Ech hydrogenase from Methanosarcina barkeri is a member of a distinct group of membrane-bound [NiFe] hydrogenases with sequence similarity to energy-conserving NADH:quinone oxidoreductase (complex I). The sequence of the enzyme predicts the binding of three [4Fe-4S] clusters, one by subunit EchC and two by subunit EchF. Previous studies had shown that two of these clusters could be fully reduced under 105 Pa of H2 at pH 7 giving rise to two distinct S½ electron paramagnetic resonance (EPR) signals, designated as the g = 1.89 and the g = 1.92 signal. Redox titrations at different pH values demonstrated that these two clusters had a pH-dependent midpoint potential indicating a function in ion pumping. To assign these signals to the subunits of the enzyme a set of M. barkeri mutants was generated in which seven of eight conserved cysteine residues in EchF were individually replaced by serine. EPR spectra recorded from the isolated mutant enzymes revealed a strong reduction or complete loss of the g = 1.92 signal whereas the g = 1.89 signal was still detectable as the major EPR signal in five mutant enzymes. It is concluded that the cluster giving rise to the g = 1.89 signal is the proximal cluster located in EchC and that the g = 1.92 signal results from one of the clusters of subunit EchF. The pH-dependence of these two [4Fe-4S] clusters suggests that they simultaneously mediate electron and proton transfer and thus could be an essential part of the proton-translocating machinery.
In recent years a novel family of membrane-bound [NiFe] hydrogenases, now called energy-converting hydrogenases, has been recognized . These enzymes form a phylogenetically distinct group within the large family of [NiFe] hydrogenases . Members of this hydrogenase family include hydrogenase 3 from Escherichia coli, CO-induced hydrogenase from Rhodospirillum rubrum, Coo hydrogenase from Carboxydothermus hydrogenoformans[3,4] and Ech hydrogenase from Methanosarcina barkeri and Thermoanaerobacter tengcongensis[5–7]. The hydrogenase large and small subunits of these enzymes show surprisingly little sequence similarity to standard [NiFe] hydrogenases, except for the conserved residues coordinating the active site and the proximal [4Fe-4S] cluster . In addition to these subunits, which are conserved in all [NiFe] hydrogenases and which are essential for H2-activation, energy-converting hydrogenases contain at least four other subunits: two hydrophilic proteins and two integral membrane proteins. These six subunits form the basic structure of energy-converting hydrogenases and are conserved in all members of this hydrogenase subfamily. Thus far the purification of only three members of this hydrogenase family has been achieved: M. barkeri Ech hydrogenase , Coo hydrogenase from C. hydrogenoformans and T. tengcongensis Ech hydrogenase .
From a biochemical perspective, the most thoroughly studied member of energy-converting [NiFe] hydrogenases is Ech hydrogenase found in the methanogenic archaeon M. barkeri. In vitro studies had shown that a low-potential, soluble two-[4Fe-4S] ferredoxin (E0′ = −420 mV) isolated from M. barkeri functions as an electron donor and electron acceptor of Ech . The biological role of Ech was studied using mutational analysis . The following conclusions were made from the data obtained: under autotrophic growth conditions, the enzyme catalyses the reduction of the low-potential ferredoxin by H2. Reduced ferredoxin generated by Ech hydrogenase donates electrons to various soluble oxidoreductases, e.g. formylmethanofuran dehydrogenase (FMD), acetyl-CoA synthase/carbon monoxide dehydrogenase complex, and pyruvate:ferredoxin oxidoreductase. FMD, for example, catalyses the reduction of methanofuran and CO2 to formylmethanofuran. The overall reduction of CO2 and methanofuran by H2 is endergonic and is driven in vivo by reverse electron transport . Reduction of ferredoxin by H2, the partial reaction catalysed by Ech hydrogenase, is the energy driven step .
Purified Ech is composed of six subunits corresponding to the products of the echABCDEF operon . The EchA and EchB subunits are predicted to be integral, membrane-spanning proteins, while the other four subunits are expected to extrude into the cytoplasm. Amino acid sequence analysis of the cytoplasmic subunits points to the presence of two [4Fe-4S] clusters in EchF and one [4Fe-4S] cluster in EchC. The EchC subunit belongs to the family of [NiFe] hydrogenase small subunits. The EchE subunit contains the characteristic binding motif for the [NiFe] centre found in the large subunits of all [NiFe] hydrogenases. Chemical analysis has revealed the presence of Ni, nonheme Fe and acid-labile S in a ratio of 1 : 12.5 : 12 , corroborating the presence of three [4Fe-4S] clusters. Characterization of the iron–sulfur clusters of the enzyme by electron paramagnetic resonance spectroscopy (EPR) showed that two of these clusters could be fully reduced under 105 Pa of H2 at pH 7 giving rise to two distinct S½ EPR signals, designated as the g = 1.89 and the g = 1.92 signal . Redox titrations at various pH values demonstrated that the midpoint potentials of the [4Fe-4S] clusters responsible for the g = 1.92 and g = 1.89 signals are pH dependent indicating that they could be involved in ion pumping. A third minor EPR signal, designated the g = 1.96 signal, was tentatively assigned to the third iron–sulfur cluster of the enzyme. Redox titrations indicated that the g = 1.96 signal has the lowest redox potential (well below −420 mV at pH 7); therefore, this cluster could only be partly reduced.
Ech hydrogenase is highly homologous to the catalytic core of complex I which is formed by the four hydrophilic subunits NuoB, C, D and I and the membrane subunits NuoH and NuoL, M, N (following the nomenclature of the E. coli enzyme). The evolutionary relationship between complex I and energy-converting hydrogenases has been addressed in recent reviews [13–16]. The catalytic core of complex I also contains three binding motifs for [4Fe-4S] centres. The characterization of these clusters has been an important issue in the complex I field in the recent years. In NuoB (the homologue of EchC) three of the four Cys residues that are known to ligate a [4Fe-4S] cluster in all [NiFe] hydrogenases are conserved. In NuoB, these Cys residues, together with a fourth unidentified residue, provide the ligands for the EPR-detectable iron–sulfur cluster N2 [17,18]. Cluster N2 exhibits a pH-dependent midpoint potential and therefore is thought to be involved in H+ pumping . Subunit NuoI shares two conserved four-Cys motifs for the binding of [4Fe-4S] clusters with EchF, which is the homologue of NuoI. In complex I, however, these clusters are not detectable by EPR spectroscopy and could only be detected by UV/Vis redox difference spectroscopy . The midpoint potentials of these clusters are pH independent. Hence, the properties of the iron–sulfur clusters present in the catalytic core of complex I seem to differ from those of the homologous clusters in Ech. In the latter enzyme two [4Fe-4S] clusters were detected by EPR, both exhibiting a pH dependent midpoint potential . To ensure that the two EPR signals detected in Ech are derived from two different clusters and to assign these clusters to distinct subunits we have performed a systematic mutagenesis study in M. barkeri in which seven of eight conserved cysteine residues in EchF were individually changed to serine. These studies were possible due to the recent development of genetic techniques for use in Methanosarcina species [21–23]. Here we apply this system for the first time to construct site-directed mutants in Methanosarcina barkeri.
Generation of echF mutants
We have constructed a set of M. barkeri mutants in which seven of eight conserved cysteine residues in EchF, expected to be involved in iron–sulfur cluster coordination, were individually replaced by serine (Fig. 1, Table 1). The construction of an M. barkeri mutant with a deletion of the echABCDEF operon has previously been described. The mutant was still able to grow on methanol as the sole energy substrate but failed to grow on H2/CO2, H2/methanol or acetate . Methanol-grown cells were therefore used for the generation of echF point mutants. The echF mutations were constructed in vitro and recombined into the chromosome as described (Fig. 2). The echF gene was sequenced from each clone to ensure that it carried the appropriate nucleotide sequence for the individual mutations.
Table 1. M. barkeri strains used in this study.
DSMZ, Braunschweig, Germany
Ech hydrogenase activity in cell extracts of the EchF mutants and the strain carrying a wild-type copy of echF was determined using the ferredoxin assay, which is specific for Ech (Table 2). In all EchF mutants Ech hydrogenase activity was strongly reduced. The highest activities were observed in cell extracts of EchF6, EchF7 and EchF8 mutants. The activity of heterodisulfide reductase, an essential enzyme of the energy metabolic pathway, was determined for internal calibration. The low Ech activity in the mutants was not due to a down-regulation of the enzyme as shown by western blot analysis of total cell extracts using an antiserum to detect the catalytic subunit EchE (Fig. 3). The serum shows cross reactivity with subunit HdrD of heterodisulfide reductase, which was used as an internal standard .
Table 2. Ech Hydrogenase and heterodisulfide reductase activity in cell extracts of EchF mutants and wild type M. barkeri. Ech hydrogenase activity was measured by following the H2- and ferredoxin-dependent reduction of metronidazole as described in methods. Heterodisulfide reductase (Hdr) activity was measured as described previously .
Cells used in this study were cultivated on methanol in single cell morphology. It was observed that Ech activity in wild-type cells grown under these conditions was approximately six- to 10-fold lower than those of Ech activity in cells grown on either methanol or acetate in low salt medium where cells grow as cell-aggregates [6,9].
Isolation of Ech hydrogenase from the echF mutant strains
Ech was purified from wild type and the echF mutant strains using a modified version of the procedure described previously . After protein solubilization, purification of Ech was carried out by chromatography on DEAE Sepharose, Q Sepharose and hydroxyapatite. The mutant enzymes studied showed the same chromatographic properties as wild-type Ech throughout all purification steps. Approximately 1.5 mg protein was obtained from 30 g of cells. The preparations thus obtained were analysed by SDS/PAGE (Fig. 4). In preparations obtained from mutant strains EchF2, EchF6 and EchF8 all six subunits of Ech were detectable in Coomassie stained gels (Fig. 4A). The preparations contained contaminating protein bands with apparent molecular masses of 63 kDa, 75 kDa and 90 kDa (only in EchF8). In the preparations obtained from mutant strains EchF1, EchF3, EchF5 and EchF7 the small subunits EchC, EchF and EchD were not clearly detectable in the Commassie stained gel (Fig. 4A), but became detectable after silver staining (Fig. 4B). Subunit EchD was only visible as a fuzzy band migrating directly below EchF. In general the purity of the enzyme from these mutants was lower than the enzyme isolated from the EchF2, EchF6 and EchF8 mutants.
Hydrogenase activity of the purified enzymes was determined by the H2-uptake assay using the M. barkeri ferredoxin, which is the physiological substrate of the enzyme, as electron acceptor. In addition, H2-uptake activity was determined with benzylviologen as an artificial electron acceptor (Table 3). As determined by both assays the EchF8 mutant had the highest activities, with approximately 10% of the activity of the wild-type enzyme. The enzymes from the EchF7, EchF5, and EchF2 mutants showed between 4% and 6% of the wild-type activity. Almost no activity was detectable in the enzymes from the EchF1, EchF3 and EchF6 mutants. The specific activities of the purified enzymes generally correlate with the activities observed in cell extracts. An exception is the EchF6 mutant in which relatively high Ech activity was determined in cell extract but almost no activity could be detected with the purified enzyme. Wild-type Ech catalysed the reduction of benzylviologen under the experimental conditions at fourfold higher rates than the reduction of ferredoxin. This activity ratio was nearly constant in the different mutant enzymes.
Table 3. Ech hydrogenase activity of purified Ech hydrogenase from EchF mutants and wild-type M. barkeri. Hydrogenase uptake activity was measured by following the H2- and ferredoxin-dependent reduction of metronidazole (Fd assay) or the H2-dependent reduction of benzyl viologen (BV assay) as described.
EPR analysis of Ech hydrogenase isolated from EchF mutant strains
The iron–sulfur centres of Ech isolated from the different EchF mutant strains and the strain carrying a wild-type copy of EchF were characterized by EPR spectroscopy (Fig. 5). Samples were reduced under an atmosphere of 100% H2 and EPR spectra were recorded at 10 K and 2 mW microwave power. The wild-type enzyme exhibited a spectrum identical to that described previously . Based on the EPR line shape and differences in temperature dependence this spectrum had been shown to be an overlap of two major EPR signals originating from S = ½ reduced [4Fe-4S] clusters, one signal with gxy = 1.921 and gz = 2.050 (designated the 1.92 signal) and the second with gxy =1.887 and gz = 2.078 (designated the 1.89 signal). EPR spectra recorded from the enzyme of the EchF8, EchF7 and EchF2 mutants show a strong signal with gxy = 1.890 and gz = 2.078, corresponding to the g = 1.89 signal found in the wild-type enzyme. The g = 1.92 signal was still detectable in these mutant enzymes but its spin concentration was strongly reduced relative to the g = 1.89 signal. The EPR spectrum obtained for the enzyme from the EchF8 mutant showed the most intense g = 1.89 signal. Both iron–sulfur signals were simulated (Fig. 6). The parameters for the g = 1.89 signal are slightly different than those reported previously  because the cluster is not involved in spin–spin interaction with other clusters. The 1.92 signal was simulated with the same parameters as before as it makes only a small contribution to the overall spectrum. The overall spin concentration in the iron–sulfur cluster region, corrected for the g =2.03/2.00 radical-like signals (see below) was 10 µm, the enzyme concentration was 19 µm. The g = 1.89 and g = 1.92 signals are present in a ratio of 9 : 1 as estimated from the simulated EPR spectra. As in the wild-type enzyme the g = 1.92 signal and the g = 1.89 signal showed a different temperature dependence. The g = 1.92 signal was optimally sharpened at 17 K whereas the g = 1.89 signal was already considerably broadened at 17 K as shown for the enzyme from the EchF8 mutant in Fig. 7.
To determine if the mutations had turned the spin of the ground state of the cluster(s) of subunit EchF to S = 3/2, EPR spectra were recorded in the low field region (50–2000 G) at low temperature (4.5 K) and high power (20 mW). Pronounced signals were observed only at g = 4.3 which could be due to adventitious Fe(III).
EPR spectra recorded from the enzyme of the EchF5 mutant showed only a weak g = 1.89 signal, but the amplitude of the g = 1.92 signal was comparable to that found in the enzyme from the EchF2, EchF7 and EchF8 mutants. The EchF6 mutant also showed a weak g = 1.89 signal but only a very weak g = 1.92 signal. In EPR spectra recorded from the EchF1 and EchF3 mutant enzymes no iron–sulfur cluster signals could be detected.
EPR spectra of the EchF mutants all showed signals with g-values at 2.033 and 2.003. The two signals showed different saturation properties (Fig. 8). The 2.03 signal could not be saturated at 4 K and 20 mW (10 dB) whereas the 2.00 signal was already saturated at 10 K and 2 mW (20 dB). The signals could be observed at temperatures up to 130 K without signal broadening. The line width of both signals was approximately 1.2 mT, which is typical for radical species. The identical line width indicates that both signals could belong to the same paramagnetic species. The g = 2.03 signal has been previously detected in wild-type Ech where it is only present at very low intensity. However, this signal is much stronger in the EchF mutants. The spin concentration of this signal was determined in the EchF1 and EchF3 mutants, which showed no iron–sulfur cluster signal. Here the spin concentration was approximately 0.7 µm, assuming an S½ species. The enzyme concentration was 19 µm. By comparing the signal amplitudes it could be estimated that the spin concentration of the g = 2.03/2.00 signals in the EchF8 mutant is 0.4–0.5 µm corresponding to 4–5% of the spin concentration of the iron–sulfur cluster signals. In general the spin concentration of the g = 2.03/2.00 signals was approximately 1.6 times higher in the enzyme from those mutants which showed no or very low intensity signals for the iron–sulfur clusters. The g = 2.03/2.00 signals shown in Fig. 8 were observed for the enzyme reduced by 100% H2. Addition of 20 mm sodium dithionite did not change the intensity of these signals. When 1 mm duroquinone (E0′ = +86 mV) was added to the enzyme under N2, the g = 1.89 signal was no longer detectable, indicating an oxidation of this iron–sulfur center. The intensities of the g = 2.03/2.00 signals were, however, not altered by duroquinone oxidation.
The characterization of the iron–sulfur clusters of Ech hydrogenase by EPR spectroscopy, performed previously, revealed the presence of two axial like EPR signals fully reducible under 100% H2. The two signals were designated as the g = 1.89 and the g = 1.92 signal. Importantly, both species have a pH-dependent midpoint potential. The E0′ value of the g = 1.92 signal decreased by 53 mV per pH unit; that of the g = 1.89 signal decreased by 62 mV per pH unit . These values are reasonably close to the theoretical value of −59 mV per pH unit for a redox titration involving a stoichiometric amount of electrons and protons. The g = 1.89 and the g = 1.92 signal showed slightly different temperature optima, the g = 1.89 signal being optimally sharpened at 12 K and the g = 1.92 signal being optimally sharpened at 17 K. At temperatures below 15 K a twofold splitting of the Nia–L signal was observed due to the interaction of the Ni-based unpaired electron with the S = ½ system of the reduced proximal [4Fe-4S] cluster. The temperature dependence of the splitting of the Nia–L signal paralleled the temperature dependence of the g = 1.89 signal. It was therefore tentatively concluded that the g = 1.89 signal is due to the reduced proximal cluster in EchC . The experiments described here substantiate this former assignment. Seven of eight cysteine residues predicted to ligate the iron–sulfur clusters in EchF were systematically changed to serine. For two of the mutant enzymes, EchF2 and EchF8, the g = 1.89 signal was the major signal and only residual spin intensities of the g = 1.92 signal were observed. The spin concentration of the g = 1.89 signal of the EchF8 mutant was highest and accounted for approximately 50% of the enzyme concentration. The determination of the spin concentration is based on the total protein concentration of the sample. Because the preparation still contained three contaminating protein bands (Fig. 4), the spin concentration is probably underestimated. EPR spectra recorded from the enzymes isolated from the EchF5, EchF6 and EchF7 mutants also contained the g = 1.89 signal, however, at lower spin intensities. The formation of the g = 1.89 and the g = 1.92 clusters was not dependent on whether the mutation was in the first (EchF2 and EchF8) or the second (EchF5 and EchF7) iron–sulfur cluster binding motif of EchF (Fig. 1). EPR spectra of the EchF2, EchF5, EchF7 and EchF8 mutant enzymes showed a weak signal in the g = 1.92 region which can be attributed to the gxy of the g = 1.92 signal. Studies with the wild-type enzyme had shown that the third cluster of the enzyme, assigned to the g = 1.96 signal, has a low redox potential and thus could only be reduced to a low extent under 100% H2. It is therefore difficult to judge if the intensity of this signal has changed in the mutant enzymes. One possible explanation for the formation of low amounts of the g = 1.92 cluster in some of the EchF mutants could be ligand exchange. Subunit EchF contains an additional free Cys residue in position 87 which could function as a ligand in some of the mutants. Likewise, the introduced Ser residues could also function as a ligand of the cluster as suggested for a Cys to Ser mutant of E. coli nitrate reductase . For the R. capsulatus NuoI mutants (see below) it was also proposed that in some of the mutants the introduced Ser residue could be a direct ligand to a [4Fe-4S] cluster .
The cluster ligating Cys residues conserved in EchF have also been mutagenized in the homologous subunit of complex I, NuoI, from R. capsulatus and E. coli. In R. capsulatus five Cys residues were individually changed to Ser. Four of these mutants had retained significant amounts of complex I activity in the membrane fraction (up to 72% of the wild-type activity). Purification of the mutant enzymes was not attempted as even the wild-type enzyme was found to be unstable upon purification. The eight cluster-ligating Cys residues of the closely related E. coli complex I were individually mutated to Ala. With the exception of the C102A mutant, which had retained 17% of the wild-type activity, all other mutants had lost complex I activity. The comparison indicates that Cys to Ser mutations are more likely to produce active enzyme in comparison with Cys to Ala mutations.
In most of the EchF mutants the intensity of the g = 1.89 signal was also reduced and in two of the mutants no signal due to an iron–sulfur cluster could be detected indicating that the mutation in EchF also had a strong effect on iron–sulfur cluster assembly in EchC. This is analogous to mutations of the cluster ligating cysteine residues of the NuoI subunit of E. coli complex I, which in most cases also resulted in a loss of the iron–sulfur cluster N2 located on subunit NuoB, a homologue of EchC . It has also been observed with other systems, e.g. two [4Fe-4S] ferredoxins, that a substitution of one of the cluster ligands in a subunit often significantly affects incorporation of the neighbouring clusters .
The characterization of the different EchF mutant strains revealed that subunit EchF, homologous to complex I subunit NuoI (or TYKY), contains an EPR-detectable iron–sulfur cluster which exhibits a pH-dependent midpoint potential. In contrast, no EPR signal could be attributed to one of the [4Fe-4S] clusters located on NuoI (or TYKY) of complex I. In studies performed with complex I from E. coli and Neurospora crassa, a redox-group was identified by means of UV/Vis spectroscopy and was assigned to the two [4Fe-4S] clusters located on NuoI (or TYKY) . A redox titration of this group, which was followed by UV/Vis spectroscopy, revealed a pH-independent midpoint potential of these clusters with an E0′ value of −270 mV. It is thought that these clusters are magnetically coupled in the reduced state and therefore are difficult to detect by EPR spectroscopy. From these results it was concluded that NuoI (or TYKY) has redox properties very similar to those of 8Fe-ferredoxins, e.g. the one from Clostridium pasteurianum. Therefore, NuoI was proposed to be involved in simple electron transfer. Other studies focusing on the characterization of the iron–sulfur clusters of NuoI where performed with the homologous protein from Paracoccus denitrificans, termed NQ09 . NQ09 was heterologously produced in E. coli. The isolated subunit was found to bind two [4Fe-4S] clusters which when reduced gave rise to a set of two relatively broad axial-type EPR signals at g = 2.08, 2.05 and 1.93 and 1.90. The two sets of EPR signals could either be derived from two distinct species of [4Fe-4S] clusters or alternatively one signal could be derived from the two S = 1/2 [4Fe-4S] clusters in NQ09 which exhibit similar EPR spectra and the second signal could arise from spin–spin interaction between the former two paramagnetic species. The midpoint potentials of these clusters were, however, < −600 mV indicating that their redox properties changed considerably in the heterologously produced subunit. In the entire complex I from P. denitrificans these signals were not observed.
In Ech the [4Fe-4S] cluster located on subunit EchC and one of the [4Fe-4S] clusters on subunit EchF exhibit a pH-dependent midpoint potential. This indicates that oxidation/reduction of these clusters depends on charge compensation of an acidic residue close to the cluster. These subunits therefore could also play a crucial role in coupling electron transfer to proton translocation. Acidic residues that could be involved in this process have been identified in multiple sequence alignments of EchF or EchC with their homologues from other energy-converting hydrogenases and the corresponding subunits of complex I from various sources. In EchF the second 4× Cys-binding motif was found to contain a Glu, Asp or His residue in all members of the protein family . In addition, a highly conserved Glu residue is found in the proximity of the second 4× Cys-binding motif (C-X(2)-C-X(2)-C-X(3)-C-P-X(8–10)-E). The cluster on EchC corresponds to cluster N2 located on the homologous subunit of complex I (NuoB or PSST). Recently, electrochemically induced FT-IR-difference spectroscopy of site-directed mutants of E. coli complex I revealed that the reduction of iron–sulfur cluster N2 is accompanied by the protonation of Y114 and Y139 of subunit NuoB [28,29]. The FT-IR data also indicated that the oxidation of cluster N2 is coupled with the protonation of one or more carboxylic amino acids. The residues corresponding to Y114 and Y139 are also conserved in Ech and other energy-converting hydrogenases. The FT-IR data also indicate that the oxidation of cluster N2 is coupled with the protonation of one or more carboxylic amino acids. Both, EchC and NuoB contain highly conserved Glu and Asp residues, which are not conserved in the homologous subunit of standard [Ni-Fe] hydrogenases suggesting a common mechanism as well.
The hydrogenase activity of the EchF mutants, with ferredoxin as well as with benzylviologen as electron acceptor, was strongly reduced. The relative hydrogenase activity, with regards to which electron acceptor was used, indicates that the EchF mutants and wild-type Ech use the same set of iron–sulfur clusters for the electron transfer from the [Ni-Fe] centre. In this context the analysis of the EchF8 mutant is of particular interest. In this mutant the spin concentration of the proximal cluster was highest and accounted for at least 50% of the enzyme concentration. The hydrogenase activity of this mutant enzyme determined with both the physiological and the artificial electron acceptor, however, was only about 10% of that of the wild-type enzyme. This indicates that the g = 1.92 cluster and probably also the g = 1.96 cluster are required not only for the reduction of the ferredoxin but also for the reduction of the artificial electron acceptor benzylviologen. Reduction of the g = 1.89 cluster by H2 was still possible in the mutant enzymes, which provides further evidence that this cluster directly interacts with the [Ni-Fe] centre and that the [Ni-Fe] centre is intact in the mutant enzymes. Hence, the electron transfer reaction mediated by the enzyme can be summarized as follows: H2 is activated at the [Ni-Fe] centre, and electrons are transferred via the proximal cluster located on subunit EchC (g = 1.89 signal) to the cluster(s) located on subunit EchF, where ferredoxin is reduced. Our studies show that at least the cluster giving rise to the g = 1.92 signal is required for ferredoxin reduction. At the current stage, the exact role of the low-potential g = 1.96 signal is not known. Even in the wild-type enzyme, the intensity of this signal is very low because of its very low redox potential. Unlike in complex I, quinones are not involved in the electron transfer reaction pathway mediated by energy-converting hydrogenases. The comparison between complex I and Ech hydrogenase rather indicates that the [Ni-Fe] centre and the quinone have complementary functions. The characterization of several complex I mutants which carry mutations at conserved positions in the NuoD (or 49 kDa) subunit has shown that this subunit of complex I carries a significant part of the quinone binding pocket and that this binding pocket could have evolved directly from the [Ni-Fe] centre binding site of the hydrogenases . The pH-dependence of the two [4Fe-4S] clusters in Ech suggests that these clusters simultaneously mediate electron and proton transfer and thus could be an essential part of the proton translocating machinery which delivers protons to proton transfer pathways formed within the membrane part. The protons required for H2 formation at the active site are thought to be delivered by a distinct proton channel located within the hydrogenase large subunit [30,31].
A question that remains is if the paramagnetic species that gives rise to the 2.03/2.00 signals is an intrinsic part of the enzyme (e.g. a yet unknown redox group) or if this species is artificially generated to a greater extent in the EchF mutants. The observation that the signals do not respond to oxidation or reduction favours the second possibility. Also the spin concentration is rather low. In freshly prepared wild-type Ech the 2.03 signal is hardly detectable (Fig. 5) whereas the signal becomes more intense upon aging of the enzyme (R. Hedderich, unpublished results). The line shape and the temperature dependence of the signals could indicate a free radical, but such high g-values have only been described for sulfur-based radicals . Such radicals are, however, very unstable and are normally observed only under presteady-state conditions. Because the g = 2.03 component of the signal could not be saturated at 4.2 K and full power, it was suggested that the signal could be due to a radical in close proximity to a very rapidly relaxing paramagnet, e.g. high spin Fe3+. EPR spectra with similar g-values (g = 2.032 and g = 2.004) and temperature behaviour have been observed for iron–nitrosyl–histidyl complexes which have for example been observed upon disassembly of the [3Fe-4S] cluster of mitochondrial aconitase upon anaerobic NO addition . If such a species is formed upon reaction of a partially assembled iron–sulfur cluster of Ech with NO, which could be reductively generated from contaminating amounts nitrate or nitrite, needs to be shown.
Plasmid and strain construction
Standard techniques were used throughout for isolation and manipulation of plasmid DNA in E. coli, using DH10B (Invitrogen, Carlsbad, CA, USA) as the host strain. All inferred plasmid sequences are available upon request. echDEF was amplified by PCR from M. barkeri chromosomal DNA using primers 5′-GGCGCGCCGGGCCCACGGAGTAGTGGCAGCACTT-3′ and 5′-GGCGCGCCCTCGAGGGAGAACATTCAGTATTGTTTTTCAAG-3′ (restriction sites are underlined), digested with ApaI and XhoI, and ligated into pBluescriptSK (Stratagene, La Jolla, CA, USA) cut with ApaI and XhoI, resulting in pAMG57. Point mutations were generated by the QuickChangeTM method (Stratagene) using pAMG57 as the PCR template. The insert of interest was sequenced to verify that the selected clones only contained the desired mutations. The downstream region of M. barkeri ech was amplified by PCR using primers 5′-GGCGCGCCCTGCAGGGTCTAAATTTGGCAGTTAAGGAA-3′ and 5′-GGCGCGCCGGATCCCCTGCACCTTTCCTGATTTT-3′, digested with BamHI and PstI (restriction sites underlined), and ligated into pJK3  cut with BamHI and PstI, resulting in pAMG77. Each of the seven point mutations generated from pAMG57 and the insert from the original pAMG57 were then subcloned into pAMG77 using the ApaI and XhoI restriction sites, resulting in pCGR1 to pCGR3 and pCGR5 to pCGR9, respectively. These plasmids were then digested with ApaI and BamHI and transformed into M. barkeri using standard techniques (Fig. 2) [21–23,34], selecting puromycin resistance. After single-colony purification, clones were screened for the correct genotype by PCR amplification and sequenced using primers 5′-ACTTATGTTACCGGGCGTCA-3′ and 5′-CCTCGAGGGAGAACATTCAG-3′, resulting in the strains listed in Table 1. M. barkeri strains were grown in single cell morphology  at 37 °C in high-salt media under strictly anaerobic conditions, as described previously . Methanol (125 mm) was added to high-salt media as carbon and energy source. Puromycin was added to 2 mg·mL−1 as appropriate.
Preparation of cell extracts and isolation of Ech hydrogenase
Ech hydrogenase was purified from methanol-grown M. barkeri echF mutant strains under strictly anaerobic conditions using a modification of the procedure described previously . Late exponential-phase M. barkeri single cells were lysed by resuspension in 50 mm Mops/NaOH pH 7.0 containing 2 mm dithiothreitol (buffer A), to which a few crystals of DNase I were added (spontaneous lysis occurs due to osmotic shock). Complete lysis was ensured by sonication four times at 200 W for 3 min. Intact cells and cell debris were removed by centrifugation at 10 000 g for 30 min at 4 °C. The cell extracts thus generated were used for activity measurements and western blot analysis. For the isolation of the membrane fraction, cell extracts were loaded on a DEAE Sephacel column (2.6 × 15 cm). The column was washed with 100 mL buffer A. The majority of Ech hydrogenase activity was recovered in the turbid void volume of the column whereas most soluble proteins were bound to the column material. This procedure resulted in higher Ech yields as compared to an ultracentrifugation step, as membranes of methanol-grown M. barkeri were difficult to sediment by ultracentrifugation, probably due to their high glycogen content. Membranes thus obtained were solubilized by dodecyl-β-d-maltoside (15 mm, 4.5 mg detergent·mg protein−1) for 12 h at 4 °C. After centrifugation at 150 000 g for 30 min the solubilized membrane proteins present in the supernatant were loaded on a DEAE–Sepharose column (2.6 × 10 cm) equilibrated with buffer A containing 2 mm dodecyl-β-d-maltoside (buffer A + DDM). The column was washed with 50 mL buffer A + DDM and proteins were eluted with NaCl in buffer A + DDM using a step gradient: 0.24 m (150 mL) and 0.4 m (150 mL). Ech hydrogenase was recovered in the fractions eluting with 0.24 m NaCl. Further purification of Ech was carried out by chromatography on Q Sepharose and hydroxyapatite as described . The enzyme eluted from the Q Sepharose column with 0.19 m NaCl and from the hydroxyapatite column with 500 mm potassium phosphate. These fractions were concentrated and the buffer was exchanged to buffer A + DDM by ultrafiltration using Amicon Ultra-4 Centrifugal Filter Devices (cut-off 50 kDa; Millipore, Eschborn, Germany). The enzyme was stored under an atmosphere of 100% H2.
Determination of Ech hydrogenase activities
All enzyme assays were carried out under anoxic conditions in 1.5-mL stoppered quartz cuvettes at 37 °C.
Hydrogen uptake activity with ferredoxin as electron acceptor was determined by following the ferredoxin-dependent metronidazole reduction at 320 nm under an H2 atmosphere (120 kPa) . The 0.8-mL assays contained 50 mm Mops/NaOH pH 7.0, 2 mm dithiothreitol, 2 mm DDM, 20 µm ferredoxin (isolated from M. barkeri as described , 150 µm metronidazole (ε320 = 9.3 mm−1 cm−1) and protein (purified Ech hydrogenase, membrane fraction or cell extract). One unit of hydrogenase activity corresponds to the oxidation of 1 µmol H2 measured by the ferredoxin-dependent reduction of 1/3 µmol of metronidazole.
Hydrogen uptake activity, with benzyl viologen as electron acceptor, was determined by following the H2-dependent benzyl viologen reduction at 578 nm under an H2 atmosphere (120 kPa) as described .
SDS/PAGE and immunodetection were performed as described . Protein concentrations were determined by the Bradford method using BSA as standard.
EPR spectroscopy measurements
EPR spectra at X-band (9 GHz) were obtained with a Bruker EMX spectrometer. All spectra were recorded with a field modulation frequency of 100 kHz. Cooling of the sample was performed with an Oxford Instruments (Oxford, UK) ESR 900 flow cryostat with an ITC4 temperature controller. Spin quantifications were carried out under nonsaturation conditions using 10 mm copper perchlorate as standard (10 mm CuSO4/2 m NaClO4/10 mm HCl). The relative intensities of EPR signals present in samples that showed more than one paramagnetic species were determined by simulating the signal of the respective species. To test if the simulated spectrum were correct they were subtracted from the original spectrum. A relative flat baseline should be left over. The simulated signals were subsequently double integrated to determine their intensity. EPR signals were simulated using noncommercial programs based on formulas described previously .
This work was supported by the Max-Planck-Gesellschaft, by the Deutsche Forschungsgemeinschaft, and by the Fonds der Chemischen Industrie. Drs S. Albracht, M. Bennati and E. Lyon are acknowledged for helpful discussions. Drs E. Duin and A. Pierik are acknowledged for their help with EPR spectroscopy and data analysis.