Electron inventory of the iron-sulfur scaffold complex HypCD essential in [NiFe]-hydrogenase cofactor assembly.

The [4Fe-4S] cluster containing scaffold complex HypCD is the central construction site for the assembly of the [Fe](CN)2CO cofactor precursor of [NiFe]-hydrogenase. While the importance of the HypCD complex is well established, not much is known about the mechanism by which the CN- and CO ligands are transferred and attached to the iron ion. We report an efficient expression and purification system producing the HypCD complex from E. coli with complete metal content. This enabled in-depth spectroscopic characterizations. The results obtained by EPR and Mössbauer spectroscopy demonstrate that the [Fe](CN)2CO cofactor and the [4Fe-4S] cluster of the HypCD complex are redox active. The data indicate a potential-dependent interconversion of the [Fe]2+/3+ and [4Fe-4S]2+/+ couple, respectively. Moreover, ATR FTIR spectroscopy reveals potential-dependent disulfide formation, which hints at an electron confurcation step between the metal centers. MicroScale thermophoresis indicates preferable binding between the HypCD complex and its in vivo interaction partner HypE under reducing conditions. Together, these results provide comprehensive evidence for an electron inventory fit to drive multi-electron redox reactions required for the assembly of the CN- and CO ligands on the scaffold complex HypCD.

Introduction

Gas-processing metalloenzymes facilitate complex and essential metabolic processes such as photosynthesis, respiration, and N2 fixation [1]. Hydrogenases catalyze the reversible conversion of H2 into protons and electrons. Catalysis occurs at bimetallic active sites that contain either nickel and iron or only iron [2]. The nature of the metallic cofactor forms the basis of the distinction of hydrogenases into three classes: [FeFe]-, [NiFe]-, and iron sulfur cluster-free hydrogenases [3,4]. [NiFe]-hydrogenase and [FeFe]-hydrogenase are unrelated at the amino acid sequence level but their cofactors share the unique feature of having diatomic ligands like carbonyl (CO) and cyanide (CN−), each associated with the iron ions. To obtain a clear understanding of these specific structural features, studies of the biosynthesis, structure, and mechanism of H2 activation at the molecular level are important [5,6]. There is also interest from applied research because it is hoped to learn from hydrogenase how a robust catalyst for reversible H2 activation can be developed [6]. The focus of this work is on the biosynthesis of [NiFe]-hydrogenase.

Standard [NiFe]-hydrogenase consists of a large subunit and a small subunit. The large subunit contains the catalytic [NiFe](CN)2CO cofactor whereas the small subunit features three iron-sulfur clusters involved in electron transfer between the active site and redox partners like ferredoxin or flavodoxin [2,4]. Additional subunits may be present to transfer the electrons to pyridine nucleotides or to anchor the multimeric enzyme to the membrane [7-9]. Despite these structural differences, the catalytic cofactor is strictly conserved among all [NiFe]-hydrogenases. The nickel ion is coordinated by four cysteinyl thiolates in the active site, while two of the thiolates additionally bind an iron ion. The iron site carries one CO and two CN− ligands [10,11]. The ligands are proposed to maintain the low-spin Fe2+ state important for H2 activation. If these ligands are not attached to the iron ion, no active hydrogenase enzyme can be synthesized [12].

A complex machinery is involved in the synthesis and insertion of the [NiFe](CN)2CO cofactor into the apo-enzyme [12,13]. Figure 1 presents our current working model for hydrogenase maturation and suggests the following sequence of events: (i) Synthesis of CO and CN−, ligation to iron, and incorporation of the fully coordinated iron into the large subunit pro-protein by the coordinated activity of HypCDEF (see below) [13-15]; (ii) Nickel insertion by concerted activity of HypA, HypB, and SlyD [14,16]. (iii) Cleavage of the C-terminal extension of the pro-protein by a specific protease that recognizes the completed metal cluster, resulting in a conformational change that locks the active site in place [13]. (iv) Finally, dimerization of the large subunit with the small subunit to yield active enzyme [12,17-19]. To date, it is not clear where the iron ion in the active site originates from, how the diatomic ligands are attached to the cofactor precursor, and how the CO ligand is synthesized. In the following, we will focus on the biosynthesis of [Fe](CN)2CO cofactor precursor. Our working hypothesis for the assembly of [Fe](CN)2CO moiety suggests the following steps (Figure 1B). (1) The HypC dimer delivers both iron and CO2, the later which represents a putative precursor of the CO ligand [20]. Subsequently, the HypCD complex is formed upon contact with HypD while a HypC monomer is released [20]. (2) The HypCD complex may catalyze an ATP-dependent reduction in CO2 to CO at the expense of 1 eq. of H2O [21]. The source of electrons required for the reduction in CO2 to CO is unknown. (3) The CN− ligands are generated by the maturation proteins HypE and HypF [22]. (4) Transfer of the CN− ligands to the CO-modified iron ion and formation of the [Fe](CN)2CO precursor on the HypCD complex [23,24]. Insertion of the cofactor precursor into the large subunit takes place by the coordinated activity of four maturation proteins, HypCDEF [12,19]. This working hypothesis is supported by the isolation and characterization of complex intermediates of Hyp-maturation proteins, e.g. HypC [Fe]CO2, HypCD [Fe]CO2, HypCD [Fe]CO, and HypCD [Fe](CN)2CO. The binding of CN−, CO, and CO2 ligands in these complexes correlates with the presence of a supernumerary iron ion along with the [4Fe-4S] cluster in HypD [24]. Replacement of conserved amino acids in HypC and HypD abolished iron binding and resulted in the loss of the ligands’ absorption bands [20,25]. These findings indicate functional coupling between Fe, CO, CN−, and CO2.

Figure 1.

Working model for maturation of the large subunit into functional [NiFe]-hydrogenase.

(A) (i) Incorporation of the Fe(CN)2CO moiety into the hydrogenase large subunit (pro-protein) by the coordinated activity of HypCDEF (dashed arrow). (ii) Nickel insertion. (iii) Proteolytic cleavage of the C-terminal peptide, associated with conformational changes of the protein. (iv) Dimerization of large and small subunit to form functional [NiFe]-hydrogenase. (B) The bottom half, biosynthesis of the [Fe](CN)2CO moiety. (1) The HypCD complex is formed upon contact with the [Fe]CO2 carrying HypC dimer. (2) The CO2 ligand may undergo an ATP-dependent reduction to CO catalyzed by HypCD complex. (3) The CN− ligands are generated by maturation proteins HypE and HypF. (4) Transfer of the CN− ligands to the CO-modified iron ion and formation of the [Fe](CN)2CO precursor on the HypCD complex.

The HypCD complex is the central synthesis platform, where CO and CN− ligands are synthesized and attached to the iron ion [24,26,27]. Cyanation would require the input of two electrons each, if the transfer takes place by an electrophilic substitution mechanism [26]. Additionally, two electrons for CO2 reduction will be required if CO2 is the metabolic precursor of CO [28]. It is likely that the HypD protein is involved in this electron transfer reaction, since HypD is the only maturation protein that carries an iron-sulfur cluster and therefore it is the only Hyp protein potentially capable of redox chemistry. However, the redox activity of the [4Fe-4S] cluster has not been demonstrated yet.

In the present work, we developed an efficient expression and purification system to produce the HypCD complex with complete cluster and cofactor content. This facilitated a comprehensive spectroscopic characterization, combining UV/Vis, resonance Raman, ATR FTIR, EPR, and Mössbauer spectroscopy. Our study indicates that the [4Fe-4S] cluster and the mononuclear iron site of the HypCD complex are redox active. Moreover, we investigate the binding affinities of HypCD to its in vivo interaction partner HypE and the HypEF complex using MicroScale thermophoresis (MST).

Materials and methods

Bacterial strains, plasmids, and growth conditions

Plasmid pT-hypDEFC [26] was used to generate pT-hypDEFCStrepXT for overproduction of the recombinant HypCD complex that has two strep-tag sequences (Twin-Strep-tag) with an internal linker region fused to HypC as C-terminal tag. The HypCD complex was overexpressed in E. coli strain MC4100 (DE3) transformed with the pT-hypDEFCStrepXT plasmid. Large scale protein purification was performed with 20 liters of anaerobic TGYEP full medium (1% tryptone, 0.5% yeast extract) at 37°C with a typical yield of 3–4 g of cells/liter. The growth medium contained 100 µg/ml ampicillin and was supplemented with 0.8% (m/v) glucose and 10 mM sodium formate. To induce gene expression, isopropyl-β-d-thiogalactopyranosid (IPTG) was added to a final concentration of 50 µM at OD600nm ∼0.3. After induction, the culture was incubated at 25°C for a further 5–8 h or after the culture reached an optical density between 0.8–1.2 (exponential growth, see Supplementary Figure S1). Cells were harvested by centrifugation at 15 000 for 20 min at 4°C. Cell pellets were used either immediately or stored at −80°C.

To produce the 57Fe-labeled HypCD complex for Mössbauer spectroscopy, the bacteria were grown anaerobically either in M9 minimal medium (M9 salt, 0.4% glucose (m/v), 0.3 mM MgSO4, 0.3 mM CaCl2, 1 µg/l biotin, 1 µg/l thiamine and trace elements) or modified minimal medium (M9*) or modified full medium (TGYEP*). The later contained only 0.2% yeast extract to reduce the natural occurring 56Fe content. At the time of gene induction, 57Fe was added to TGYEP* medium to a final concentration of 20 µM or 100 µM (the growth curve in Supplementary Figure S1 shows TGYEP* + 20 µM 57Fe). The M9* medium contained the following compounds: 34 mM Na2HPO4, 22 mM KH2PO4, 8.5 mM NaCl, 9.35 mM NH4Cl, 0.4% (m/v) glucose, 0.3 mM MgSO4, 0.3 mM CaCl2, 1 µg/l biotin, 1 µg/l thiamine, 0.2% peptone, and 10 mM Na formate. The following trace elements were added: 134 µM EDTA, 6 µM ZnCl2, 0.76 µM CuSO4, 0.42 µM CoCl2, 1.6 µM H3BO3, and 0.08 µM MnCl2. At the time of gene induction, 20 µM 57Fe was added to the medium. Cells were harvested in the exponential growth phase and washed twice with 100 mM Tris–HCl buffer pH 8 containing 150 mM NaCl and 1 mM EDTA.

Preparation of crude extracts and protein purification

All steps were carried out at 4°C under an atmosphere of N2 and 1% H2 in an anaerobic chamber (Coy Laboratories), unless stated otherwise. Wet cell paste of MC4100 (DE3) transformed cells containing Twin-Strep-tagged HypC in complex with HypD was suspended at a ratio of 1:5 (m/v) in breaking buffer (100 mM Tris–HCl pH 8, 150 mM NaCl, 10 mM Mg acetate, 5 mg/l DNase, and 0.2 mM PMSF). Ice-cooled cells were disrupted by sonication (70 W for 30 min with 0.5 s pulses). Unbroken cells and debris were removed by centrifugation for 30 min at 15 000 at 4°C. The supernatant or crude extract derived from 40–60 g of wet cells was used for anaerobic purification of the HypCD complex. Proteins were isolated by chromatography on a 20 ml Strep-Tactin-XT column (IBA) using gravity-flow. Unbound proteins were removed by washing the column with 15 column volumes of buffer W (100 mM Tris–HCl pH 8, 150 mM NaCl, 10 mM magnesium acetate, and 1% glycerol). Recombinant protein was eluted with the same buffer but including 50 mM biotin. The protein was desalted on a Sephadex G-25 column (GE Healthcare) in buffer W. Purified protein was concentrated by centrifugation at 5000 using centrifugal filters (Amicon Ultra, 30 kDa cut-off, Millipore) and applied to a Superdex 200 gel filtration column (2.6 × 60 cm) equilibrated with the same buffer.

The HypCD complex was adjusted to 40–50 mg/ml for spectroscopy (1–2 mg/ml for UV/Vis spectroscopy). Such protein preparations are colored (reddish-brown at >10 mg/ml) and referred to ‘as-isolated’. Reduction was achieved in the presence of 1–10 mM NaDT and oxidation was achieved with 20 or 40 eq. phenazine methosulfate (PMS) or 0.2–2% v/v H2O2. To ‘recover’ the as-isolated state of the HypCD complex, NaDT-, PMS- or H2O2-containing buffer was exchanged using Viva-spin centrifugal concentrators (30 kDa MWCO). The protein sample was diluted 120× with protein buffer at a pH 8, 9, or 10 (as indicated) and concentrated at 4000 for 30 min. The concentrated sample was diluted 120× again with fresh protein buffer and re-adjusted to the original concentration of 40–50 mg/ml.

His-tagged HypF and Strep-tagged HypE were purified from BL21(DE3) cells transformed with the appropriate plasmid derivative as described previously [12,19]. A novel protocol was developed for reconstitution of a homogenous HypEF complex on an affinity column ensuring that the HypEF complex is ‘saturated’ with its interaction partner. The Strep-tagged HypE protein was loaded on a Strep-Tactin gravity flow column. After washing with three column volumes of buffer W, His-HypF was loaded in two-fold molar excess, followed by three washing steps with the same buffer. HypEF complex was eluted with buffer W containing 50 mM biotin. The eluted protein was collected and loaded on a gel filtration column equilibrated with buffer W containing 1 mM NaDT and 5 mM DTT. The HypEF complex was eluted in one major peak.

Polyacrylamide gel electrophoresis and metal quantification

To check for protein purity, 10–20 µg of the HypE, HypCD or HypEF complexes were separated by SDS–PAGE (NuPAGE 4 to 20% Tris-glycine, Invitrogen). Protein identity was confirmed by protein mass spectrometry as described previously [29]. A defined amount of crude extracts derived from a HypD deletion mutant (DHP-D) [30] that was complemented with pT-hypDEFCStrepXT plasmid was analyzed for hydrogenase activity via native PAGE (NuPAGE 4–12% bis-tris glycine) under non-denaturing conditions. Protein samples were incubated with 4% (w/v) Triton X-100 prior to application to the gels. The gel was subsequently stained for hydrogenase enzyme activity as reported previously [31]. The protein concentration was determined as described by Lowry and co-workers [32]. Metal analysis was done by inductively coupled plasma mass spectrometry (ICP-QqQ-MS) using a high resolution 8800 ICP-QqQ-MS device (Agilent Technologies) as described previously [33]. The instrument was run with H2 as reduction gas in the Octopole Reaction System to reduce polyatomic interferences and in direct infusion mode via a nebulizer. ICP multi-element standard solution XVI (Merck) and rhodium was used as internal standards.

UV/Vis spectroscopy

Absorption spectra were recorded under anaerobic conditions in septum-sealed 1 mm quartz cuvettes at room temperature (∼300 K), using a Shimadzu UV-2600 spectrophotometer. The protein solution was diluted to 1–2 mg/ml. Spectra of reduced HypCD (with 1 mM NaDT) or oxidized HypCD (with 0.2% H2O2) were recorded until no further changes were observed (Supplementary Figure S2). Due to the intense color of PMS, no UV/Vis spectra of PMS-oxidized HypCD were recorded.

Resonance Raman and FTIR spectroscopy

As-isolated, reduced (with 10 mM NaDT), oxidized (with 2% H2O2), and ‘recovered’ samples of HypCD (40–50 mg/ml) were prepared as described above. For each sample, 5 µl protein solution were squeezed between two glass coverslips using a 0.5 mm spacer and sealed with vacuum grease to avoid dehydration and contact with air. Resonance Raman (RR) spectroscopy was performed at room temperature (∼300 K) on a LabRAM spectrometer (HORIBA, Kyoto, Japan) as described previously [34] but using a frequency-stabilized diode-pumped solid-state laser (CNI) emitting at 457 nm. Laser power at the sample was ∼5 mW. The recorded spectra were baseline corrected using the rubber band method implemented in the OPUS software and normalized to the amide I vibration at 1670 cm−1 (Supplementary Figure S3). Due to the intense color of PMS, no RR spectra of PMS-oxidized HypCD were recorded.

Attenuated total reflection (ATR) Fourier-transform infrared (FTIR) spectroscopy was performed in an anaerobic gas chamber (Coy Laboratories) in a continuously purged and effectively water-free atmosphere of N2 and 1% H2. All spectroscopic experiments were performed at room temperature (∼300 K) on a Tensor27 FTIR spectrometer (Bruker Optik) equipped with a three-reflection ZnSe/silicon crystal ATR cell (Smiths Detection) as described previously [35]. Beam path, spectrometer, and anaerobic chamber were purged with dry N2 gas as provided by an Inmatec nitrogen generator. For each experiment, 1 µl of HypCD protein sample (40–50 mg/ml, as-isolated, NaDT-reduced, oxidized with H2O2 or PMS, and ‘recovered’ as indicated above) was pipetted onto the silicon crystal, dried under N2, and re-hydrated in the presence of an aerosol by running the gas mixture through a wash bottle with A. bidest [35,36]. Absorbance spectra were recorded with a spectral resolution of 2 cm−1 and up to 10.000 interferometer scans at 80 kHz scanner velocity.

EPR and Mössbauer spectroscopy

For electron paramagnetic resonance (EPR) spectroscopy different preparations of the HypCD complex were prepared (40–50 mg/ml, as-isolated, NaDT-reduced, oxidized with H2O2 or PMS, and ‘recovered’ as indicated above). Data were recorded with a Bruker Elexsys E580 X-band spectrometer, equipped with an Oxford Instruments ESR900 helium flow cryostat or Bruker ER 167FDS-Q liquid nitrogen finger dewar. A modulation amplitude of 1.0 mT at a frequency of 100 kHz was used.

For Mössbauer spectroscopy, different preparations of the 57Fe-labeled HypCD complex were prepared (40–50 mg/ml, as-isolated, NaDT-reduced, oxidized with H2O2 or PMS). Mössbauer spectra were recorded in the constant acceleration mode with a conventional spectrometer from Wissel GmbH with a multi-channel analyzer in the time-scale mode (WissEL GmbH). Experiments at 77 K were conducted with an LN2 bath cryostat (Oxford Instruments). High-field, low temperature spectra were measured with the same type of spectrometer in a helium closed-cycle cryostat equipped with a superconducting magnet (CRYO Industries of America Inc.) operating with the applied field parallel to the γ-rays. Isomer shifts δ are given relative to α-iron at room temperature. Data were transferred from the multi-channel analyzer to a PC for analysis employing the public domain program Vinda running on an Excel 2003 platform. Magnetically split spectra were simulated with the spin Hamiltonian formalism. Spectra were analyzed by least-squared fits using Lorentzian line shapes.

Microscale thermophoresis

For protein labeling and MST [37] measurements, 100 mM Tris–HCl (pH 8) containing 150 mM NaCl, 10 mM MgCl2, 1% glycerol, and 0.05% Tween 20 was used. Experiments were performed on a Monolith NT.115 (NanoTemper Technologies). The protein sample was labeled using the companys’ protein labeling kit RED-NHS. The labeling reaction was performed according to the manufacturers’ instructions applying a concentration of 40 µM of HypCD complex (molar dye:protein ratio ≈ 3 : 1) at room temperature (∼300 K) for 30 min in the dark. Unreacted dye was removed with the supplied dye removal column. The presence of the [Fe](CN)2CO moiety on the dye-treated HypCD complex was routinely verified by FTIR. Labeled HypCD complex was adjusted to 40 nM with the buffer mentioned above. Then, a series of 16 consecutive 1 : 1 dilutions of the HypE or the HypEF complex (‘ligand’) was prepared using the same buffer, producing ligand concentrations ranging from 10 nM to 1000 µM. For the measurement, each ligand dilution was mixed with one volume of labeled HypCD complex (40 nM). After 20 min incubation followed by centrifugation at 20 000 for 15 min, the samples were loaded onto standard capillaries. Monolith instrument parameters were adjusted to 40–60% LED power and 90–100% MST power. Data of at least three independent measurements were analyzed by NanoTemper analysis software. MST signals with a ‘laser-on’ time of 30 s and a ‘laser-off’ time of 5 s were applied for analysis. For affinity measurements under anaerobic conditions, in the presence of 1 mM NaDT, all samples and dilution series were prepared anaerobically in a COY anaerobic chamber.

Results

Production of the HypCD complex and incorporation of 57Fe

To produce the mature HypCD complex with complete cluster and cofactor content in preparative amounts, an E. coli expression system and an optimized purification protocol were developed. For overproduction of the recombinant HypCD complex, we constructed the plasmid pT-hypDEFCStrepXT with two Strep-tag sequences fused to HypC. The Twin-Strep-tag has much higher affinity to Strep-Tactin-XT resins than the single Strep-tag and thus enabling efficient purification of the HypCD complex with high purity and structural integrity [38]. To confirm the catalytic competence of the pT-hypDEFCStrepXT product for its ability to restore the hydrogenase activity in vivo, the plasmid was introduced into a hypD deletion mutant (DHP-D) and the restored hydrogenase activity was confirmed by specific staining after separation of the enzyme by native-PAGE. The stained bands corresponding to active [NiFe]-hydrogenases from E. coli Hyd-1, Hyd-2, and Hyd-3 are indicated in Figure 2A. Furthermore, we optimized the growth conditions to obtain high yields of the mature HypCD complex (Supplementary Figure S1). Several HypCD preparations were isolated from transformed cells grown in different media. Protein preparations were routinely probed by SDS–PAGE (Figure 2B) and ATR FTIR spectroscopy to identify the [Fe](CN)2CO cofactor precursor (data not shown). We found that both media composition and growth conditions are important for cofactor integrity and protein yield. High yields of partially mature protein were obtained using full medium (TGYEP, 1 g wet cells with ∼12 mg recombinant protein) while M9 minimal medium delivered lower yields but fully functional protein (1 g wet cells contained ∼4 mg recombinant protein). Metal analysis of the mature HypCD complex performed by ICP-MS revealed that protein preparations typically contain five equivalents of iron per protein molecule. The amount of nickel, copper, and zinc were not significant (Table 1). For Mössbauer spectroscopy, two growth media were used to produce the 57Fe-labeled HypCD complex. Both, M9* and TGYEP* allowed good growth rates (Supplementary Figure S1) and delivered similar yields with a 57Fe enrichment of more than 99%. For this, a concentration of 20 µM and 100 µM 57Fe were necessary in the M9* and TGYEP* media, respectively. Only ∼50% enrichment of 57Fe was detected when 20 µM 57Fe were added to TGYEP* medium or when the cells were grown in TGYEP full medium (Table 1). Thus, a full enrichment of the 57Fe-labeled HypCD complex can be achieved with five-fold lower 57Fe concentration by using our chemically defined M9* minimal medium, suggesting that the approaches developed herein can be generally applied for 57Fe enrichment of other metalloproteins.

Figure 2.

Production of the functional HypCD complex.

(A) Native PAGE confirms the catalytic competence of HypCD. Strain DHP-D (ΔhypD) can be complemented in trans by plasmid pT-hypDEFCStrepXT. Samples of crude extracts (∼30 µg of protein) were separated by native PAGE under non-denaturing conditions and stained for hydrogenase activity. Lane 1: complemented DHP-D strain. Lane 2: DHP-D strain. The stained bands corresponding to active Hyd-1, Hyd-2 and Hyd-3 are indicated. A weak hydrogenase-independent activity due to formate dehydrogenase (Fdh) acts as a loading control (*). (B) SDS–PAGE confirms the purity of the strep-tagged HypCD preparations (7 µg of each) from cells grown in different media. M9* (lane 1) and TGYEP* (lane 2) delivered similar yields of fully charged HypCD. TGYEP full medium (lane 3) delivered higher yields but only partially charged HypCD. M, 6 µl of PS10 Plus pre-stained marker.

Table 1

Metal contents of HypCD preparations from cells grown in different media

	mol metal/mol HypCD
	20 µM 57Fe		100 µM 57Fe
	M9*	TGYEP*	TGYEP	TGYEP*
57Fe	4.88 ± 0.31 (n = 5) (99.7%)	4.59 ± 0.20 (n = 5) (49.3%)	4.38 ± 0.18(n = 5) (50.3%)	4.79 ± 0.22 (n = 5) (99.5%)
Ni	0.002 (n = 4)	0.001 (n = 4)	ND (n = 3)	ND (n = 3)
Zn	0.002 (n = 3)	0.002 (n = 3)	0.001 (n = 3)	ND (n = 3)
Cu	0.0003 (n = 3)	ND (n = 3)	ND (n = 3)	0.0002 (n = 3)

The metal content was quantified by ICP-MS. The molar ratio is calculated based on the molecular mass of the HypCD complex (50 kDa). Values in brackets represent percentage of 57Fe-enrichment. All samples were measured with two independent technical replicates. Data are averages of at least three independent determinations (n) and reported as the mean ± SD (standard deviation). ND, not detected.

The redox changes of HypCD are reversible

A dilute solution of the HypCD scaffold complex in the ‘as-isolated’ state (see above) showed the typical UV/Vis absorbance band at 410 nm characteristic for iron-sulfur proteins (Figure 3A) [39]. Resonance Raman spectroscopy confirmed the presence of a [4Fe-4S]+2 cluster (Figure 3B) [40]. Incubation of HypCD with 0.2% H2O2 or 1 mM NaDT led to a sudden decrease in absorbance hinting at an oxidation or reduction event, respectively. We found the effect of H2O2 to be irreversible in a previous study [25]; however, upon buffer exchange it was possible to restore the ‘as-isolated state’. While a mild re-reduction with dithiothreitol (DTT) was necessary to recover the UV/Vis and RR spectra of H2O2-treated HypCD, removal of NaDT immediately restored its absorption properties (Supplementary Figures S2, S3). These results suggest a qualitatively reversibly conversion of redox states, including a NaDT-reduced and H2O2-oxidized iron-sulfur cluster. EPR spectra of the H2O2-incubated sample revealed a [3Fe-4S]+ species (Supplementary Figure S4) that has been discussed as oxidative damage to the cluster previously [26]. Therefore, we screened for an alternative oxidant. Investigating the [Fe](CN)2CO cofactor precursor by ATR FTIR spectroscopy we observed similar band shifts in the presence of H2O2 and PMS (Figure 4 and Supplementary Figures S5, S6). Second derivative spectra of HypCD reveal the typical CO/ CN− signature of the cofactor precursor in the ‘as-isolated’ state (2100, 2072, and 1952 cm–1). Oxidation with 40 eq. PMS (blue traces in Figure 4A) resulted in a shift to higher frequencies (2115, 2092, and 1996 cm–1), indicative of an oxidized [Fe](CN)2CO cofactor precursor [24,27]. The effect of PMS and H2O2 was found to be reversible after buffer exchange and reduction, similar to what was observed by UV/Vis and RR spectroscopy (Supplementary Figures S2, S3). The intense CO2 band that hints at oxidative damage in the presence of H2O2 was not observed with PMS (Supplementary Figure S4). Moreover, we noted a pronounced decline of SH signals [41] in H2O2- and PMS-oxidized sample at 2550 cm–1 compared with as-isolated, NaDT-reduced, and NaDT-restored preparations (Figure 4B and Supplementary Figure S6B). This is likely to reflect disulfide bond formation under oxidizing conditions.

Figure 3.

UV/Vis and Raman spectroscopy.

(A) UV/Vis spectra of the HypCD scaffold complex at pH 9 in as-isolated form (black) and in the presence of 1 mM NaDT (red). Note the decrease in the [4Fe-4S] cluster band at 410 nm upon reduction. UVA bands include 280 nm (* to protein) and 315 nm (** to NaDT). (B) Resonance Raman spectra of the HypCD complex at pH 9 in as-isolated form (black) and in the presence of 10 mM NaDT (red). Note the decrease in the terminal (400 and 360 cm−1) and bridging Fe-S bands (335 and 245 cm−1) upon reduction. These feature confirm the assignment of a [4Fe-4S]2+ cluster on as-isolated HypCD.

Figure 4.

ATR FTIR spectroscopy.

(A) Second derivative ATR FTIR spectra of the HypCD complex in as-isolated form (black), in the presence 40 eq. PMS (blue), or restored with 10 mM NaDT after reduction (red). Note the shift of the CO/CN− bands to higher frequencies upon oxidation with PMS, that hints at oxidation of the cofactor precursor from [Fe]2+ to [Fe]3+. As-isolated and NaDT-restored samples show the same band positions. (B) As-isolated (black) and NaDT-reduced HypCD (red) shows a clear band at 2550 cm−1 that can be assigned to SH vibration of cysteine residues. In the presence of 20 or 40 eq. PMS the absorbance decreased (cyan and blue traces), indicative of disulfide bond formation. This process is reversible, as highlighted by the recovery of the SH signal in NaDT-restored sample (dashed traces).

Summing up, our analysis of the HypCD scaffold complex suggests the existence of at least three redox states. We assume that the iron-sulfur cluster may interconvert between [4Fe-4S]+2 (as-isolated) and [4Fe-4S]+ (NaDT-reduced) whereas the mononuclear iron center may change between [Fe]2+ (as-isolated) and [Fe]3+ (PMS-oxidized). Additionally, EPR spectroscopy identified an ‘overoxidized’ [3Fe-4S]+ cluster in some preparations, which hints at damaged HypD. In order to gain direct evidence for the existence of intact reduced and oxidized metal centers, we performed further experiments.

Characterization of the reduced and oxidized metal centers

The [4Fe-4S]2+ cluster of the HypCD scaffold complex is diamagnetic and therefore EPR-inactive. In an earlier attempt to characterize the HypD monomer by Blokesch and colleagues, traces of a [3Fe-4S]+ cluster were detected [26], in agreement with our data (Supplementary Figure S4). While these signals vanished upon reduction with 50 mM NaDT at pH 8, no S = 1/2 or high spin EPR signal of the reduced [4Fe-4S]+ cluster was observed [26]. Being aware that the redox potential of NaDT is pH-dependent [42], we adjusted the sample to pH 10 and probed the HypCD complex in the absence and presence of 10 mM NaDT (Figure 5A). X-band EPR spectroscopy at low temperature now revealed a sharp rhombic EPR signal with g-values 2.062, 1.935, and 1.911, which is typical for a [4Fe-4S]+ cluster [43]. Only minor reduction was observed at pH 8. Moreover, we investigated a sample of the HypCD complex in the presence of 40 eq. PMS (Figure 5A). Only an intense sharp isotropic signal, which under the conditions employed for the detection of Fe/S EPR signals does not show hyperfine couplings due to the modulation amplitude presumably from PMS radicals [44] was detected. No traces of iron-sulfur clusters like [3Fe-4S]+ or [4Fe-4S]3+ were observed [26,45]. This confirms the utility of PMS as a site-selective oxidant of the mononuclear iron center (Figure 4A) and suggests that the ‘overoxidized’ iron-sulfur cluster is biologically not relevant. The lack of damage to the [4Fe-4S] cluster in the presence of PMS is additionally highlighted in Figure 5B. Here, the [4Fe-4S]+ spectrum of HypCD after direct reduction (red) is compared with the [4Fe-4S]+ spectrum after PMS-mediated oxidation and re-reduction (dashed traces). With 20 eq. of PMS, full recovery was observed.

Figure 5.

EPR spectroscopy.

(A) X-band EPR spectra at 10 K of the HypCD complex in as-isolated form (diamagnetic, black) and NaDT-reduced form at pH 8 (dashed red) and pH 10 (red). The g-values of the reduced [4Fe-4S] cluster are annotated. In the presence of PMS, only the phenanzine free radical is observed (blue). (B) Comparison of NaDT-reduced (red trace) and NaDT-restored HypCD (dashed traces) at pH 9. The former was reduced from as-isolated sample directly. The later was treated with PMS (20 or 40 eq.), washed, and reduced with NaDT to restore the [4Fe-4S]+ state. The signal at g = 2.007 hints at a dithionite radical.

For the first time, EPR spectroscopy identified the reduced form of the iron-sulfur cluster on the HypCD scaffold complex, corroborating its redox activity. However, while ATR FTIR spectroscopy suggested at least two possible redox states of the [Fe](CN)2CO cofactor precursor (Figure 4), no direct identification of its redox states has been obtained yet. To provide a comprehensive description of all metal sites on the functional HypCD complex, we enriched the protein in 57Fe and performed Mössbauer spectroscopy in the temperature range from 4.2–200 K and within an external B-field of 0.0–5.0 T (Supplementary Figures S7, S8). Figure 6A shows the Mössbauer spectrum of the as-isolated HypCD complex at 77 K and 0 T, revealing the presence of two distinct species. The main part of the spectrum is a quadrupole doublet (black traces) comprised of two components each representing a Fe2.5+-Fe2.5+ pair of a [4Fe-4S]2+ cluster [46]. Mössbauer spectra from 4.2–200 K (Supplementary Figure S8) show a temperature dependent quadrupole splitting of the valence delocalized Fe2.5+-Fe2.5+ pairs of the [4Fe-4S] cluster as observed by Middleton and colleagues [47] and subsequently reported for multiple [4Fe-4S] clusters [48]. While our data are in excellent agreement with the Mössbauer spectra of HypCD reported by Blokesch and colleagues [26], an additional component was included in the description, accounting for a shoulder at ∼–0.5 mm/s (magenta trace). The assignment of the [Fe](CN)2CO cofactor is in accordance with a similar species in the [NiFe]-hydrogenase of Ralstonia eutropha [49]. Mössbauer spectroscopy in an external B-field of 5.0 T (Figure 6B) confirms the diamagnetic state of the [4Fe-4S]2+ cluster and identifies the mononuclear iron site as a diamagnetic low-spin [Fe]2+ with δ = 0.22 mm/s and ΔEQ = 1.62 mm/s. Upon addition of 10 mM NaDT, the [4Fe-4S]2+ cluster was reduced to a paramagnetic [4Fe-4S]+ cluster (Figure 6C,D), whereas the [Fe](CN)2CO moiety showed no detectable changes. The two components representing the [4Fe-4S]+ cluster (red traces) can be assigned to a Fe2.5+-Fe2.5+ pair (δ = 0.43 mm/s and ΔEQ = 1.10 mm/s) and a Fe2+-Fe2+ pair (δ = 0.59 mm/s and ΔEQ = 1.88 mm/s). No oxidation of the [4Fe-4S]2+ cluster was observed in the presence of 40 eq. PMS (Figure 6E,F), in agreement with our EPR data. However, in an external B-field of 5.0 T, we were able to identify a component that resembles a high-spin [Fe]3+ site (δ = 0.26 mm/s and ΔEQ = 0.81 mm/s). The Mössbauer parameters are summarized in Table 2.

Figure 6.

Mössbauer spectroscopy.

Spectra of the 57Fe-labeled HypCD complex (40–50 mg/ml at pH 9) in as-isolated (A and B), NaDT-reduced (C and D), and PMS-oxidized form (E and F) obtained at 77 K (A,C,E) or 4.2 K and an external B-field or 5.0 T (B,D,F). Legend: experimental data (open circles), sum of simulations (orange trace), [4Fe-4S]2+ (1 + 2, black), [4Fe-4S]+ (3 + 4, red), [Fe]2+ (5, magenta), and [Fe]3+ (6, blue). See Table 2 for further details.

Table 2

Mössbauer parameters

Metal center	[4Fe-4S]2+						[4Fe-4S]1+						[Fe]
	Fe2.5+-Fe2.5+			Fe2.5+-Fe2.5+			Fe2+			Fe2+			Fe2+	Fe3+
Subspectrum	1			2			3			4			5	6
T (K)	200	77	4.2	200	77	4.2	200	77	4.2[1]	200	77	4.2[1]	n.d.[2]	n.d.[2]
δ (mm s−1)	0.36	0.41	0.41	0.36	0.42	0.42	0.40	0.43	0.44	0.54	0.59	0.61	0.24	0.26
ΔEQ (mm s−1)	0.52	0.70	0.76	0.81	1.08	1.25	1.05	1.10	0.76	1.60	1.88	1.25	1.62	0.81
Γ (mm s−1)	0.32	0.36	0.50	0.34	0.36	0.50	0.38	0.46	0.46	0.40	0.46	0.37	0.35	0.61

The spectral components of the [4Fe-4S]1+ cluster at 4.2 K were simulated using an effective spin S = 1/2 and the spin-Hamiltonian formalism in the slow relaxation limit. The hyperfine coupling tensor /µNgN at 4.2 K for subspectrum 3 has been taken as −230, −245, and −220 kG and for subspectrum 4 a 210, 110, and 85 kG. For the field spectra the simulations representing the [Fe](CN)2CO moiety and the [4Fe-4S]2+ cluster were obtained assuming a diamagnetic state;

A temperature dependence for the mononuclear iron site could not be detected. The Fe3+ component was simulated with a zero field splitting D = 2 cm−1 and a rhombicity E/D = 0.33 as well as /µNgN = (−200, −200, −200) kG in the slow relaxation limit for B = 5T and in the fast relaxation limit at B = 0.1T (Supplementary Figure S7);

Mössbauer parameters of the metal centers of the as-isolated, reduced, and oxidized form of the HypCD complex as obtained from the simulations given in Figure 6 and Supplementary Figure S7, S8. δ: Isomer shift; ΔE: Quadrupole splitting; Γ line width at half maximum; A: Hyperfine coupling tensor.

The affinity between HypCD and HypE(F) dependents on the redox state

MicroScale thermophoresis was applied to determine the binding affinities of the HypCD complex to its in vivo interaction partner HypE. The concentration of dye-labeled HypCD complex was kept constant (40 nM) and mixed with increasing concentrations of the binding partner HypE. The binding affinities were determined under aerobic or reducing conditions (1 mM NaDT). Figure 7A depicts the HypE dependence of the MST signal for dye-labeled HypCD. The MST signals were fitted according to Michaelis–Menten kinetics to obtain the apparent binding affinity between HypCD and HypE (solid lines). The data show that HypCD displayed a higher binding affinity to HypE under reducing conditions (Kd = 4.67 ± 1.5 µM) compared with aerobic conditions (Kd = 19.5 ± 3.7 µM). Similar results were obtained for the interaction between HypCD and the HypEF complex (Figure 7B), indicating preferable binding under reducing conditions with Kd = 9.93 ± 1.9 µM compared with aerobic conditions (Kd = 126 ± 32 µM). Based on the crystal structure of the HypCDE [50] and HypEF complexes [51], it seems that interface between HypE and HypF overlaps with that between HypE and HypCD, which may explain the approximately six-fold higher binding affinity of HypE to HypCD in the absent of HypF under aerobic conditions. Conceivably, the regulation of the maturation process is based on specific molecular interactions between individual subunits. The difference in Kd is less pronounced under reducing conditions, but the trend is the same. Although the exact nature of the interactions and the temporal sequence of events has not been resolved in detail, the binding and release of HypE from the HypCD complex must be initiated by an interaction with other maturation protein(s) or by the delivery of the [Fe](CN)2CO cofactor precursor to the pro-protein.

Figure 7.

Redox dependency of the binding affinities between HypCD and HypE(F).

(A) HypE dependence of the MST signal for the fluorescent-labeled HypCD complex under aerobic conditions (black) or reducing conditions (1 mM NaDT, red). (B) HypEF dependence of the MST signal for the fluorescent-labeled HypCD complex under aerobic conditions (black) or reducing conditions (1 mM NaDT, red). The MST signals of at least 12 different concentrations of HypE or HypEF ranging from 0.01–1000 µM were fitted according to Michaelis–Menten kinetics. The insets in panel (A) and (B) depict the obtained binding affinities, Kd. Data of at least three independent measurements (n) were combined to create one dataset for which the averaged data points as the mean ± SD (standard deviation) were fitted by Nano Temper analysis software using a Kd-model. Our data show significantly higher binding affinities of HypCD to HypE and HypEF under reducing conditions.

Discussion

The biosynthesis of the [Fe](CN)2CO cofactor precursor is a complex and dynamic process catalyzed by the coordinated activity of HypCDEF [12,13]. While HypE and HypF catalyze the synthesis of the CN− ligand through carbamoylation and cyanation of the C-terminal cysteine of HypE [22], the maturation proteins HypC and HypD form the HypCD scaffold complex, where the CN− and CO ligands are attached to the iron ion to form the [Fe](CN)2CO cofactor precursor [12,27]. It has been hypothesized that the input of two electrons is a prerequisite for the reductive transfer of each CN− ligand from HypE to the HypCD complex [26,28,52], and minimally two further electrons would be required for the generation of CO if the substrate is CO2 [20]. Among the four maturation proteins only HypD contains an iron-sulfur cluster and is therefore the only Hyp protein capable of performing redox chemistry; however, not much is known about the mechanism by which the CN− and CO ligands are transferred and attached to the iron atom on HypCD. Therefore, an efficient expression and purification system was developed to produce the HypCD complex with complete cluster and cofactor content. The successful enrichment with 57Fe suggests that the approaches developed herein can be generally applied for other metalloproteins.

The crystal structure of HypD from Thermococcus kodakaraensis (TkHypD) revealed the existence of a redox cascade composed of an [4Fe-4S] cluster and two pairs of disulfide motifs SS1 and SS2 [50,53]. According to the computational model from Albareda and colleagues, the cascade also involves the [Fe](CN2)CO moiety at the interface of HypD and HypC [54]. Although the disulfide bond SS1 is not conserved in HypD from E. coli [55], the distance between the [4Fe-4S] cluster and the conserved disulfide motif SS2 is short enough to facilitate electron transfer along the redox cascade (Figure 8A), e.g. facilitating reductive ligand transfer to the iron site [54]. This redox cascade has been proposed to provide reducing equivalents for the synthesis and attachment of ligands to the cofactor precursor, in agreement with our infrared data and the faradaic currents measured upon oxidation and reduction in disulfide motif SS2. Formation and excision of SS2 on the HypCD complex from E. coli is indicative of a two-electron reaction that comprises two consecutive one-electron transfer steps [56,57]. Our data now confirm the redox activity of the [4Fe-4S] cluster and suggest an electron confurcation mechanism that involves two one-electron transfer steps to form the reduced disulfide motif SS2red (Figure 8B). As discussed earlier [50,52,53], oxidation of SS2red would supply two electrons for the synthesis of each diatomic ligand (Figure 8C).

Figure 8.

Proposed mechanism for the assembly of the [Fe](CN)2CO cofactor precursor.

(A) The distance between [4Fe-4S] cluster, disulfide motif SS2, and mononuclear iron site allows for efficient electron transfer. The disulfide bond (SS2ox) correlates with an oxidized iron site, [Fe]3+. (B) Reduction in the [4Fe-4S]2+ cluster by an external electron donor to form [4Fe-4S]+ for the reduction in disulfide bridge SS2ox to SS2red (red) in two consecutive one-electron transfer steps. (C) Modification of the iron site with one CO and two CN− ligands at the expense of SS2red to form the [Fe]2+(CN)2CO cofactor precursor. Per ligand, two electrons are consumed but the exact number of electrons is unclear.

Previous studies demonstrated that the CN− transfer did not occur when the HypCD complex was treated with oxidizing agents or was purified under aerobic conditions, in which the disulfide bond was formed [26,28,52]. The inhibition of CN− transfer under oxidizing conditions seems to be due to the formation of disulfide bond SS2ox, which would interrupt the redox cascade (Figure 8). This result is in agreement with our biomolecular interaction studies that indicate up to 12-fold higher binding affinity between HypCD and HypEF under reducing conditions. The preferable binding under reducing conditions is likely to facilitate electron transfer during [Fe](CN)2CO cofactor synthesis. Furthermore, in the absence of reductant or under aerobic conditions, the cofactor of [NiFe]-hydrogenase cannot be assembled in vitro [12,19,58]. We conclude that NaDT serves as sacrificial reductant in the in vitro assay, replacing the role of the physiological electron donor of HypD.

The finding that HypC coordinates an iron-bound CO2 molecule suggests that the HypC superfamily may deliver both iron and CO2 for cofactor biosynthesis [20]. Our working hypothesis comprises reduction in CO2 with 2 e− and 2 H+ to CO and H2O as catalyzed by the HypCD complex; however, the proposed reaction requires a low redox potential (E0′ ≈ −530 mV). It has been demonstrated that the HypCD complex catalyzes the hydrolysis of ATP [21], which would help to overcome such thermodynamic barriers, not unlike CO dehydrogenase [59], the ‘iron protein’ of nitrogenase [60], or enoyl-CoA reductase [61]. Under aerobic conditions, the CO ligand is generated by decarbonylating formyltetrahydrofolate [62,63]. However, the metabolic source in strict or facultative anaerobes must be different. To this end, our spectroscopic evaluation of the HypCD scaffold complex helps to define the different redox states of the [4Fe-4S] cluster and the [Fe](CN)2CO cofactor precursor. The data demonstrate that the [4Fe-4S] cluster is redox active and provides comprehensive evidence for an electron inventory fit to drive multi-electron redox reactions required for the assembly of the CN− and CO ligands on the HypCD complex.