] state (green). The Mo center (magenta) includes the active site residues SeCys140, His141, and Arg333 and the pterin portions (MPT) of the Mo cofactors. [Prepared with MOLSCRIPT and RASTER3D (26).] Fig. 4 (right). The nitrite-binding site. A 4σ Fobs − Fcalc electron density map calculated to 2.9 Å using phases from the uncomplexed oxidized FDHH reveals the binding of the inhibitor nitrite. [Prepared with SETOR (27).] The coordinates for the reduced, oxidized, and nitrite-bound forms of FDHH have been deposited in the Protein Data Bank (accession numbers 1aa6, 1fdo, and 1fdi, respectively).

The binding of the nitrite inhibitor (2, 3) to the oxidized FDHH is clearly visible in the initial Fobs - Fcalc electron density map as a crescent-shaped 4σ peak indicating displacement of the hydroxyl ligand originally at the same position (Fig. 4). One nitrite oxygen is bound to the Mo (bond length 2.5 Å); the other is hydrogen-bonded to both the main chain amide of His141 and the side chain of Arg333, which moves 0.5 Å closer to Mo in order to accommodate this hydrogen bond. Nitrite also displaces the H2O64 observed in oxidized FDHH. Both Arg333 and His141 are strictly conserved in all Mo-dependent formate dehydrogenases (11). Apart from the nitrite-binding site, the structure of inhibitor-complexed oxidized FDHH is essentially the same as that of the benzyl viologen-oxidized enzyme. When formate is modeled into the nitrite-binding site, the α proton of formate is located less than 1.5 Å from the selenium of SeCys140, poising it for abstraction by selenium. The side chain of His141 is positioned on the other side of the selenium, opposite the formate. This putative substrate-binding site lies at the bottom of a deep crevice located between domains II and III, with Arg333 at the base of the crevice providing both a positive charge and a critical hydrogen bond for orienting the substrate for catalysis.

Kinetic experiments have suggested that the two electrons from oxidized formate leave the enzyme one at a time through a ping-pong mechanism (3). The Fe4S4 cluster located just below the protein surface in domain I provides a one-electron sink for a downstream electron acceptor. The observed reduction of the Fe4S4 cluster upon substrate binding indicates that electrons must travel from the Mo center to the Fe4S4 cluster. The most direct path between the Mo atom and the Fe4S4 cluster is through the partially conjugated ring system of MGD802, exiting through N20 and following the hydrogen bond pathway from H2O30 to the Nζ of Lys44 to the S1 of the Fe4S4 cluster (Fig. 5). Both H2O30 and the Nζ of Lys44 have well-ordered electron densities with temperature factors in the reduced form of 21.7 and 27.9 Å2, respectively. Lys44, although strictly conserved in the family of MGD-containing formate dehydrogenases (11), is completely buried in the interior of the protein and has no countercharge partner with which to interact.

Fig. 5.

The proposed reaction mechanism for FDHH. BV, benzyl viologen.

Catalysis begins with formate replacing the Mo-bound hydroxyl in the [Mo(VI), Fe4S4(ox)] state of the enzyme, presumably being stabilized and oriented by hydrogen bonding through a carbonyl oxygen of formate to both Arg333 and the amide nitrogen of His141 (Fig. 5). The subsequent oxidation of formate to carbon dioxide and the transfer of two electrons to the Mo center (Fig. 5) may occur either by a direct two-electron transfer through the oxygen of formate to Mo or by a direct hydride transfer to Mo. Upon Mo reduction, the α proton of formate is released to the nearby His141 through protonation of SeCys140. The involvement of a histidine residue is consistent with known pH dependencies in the catalytic activity of both the SeCys140 → Cys140 mutant and wild-type FDHH (13), and the protonation of His141 by the α proton of formate is supported by electron paramagnetic resonance (EPR) observations (14). The next step is to shuttle electrons from the Mo(IV) to a downstream electron acceptor through the Fe4S4 cluster. As the first electron is transferred through Lys44 to the Fe4S4 cluster, it produces an intermediate [Mo(V), Fe4S4(red)] that is easily observed in EPR experiments (14). Meanwhile, the transfer of a formate-derived proton to His141 will lead to hydrogen bond formation between the imidazole of His141 and the selenium of SeCys140 while the selenium is coordinated to Mo(V) (15). Although the nature of the in vivo electron acceptor for FDHH remains unknown, the reoxidation of the Fe4S4 can be achieved with benzyl viologen, a one-electron acceptor. Once the Fe4S4 cluster is reoxidized, a second electron can be transferred from Mo(V) to the Fe4S4 cluster and the enzyme returns to its initial state after the second oxidation of the Fe4S4 cluster. The oxidation of Mo(V) to Mo(VI) would cause the hydrogen bond between SeCys140 and His141 to break, thereby releasing the proton of His141 to solvent.

A common feature among many MPT-containing enzymes is the coupling of the redox state of Mo with the substrate oxidation-reduction process. In FDHH, the reduction of Mo(VI) to Mo(IV) profoundly affects the Mo coordination geometry and thus the conformation of MPTs. Such changes, which are also observed in model compounds (16), may represent a general feature associated with MPT-dependent Mo- and W-containing enzymes. In contrast, the incorporation of a SeCys in FDHH, as compared with incorporation of a Cys or Ser in other di(MPT)-dependent enzymes, appears to correlate with the usage of a hydroxyl ligand as opposed to sulfido or oxo ligands to Mo (7, 17, 18). This is also evident in extended x-ray absorption fine structure (EXAFS) studies of FDHH where a terminal oxo ligand to Mo is observed in a SeCys140 → Cys140 mutant but not in wild-type FDHH (19). This mutation results in a much lower initial rate of substrate oxidation, 1/300 that of the wild type (13). Thus, the choice of a SeCys, Cys, or Ser ligand to Mo may serve to fine-tune the coordination of a particular cis-ligand and hence set the substrate preference. This suggests a new role of selenium in biology, involving ligation to a metal and proton transfer during catalysis.

The combination of the MPT redox center, SeCys140, and the Fe4S4 cluster, each precisely positioned, results in an enzyme that not only catalyzes the oxidation of formate but also effectively couples the oxido-reduction to an electron acceptor in the formate hydrogen lyase complex. This suggests that the MPT moiety, in addition to providing a structural framework for Mo coordination, also functions as part of an electron transfer path and potentially as an electron sink.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
    The overall fold of Rhodobacter sphaeroides DMSO reductase is similar to that of FDHH.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
    The results of EPR studies of reduced and oxidized FDHH crystals are consistent with the presence of a [Mo(IV), Fe4S4(red)] state and a [Mo(VI), Fe4S4(ox)] state, respectively.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
View Abstract

Stay Connected to Science

Navigate This Article