The Three-Dimensional Structure of [NiFeSe] Hydrogenase from Desulfovibrio vulgaris Hildenborough: A Hydrogenase without a Bridging Ligand in the Active Site in Its Oxidised, “as-Isolated”

This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/copyright

Author's personal copy

The Three-Dimensional Structure of [NiFeSe]Hydrogenase from Desulfovibrio vulgarisHildenborough: A Hydrogenase without aBridging Ligand in the Active Site in Its Oxidised,“as-Isolated” State

Marta C. Marques1, Ricardo Coelho1, Antonio L. De Lacey2,Inês A. C. Pereira1 and Pedro M. Matias1⁎1Instituto de TecnologiaQuímica e Biológica,Universidade Nova de Lisboa,Apartado 127, 2781-901 Oeiras,Portugal2Instituto de Catálisis yPetroleoquímica, CSIC, c/ MarieCurie 2, 28049 Madrid, Spain

Received 29 September 2009;received in revised form3 December 2009;accepted 10 December 2009Available online21 December 2009

Hydrogen is a good energy vector, and its production from renewablesources is a requirement for its widespread use. [NiFeSe] hydrogenases(Hases) are attractive candidates for the biological production of hydrogenbecause they are capable of high production rates even in the presence ofmoderate amounts of O2, lessening the requirements for anaerobicconditions. The three-dimensional structure of the [NiFeSe] Hase fromDesulfovibrio vulgaris Hildenborough has been determined in its oxidised“as-isolated” form at 2.04-Å resolution. Remarkably, this is the firststructure of an oxidised Hase of the [NiFe] family that does not containan oxide bridging ligand at the active site. Instead, an extra sulfur atom isobserved binding Ni and Se, leading to a SeCys conformation that shieldsthe NiFe site from contact with oxygen. This structure provides severalinsights that may explain the fast activation and O2 tolerance of theseenzymes.

© 2009 Elsevier Ltd. All rights reserved.

Edited by R. Huber Keywords: hydrogenase; O2 tolerance; biohydrogen production; structure

Introduction

Hydrogen is an attractive candidate for storingand transporting energy due to its high energeticcontent and lack of contribution to the greenhouseeffect. Currently, hydrogen is producedmainly fromfossil fuels in processes that are not environmentallyfriendly. Production of hydrogen from renewablesources is of fundamental importance to its viabilityas an alternative fuel, and the use of microorganisms(biohydrogen production) is one of the mostinteresting options.1 Biological hydrogen conver-sion is carried out by hydrogenases (Hases), thenatural “molecular reactors” that are extremely

efficient catalysts for H2 production and oxidationand that are widely found in microorganisms,particularly in anaerobic bacteria.2

Both main groups of Hases, the [NiFe] Hases andthe [FeFe] Hases, contain a complex metal-basedcatalytic centre with two metals (Ni and Fe or twoFe's) coordinatedby thiolates, COandCN− ligands.3,4

The unique active-site architecture, with strongsimilarities for both kinds of Hases, is able to carryout the difficult reactions of H2 production/oxidationat ambient conditions without involving noblemetals, such as platinum. It is vital to understand inmolecular detail the reactivity of these active sites,which present a complex set of redox states, in orderto engineermore robust enzymes/microorganisms ordevelop biomimetic catalysts.A crucial issue in Hase catalysis is the reactivity of

Hases with oxygen, which leads to inactivation andintroduction of oxygen species at the active site andis a major problem for the technological use of theseenzymes. In the [FeFe] Hases, reaction with O2 leadsto irreversible damage, whereas in [NiFe] Hases, the

*Corresponding author. E-mail address:[email protected] used: Hase, hydrogenase; EPR, electron

paramagnetic resonance; FTIR, Fourier transform infraredspectroscopy; MAD, multiple-wavelength anomalousdispersion.

doi:10.1016/j.jmb.2009.12.013 J. Mol. Biol. (2010) 396, 893–907

Available online at www.sciencedirect.com

0022-2836/$ - see front matter © 2009 Elsevier Ltd. All rights reserved.

Author's personal copy

inactivation is reversible.3,5–7 Upon exposure to O2,the [NiFe] Hases from anaerobic bacteria form amixture of two inactive species known as Ni-A andNi-B.3,5,8 Nickel is in a Ni(III) oxidation state in bothspecies, so they are paramagnetic and can bedetected by electron paramagnetic resonance (EPR)spectroscopy, in contrast to Ni(II) species, which areEPR silent. In the Ni-B species, a mono-oxide ligand,most likely a hydroxide, is found bridging the Niand the Fe.9,10 There is still some controversyregarding the structure of the Ni-A species,3,8 butfurther electron density at the active site indicatesthe presence of a di-oxo peroxide species bridgingthe Ni and Fe or modification of one or more active-site cysteine thiolates to sulfenates.9–12 Ni-A requiresthe presence of oxygen to be formed, whereas Ni-Bcan also be obtained under anaerobic oxidativeconditions.7,13,14 Reduction with H2 leads to reacti-vation and removal of these oxygen species fromboth Ni-A and Ni-B,15,16 but the kinetics of thisprocess differ significantly between them. Thereactivation of Ni-A is a very slow process that cantake hours (Ni-A is thus also named the “Unready”state), whereas Ni-B takes only seconds to beconverted to a catalytically active state (and thus isalso named “Ready”).17

However, a few [NiFe] Hases display a muchhigher tolerance to O2. Several examples are foundin the chemolithotrophic Knallgas bacteria, whichcan grow aerobically by H2 oxidation.18 In themodel organism Ralstonia eutropha H16, three kindsof Hases can catalyze H2 oxidation even in thepresence of atmospheric oxygen levels.19,20 Signifi-cantly, the membrane-bound Hase from this organ-ism was recently shown to form Ni-B, but no Ni-A,upon exposure to oxygen,21 which allows for its fastreactivation.22 The other two Hases from R. eutrophaalso do not form Ni-A and can be reactivatedquickly,23,24 so it is likely that the prevention of Ni-Aformation is a requirement for the O2 tolerance of[NiFe] Hases. A downside of the O2-tolerant [NiFe]Hases is that they are much less active than typical[NiFe] Hases, particularly in H2 production due toproduct inhibition.25

[NiFeSe] Hases, another subgroup of the [NiFe]Hases, which contain a selenocysteine (SeCys) as aligand to the nickel, also display tolerance to oxygeninactivation in H2 production but present muchhigher catalytic activities.26,27 The [NiFeSe] Hasesare isolated aerobically in a Ni(II) oxidation state27–30

and do not form Ni-A or Ni-B species even afterreduction and exposure to oxygen.27,31 In the sulfate-reducing bacterium Desulfovibrio vulgaris Hildenbor-ough, whose genome encodes six Hases, the [NiFeSe]Hase becomes the major Hase expressed when Se ispresent in the growth medium, since this leads todown-regulation of the [NiFe] and [FeFe] Hases.32

This [NiFeSe] Hase contains an N-terminal hydro-phobic group in the large subunit that attaches it tothemembrane,33 a property that may be exploited forits immobilisation in bio-nanostructured electrodes.This is of great interest as it has potential applicationsfor development of fuel cells, biological water

electrolyzers and electrochemical biosensors.7,34,35

The D. vulgaris [NiFeSe] Hase is isolated aerobicallyin an inactive state but converts very quickly to anactive state upon reduction with H2 or a low redoxpotential.36 Interestingly, oxygen exposure of thereduced enzyme leads to a form of the active site thatis distinct from the “as-isolated” form but is not anoxygenated Ni(III) species.27,36

A recent electrochemical study of the [NiFeSe]Hase from Desulfomicrobium baculatum also showedthat the aerobically isolated enzyme is very rapidlyreactivated under reducing conditions, albeit at lowredox potentials.26 In addition, this Hase was shownto sustain partial H2-producing activity in thepresence of 1% O2 and to suffer less productinhibition than [NiFe] Hases. Thus, [NiFeSe] Hasesare attractive catalysts for H2 production in thepresence of low levels of O2. This was recentlydemonstrated in a system using the D. baculatumenzyme adsorbed on TiO2 nanoparticles wherevisible light-driven H2 production from water wasachieved under non-strict anaerobic conditions.35

Two properties seem to distinguish the [NiFeSe]Hases from other [NiFe] O2-tolerant Hases: nooxidised Ni(III) species (Ni-A/Ni-B) is formed inthe selenium enzymes, and their reactivation is fastbut requires a low redox potential.26,36 This suggeststhat different species may be formed upon reactionof the [NiFeSe] Hases with O2. In contrast to [NiFe]Hases, for which a wealth of structural informationis available (Refs. 3 and 4 and references therein),comparatively little information is available for the[NiFeSe] Hases, with only one three-dimensionalstructure reported so far15 for the reduced state ofthe D. baculatum [NiFeSe] Hase.In this article, we present the three-dimensional

structure of the [NiFeSe] Hase from D. vulgarisHildenborough as isolated from an aerobic prepa-ration and crystallization. This is the first structureof an oxidised Hase from the [NiFe] family where abridging oxide ligand is not present at the NiFeactive site, providing definitive evidence that theoxidised species of [NiFeSe] Hases are distinct fromthose of standard [NiFe] Hases. This structure alsoprovides important clues regarding the oxygentolerance of [NiFeSe] Hases, which allows them toproduce H2 in the presence of low levels of O2.

Results

The structure of D. vulgaris [NiFeSe] Hase

The soluble form of the [NiFeSe] Hase from D.vulgaris, lacking the first 11 residues in the largesubunit and the lipidic group,27,33 was isolated andcrystallized in aerobic conditions to obtain anoxidised form of the protein.37 The crystals containa single heterodimer in the asymmetric unit. Thedomain structure of both subunits is typical of two-subunit periplasmic-facing [NiFe{Se}] Hases.4 Thesmall subunit (chain A) contains three Fe4S4

894 3D Structure of D. vulgaris [NiFeSe] Hydrogenase

Author's personal copy

clusters, whereas the large subunit (chain B) bindsthe binuclear NiFe active site (Fig. 1a). Theproximal iron–sulfur cluster (with regard to theNiFe active site) is coordinated by the Sγ atoms ofcysteine residues 18A, 21A, 121A and 159A; themesial iron–sulfur cluster is ligated to the Sγ atomsof cysteine residues 247A, 259A, 265A and 268A,while the distal iron–sulfur cluster is bonded to theNδ1 atom of His208A and the Sγ atoms of cysteineresidues 211A, 232A and 238A. This coordinationscheme is identical with that found in the D.baculatum [NiFeSe] Hase15 and with that found inthe [NiFe] Hases from sulfate-reducing bacteria,with the notable difference that the mesial iron–sulfur cluster is an Fe3S4 centre in the [NiFe]Hases.3,4 The NiFe active site is coordinated bytwo cysteine residues, 78B and 492B, formingthiolate bridges between the two metals, while theNi atom is further ligated by Cys75B andSeCys489B and the Fe atom has the typical oneCO and two CN− diatomic ligands (Fig. 1b).

The active site

A first conspicuous difference relative to standard[NiFe] Hases is the absence of a bridging oxideligand between the Ni and the Fe (Fig. 1b). Thepresence of such ligand species is always observedin these Hases3 and corresponds to the inactive Ni(III) states that are not detected in [NiFeSe] Hases.The absence of the oxide ligand is accompanied by arelatively short distance between the Ni and Fe

atoms (2.5 versus 2.7–2.9 Å in [NiFe] Hases), whichprobably changes very little upon enzyme reduction(e.g., 2.53 Å in the fully reduced D. baculatum[NiFeSe] Hase).The electron density corresponding to SeCys489B

terminally attached to the Ni atom could not beinterpreted in terms of a single conformation of thisresidue (Fig. 2b). After several unsatisfactoryattempts, also involving refinement of anisotropicthermal motion parameters for the Se atom inSeCys489B, the electron density was interpreted ascorresponding to three conformations: SeCys ispartly persulfurated [S-SeCys, 3-(sulfonylselanyl)-L-alanine] in two of them (Fig. 2b and c), whereas itis not persulfurated in the third (Fig. 2d). Persulfura-tion was chosen in favor of oxidation due togeometric and crystallographic considerations: theSeγ–Xδ bond distances (2.13 and 2.15 Å) and the Cβ–Seγ–Xδ bond angles (102.1° and 103.6°) are closer tothe typical values for X=S (2.18–2.20 Å, 101.3°–105.9°) than those for X=O (1.87 Å, 98.1°) as seenfrom a search of the Cambridge StructuralDatabase;38 also, with X=O, its isotropic thermalmotion parameter became abnormally low incomparison with that of nearby atoms, an indicationof X being assigned an atomic number too low. Thisdisorder scheme was refined with occupancies of0.70, 0.15 and 0.15 based on achieving similar valuesfor the refined isotropic thermal displacementparameters of the Seγ and Sδ atoms and yielded an|Fo|−|Fc| electron density map without anysignificant positive or negative features (Fig. 2d).

Fig. 1. The structure of D. vulgaris [NiFeSe] Hase. (a) Cα tube diagram representation of the overall structure, with theco-factors represented as spheres; the small subunit (chain A) is shown in gold, whereas the large subunit (chain B) isshown in dark cyan. Atom colors are light blue for carbon, blue for nitrogen, red for oxygen, yellow for sulfur, brown foriron, cyan for nickel and green for Cl−. (b) Detailed view of the [NiFeSe] active site. The NiFe(CN)2(CO) cluster isrepresented as spheres, and the cysteine and selenocysteine residues that bind it to the protein chain of the large subunitare drawn in ball-and-stick representation. Only themajor conformer of S-SeCys489B is shown. The coloring scheme is thesame as that in (a), with the selenium atom shown in purple.

8953D Structure of D. vulgaris [NiFeSe] Hydrogenase

Author's personal copy

Fig. 2 (legend on next page)

896 3D Structure of D. vulgaris [NiFeSe] Hydrogenase

Author's personal copy

The three conformers for S-SeCys489B observed arerepresented in Fig. 3. In both S-SeCys489B con-formers, the positions of the extra Sδ atom are verynearly the same, with the difference being theposition of the Seγ atom. In the predominantconformer (I), both Seγ and Sδ atoms coordinate Ni(2.34 and 2.13 Å, respectively), whereas in the minorS-SeCys489B conformer (II), only the Sδ atom isbonded to Ni (2.14 Å). In the SeCys489B minorconformer (III), Seγ binds Ni (2.22 Å). Figure 4 alsoillustrates this, with the help of the anomalousFourier map calculated using the peak data set. Thepeak corresponding to S-SeCys489B Seγ is clearlyelongated, and a hint as to the sulfur chemicalnature of its terminal Sδ atom is also given by theshape of this peak. Conformer I is likely tocorrespond to its oxidised state. Conformer III issimilar to that in the fully reduced structure of thehomologous D. baculatum Hase.15 Conformer IIprobably corresponds to an intermediate enzymestate between the other two conformers, which isvery likely unique to [NiFeSe] Hases.We initially suspected that the active site of the

[NiFeSe] D. vulgaris Hase might have been partiallyreduced due to photoelectrons39 produced by theintense X-ray beam at the Diamond Light Source.However, refinement of the structure against apreviously measured37 2.4-Å in-house data set(results not shown) produced identical results (i.e.,three conformations for SeCys489, two of whichwere persulfurated and had similar occupationfactors). Therefore, since both crystals were obtainedfrom the same protein batch, we concluded that thisheterogeneity most likely results from the proteinpreparation.Another noteworthy feature of the active site of

the as-isolated D. vulgaris [NiFeSe] Hase is that theterminal Cys75B is present in a doubly oxidised Sγ

(cysteine-S-dioxide) state but still forms a bond to

the Ni atom (2.27 Å). This interpretation stems fromthe observation of a wide electron density shapeassociated with the side chain of Cys75B coupledwith a large thermal displacement parameter atCys75B Sγ and two strong symmetrical peaks in the|Fo|−|Fc| electron density map located on eitherside of the Sγ atom (Fig. 2a). Refinement ofanisotropic thermal motion parameters for Cys75BSγ attenuated these features without eliminatingthem, and neither did a disorder scheme solelybased on Cys75B Sγ. Cysteine oxidation at theactive site has been previously reported for severalhigh-resolution crystal structures of [NiFe] Hasesfrom D. vulgaris Miyazaki F9 and Desulfovibriofructosovorans.3,10 However, it usually involves asingle oxygen atom at either the bridging cysteinecorresponding to Cys78B or the terminal cysteineequivalent to SeCys489B in D. vulgaris [NiFeSe]Hase. In the reported structure of the Ni-A state of[NiFe] Hase from D. vulgaris Miyazaki F,9 bothcysteines are oxidised. To our knowledge, the kindof cysteine modification observed here has neverbeen encountered before in Hases or other proteins.However, sulfonyl groups bound to carbon and atransition metal (R-CX2SO2M) are found in small-molecule compounds, as revealed in a search of theCambridge Structural Database:38 of the 123 com-pounds found with similar sulfur coordination andbond geometry, 26 have Ni as the transition metal.

The proximal iron–sulfur cluster

Another modification of the [NiFeSe] Hase co-factors that was noticed during refinement occurs atthe proximal iron–sulfur cluster in the small subunit.The isotropic thermal motion parameter of the Featom bonded to Cys21A refined to a value signif-icantly higher (33.3 Å2) than that of its neighboringatoms (17.2–23.6 Å2). Also, a negative peak was

Fig. 3. S-SeCys489B conformers in the active site of D. vulgaris [NiFeSe] Hase.

Fig. 2. Modelling the active site of D. vulgaris [NiFeSe] Hase. Stereoviews showing the 2|Fo|−|Fc| and |Fo|−|Fc|electron density maps at several refinement stages overlaid onto Cα tube diagram representations of the large subunit,with the NiFe(CN)2(CO) cluster and the side-chain atoms of the cysteine and selenocysteine residues that bind it to theprotein chain drawn in ball-and-stick representation. The coloring scheme is the same as that in Fig. 1. The 2|Fo|−|Fc|maps are represented as chicken wire, drawn at the 1σ (map r.m.s.) level and shown in orange, while the |Fo|−|Fc|maps are represented as semitransparent surfaces and drawn at the 3σ (green) and −3σ (red) levels. (a) After initialrefinement assuming unmodified Cys75A and SeCys489B. (b) After the second refinement with doubly oxidised Cys75Aand single conformation for S-SeCys489B. (c) After the third round of refinement with a second conformer added for S-SeCys489B. (d) Final electron density maps with doubly oxidised Cys75A, two S-SeCys489B conformers and a thirdSeCys489B conformer.

8973D Structure of D. vulgaris [NiFeSe] Hydrogenase

Author's personal copy

observed near this Fe atom in the |Fo|−|Fc|electron density map and positive peaks could beseen near the side chain of Cys21A (Fig. 5a). Thiswas interpreted as evidence of partial oxidation ofthis iron–sulfur cluster to Fe4S3O3, whereby this Featom becomes bonded to three oxygen atoms, one ofwhich binds Cys21A. The best refinement results interms of residual peaks in the |Fo|−|Fc| electrondensity map were obtained by postulating a 40%oxidation for this iron–sulfur cluster (Fig. 5b).However, it was not possible to satisfactorily refinean alternate conformation for the side chain ofCys21A, possibly due to the data resolution and theoverlap between atoms from both forms of theproximal iron–sulfur cluster, and a residual peak inthe |Fo|−|Fc| electron density map remained atthe end of the refinement (Fig. 5b).This interpretation of the electron density for this

cluster is also supported by the anomalous Fouriermaps calculated using the final phases from therefinement and the anomalous differences from thepeak and remote data sets as coefficients (Fig. 4b). A

similar oxidation pattern of the proximal iron–sulfurcluster was observed in the structure refinementagainst the in-house data set (results not shown).This process is quite surprising since the proximaliron–sulfur cluster is deeply buried inside theprotein core, although close to the active site, andthe actual oxidation process is at present unknown.This kind of modification of the proximal iron–sulfur cluster has only been previously observed inthe three-dimensional structure of the [NiFe] Hasefrom D. desulfuricans ATCC 2777440 and is likely toresult from O2 attack during aerobic purification ofthe enzyme since this modification was no longerobserved in later structural studies using anaerobi-cally purified protein from D. desulfuricans (ourunpublished results).

Comparison with Fourier transform infraredspectroscopy results

Fourier transform infrared spectroscopy (FTIR)characterization of D. vulgaris [NiFeSe] Hase

Fig. 4. Anomalous electron density maps at the active site and the proximal iron–sulfur cluster of D. vulgaris [NiFeSe]Hase. Stereoviews showing the anomalous Fourier maps calculated using the phases from the final refinement and theanomalous differences from the peak (1.7244 Å, orange) and remote (1.0332 Å, green) wavelengths overlaid onto Cα tubediagram representations of the protein chains,with theNiFe(CN)2(CO) cluster [in (a)], the proximal iron–sulfur cluster [in (b)]and the side-chain atoms of the cysteine and selenocysteine residues that bind them to the protein chain drawn in ball-and-stick representation. Bothmaps are drawn at the 3σ (map r.m.s.) level. The protein coloring scheme is the same as that in Fig. 1.

898 3D Structure of D. vulgaris [NiFeSe] Hydrogenase

Author's personal copy

showed that the active site is quite heterogeneousas two isoforms could be detected for each redoxstate of the enzyme.36 Figure 6 shows the spectrumfor the Hase isolated aerobically. Two bandscorresponding to the terminal CO ligand of eachisoform are observed at 1939 and 1904 cm−1. Thelowest frequency band has considerably higherintensity than the other one; thus, it may beassigned to the predominant S-SeCys484B conform-er I observed in the crystallographic structure. Thebidendate coordination of both S and Se to the Niatom could explain this quite low frequency(1904 cm− 1) for the stretching vibration of aterminal CO ligand. The additional coordinationby Se, which is less electronegative than S, shouldincrease the π-donating effect from its d orbitals tothe 2pπ⁎ orbital of the CO ligand.41 In addition, theCO band at 1939 cm−1 may be assigned to theminor S-SeCys489B conformer II because it haslower intensity and its frequency is similar to thefrequencies of standard NiFe Hases in theiroxidised states.5 The latter only have S coordinationto the Ni, as observed for the minor S-SeCys484Bconformer in this work. Concerning the minor

conformer III, SeCys484B, its CO band may overlapwith that of the major conformer I.

Additional ligands

In addition to the iron–sulfur clusters and the NiFeactive site, [NiFe{Se}] Hases normally contain anadditional metal site in the large subunit, identifiedas Mg2+ in [NiFe] Hases and Fe2+/3+ in the [NiFeSe]Hase from D. baculatum. In the as-isolated structureof D. vulgaris [NiFeSe] Hase, and similarly to thoseother enzymes, this metal ion is octahedrallycoordinated by three water molecules, the carbox-ylate Oɛ2 of Glu56B, the carbonyl O of Ile441B andimidazolate Nɛ2 of the C-terminal His495B. Anom-alous electron density maps, calculated using theanomalous differences for peak and remote data setsas coefficients and phases obtained from the finalmodel retarded by π/2, show significant peaks atthe metal position (Fig. 4a). However, structurerefinement with that metal position assigned toFe2+/3+ yielded an isotropic thermal motion param-eter for this atom significantly higher than that forthe surrounding atoms, as well as a negative peak at

Fig. 5. The native and oxidised forms of the proximal Fe4S4 cluster. Stereoviews showing the 2|Fo|−|Fc| and|Fo|−|Fc| electron density maps at two refinement stages overlaid onto Cα tube diagram representations of the largesubunit, with the proximal iron–sulfur cluster and the side-chain atoms of the cysteine and selenocysteine residues thatbind it to the protein chain drawn in ball-and-stick representation. The coloring scheme is the same as that in Fig. 1. The2|Fo|−|Fc| maps are represented as chicken wire, drawn at the 1σ (map r.m.s.) level and shown in orange, while the|Fo|−|Fc| maps are represented as semitransparent surfaces and drawn at the 3σ (green) and −3σ (red) levels. (a)After initial refinement assuming unmodified Fe4S4. (b) After final refinement assuming 60% Fe4S4 and 40% Fe4S3O3.

8993D Structure of D. vulgaris [NiFeSe] Hydrogenase

Author's personal copy

its position in the residual |Fo|−|Fc| electrondensity map. While the coordination distances tothe water molecules and protein atoms wererestrained during refinement, a test refinementwithout restraints gave Fe2+/3+ coordination dis-tances not inconsistent with typical Fe2+/3+…N(2.16 Å to His495B Nɛ2) and Fe2+,3+…O (2.09 Å toGlu56B Oɛ2 and 2.13 Å to Ile441B O) coordinationvalues. Therefore, we assumed this metal site to bepartially depleted and refined the Fe2+/3+ site withan occupation factor of 0.75, which lowered itsisotropic thermal parameter to values consistentwith those of its neighboring atoms and cleared anyresidual peaks above 3 r.m.s.d. units in the residual|Fo|−|Fc| electron density map. However, thisapparent depletion could be an artifact of theanomalous dispersion properties of this iron atom.The peak data were used in the structure refinementbecause they had the best quality statistics, and fromTable 1, the estimated Δf′ at the peak energy is−4.12e−. This term represents almost 15% of the totalatomic number of iron and adds to the real part of itsatomic scattering factor, lowering it and possiblycausing an apparent lower occupation for the atom.Nevertheless, at 2.04-Å resolution, this effect was notnoticeable for the other iron atoms in the structure.A second ligand species present in the crystal

structure of as-isolated D. vulgaris [NiFeSe] Hase islocated close to the NiFe active site and coordinatedby the Oγ hydroxyl of Thr80B (3.07 Å), the peptidenitrogen of Ala81B (3.25 Å) and the amide Nδ2 ofAsn113B (3.14 Å). There is no ambiguity in theassignment of the side-chain conformation of thisasparagine residue since its Nδ2 atom is hydrogen-

bonded to the carbonyl oxygen of Val77B (2.97 Å),whereas its Oδ1 atom is in turn hydrogen-bonded toNɛ2 from His173B (2.87 Å). In the structure of the[NiFeSe] Hase from D. baculatum,15 this ligand wasrefined as a H2S molecule. However, in ourstructure, it might also be a Cl− ion from the Tris–HCl crystallization buffer. These two species areisoelectronic, although the refined isotropic thermalmotion parameter is slightly lower for H2S (21.5 Å2)

Fig. 6. FTIR spectrum of 0.12 mM D. vulgaris [NiFeSe]Hase as isolated aerobically.

Table 1.Data collection, processing and phase refinementstatistics

Peak Inflection Remote

Crystallographic processingDiamond Light

Source beamlineI04

Detector ADSC Q315 CCDWavelength (Å) 1.7244 1.7434 1.0332Data processing XDSSpace group P21Unit cell parametersa (Å) 60.60 60.64 60.80b (Å) 91.22 91.28 91.49c (Å) 66.75 66.73 66.90β (°) 101.73 101.68 101.66Resolution (Å) 19.78–2.04

(2.16–2.04)19.78–2.06(2.19–2.06)

19.82–2.07(2.19–2.07)

No. of observations 231,058(28,237)

147,057(18,313)

152,907(19,171)

Unique reflectionsa 86,597(12,124)

80,025(10,834)

81,880(11,936)

Completeness (%)a 96.4 (83.8) 92.6 (78.1) 95.2 (86.8)Rmerge (%)a,b 6.1 (17.5) 4.6 (14.8) 5.1 (31.8)⟨I/σ(I)⟩ 12.6 (4.8) 11.4 (4.0) 10.3 (2.0)Rmeas (%)c 7.4 (21.9) 6.3 (19.9) 7.0 (43.3)Zd 1Estimated Vm 2.1Estimated solvent

content (%)41.0

Phasing statisticsΔf″ 6.04e 3.37e 2.29f

Δf′ −4.12e −11.37e 0.20f

Acentric FOM 0.516Centric FOM 0.383Acentric phasing

power— 1.091 0.582

Centric phasingpower

— 1.005 0.491

Anomalousphasing power

1.558 0.892 0.350

Acentric Rcullis — 0.593 0.824Centric Rcullis — 0.538 0.791Anomalous Rcullis 0.704 0.858 0.972Overall |E2|

correlationg0.699

FOM after finalDM rung

0.877

a Considering Bijvoet pairs as separate observations.b Rmerge=merging R-factor: (∑hkl∑i|Ii(hkl)− ⟨I(hkl)⟩|)/(∑hkl∑iI

(hkl))×100.c Rmeas=redundancy independent R-factor: ∑hkl[N/(N−1)]1/2

∑i|Ii(hkl)− ⟨I(hkl)⟩|/(∑hkl∑iIi(hkl))×100.42 For each unique Bragg

reflection with indices (hkl), Ii is the ith observation of its intensityand N is its multiplicity.

d No. of molecules in the asymmetric unit according to theMatthews coefficient.43

e Estimated with CHOOCH.44f From the tables presented by Sasaki.45g From the SHARP density modification procedure

(SOLOMON followed by final DM run).

900 3D Structure of D. vulgaris [NiFeSe] Hydrogenase

Author's personal copy

than for Cl− (24.8 Å2). In addition, the Δf″ values forboth elements at 1.73 Å are quite similar (0.69e− for Sand 0.87e− for Cl),45 and thus the peak heightobserved at this position in an anomalous differenceFourier map (Fig. 3a) is not conclusive. The onlyindication favoring its interpretation as a chlorideion comes from the coordinating distances, whichare slightly larger in our structure than thecorresponding values in the D. baculatum [NiFeSe]Hase: 2.80 Å to Thr75L Oγ and 2.98 Å to Asn108LNδ2, consistent with the larger ionic radius of thechloride ion versus the atomic radius of sulfur.Therefore, the present data are insufficient to

conclusively determine the chemical nature of thisspecies in the crystal structure of as-isolated D.vulgaris [NiFeSe] Hase.

Insights into oxygen tolerance

Similar to all other [NiFe{Se}] Hases, the as-isolated structure of D. vulgaris [NiFeSe] Hasecontains a large hydrophobic tunnel leading outfrom the NiFe active site and branching into threetunnels as it approaches the protein surface. Up tothe first branching point, the tunnel is defined by theinterface between the small and large subunits.

Fig. 7. Hydrophobic tunnels in the structure of [NiFeSe] Hase from D. vulgaris Hildenborough and oxygen tolerance.Molecular surface and Cα tube diagram representations of the overall structure, with the co-factors shown in ball-and-stick representation or as spheres; the small subunit (chain A) is shown in gold, whereas the large subunit (chain B) isshown in dark cyan. Atom colors are as those in Fig. 1. From top to bottom: (a) major S-SeCys conformer, representing theoxidised state; (b) minor S-SeCys conformer; and (c) minor SeCys conformer, representing the fully reduced state. Fromleft to right: MSMS46 molecular surfaces calculated with probe radii of 1.4, 1.2 and 1.0 Å, respectively. The surfaces arecolored according to their respective contributing active-site atoms: purple for selenium, red for oxygen, light blue forcarbon and cyan for nickel.

9013D Structure of D. vulgaris [NiFeSe] Hydrogenase

Author's personal copy

When this tunnel is analysed using molecularsurfaces calculated with decreasing probe radii(Fig. 7), it can be seen that the conformations of S-SeCys489B in conformers I and II (Fig. 7a and b)have the effect of shielding the Ni atom in active site,blocking its access by, for example, O2. Only in themore reduced conformer III (Fig. 7c) does the side-chain conformation of SeCys489B expose the Niatom. Interestingly, the extra sulfur atom in S-SeCys489B does not seem to play a direct role in thisshielding. This situation is totally unlike whathappens in [NiFe] Hases, where the enzyme in theoxidised state has the bridging oxide ligand and theside-chain conformation of the cysteine residueequivalent to SeCys489B remains the same in bothoxidised and reduced forms of the enzyme, thusallowing the NiFe active site to be exposed toreaction with O2 at that position.3,47 On the otherhand, Fig. 7 also illustrates that an O2 molecule caninteract with Cys75B Sγ, accounting for the modifi-cation of that residue as cysteine-S-dioxide.

Discussion

The structure of the as-isolated [NiFeSe] Hasefrom D. vulgaris has revealed several interestingfeatures relevant to its biological and spectroscopicproperties. A striking property of the active site isthe absence of (hydr)oxo or (hydro)peroxo bridgingligands, which are invariably found at the active siteof the oxidised [NiFe] Hases3 and are responsible forthe EPR Ni-A and Ni-B signals of the inactive states.In the D. vulgaris [NiFeSe] Hase, the selenocysteineligand to Ni, SeCys489B, is seen to be partiallypersulfurated and in three side-chain conformationswith relative contributions of 70:15:15. In the majorS-SeCys conformer I, corresponding to the oxidisedstate of the enzyme, both Se and S atoms are bondedto the Ni atom in the binuclear metal centre, thusmaking it 5-fold coordinated. A 5- or 6-foldcoordination of Ni in oxidised [NiFeSe] Hases hasbeen previously proposed based on extended X-rayabsorption fine structure studies.48 In the minor S-SeCys conformer II, only the sulfur atom is bondedto Ni, and we believe this corresponds to anintermediate state between the oxidised conformerI and the more reduced conformer III, which, todate, has never been described or characterized. Inconformer III, the extra sulfur atom is absent and theSe atom is bonded to Ni, similarly to the fullyreduced structure of D. baculatum [NiFeSe] Hase,15and so this conformation likely corresponds to theconformation in the fully reduced state ofD. vulgaris[NiFeSe] Hase. These structural features at the activesite of the enzyme were not an artifact of photoelec-tron reduction caused by the intense synchrotronX-ray beam because similar results were obtainedfrom a diffraction data set measured in-house. How-ever, it seems unlikely that a small but significantpart of the enzyme would remain in a reduced orsemireduced state for the several weeks it takes topurify the protein and obtain the crystals used in the

diffraction data measurements, especially when thenearby Cys75B, which is coordinated to Ni, is fullyoxidised. It is more plausible that, despite oxidationof the NiFe active site, a minor proportion of theprotein still retains the NiSe conformation of thereduced state (conformer III).Based on these observations, we propose that

oxidation of the reduced [NiFeSe] Hase involvesoxidation of Cys75B, which is accompanied withbinding of an exogenous sulfur atom to the Se andNi atoms, together with a change in the side-chainconformation of SeCys489B to yield the oxidisedconformation I. This active-site rearrangement maybe kinetically or thermodynamically restrained,allowing for a relatively stable intermediate confor-mation II. This proposal is supported by theobservation of a binding place for a monatomic orsmall molecular species near the active site (ca 7 Åaway from the S atom of S-SeCys489), which wasrefined as H2S in the fully reduced structure of D.baculatum [NiFeSe] Hase.15 In the D. vulgaris[NiFeSe] Hase structure, it can equally well be Cl−

from the Tris–HCl buffer used in crystallization.Although the chemical identification of this speciescannot be unequivocally made with our presentdata, we propose that this binding site is the storageplace for the H2S/HS− released upon enzymereduction, ready to re-form the S-SeCys once theenzyme becomes reoxidised and inactive. If thispicture is correct, then this site should be indeedoccupied by H2S/HS− in the reduced (active)structure of D. baculatum [NiFeSe] Hase and by Cl−

in our oxidised structure, with the chloride ion fromthe crystallization buffer being displaced by H2S/HS− once the enzyme is reduced and becomesactive. Work aimed at the structure determination oftheD. vulgaris [NiFeSe] Hase in its fully reduced andreoxidised forms, which could further clarify thismatter, is underway.SinceD. vulgaris andD. baculatum are both sulfate-

reducing bacteria whose metabolism produces largeamounts of sulfide, it is not too surprising that aH2S/HS− species is found in [NiFeSe] Hase.Interaction of the active site in [NiFe] Hases withsulfide has been previously reported. The [NiFe]Hase from Allochromatium vinosum, also a sulfur-metabolising bacterium, is reported to be occasion-ally isolated in an EPR-silent state displaying a COband in FTIR at 1909 cm−1 and CN− bands at 2066and 2057 cm−1.49 This form of the enzyme can beprepared by adding excess Na2S to enzyme in theactive Ni-C state, followed by anaerobic oxidationwith ferricyanide, suggesting that the EPR-silentform of the enzyme contained an extra sulfur speciesat the active site. A cyclic voltammetry studyreported that reaction of the [NiFe] Hases withsulfide is dependent on the redox potential and thatthe Ni-A and Ni-B species do not react withsulfide.50 The sulfide ligand is displaced by reduc-tion and is only retained if the enzyme is firstoxidised. The sulfide adduct can also react furtherwith oxygen to form a species that is very difficult toreactivate. These properties indicate a different

902 3D Structure of D. vulgaris [NiFeSe] Hydrogenase

Author's personal copy

behaviour of [NiFe] and [NiFeSe] Hases and suggestthat in the former enzymes the sulfur species bindsin a similar bridging position as the oxide species inNi-A/Ni-B.Our structural analysis shows that the Se atom

of S-SeCys489B in its conformers I and II caneffectively shield the Ni atom (and by consequencethe bridging position) from access by small molec-ular species, such as O2, conferring to this enzyme asignificant degree of O2 tolerance. In both oxidisedand reduced forms of [NiFe] Hases, the sulfur atomof the equivalent cysteine residue is always locatedin a position comparable with that of Se atom in thereduced form of D. vulgaris [NiFeSe] Hase (confor-mation III) and thus does not block the access ofsmall diatomic ligands, such as O2 and CO, to the Niatom and the bridging position. In contrast, inconformers I and II of S-SeCys489B in D. vulgaris[NiFeSe] Hase, the Se atom is located in completelydifferent positions, corresponding to the other tworotamers of the SeCys side chain (±120° away, aboutthe Cα–Cβ bond). These positions, rather than thelarger atomic radius of Se compared with that of S in[NiFe] Hases, combined with a shorter distancebetween the Ni and Fe atoms in the binuclear metalcentre (which is also related to the absence of abridging ligand) seem to be the main structuralfactors responsible for the fast reactivation in D.vulgaris [NiFeSe] Hase and quite likely other[NiFeSe] Hases as well. Thus, binding of the extraS atom to the active site, leading to a differentposition of the Se atom, effectively acts as a switch toprevent direct reaction of O2 at the bridging positionof the NiFe site, which forms the inactive Ni(III)states.However, this protection of the bridging position

of the NiFe site does not preclude oxidation at othersites of the enzyme as observed by the presence ofthe terminal, Ni-binding Cys75B as a cysteine-S-dioxide and the proximal Fe4S4 cluster that is ca 40%oxidised to Fe4S3O3. The presence of differentoxidised species in the [NiFeSe] Hase agrees withthe fact that the reactivation of [NiFeSe] Hases fromD. vulgaris36 and D. baculatum26 occurs at lowerpotentials than [NiFe] Hases, indicating that, ratherthan (hydr)oxo or (hydro)peroxo bridging ligands,different species are formed in [NiFeSe] Hases uponO2 attack. While an oxidised proximal Fe4S4 clusterhas been previously observed in the structure of D.desulfuricans ATCC 27774,40 a cysteine-S-dioxide isunprecedented. Since the oxidised state of [NiFeSe]Hases is EPR silent, the Cys75B and SeCys489Bmodifications may contribute to the stabilisation ofNi(II). The structure of the reduced form of thisenzyme will clarify whether the oxidation of Cys75Bresults from O2 attack or is an essential structuralmodification of an active-site ligand. The ability ofthe proximal iron–sulfur cluster to become revers-ibly oxidised upon O2 attack may also constitute amechanism for protecting the active site. In themembrane Hase from R. eutropha, peculiar spectro-scopic features also suggest a modification of theproximal iron–sulfur cluster, which could be similar

to that observed here, and led the authors to suggestthat this cluster is functional in preventing theformation of the Ni-A state.21 The oxidised speciesin [NiFeSe] Hase are apparently rapidly reduced.The previous FTIR studies showed that oxidation ofthe reduced enzyme leads to a form of the enzymethat is different from the as-isolated form,36 suggest-ing that some of the processes associated withoxidation (possibly binding of the extra S atom and/or oxidation of the Fe4S4 cluster) may have kineticbarriers. Moreover, in the [NiFeSe] Hase from D.baculatum, two oxidised inactivated species thatreactivate at completely different potentials weredetected.26 The low-potential species is formed onlyin the presence of O2, whereas the minor high-potential one can also be formed after anaerobicinactivation. Based on our observations, the high-potential species may correlate with binding of the Satom at the active site and the low-potential speciesmay correlate with introduction of the oxygenspecies at Cys75B and/or the proximal FeS cluster.In conclusion, the structure reported herein

provides crucial insights into the remarkable prop-erties of O2 tolerance exhibited by [NiFeSe] Hasesand reveals that important differences are presentbetween the oxidised states of these and standard[NiFe] Hases. Two novel conformations of the activesite, in which an exogenous sulfur atom is bindingthe Ni and Se atoms and where a bridging ligand isnot present even though the enzyme is oxidised, aredescribed. This sulfur atom causes a change in theside-chain conformation of SeCys489B that effec-tively blocks access of O2 to Ni. The lack of theoxygen bridging ligand is a likely explanation forthe faster activation of oxidised [NiFeSe] Hases incomparison with oxidised [NiFe] Hases. Theseresults highlight the need for additional structuralstudies to further characterize the [NiFeSe] Hases,since these are the most promising class of Hases forthe biological production of H2.

Experimental Procedures

Data collection and structure determination

Pure soluble form of [NiFeSe] Hase from D. vulgarisHildenborough was obtained and crystallized aerobicallyas previously described.37 A three-wavelength multiple-wavelength anomalous dispersion (MAD) data set to ca2.0-Å resolution was measured at the iron absorption edgeK from a flash-cooled crystal on Diamond Light Sourcebeamline I04. The peak and inflection point wavelengthswere chosen to maximize the anomalous and dispersivesignals from an X-ray fluorescence scan near the edgeusing CHOOCH.44 The remote wavelength was chosenaround 1.0 Å to minimize absorption effects andmaximizedispersive differences during phasing. Diffraction imageswere processed with the XDS Program Package.51 Asummary of the data collection statistics is presented inTable 1.The three-dimensional structure was solved using the

MAD method. One Hase heterodimer was predicted to bepresent in the asymmetric unit, and therefore 13 Fe sites

9033D Structure of D. vulgaris [NiFeSe] Hydrogenase

Author's personal copy

were expected. Using the HKL2MAP graphical userinterface,52 we analysed the MAD data set with SHELXC(G. M. Sheldrick, personal communication), determinedthe heavy-atom substructure with SHELXD53 and solvedthe phase problem with SHELXE.54 SHELXD found manypossible solutions out of 100 trials, with the best having acorrelation coefficient of 33.8% and containing 16 possibleiron sites in the asymmetric unit of the crystal structure.However, the discrimination between the correct and theinverted substructures with SHELXE was not clear andthe electron density map was not of sufficient quality toallow interpretation and model building. The top 11 ironsites (with occupation factors greater than 0.5) locatedwith SHELXD were then input to a maximum-likelihoodheavy-atom parameter refinement using SHARP.55 TheSHARP calculations allowed the location of the tworemaining sites. Density modification was performedwith SOLOMON56 and DM.57 Automated model buildingwith ARP/wARP58 produced a nearly complete model,with 690 residues out of the expected 770 built and dockedinto sequence, with R=0.180 and Rfree=0.250. Rfree

59 wasbased on a random 5% sample of the reflection data usedthroughout model building and structure refinement. Theremaining residues (except for a few disordered N-terminal residues in both subunits) and the Fe4S4 andNiFe co-factors were modelled with COOT.60

Structure refinement

The structure was refined against the 2.04-Å peak dataset in a maximum-likelihood procedure with REFMAC,61

as implemented through the CCP4i graphics userinterface.62 In the final stages, a translation–libration–screw rigid-body refinement63 was carried out prior torestrained refinement of atomic positions and isotropicthermal motion parameters. One rigid body was definedfor each of the two subunits in the heterodimer, includingtheir respective co-factors: the three Fe4S4 clusters in thesmall subunit (chain A) and the NiFe active site in thelarge subunit (chain B). Electron density was observed forthe aliphatic chains of four sulfobetaine 3-12 detergentmolecules in two hydrophobic surface pockets, and thevisible atoms were included in the refinement. Solventmolecules were located with ARP/wARP64 and byinspection 2|Fo|−|Fc| and |Fo|−|Fc| electron densitymaps with COOT and included in the refinement. OneFe2+/3+ ion with partial occupancy (0.75) and one Cl− ionwere also included in the refinement. The refinementprocedure gave a well-refined model, with R=0.144 andRfree=0.201. In the final model, continuous electrondensity is present for residues 7–283 (C-terminal) in thesmall subunit and for residues 15–495 (C-terminal) in thelarge subunit. The final refinement statistics are listed inTable 2. In total, 24 side-chain protein atoms in ninesurface Lys and Glu residues could not be modelledsatisfactorily into the electron density and were given azero occupancy factor. The side chain of Cys75B wasrefined with a doubly oxidised Sγ atom (cysteine-S-dioxide), the side chain of SeCys489B was refined as amostly persulfurated selenocysteine with 3-fold disorder(with populations of 70%, 15% and 15%) and theproximal Fe4S4 cluster was refined as partly oxidised(40%) to Fe4Se3O3. The structure was analysed withPROCHECK,65 and its stereochemical quality parameterswere within their respective confidence intervals. Onlytwo non-glycine and non-proline protein residues werelocated outside the most favored regions of a Ramachan-dran φ,ϕ plot66: Ala263A and His185B. However, their

unusual φ,ϕ conformation is supported by their well-defined electron density. Figures 1 through 3, 5 and 7were prepared with DINO†.

Accession numbers

The final refined coordinates and the structure factorshave been deposited with the Protein Data Bank67 withaccession numbers 2wpn and r2wpnsf.

Acknowledgements

This work was carried out with the support of theDiamond Light Source. The research leading tothese results received funding from the EuropeanCommunity's Seventh Framework Programme(FP7/2007-2013) under grant agreement no.226716, from Fundação para a Ciência e Tecnologia(Ministério da Ciência, Tecnologia e Ensino Supe-rior, Lisboa, Portugal) and FEDER Program underresearch grant no. PTDC/BIA-PRO/70429/2006,from Ministerio de Ciencia e Innovacion (Madrid,Spain) under research grant no. CTQ2006-12097and from Conselho de Reitores das UniversidadesPortuguesas and Ministerio de Educación y Cienciathrough a Luso–Spanish Joint Action. We thank theI04 beamline staff, in particular Dr. MichaelMrosek, for providing help and technical advice.We also thank João Carita for growing the bacterialcells and Isabel Pacheco for helping with the proteinpurification.

Table 2. Refinement statistics

Data set Diamond Light Source I04 peak

Resolution limits (Å) 19.8–2.04No. of reflections in working

set/test set42,026/2223

Rwork/Rfree 0.144/0.201No. of atomsProtein 5832Ligand/ion 44Water 331Detergent (partial chains) 44Average B-factors (Å2)Protein main chain/side chain 33.4/36.4Ligand/ion 29.1Water 23.6Detergent 41.8r.m.s.d. values from ideal valuesBond lengths (Å) 0.017Bond angles (°) 1.75

Rfree is calculated from a random sample containing 5% of thetotal number of independent reflections measured. Average B-factors were calculated from equivalent isotropic B-values,including the translation–libration–screw contribution for theprotein atoms.

†http://www.dino3d.org

904 3D Structure of D. vulgaris [NiFeSe] Hydrogenase

Author's personal copy

References

1. Levin, D. B., Pitt, L. & Love, M. (2004). Biohydrogenproduction: prospects and limitations to practicalapplication. Int. J. Hydrogen Energy, 29, 173–185.

2. Vignais, P. M. & Billoud, B. (2007). Occurrence,classification, and biological function of hydroge-nases: an overview. Chem. Rev. 107, 4206–4272.

3. Fontecilla-Camps, J. C., Volbeda, A., Cavazza, C. &Nicolet, Y. (2007). Structure/function relationships of[NiFe]- and [FeFe]-hydrogenases. Chem. Rev. 107,4273–4303.

4. Matias, P. M., Pereira, I. A., Soares, C. M. & Carrondo,M. A. (2005). Sulphate respiration from hydrogen inDesulfovibrio bacteria: a structural biology overview.Prog. Biophys. Mol. Biol. 89, 292–329.

5. De Lacey, A. L., Fernandez, V. M., Rousset, M. &Cammack, R. (2007). Activation and inactivation ofhydrogenase function and the catalytic cycle: spectro-electrochemical studies. Chem. Rev. 107, 4304–4330.

6. Ghirardi, M. L., Posewitz, M. C., Maness, P. C.,Dubini, A., Yu, J. & Seibert, M. (2007). Hydro-genases and hydrogen photoproduction in oxygenicphotosynthetic organisms. Annu. Rev. Plant Biol. 58,71–91.

7. Vincent, K. A., Parkin, A. & Armstrong, F. A. (2007).Investigating and exploiting the electrocatalytic prop-erties of hydrogenases. Chem. Rev. 107, 4366–4413.

8. Lubitz, W., Reijerse, E. & van Gastel, M. (2007).[NiFe] and [FeFe] hydrogenases studied by advancedmagnetic resonance techniques. Chem. Rev. 107,4331–4365.

9. Ogata, H., Hirota, S., Nakahara, A., Komori, H.,Shibata, N., Kato, T. et al. (2005). Activation process of[NiFe] hydrogenase elucidated by high-resolutionX-ray analyses: conversion of the ready to the unreadystate. Structure, 13, 1635–1642.

10. Volbeda, A., Martin, L., Cavazza, C., Matho, M.,Faber, B. W., Roseboom, W. et al. (2005). Structuraldifferences between the ready and unready oxidizedstates of [NiFe] hydrogenases. J. Biol. Inorg. Chem. 10,239–249.

11. Lamle, S. E., Albracht, S. P. & Armstrong, F. A. (2004).Electrochemical potential-step investigations of theaerobic interconversions of [NiFe]-hydrogenase fromAllochromatium vinosum: insights into the puzzlingdifference between unready and ready oxidizedinactive states. J. Am. Chem. Soc. 126, 14899–14909.

12. van Gastel, M., Stein, M., Brecht, M., Schroder, O.,Lendzian, F., Bittl, R. et al. (2006). A single-crystalENDOR and density functional theory study of theoxidized states of the [NiFe] hydrogenase fromDesulfovibrio vulgaris Miyazaki F. J. Biol. Inorg. Chem.11, 41–51.

13. de Lacey, A. L., Hatchikian, E. C., Volbeda, A., Frey,M., Fontecilla-Camps, J. C. & Fernandez, V. M. (1997).Infrared-spectroelectrochemical characterization ofthe [NiFe] hydrogenase of Desulfovibrio gigas. J. Am.Chem. Soc. 119, 7181–7189.

14. Jones, A. K., Lamle, S. E., Pershad,H. R., Vincent, K. A.,Albracht, S. P. & Armstrong, F. A. (2003). Enzymeelectrokinetics: electrochemical studies of the anaero-bic interconversions between active and inactive statesof Allochromatium vinosum [NiFe]-hydrogenase. J. Am.Chem. Soc. 125, 8505–8514.

15. Garcin, E., Vernede, X., Hatchikian, E. C., Volbeda, A.,Frey, M. & Fontecilla-Camps, J. C. (1999). The crystalstructure of a reduced [NiFeSe] hydrogenase provides

an image of the activated catalytic center. Structure, 7,557–566.

16. Higuchi, Y., Ogata, H., Miki, K., Yasuoka, N. & Yagi,T. (1999). Removal of the bridging ligand atom at theNi–Fe active site of [NiFe] hydrogenase upon reduc-tion with H2, as revealed by X-ray structure analysis at1.4 Å resolution. Structure, 7, 549–556.

17. Fernandez, V. M., Hatchikian, E. C., Patil, D. S. &Cammack, R. (1986). ESR-detectable nickel and iron–sulphur centres in relation to the reversible activationof Desulfovibrio gigas hydrogenase. Biochim. Biophys.Acta, Gen. Subj. 883, 145–154.

18. Friedrich, B. & Schwartz, E. (1993). Molecular biologyof hydrogen utilization in aerobic chemolithotrophs.Annu. Rev. Microbiol. 47, 351–383.

19. Burgdorf, T., Lenz, O., Buhrke, T., van der Linden, E.,Jones, A. K., Albracht, S. P. & Friedrich, B. (2005).[NiFe]-hydrogenases of Ralstonia eutropha H16: mod-ular enzymes for oxygen-tolerant biological hydrogenoxidation. J. Mol. Microbiol. Biotechnol. 10, 181–196.

20. Vincent, K. A., Cracknell, J. A., Lenz, O., Zebger, I.,Friedrich, B. & Armstrong, F. A. (2005). Electrocata-lytic hydrogen oxidation by an enzyme at high carbonmonoxide or oxygen levels. Proc. Natl Acad. Sci. USA,102, 16951–16954.

21. Saggu, M., Zebger, I., Ludwig, M., Lenz, O., Friedrich,B., Hildebrandt, P. & Lendzian, F. (2009). Spectro-scopic insights into the oxygen-tolerant membrane-associated [NiFe] hydrogenase of Ralstonia eutrophaH16. J. Biol. Chem. 284, 16264–16276.

22. Ludwig, M., Cracknell, J. A., Vincent, K. A., Arm-strong, F. A. & Lenz, O. (2009). Oxygen-tolerant H2oxidation by membrane-bound [NiFe] hydrogenasesof Ralstonia species. Coping with low level H2 in air.J. Biol. Chem. 284, 465–477.

23. Buhrke, T., Lenz, O., Krauss, N. & Friedrich, B. (2005).Oxygen tolerance of the H2-sensing [NiFe] hydroge-nase from Ralstonia eutropha H16 is based on limitedaccess of oxygen to the active site. J. Biol. Chem. 280,23791–23796.

24. Burgdorf, T., Loscher, S., Liebisch, P., Van der Linden,E., Galander, M., Lendzian, F. et al. (2005). Structuraland oxidation-state changes at its nonstandard Ni–Fesite during activation of the NAD-reducing hydrog-enase from Ralstonia eutropha detected by X-rayabsorption, EPR, and FTIR spectroscopy. J. Am.Chem. Soc. 127, 576–592.

25. Goldet, G., Wait, A. F., Cracknell, J. A., Vincent, K. A.,Ludwig, M., Lenz, O. et al. (2008). Hydrogenproduction under aerobic conditions by membrane-bound hydrogenases from Ralstonia species. J. Am.Chem. Soc. 130, 11106–11113.

26. Parkin, A., Goldet, G., Cavazza, C., Fontecilla-Camps,J. C. & Armstrong, F. A. (2008). The difference a Semakes? Oxygen-tolerant hydrogen production by the[NiFeSe]-hydrogenase from Desulfomicrobium bacula-tum. J. Am. Chem. Soc. 130, 13410–13416.

27. Valente, F. M., Oliveira, A. S., Gnadt, N., Pacheco, I.,Coelho, A. V., Xavier, A. V. et al. (2005). Hydrogenasesin Desulfovibrio vulgaris Hildenborough: structuraland physiologic characterisation of the membrane-bound [NiFeSe] hydrogenase. J. Biol. Inorg. Chem. 10,667–682.

28. Pereira, A. S., Franco, R., Feio, M. J., Pinto, C.,Lampreia, J., Reis, M. A. et al. (1996). Characterizationof representative enzymes from a sulfate reducingbacterium implicated in the corrosion of steel.Biochem. Biophys. Res. Commun. 221, 414–421.

29. Teixeira, M., Fauque, G., Moura, I., Lespinat, P. A.,

9053D Structure of D. vulgaris [NiFeSe] Hydrogenase

Author's personal copy

Berlier, Y., Prickril, B. et al. (1987).Nickel–[iron–sulfur]–selenium-containing hydrogenases from Desulfovibriobaculatus (DSM 1743)—redox centers and catalyticproperties. Eur. J. Biochem. 167, 47–58.

30. Wang, C. P., Franco, R., Moura, J. J., Moura, I. & Day,E. P. (1992). The nickel site in active Desulfovibriobaculatus [NiFeSe] hydrogenase is diamagnetic. Multi-field saturation magnetization measurement of thespin state of Ni(II). J. Biol. Chem. 267, 7378–7380.

31. Medina, M., Hatchikian, E. C. & Cammack, R. (1996).Studies of light-induced nickel EPR signals in hy-drogenase: comparison of enzymes with and withoutselenium. Biochim. Biophys. Acta, Bioenerg. 1275,227–236.

32. Valente, F. M., Almeida, C. C., Pacheco, I., Carita, J.,Saraiva, L. M. & Pereira, I. A. (2006). Selenium isinvolved in regulation of periplasmic hydrogenasegene expression in Desulfovibrio vulgaris Hildenbor-ough. J. Bacteriol. 188, 3228–3235.

33. Valente, F. M., Pereira, P. M., Venceslau, S. S., Regalla,M., Coelho, A. V. & Pereira, I. A. (2007). The [NiFeSe]hydrogenase from Desulfovibrio vulgaris Hildenbor-ough is a bacterial lipoprotein lacking a typicallipoprotein signal peptide. FEBS Lett. 581, 3341–3344.

34. Cheng, S. & Logan, B. E. (2007). Sustainable andefficient biohydrogen production via electrohydro-genesis. Proc. Natl Acad. Sci. USA, 104, 18871–18873.

35. Reisner, E., Fontecilla-Camps, J. C. & Armstrong, F. A.(2009). Catalytic electrochemistry of a [NiFeSe]-hy-drogenase on TiO2 and demonstration of its suitabilityfor visible-light driven H2 production. Chem. Commun.45, 550–552.

36. De Lacey, A. L., Gutierrez-Sanchez, C., Fernandez, V.M., Pacheco, I. & Pereira, I. A. (2008). FTIR spectro-electrochemical characterization of the Ni–Fe–Sehydrogenase from Desulfovibrio vulgaris Hildenbor-ough. J. Biol. Inorg. Chem. 13, 1315–1320.

37. Marques, M., Coelho, R., Pereira, I. A. C. & Matias,P. M. (2009). Purification, crystallization and prelim-inary crystallographic analysis of the [NiFeSe]hydrogenase from Desulfovibrio vulgaris Hildenbor-ough. Acta Crystallogr., Sect F: Struct. Biol. Cryst.Commun. 65, 920–922.

38. Allen, F. H., Bellard, S., Brice, M. D., Cartwright, B.,Doubleday, A., Higgs, H. et al. (1979). CambridgeStructural Database. Acta Crystallogr., Sect. B: Struct.Sci. 35, 2331–2339.

39. Carugo, O. &Djinovic Carugo, K. (2005). When X-raysmodify the protein structure: radiation damage atwork. Trends Biochem. Sci. 30, 213–219.

40. Matias, P. M., Soares, C. M., Saraiva, L. M., Coelho, R.,Morais, J., Le Gall, J. & Carrondo, M. A. (2001). [NiFe]hydrogenase from Desulfovibrio desulfuricans ATCC27774: gene sequencing, three-dimensional structuredetermination and refinement at 1.8 Å and modellingstudies of its interaction with the tetrahaem cyto-chrome c3. J. Biol. Inorg. Chem. 6, 63–81.

41. Nakamoto, K. (1997). Complexes of carbon monoxideand carbon dioxide. In Infrared and Raman Spectra ofInorganic and Coordination Compounds (Part B),5th edit., pp. 126–148, John Wiley & Sons, NewYork, NY.

42. Diederichs, K. & Karplus, P. A. (1997). Improved R-factors for diffraction data analysis in macromolecularcrystallography. Nat. Struct. Biol. 4, 269–275.

43. Matthews, B. W. (1968). Solvent content of proteincrystals. J. Mol. Biol. 33, 491–497.

44. Evans, G. & Pettifer, R. F. (2001). CHOOCH: aprogram for deriving anomalous-scattering factors

from X-ray fluorescence spectra. J. Appl. Crystallogr.34, 82–86.

45. Sasaki, S. (1989). Numerical tables of anomalousscattering factors calculated by the Cromer andLiberman method. pp. 1–136, National Laboratoryfor High Energy Physics, Tsukuba, Japan.

46. Sanner, M. F., Olson, A. J. & Spehner, J. C. (1996).Reduced surface: an efficient way to compute molec-ular surfaces. Biopolymers, 38, 305–320.

47. van der Zwaan, J. W., Coremans, J. M. C. C., Bouwens,E. C. M. & Albracht, S. P. J. (1990). Effect of 17O2 and13CO on EPR spectra of nickel in hydrogenase fromChromatium vinosum. Biochim. Biophys. Acta, ProteinStruct. Mol. Enzymol. 1041, 101–110.

48. Eidsness, M. K., Scott, R. A., Prickril, B. C., Dervarta-nian, D. V., Legall, J., Moura, I. et al. (1989). Evidencefor selenocysteine coordination to the active-sitenickel in the [NiFeSe] hydrogenases fromDesulfovibriobaculatus. Proc. Natl Acad. Sci. USA, 86, 147–151.

49. Bleijlevens, B., van Broekhuizen, F. A., De Lacey, A. L.,Roseboom, W., Fernandez, V. M. & Albracht, S. P. J.(2004). The activation of the [NiFe]-hydrogenase fromAllochromatium vinosum. An infrared spectro-electro-chemical study. J. Biol. Inorg. Chem. 9, 743–752.

50. Vincent, K. A., Belsey, N. A., Lubitz, W. & Armstrong,F. A. (2006). Rapid and reversible reactions of [NiFe]-hydrogenases with sulfide. J. Am. Chem. Soc. 128,7448–7449.

51. Kabsch, W. (1993). Automatic processing of rotationdiffraction data from crystals of initially unknownsymmetry and cell constants. J. Appl. Crystallogr. 26,795–800.

52. Pape, T. & Schneider, T. R. (2004). HKL2MAP: agraphical user interface for macromolecular phasingwith SHELX programs. J. Appl. Crystallogr. 37,843–844.

53. Schneider, T. R. & Sheldrick, G. M. (2002). Substruc-ture solution with SHELXD. Acta Crystallogr., Sect. D:Biol. Crystallogr. 58, 1772–1779.

54. Sheldrick, G. M. (2002). Macromolecular phasing withSHELXE. Z. Kristallogr. 217, 644–650.

55. La Fortelle, E. D. & Bricogne, G. (1997). Maximum-likelihood heavy-atom parameter refinement for themultiple isomorphous replacement and multiwave-length anomalous diffraction methods. MethodsEnzymol. 276, 472–494.

56. Abrahams, J. P. & Leslie, A. G. W. (1996). Methodsused in the structure determination of bovine mito-chondrial F1 ATPase. Acta Crystallogr., Sect. D: Biol.Crystallogr. 52, 30–42.

57. Cowtan, K. (1994). ‘dm’: an automated procedure forphase improvement by density modification.pp. 34–38, Daresbury Laboratory, Warrington, UK.

58. Perrakis, A., Morris, R. & Lamzin, V. S. (1999).Automated protein model building combined withiterative structure refinement. Nat. Struct. Biol. 6,458–463.

59. Brünger, A. T. (1992). Free R value: a novel statisticalquantity for assessing the accuracy of crystal struc-tures. Nature, 355, 472–474.

60. Emsley, P. & Cowtan, K. (2004). Coot: model-buildingtools for molecular graphics. Acta Crystallogr., Sect. D:Biol. Crystallogr. 60, 2126–2132.

61. Murshudov, G. N., Vagin, A. A. & Dodson, E. J.(1997). Refinement of macromolecular structures bythe maximum-likelihood method. Acta Crystallogr.,Sect. D: Biol. Crystallogr. 53, 240–255.

62. Potterton, E., Briggs, P., Turkenburg, M. & Dodson, E.(2003). A graphical user interface to the CCP4 program

906 3D Structure of D. vulgaris [NiFeSe] Hydrogenase

Author's personal copy

suite. Acta Crystallogr., Sect. D: Biol. Crystallogr. 59,1131–1137.

63. Winn, M. D., Isupov, M. N. & Murshudov, G. N.(2001). Use of TLS parameters to model aniso-tropic displacements in macromolecular refine-ment. Acta Crystallogr., Sect. D: Biol. Crystallogr.57, 122–133.

64. Lamzin, V. S. & Wilson, K. S. (1993). Automatedrefinement of protein models. Acta Crystallogr., Sect.D: Biol. Crystallogr. 49, 129–147.

65. Laskowski, R. A., MacArthur, M. W., Moss, D. S. &Thornton, J. M. (1993). PROCHECK—a program tocheck the stereochemical quality of proteins. J. Appl.Crystallogr. 26, 283–291.

66. Ramachandran, G. N. & Sasisekharan, V. (1968).Conformation of polypeptides and proteins. Adv.Protein Chem. 23, 283–438.

67. Berman, H. M., Westbrook, J., Feng, Z., Gilliland, G.,Bhat, T. N., Weissig, H. et al. (2000). The Protein DataBank. Nucleic Acids Res. 28, 235–242.

9073D Structure of D. vulgaris [NiFeSe] Hydrogenase