The Crystal Structure of the [NiFe] Hydrogenase

The Crystal Structure of the [NiFe] Hydrogenasevinosum: Characterization of the Oxidized Enzyme Ni-A State Max-Planck-Institut für Bioanorganische Chemie, Stiftstrasse 34–36, D-45470 Mül

The Crystal Structure of the [NiFe] Hydrogenase

vinosum: Characterization of the Oxidized Enzyme

(Ni-A State)

Max-Planck-Institut für

Bioanorganische Chemie,

Stiftstrasse 34–36, D-45470

Mülheim an der Ruhr, Germany

Received 12 May 2010;

received in revised form

19 July 2010;

accepted 20 July 2010

Available online

29 July 2010

The crystal structure of the membrane-associated [NiFe] hydrogenase from Allochromatium vinosum has been determined to 2.1 Å resolution Electron paramagnetic resonance (EPR) and Fourier transform infrared spectroscopy

on dissolved crystals showed that it is present in the Ni-A state (N90%) The structure of the A vinosum [NiFe] hydrogenase shows significant similarities with [NiFe] hydrogenase structures derived from Desulfovibrio species The amino acid sequence identity is ∼50% The bimetallic [NiFe] active site is located in the large subunit of the heterodimer and possesses three diatomic non-protein ligands coordinated to the Fe (two CN−, one CO) Ni is bound to the protein backbone via four cysteine thiolates; two of them also bridge the two metals One of the bridging cysteines (Cys64) exhibits a modified thiolate in part of the sample A mono-oxo bridging ligand was assigned between the metal ions of the catalytic center This is in contrast to a proposal for Desulfovibrio sp hydrogenases that show a di-oxo species in this position for the Ni-A state The additional metal site located

in the large subunit appears to be a Mg2+ ion Three iron–sulfur clusters were found in the small subunit that forms the electron transfer chain connecting the catalytic site with the molecular surface The calculated anomalous Fourier map indicates a distorted proximal iron–sulfur cluster in part of the crystals This altered proximal cluster is supposed to be paramagnetic and is exchange coupled to the Ni3+ ion and the medial [Fe3S4]+ cluster that are both EPR active (S = 1/2 species) This finding

of a modified proximal cluster in the [NiFe] hydrogenase might explain the observation of split EPR signals that are occasionally detected in the oxidized state of membrane-bound [NiFe] hydrogenases as from

A vinosum

© 2010 Elsevier Ltd All rights reserved Edited by R Huber

Keywords: [NiFe] hydrogenase; Allochromatium vinosum; photosynthetic purple-sulfur bacterium; iron–sulfur cluster; Ni-A state

*Corresponding authors E-mail addresses:ogata@mpi-muelheim.mpg.de;lubitz@mpi-muelheim.mpg.de

Present address: P Kellers, Department of Photochemistry and Molecular Science, The Ångström Laboratories, Uppsala University, SE-751 20 Uppsala, Sweden

Abbreviations used: EPR, electron paramagnetic resonance; FTIR, Fourier transform infrared; ICP-OES, inductively coupled plasma with optical atomic emission spectrometry; MAD, multiwavelength anomalous dispersion; D vulgaris H, Desulfovibrio vulgaris Hildenborough; D vulgaris MF, Desulfovibrio vulgaris Miyazaki F; PDB, Protein Data Bank

Available online at www.sciencedirect.com

0022-2836/$ - see front matter © 2010 Elsevier Ltd All rights reserved.

Hydrogenases catalyze the reversible oxidation of

molecular hydrogen They play a key role in the

energy metabolism of various microorganisms.1

Hydrogenases can be divided into three groups

according to the metal content of their active site:

[NiFe], [FeFe], and [Fe] hydrogenases.2–7An

alterna-tive distinction of hydrogenases can be made based

on their function: uptake (hydrogen splitting),

pro-duction (hydrogen generation), bidirectional, and

sensory/regulatory hydrogenases.8 They appear as

either membrane-bound/associated or as soluble

enzymes.1 Several organisms possess not only one

but often more and even different hydrogenases

participating in a variety of metabolic processes

comprising variable catalytic activity levels, oxygen,

and temperature tolerance.9,10 These differences

determine the eligibility of particular hydrogenases

for special biotechnological applications

To date, four crystal structures of O2-sensitive

hydrogen uptake [NiFe] hydrogenases from

sulfate-reducing Desulfovibrio species have been determined

but not a single one from photosynthetic bacteria or

other microorganisms.2,11–18Their crystal structures

revealed that [NiFe] hydrogenases are composed of

two subunits forming a heterodimer by hydrophobic

interaction These contain three iron–sulfur clusters

and a bimetallic active site, located in the small and

the large subunit, respectively The catalytic center is

composed of a Ni and an Fe atom, which are bridged

by two thiolates of cysteines from the protein

backbone Three diatomic non-protein ligands are

bound to the Fe They were assigned as two CN−and

one CO based on Fourier transform infrared (FTIR)

results.19,20 Furthermore, two other cysteine

thio-lates coordinate the Ni in a terminal fashion A third

bridging ligand between these metals is temporarily

present depending on the respective apparent

oxida-tion state.21,22Additionally, there is another metal site

existent in the large subunit It was identified as Mg2+

and is assumed to participate in mediating the proton

transfer through the protein matrix The substrate

(H2) crosses the large subunit via adequate

hydro-phobic channels.14,21,23,24 It has also suggested that

the Mg2+ion functions as recognition site during the

maturation steps.25

In addition, two crystal structures of [NiFeSe]

hydrogenases are also available These hydrogenases

were derived from proteins isolated from the

sulfate-reducing bacteria Desulfomicrobium baculatum26 and

Desulfovibrio vulgaris Hildenborough (D vulgaris

H).27 [NiFeSe] hydrogenases form a subdivision of

[NiFe] hydrogenases that exhibit a selenocysteine

instead of a cysteine involved in the coordination of

the nickel atom All iron–sulfur clusters of [NiFeSe]

hydrogenases are of the [Fe4S4] type, which is in

contrast to [NiFe] hydrogenases that possess both

[FeS] and [FeS] clusters Finally, the extra metal, a

Mg2+ in [NiFe] hydrogenases, is in this case most likely replaced by an Fe2+/3+ion.26,27

The [NiFe] enzyme passes through various redox states during its catalytic cycle The “as-isolated” protein is in many cases a mixture of the two most oxidized but inactive states Ni-A (unready) and Ni-B (ready) Both are paramagnetic (Ni3+, S = 1/2) but exhibit different g-values (gx= 2.32, gy= 2.24, and

gz= 2.01 for the Ni-A state and gx= 2.33, gy= 2.16, and

gz= 2.01 for the Ni-B state) in electron paramagnetic resonance (EPR)28and different activation rates.29,30 Crystallographic studies on these states of the hydrogenases from D vulgaris Miyazaki F (D vulgaris MF), Desulfovibrio gigas, and Desulfovibrio desulfuricans indicated that the third bridging ligand

is probably not the same in Ni-A and Ni-B.16,17In the case of the Ni-B state, a hydroxide (OH−) was identified,31 whereas in the Ni-A state, a di-oxo species in the form of a hydroperoxy (OOH−) ligand has been discussed.16,17Furthermore, a modification

of cysteine residues was observed in some hydro-genases in their oxidized states In the case of the [NiFe] hydrogenases from D vulgaris MF and Desulfovibrio fructosovorans, oxidation of the bridging cysteine residues Cys84 and Cys75, respectively, was found.16,17 The [NiFeSe] hydrogenase from D vulgaris H showed a terminal cysteine residue (Cys75) binding two oxygen atoms.27

The photosynthetic purple-sulfur bacterium Allochromatium vinosum (strain DSM 185) is phy-logenetically distinct from sulfate-reducing orga-nisms Its [NiFe] hydrogenase is a periplasmic, membrane-associated, heterodimeric protein of about 91 kDa†.32 As the other well-studied hydro-genases, such as the periplasmic soluble [NiFe] hydrogenase from D gigas, spectroscopic investiga-tions of A vinosum hydrogenase suggest similar characteristics of the active site.28,32,33 In the small subunit, this protein contains two types of iron– sulfur clusters, namely, [Fe3S4]1+/0and [Fe4S4]2+/1+,

as derived from Mössbauer studies.34,35 Earlier EPR studies at low temperatures revealed a complex split EPR signal around g = 2 due to an [Fe3S4]1+ cluster coupled to an Fe-containing moiety.29,34,36,37 This additional paramagnet is supposed to be only present in the oxidized form

of the enzyme and probably mediates a magnetic coupling between the medial [Fe3S4]1+ cluster and the Ni3+ of the active site Either an extra Fe3+ species located close to the proximal iron–sulfur cluster or a modification of the proximal cluster itself has been suggested.29,32,36,38 The above-described

† It is known that A vinosum has at least two hydrogenases One is a hydrogenase with two subunits that is studied in this article and another is a five-subunit hydrogenase that is known as a HoxEFUYH type.50

phenomenon was also reported for other

hydro-genases such as the membrane-associated [NiFe]

hydrogenases of Ralstonia eutropha10,37 and Aquifex

aeolicus.39 So far, a consistent interpretation of the

complex EPR split signal has not been achieved, the

main reason being lack of structural information

The electrons that are released during the

hetero-lytic cleavage of hydrogen are transferred via the

electron transfer chain, formed by the iron–sulfur

clusters, to an external electron acceptor In the case of

the membrane-associated uptake [NiFe] hydrogenase

(Hup) from the closely related purple-sulfur

bacte-rium Thiocapsa roseopersicina, an interaction partner, namely HupC, has been identified.40 This protein is suggested to be another unit of the hydrogenase complex catalyzing the H2-dependent reduction of quinones A similar mechanism for the A vinosum [NiFe] hydrogenase appears possible The electron acceptor may differ from the one identified in sulfate reducers (i.e., cytochrome c),13,41,42since the physio-logical properties of the A vinosum hydrogenase are

so far not established Recently, the complete genome

of A vinosum (strain DSM 180) has been determined and cytochrome b was identified in the genome sequence—as found in other membrane-bound hydrogenases, where it acts as electron acceptor in the membrane

No crystal structures of [NiFe] hydrogenases from species other than Desulfovibrio and Desulfomicrobium have been determined so far In the present work, the spectroscopically well-studied19,20,29,34,36,43–46A vino-sum hydrogenase has been crystallized and applied to X-ray crystallographic analysis The derived structure has a resolution at 2.1 Å and allows the direct comparison with other hydrogen-converting enzymes The crystallized enzyme is shown to be in the (oxygen-inhibited) oxidized Ni-A state by EPR and FTIR spectroscopy This opens the possibility to examine the structural details and elucidate the peculiar spectroscopic features of this important species that is considered to be the oxygen-inhibited state of the enzyme, since all oxygen-tolerant [NiFe] hydrogenases from aerobic bacteria investigated so far do not show the Ni-A state.10,47

Results

Preparation and spectroscopic characterization The protein preparation was conducted aerobically

as described earlier.48 The cells were extensively washed with cold organic solvent and the protein was extracted from the obtained membranes with

Fig 1 (a) cw-EPR spectrum from [NiFe] hydrogenase from A vinosum at T= 100 K The sample was prepared using dissolved crystals combined from different crystallization batches in 50 mM Tris–HCl buffer, pH 8.0 Experimental details: X-band, microwave frequency=9.636 GHz, mod-ulation amplitude=0.30 mT, mw power= 19.97 mW (b) cw-EPR spectrum recorded at T= 10 K The enlarged spectrum of the Ni region is shown in the inset Experimental details: microwave frequency = 9.635 GHz, modulation amplitude = 0.50 mT, mw power = 0.66 mW (c) FTIR spectrum recorded at T = 288 K The same sample was used to measure the spectra depicted in (a) and (b) The signals of 1945 cm−1 (CO) and 2082 cm−1 and 2093 cm−1 (both CN− ) belong to the Ni-A state An extra signal (1951 cm−1) is present marked with an asterisk (see text)

detergent-containing buffer The solubilized enzyme

was further purified by five different column

chro-matographic steps.48The last two polishing steps (ion

exchange and gel filtration) were repeated and only

the purest fractions were selected to achieve the

highest degree of purity for crystallization

These samples were characterized by cw-EPR and

FTIR spectroscopy prior to the crystallographic

experiments The EPR and FTIR spectra of

as-isolated samples are shown in Fig S1 The EPR

spectrum recorded at 100 K demonstrated that the

sample contained dominantly the Ni-A state (∼90%)

and a small fraction of the Ni-B state (∼10%) At low

temperature (10 K) in part of the sample, signals of

the coupled [Fe3S4] cluster and the Ni center were

observed (Fig S1b) as described earlier by Albracht

et al.29,36,38 Furthermore, the FTIR spectrum

pre-sented inFig S1c showed the typical Ni-A signals:

CO stretching band at 1944 cm− 1 and two CN−

vibration bands at 2082 and 2091 cm− 1 In addition,

a shoulder at 1951 cm− 1 was present and bands at

1909 (shoulder at 1919), 2056, and 2066 cm− 1 were

observed, which are probably related to that part of

the sample showing the split signals in the EPR

spectrum (seeFig S1b).49

Only few crystals were suitable for the

high-resolution X-ray diffraction experiments The other

crystals were dissolved in 50 mM Tris–HCl, pH 8.0,

buffer and characterized by EPR and FTIR

spectros-copy (Fig 1) The sample obtained by dissolving the

crystals (seeMaterials and Methods) demonstrated

that the Ni-A state was dominant (∼90%) in the EPR

spectra (Fig 1a and b) However, no extra splitting

has been observed for Ni-A and for the [Fe3S4]+

cluster at low temperature In the FTIR spectra

obtained from the same sample, only the Ni-A state

and no other signals from the EPR-silent states were

observed (Fig 1c) One additional signal was

observed in the region of the CO stretching band

(at 1951 cm− 1) without a counterpart at the CN−

vibrational region For this band, no clear

assign-ment was achieved

The overall structure

The crystal structure of the A vinosum [NiFe]

hydrogenase in its Ni-A state has been determined at

2.1 Å resolution After calculating the phases from the

Fe-multiwavelength anomalous dispersion (MAD)

data sets (5.0 Å resolution), the electron density peaks

of the iron–sulfur clusters and the [NiFe] active site were clearly confirmed Four molecules (Molecules 1–4) were observed in the asymmetric unit Two of them formed a dimer with a 2-fold non-crystallo-graphic symmetry (Fig 2a) The contact surface area between the small subunit and large subunit was determined to approximately 3800 Å2 The con-tact surface area between the molecules in the dimer was determined to approximately 2800 Å2 and the solvent-accessible area of each molecule to approximately 25,500 Å2(Fig 2b) Rmsd's of the Cα atoms of the four molecules in the asymmetric unit showed values of 0.15–0.19 Å (Table S1andS2) The comparison of the structures from the A vinosum and

D vulgaris MF hydrogenases bears striking

homolo-gy (Fig 2c) The electrostatic potential of the contact surface of the dimer with the 2-fold non-crystallo-graphic symmetry showed a relatively neutral charge distribution (Fig 2d and e)

Imidazole molecules that were part of the crystal-lization solution were also identified in the electron density map Some of them form a hydrogen-bond network including some amino acids, which most likely serves to stabilize the molecule

Sequence alignments of selected [NiFe] hydrogenases

The amino acid sequence alignment of the small subunit from selected [NiFe] hydrogenases is depicted in Fig S2 Cysteines that coordinate the active site and the iron–sulfur clusters are highlighted The amino acid identity values and rmsd values of the carbon atoms (Cα) between these structures are summarized in Table S3 Rmsd's of the small subunits between A vinosum and the other hydrogenases show values of 0.75– 0.84 Å The large subunits exhibit similar devia-tions (0.78–0.85 Å)

The determination of the C-terminal amino acid sequence of the small subunit up to Leu271 was based on the electron density map of Molecule 2 of the asymmetric unit The respective region of the three remaining molecules (Molecules 1, 3, and 4) could be defined up to the residue Pro270 and their temperature factors were refined to an average value of 52.0 A2 Due to the crystallographic packing, they show a slightly different conformation (Fig 3) This exposed C-terminal end of the small subunit, consisting of Val267-Ser268-Val269-Pro270

Fig 2 (a) Overall crystal structure of the dimeric [NiFe] hydrogenase from A vinosum In one-half the iron–sulfur clusters (small subunit), the [NiFe] active site and the extra Mg metal (large subunit) are indicated (b) Molecular surface representation of the dimeric form The two molecules are represented in green and blue (the small subunit is shown in light color) (c) Superimposed ribbon representation of the structures of A vinosum (blue) and D vulgaris MF (red) Rmsd

of the A vinosum and D vulgaris MF hydrogenases were 0.78 Å (small subunit) and 0.82 Å (large subunit), respectively (Table S3) (d) The electrostatic potential of one molecule depicting the dimerization contact region Positively charged surface is blue and negatively charged surface is red (e) The electrostatic potential after a 180° rotation of figure (d)

Fig 2 (legend on previous page)

and Leu271, appears to have a neutral electrostatic

potential (Fig 2d and e)

At the N-terminus of the large subunit, Long et al

recently identified five extra amino acids

(Glu-Gln-Ala-Arg-Arg).50Three of these amino acid residues

(Ala-Arg-Arg) were confirmed in the electron

density map of A vinosum hydrogenase The results

of matrix-assisted laser desorption/ionization

time-of-flight mass spectrometry showed that the small

subunit of the purified enzyme has a molecular mass

of 29.7 kDa (± 0.4%).48 This is in good agreement

with the results obtained from the crystal structure

presented here, which has a calculated molecular

mass of 29.1 kDa The slightly smaller value is

probably due to unidentified amino acid residues in

the electron density map

At the C-terminus of the large subunit, 15 amino

acids after His561 exist in the precursor sequence of

the A vinosum hydrogenase (Fig S3) His561 is the

last amino acid of the conserved C-terminal [NiFe]

cluster binding motif (DPCxxCxxH) These 15

amino acids are missing in the electron density

map This can be explained by the final maturation

step of the hydrogenase, where the tail after His561

is cleaved off by a specific endopeptidase, as has

been shown for other hydrogenases.51

The [NiFe] active site in the Ni-A state

In the electron density omit map, no diatomic

ligand was observed at the third bridging position

(indicated as X in Fig 4a and b) When a single

oxygen atom was assigned as bridging ligand, the

temperature factor of the oxygen atom converged to

approximately 19.4 Å2 The distance between Ni and

Fe exhibited a slightly larger value (3.0 Å) than that

determined for the D vulgaris MF hydrogenase

(2.70–2.80 Å in the oxidized states) The distances of

Fe-O and Ni-O were very similar, that is, the

bridging ligand was located almost symmetrically

between the Ni and Fe atoms The Fe-O-Ni angle

was refined to approximately 102° Earlier reports

about the hydrogenase from Desulfovibrio sp

sug-gested that the bridging ligand in the Ni-A state is

assigned to a hydroperoxide ligand.16,17 When a

bridging peroxide was modeled in A vinosum

hydrogenase, a negative electron density at the second

oxygen position of the hydroperoxide ligand was

observed in the difference Fobs−Fcalc Fourier map

Higher temperature factors (24.3 and 36.9 Å2, respec-tively) were also observed Furthermore, the possibil-ity of a half-occupied sulfido ligand was examined This turned out to be not reasonable for a bridging ligand The sulfido ligand was refined with a bond length of rNi-S= 2.1 Å and rFe-S= 2.1 Å, an angle of Ni-S-Fe = 92°, and a temperature factor of 11.5 Å2 However, a residual electron density was observed

in the difference Fobs−FcalcFourier map Therefore, the possibility of a diatomic ligand or a sulfido ligand at the bridging position between the metals was excluded for A vinosum

Modifications of bridging cysteine residues by the coordination of oxygen atoms were observed in the Ni-A/Ni-SU state of Desulfovibrio hydrogenases.16,17

A small residual electron density near a sulfur atom

of the bridging cysteine (Cys64) was also observed

in A vinosum (Fig 4b, labeled Y) The distances between the sulfur of the bridging cysteine residues (Cys64 and Cys558) and the Ni atom were slightly longer than that of the sulfur of the terminal cysteine residues (Cys61 and Cys555) The glutamate residue (Glu14) near the terminal cysteine (Cys555) has been proposed to be the first residue accepting the proton

in the proton transfer pathway.52 The carboxyl oxygen of Glu14 points towards the sulfur of Cys555 and lies within a distance of about 3.2 Å [S (Cys555)–Oɛ2 (Glu14)]

Assignment of the additional metal site

In order to clarify and assign the electron density originating from a possible metal site located at the C-terminus (Fig S4), we applied the Fe-MAD method After the calculation of the anomalous difference Fourier map using the phases from the model and a data set collected at the Fe K-edge (λ=1.74014 Å), no electron density peak corresponding to iron was found in this position Furthermore, inductively coupled plasma with optical atomic emission spec-trometry (ICP-OES) was carried out to determine the metal content of the enzyme The results confirmed that the A vinosum hydrogenase contains Mg, Ca, Ni, and Fe, respectively When the electron density peak was refined as a Mg2+ ion, the temperature factor converged to 16.0± 2.7 Å2 The averaged temperature factor of the surrounding amino acid residues (Glu42, Glu326, Ala506, and His561; seeFig S4) bound to Mg showed a value of 17.5 Å2 The possibility of Ca in the

Fig 3 (a) C-terminus of the small subunit: Superposition of four independently refined molecules observed in the asymmetric unit (Mol1, chains A and B; Mol2, chains C and D; Mol3, chains E and F; Mol4, chains G and H in PDB ID: 3MYR) Val267-Leu271 from molecule 2 (Mol2 in green, chain C) is depicted in stick representation The C-terminus of the remaining molecules Mol1 (blue), Mol3 (yellow), and Mol4 (purple) is shown as ribbon diagram The C-terminus of Mol2

is exposed to the solvent region Due to the crystallographic packing, the C-terminus of Mol1, Mol3, and Mol4 are contacted to neighboring molecules (b) Ribbon and surface representations of four independently refined models in the asymmetric unit Mol4 is shown as surface representation The C-terminus region of Mol1 and Mol2 is indicated with arrows In Mol1, the C-terminus region contacts another molecule (Mol4) due to the crystallographic packing

Fig 3 (legend on previous page)

C-terminus was excluded since the crystallographic

parameters converged with a relatively high

temper-ature factor of 25.5 ± 3.5 Å2 and a negative residual

electron density in the Fobs−Fcalcmap was observed after the refinement Furthermore, no electron density corresponding to Ca was observed in the anomalous

Fig 4 (a) Stereoview of the electron density omit map (Fobs−Fcalc, 4σ cutoff) including the refined model of the active site of the [NiFe] hydrogenase from A vinosum The third bridging ligand is labeled X (b) Stereoview of the electron density omit map (Fobs−Fcalc, 4σ cutoff) including the refined model without the bridging ligand and the diatomic ligand (CO) of the Fe atom The three meshed spheres show the electron density of the bridging ligand assigned as an oxygen, the diatomic ligand (CO), and the modification of the thiolate of Cys64 Only a small amount of density was observed near the thiolate of Cys64 (labeled Y) (c) Stick representation of the [NiFe] active site and possible hydrogen-bond network (black dotted lines) with neighboring amino acids The red broken line between Cys555 and Glu14 illustrates part of a possible proton pathway (see text).52

difference Fourier map from the Fe K-edge data set.

Therefore, the metal at the C-terminus was assigned

to a Mg2+ ion It shows a distorted octahedral

coordination and binds to Glu42, His561, Ala506,

and three water molecules (Fig S4), which form

additional hydrogen bonds to Gln505, Glu326, Glu42,

and Lys377 No Ca ion could be identified in the

electron density map

The iron–sulfur clusters

The anomalous difference Fourier map of the iron–

sulfur clusters, calculated with the phase of the

refined model and the data set collected at 1.0 Å of

the X-ray wavelength, is presented in Fig S5 It

clearly shows that the distal iron–sulfur cluster

(Fig S5a) has four iron atoms and the medial

cluster (Fig S5b) has three iron atoms The distal

[Fe4S4] cluster in A vinosum is coordinated by three

cysteines (Cys190, Cys215, and Cys221) and one

histidine (His187), which are also highly conserved

in Desulfovibrio species (Fig S2) The medial iron–

sulfur cluster was confirmed to be an [Fe3S4] type

coordinated by three cysteine residues (Cys230,

Cys249, and Cys252)

The four highly conserved cysteines, which are

known to coordinate the proximal cluster in

sulfate-reducing species (Fig S2), suggested that the proximal cluster of the A vinosum hydrogenase

is of [Fe4S4] type, too The anomalous Fourier maps calculated using the final phases from the refined model and the anomalous differences from the remote data set (X-ray wavelength of 1.0 Å) showed an ambiguous moiety at the proximal iron–sulfur cluster (Fig S5c) These results indicate that the iron (Fe4) that is coordinated by sulfur atom (Fig S2) could exist in two different conformations The best interpretation of the electron density map shown inFig 5was achieved

by refining it as a superposition of two confor-mers One is the standard cubane [Fe4S4] cluster (Fig 5b) and the other is a distorted [Fe4S4] cluster (Fig 5a) with an estimated occupancy of approx-imately 40% Another data set collected at 2.35 Å resolution of the A vinosum hydrogenase showed similar results in the anomalous Fourier map The proximal iron–sulfur cluster was also refined as a distorted [Fe4S4] cluster with an estimated occu-pancy of approximately 90% and the rest of the electron density fitted as a standard cubane [Fe4S4] cluster (data not shown) This shows that there is a variation in the proportion of the two conformers Furthermore, Asp75 was found in close vicinity

to Fe4 of the distorted cluster within a distance of

Fig 5 Stereoviews of models using the Fobs−Fcalcomit map (4σ cutoff ) of the proximal iron–sulfur cluster in its two possible conformations (a) The distorted iron–sulfur cluster showing the Fe4 shifted towards Asp75 (b) The standard cubane [Fe4S4] cluster

2.3 Å (Figs 5a and 6) This Fe4 atom was still

bound to the sulfur atom of Cys19 within a

distance of 2.3 Å and the coordinating sulfur (Fig

S2) lies in a distance of 2.3 Å from Fe4 Asp75 is

present in the A vinosum hydrogenase, while a

glutamate residue was found in hydrogenases from

sulfate reducers Additionally, a small amount of a

residual electron density in the Fobs−Fcalcomit map

was identified between the Fe4 and Fe2 atom of the

distorted cluster An oxygen or sulfur atom

inserted at this position gave the best fit to the

electron density but had a low occupancy (∼0.1)

Therefore, it was not modeled in the structure

presented here

Discussion

Spectroscopic characterization

It was reported in earlier spectroscopic studies

that the activity and spectroscopic properties of the

as-isolated samples of A vinosum hydrogenase vary

from preparation to preparation.29,36,38 Activity

measurements as well as EPR and FTIR spectra

obtained here also showed similar problems with

the sample after extensive aerobic purification of the

oxidized enzyme before crystallization (Fig S1) The

enzyme seems to be irreversibly inactivated, as a

comparatively low hydrogen uptake activity was

found in preparations of A vinosum, accounting for

values as low as 10 μmol H2 min− 1 mg− 1

hydrogenase An average H2uptake specific activity

of ∼70 μmol H2 min− 1 mg− 1 hydrogenase was

reported in earlier investigations,53 a several times higher value Most probably, the enzyme is quite sensitive to oxygen leading to structural alterations that causes differences in the spectra and in the enzyme activity Surprisingly, dissolved crystals (that were not exposed to X-rays) showed a typical Ni-A signal both in EPR and in FTIR At low temperature, the unusual spin coupling was not observed (seeFig 1b) The only irregular feature in the IR spectra was the band at 1951 cm− 1 in a fraction of the sample This can therefore not be assigned to the spin-coupled Ni-A but may be related to another side product, for example, oxygenation of the cysteine sulfurs Obviously, a selective crystallization of only those enzyme molecules that had the intact (original) Ni-A state took place Since the crystallization required several weeks (sometimes even months) to form the crystals under aerobic conditions, the modified enzyme molecules with spin-coupled species, which proba-bly involves irreversible inactivation of the enzyme, were obviously degraded

An additional problem with respect to the comparison between spectroscopic and crystallo-graphic data might arise from the fact that in the aerobically grown crystals, radiation damage might occur during the prolonged data collection in the high-brilliance X-ray beam of the synchrotron even

at cryogenic temperatures This is expected to result

in reactive oxygen species leading to (photo)reduc-tion of some of the sensitive cofactors in the protein.54,55 Prominent targets for such reactions are the [NiFe] center and the [FeS] clusters, in particular the proximal [Fe4S4] cluster

Fig 6 Comparison of different proximal iron–sulfur clusters in hydrogenases including the possible interaction partners of the respective shifted Fe atom (Fe4): A vinosum, distorted conformation (orange); D vulgaris MF, standard cubane [FeS] cluster (green); and D desulfuricans, superoxidized [FeSO] cluster (yellow)