Article pubs.acs.org/IC
Cite This: Inorg. Chem. XXXX, XXX, XXX−XXX
Differential Protonation at the Catalytic Six-Iron Cofactor of [FeFe]Hydrogenases Revealed by 57Fe Nuclear Resonance X-ray Scattering and Quantum Mechanics/Molecular Mechanics Analyses Stefan Mebs,*,† Jifu Duan,§ Florian Wittkamp,∥ Sven T. Stripp,‡ Thomas Happe,§ Ulf-Peter Apfel,*,∥,⊥ Martin Winkler,*,§ and Michael Haumann*,†
Downloaded via MACQUARIE UNIV on February 27, 2019 at 10:16:39 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
†
Department of Physics, Biophysics of Metalloenzymes and ‡Department of Physics, Experimental Molecular Biophysics, Freie Universität Berlin, Arnimallee 14, 14195 Berlin, Germany § Department of Biochemistry of Plants, Photobiotechnology and ∥Department of Chemistry and Biochemistry, Inorganic Chemistry I, Ruhr-Universität Bochum, Universitätsstrasse 150, 44801 Bochum, Germany ⊥ Fraunhofer UMSICHT, Osterfelder Straße 3, 46047 Oberhausen, Germany S Supporting Information *
ABSTRACT: [FeFe]-hydrogenases are efficient biological hydrogen conversion catalysts and blueprints for technological fuel production. The relations between substrate interactions and electron/proton transfer events at their unique six-iron cofactor (H-cluster) need to be elucidated. The H-cluster comprises a four-iron cluster, [4Fe4S], linked to a diiron complex, [FeFe]. We combined 57Fe-specific X-ray nuclear resonance scattering experiments (NFS, nuclear forward scattering; NRVS, nuclear resonance vibrational spectroscopy) with quantum-mechanics/molecular-mechanics computations to study the [FeFe]-hydrogenase HYDA1 from a green alga. Selective 57Fe labeling at only [4Fe4S] or [FeFe], or at both subcomplexes was achieved by protein expression with a 57Fe salt and in vitro maturation with a synthetic diiron site precursor containing 57Fe. H-cluster states were populated under infrared spectroscopy control. NRVS spectral analyses facilitated assignment of the vibrational modes of the cofactor species. This approach revealed the H-cluster structure of the oxidized state (Hox) with a bridging carbon monoxide at [FeFe] and ligand rearrangement in the CO-inhibited state (Hox-CO). Protonation at a cysteine ligand of [4Fe4S] in the oxidized state occurring at low pH (HoxH) was indicated, in contrast to bridging hydride binding at [FeFe] in a one-electron reduced state (Hred). These findings are direct evidence for differential protonation either at the four-iron or diiron subcomplex of the H-cluster. NFS time-traces provided Mössbauer parameters such as the quadrupole splitting energy, which differ among cofactor states, thereby supporting selective protonation at either subcomplex. In combination with data for reduced states showing similar [4Fe4S] protonation as HoxH without (Hred′) or with (Hhyd) a terminal hydride at [FeFe], our results imply that coordination geometry dynamics at the diiron site and proton-coupled electron transfer to either the four-iron or the diiron subcomplex discriminate catalytic and regulatory functions of [FeFe]-hydrogenases. We support a reaction cycle avoiding diiron site geometry changes during rapid H2 turnover.
■
INTRODUCTION [FeFe]-hydrogenases are nature′s most efficient hydrogen (H2) forming enzymes.1,2 Reversible H2 oxidation and proton reduction (2H+ + 2e− ↔ H2) are catalyzed at their unique sixiron cofactor (H-cluster).3−5 The H2 conversion chemistry at the H-cluster needs to be understood to promote biotechnological application of the enzymes6,7 and development of improved synthetic catalysts for hydrogen fuel production.8−10 The cofactor consists of a four-iron cubane cluster, [4Fe4S], cysteine-linked to a diiron complex, [FeFe] (Figure 1). [FeFe] shows coordination of its proximal (Fep) and distal (Fed) iron ions by two terminal cyanide and carbon monoxide ligands and by a metal-bridging CO (μCO) in the oxidized resting state of the cofactor (Hox).11−14 Hox comprises mixed-valence [4Fe4S] and [FeFe] sites ([4Fe4S]2+-[FeFeI,II]).15,16 The © XXXX American Chemical Society
iron centers of [FeFe] are bridged by an azadithiolate group (adt = (SCH2)2NH).17−19 A vacancy at Fed, which is at the apical position in crystal structures of Hox, is the active site of H2 conversion as well as the target of inhibitory reactions induced by CO or O2 gas exposure (Figure 1).1,2,20 Hox represents the starting point of the catalytic cycle of H2 conversion.1,2,21 Reduction of the H-cluster, i.e. by using H2, redox agents, or electrochemical approaches, populates a variety of cofactor states.1,2,22−24 Each two one-electron (Hred′, Hred) or two-electron (Hhyd, Hsred) reduced states (relative to Hox) were spectroscopically discriminated.16,25−36 Hox and Hred′ are inhibited due to CO gas exposure to form Received: January 11, 2019
A
DOI: 10.1021/acs.inorgchem.9b00100 Inorg. Chem. XXXX, XXX, XXX−XXX
Article
Inorganic Chemistry
clusters besides the H-cluster.3,4 Purified HYDA1 apoprotein binding only [4Fe4S] is obtained after heterologous overexpression in Escherichia coli.49,50 Expression of apoprotein with 56Fe or 57Fe salts and in vitro maturation with a synthetic, i.e., 57Fe labeled diiron site analogue (2Feadt = Fe2(μadt)(CO)4(CN)2) facilitates the assembly of the fully active H-cluster18,29,51−53 and results in selective insertion of 57Fe instead of 56Fe at [4Fe4S], [FeFe], or both subcomplexes.32,54 Various H-cluster states can then be populated in HYDA1 under solution conditions using, e.g., H2 or CO gas exposure, as well as pH variation.32,44,55−58 Nuclear resonance scattering (NRS) is a synchrotron-based spectroscopic technique relying on the Mössbauer effect, that is, resonant excitation of 57Fe nuclei using ∼14.4 keV Xrays.59−62 The excitation creates a 1s level photoelectron due to nuclear excited state decay followed by core hole refill from higher electron levels (e.g., Fe-2p) and X-ray fluorescence photon emission (e.g., Fe Kα at ∼6.4 keV), which is employed to monitor the NRS effects in a time-resolved (nanoseconds) detection scheme including a high-resolution (meV) monochromator. Site-selective studies on the H-cluster are facilitated by the exclusive 57Fe sensitivity of NRS.28,32,54,63−65 Nuclear forward scattering (NFS) probes coherent emission interference during decay (lifetime 141 ns) of the I1/2 and I3/2 57Fe excited nuclear spin levels to access Mössbauer parameters (quadrupole splitting energy, ΔEQ, and line width, Γ).66 Nuclear resonance vibrational spectroscopy (NRVS) probes excitation or annihilation of phonons in the Stokes or antiStokes energy regions close to the 57Fe resonance to monitor vibrational modes.60,61,67,68 NRVS is versatile because it is not limited by the Raman or infrared spectroscopy selection rules, but shows all modes with 57Fe ion contributions, thereby probing the primary iron-ligand vibrations as well as cofactorprotein modes. NRS on dilute biological samples (mM metal concentration) is technically demanding because of low X-ray emission yields. Assignment of vibrational bands in NRVS spectra requires normal-mode analysis (spectral calculations) based on molecular models as derived from quantum mechanics/molecular mechanics (QM/MM) or density functional theory (DFT) approaches, which was established for [FeFe]-hydrogenases earlier.28,32,63,64 Here, [4Fe4S], [FeFe], or both subcomplexes in [FeFe]hydrogenase HYDA1 were labeled with 57Fe. Various H-cluster states were populated (Hox, HoxH, Hred) and NFS time traces as well as NRVS spectra were acquired. Mössbauer parameters of the H-cluster states were derived from the NFS data. QM/MM provided assignments for NRVS vibrational modes and cofactor structures. The NRVS spectral changes relative to Hox were consistently described by [4Fe4S] protonation in HoxH in contrast to bridging hydride binding at [FeFe] in Hred. These findings support a rapid H2 conversion cycle of [FeFe]-hydrogenases including only Hcluster species with a similar cofactor geometry (Hox/HoxH, Hred′, Hhyd) and reversible [4Fe-4S] cluster as well as diiron site protonation.
Figure 1. Crystal structure of the H-cluster in [FeFe]-hydrogenases. The structure is for Clostridium pasteurianum enzyme (CPI) in the Hox state (Protein Databank (PDB) entry 4XDC, 1.63 Å resolution11). Amino acids are indicated, which contribute to the proton paths leading to [FeFe] or [4Fe4S], or neighbor the CO/CN− ligands (CPI numbering; Ser232 is Ala and Cys299 is Cys169 in [FeFe]-hydrogenase protein from C. reinhardtii (HYDA1). The bridging group is aminodithiolate (adt = (SCH2)2NH). Red spheres show water species.47 The arrow points to the open coordination site at Fed.
Hox-CO and Hred′-CO.37−41 We have shown that the COinhibited states bind an apical CN− ligand at Fed37,38 as opposed to an earlier suggested apical CO ligand,20,39 which emphasizes the configurational flexibility of the diiron site. The structure of the H-cluster regarding the [FeFe] geometry and cofactor protonation in the reduced states is much debated.1,2,21,42,43 Redox titrations have shown a decrease of the midpoint potential of about 60 mV per pH unit in Pourbaix diagrams of the Hox to Hred′ and Hox to Hred transitions, meaning that they represent one electron/one proton reactions.27,35 Our earlier spectroscopic and computational work has suggested that Hred and Hsred carry a metalbridging hydride (μH−) at the diiron site,36,44,45 which contrasts geometries that lack μH− suggested for these states by other authors.25,27 In Hhyd, an apical hydride at Fed was consistently assigned.16,28,30−33,46 However, our data implied additional protonation at a cysteine ligand of [4Fe4S] in Hhyd (Figure 1)32 and similar protonation was suggested for an oxidized state (HoxH) as well as for Hred′ and Hred′CO.35,37,47 HoxH may represent an intermediate in Hhyd to Hox conversion after H2 release, which accumulates when [4Fe4S] deprotonation is impaired at increased proton concentration (low pH) due to preprotonation of the adjacent base (possibly a water molecule), which serves as proton acceptor/donor of the [4Fe-4S] cluster (for detailed discussion see refs.2,47). Clarification of the H-cluster structures, in particular with regard to protonation events at the two subcomplexes, is a central point to judge on the role of the various states in the catalytic cycle or in potential regulatory and sensory reactions of [FeFe]-hydrogenases, for which conflicting hypotheses have been put forward.1,2,30,33,42,48 The [FeFe]-hydrogenase HYDA1 from the green alga Chlamydomonas reinhardtii is superior for focusing on the catalytic cofactor using (iron-based) spectroscopy because, in contrast to bacterial enzymes, it carries no further iron−sulfur
■
MATERIALS AND METHODS
Protein Sample Preparation. [FeFe]-hydrogenase HYDA1 apoprotein (wild-type) containing only [4Fe4S] was overexpressed in Escherichia coli BL21 ΔiscR,69 without the maturases HydE/F/G, as described previously.51 For 57Fe labeling of [4Fe4S], 51 mg 57FeBr2 was added as iron precursor per liter cell culture immediately before induction instead of 2 mM ferric ammonium citrate. The [457Fe4S] or B
DOI: 10.1021/acs.inorgchem.9b00100 Inorg. Chem. XXXX, XXX, XXX−XXX
Article
Inorganic Chemistry
(I = NFS amplitude, Ai = amplitude scaling factor, τι = effective decay lifetime, υi = modulation frequency, φ = phase shift, B = detector count offset) and calculated according to ΔEQ = hυ (h = 4.135 × 10−6 neV s).73−75
[456Fe4S] containing apo-proteins were maturated to yield active holo-protein in vitro, using either 256Feadt or 257Feadt (2Feadt = Fe2(μadt)(CO)4(CN)2), which yielded either [457Fe4S]-[56FeFe] or [457Fe4S]-[57FeFe] or the corresponding [456Fe4S]-[57/56FeFe] variants, respectively.18,32,49,50,68 2Feadt was synthesized with 56Fe or 57 Fe by following established routes.18,51,70 The protein preparation and handling procedures were carried out under strictly anoxic conditions and dim light. All HYDA1 preparations showed similar H2 evolution activities (∼800 μmol H2 mg−1 min−1) and close to six iron ions per protein, as previously reported,32,71 indicating quantitative incorporation of the complete H-cluster. H-cluster states in HYDA1 samples (∼40 μL, ∼2 mM protein) were populated at three different pH values in NRS sample holders under infrared spectroscopy control32 using varying gas compositions in atmospheres humidified in an anoxic glovebox with buffers adjusted to the corresponding pH.47 Hox or HoxH were gained by flushing of as-isolated HYDA1 suspended in buffers at pH 8.0 or 6.0 for ∼1 h with N2 gas. Hox-CO resulted from flushing of HYDA1 poised initially in Hox (pH 8.0) for ∼30 min with CO gas. Hred was accumulated by flushing as-isolated HYDA1 (pH 7.0) for ∼1.5 h with H2 gas. Immediately after the treatments, samples were rapidly frozen in liquid nitrogen within holders for NRS in the glovebox. FTIR Spectroscopy. Fourier-transform infrared spectroscopy (FTIR) was performed on hydrogenase protein films (1 μL) in attenuated total reflection geometry (ATR) on a Tensor27 spectrometer (Bruker) placed in an anoxic glovebox.38,47 H-cluster state population in NRS samples was monitored according to the vibrational bands of the CO and CN− ligands by taking aliquots at increasing time intervals during the gas exposure procedures and attenuated total reflection Fourier-transform infrared spectroscopy (ATR-FTIR) data collection (2 cm−1 resolution) in less than 1 min. FTIR spectra were evaluated using in-house software and leastsquares fit algorithms to quantify the relative H-cluster state populations in the NRS samples (Supporting Information, Figure S1).38,47 Notably, while the IR-detected Hox, HoxH, and Hox-CO populations most likely represent the actual state populations in the NRS samples, the Hred population may represent the lower limit, due to partial conversion to other states (mostly Hox) during transfer of the protein solution from the NRS sample holder to the ATR cell of the FTIR spectrometer. Nuclear Resonance Spectroscopy. NRS data were collected at undulator beamline ID18 at the European Synchrotron Radiation Facility (ESRF, Grenoble, France)72 using the previously described setup including a heat-load monochromator, a high-resolution monochromator (fwhm ∼0.7 meV), gated APD detectors (∼1 cm2 active area) for prompt and delayed inelastic and forward scattering detection, and a coldfinger liquid-helium cryostat.32,68 The storage ring was operated in 16-bunch mode (∼90 mA). A sample temperature during the measurements of 50 ± 15 K was estimated from the ratio of NRVS counts in ±3−7 meV windows around the resonance.68 NFS time traces were collected in a ∼160 ns time window within ∼30 min with 4 stacked APD detectors at ∼2 m behind the sample. NRVS counts were detected with an APD at 90° to the incident X-ray beam and at ∼4 mm distance from the sample. The energy axis of the high-resolution monochromator was calibrated using the CN− vibrational band at 70.0 meV of a (NH4)2Mg57Fe(CN)6 powder sample as a reference (Figure S4). NRVS spectra were collected in a −15 to 110 meV energy region around the resonance (0.2 meV steps, 3 s per data point, spot size on sample ∼1.5 × 0.5 mm2) and up to 40 scans were averaged for signal-to-noise ratio improvement (5 scans of ∼30 min per sample spot). NRVS data were processed and the partial vibrational density of states (PDOS) was calculated with the software package available at ID18 (Figure S4). Final NRVS spectra were derived by smoothing of averaged, raw spectra by adjacent averaging over 9 data points (1.8 meV) and interpolation to a 0.4 meV energy step axis, followed by energy to frequency axis conversion (1 meV = 8.06554 cm−1) to give an effective band resolution of ∼15 cm−1 (full width at half-maximum, fwhm). Quadrupole splitting energies (ΔEQ) were determined from fit analysis of NFS traces using eq 1 with one or two components (i)
INFS(t ) =
∑ Ai exp(−ti/τi) cos2(πνit + φ) + B
(1)
The apparent Mössbauer line width (Γ) was calculated from τ using eq 2 (excited state lifetime, τ0 = 141 ns; 1 mm s−1 = 4.8 × 10−5 meV):
Γ = h/2π(1/τ0 + 1/τ )
(2)
QM/MM Calculations. Calculations were carried out on the Soroban computer cluster of Freie Universität Berlin. They involved model structures as constructed using the crystal structure of [FeFe]hydrogenase (CPI) from Clostridium pasteurianum (PDB entry 4XDC, 1.63 Å resolution)11 as previously described47 (Figure S6). A quantum mechanics/molecular mechanics (QM/MM) approach including ONIOM76,77 and the universal force field as implemented in the Gaussian09 program78 were used for the MM treatment of the protein environment (low-layer) and the BP86 functional with the TZVP basis-set79−82 was used for the QM core (high-layer including the H-cluster and adjacent amino acids) for unconstraint geometryoptimization. A broken-symmetry approach and proper assignment of molecular fragments were used for calculation of antiferromagnetic spin couplings.32,36,47 For Hox, the total spin multiplicity (M = 2S+1) was 2 and the total charge of the H-cluster was −3. For other species, multiplicities and charges reflected the number of added electrons and protons. Vibrational frequencies in the IR and NRVS regions were derived from normal-mode analysis of relaxed structures using Gaussian09. NRVS and normalized PDOS spectra were calculated using NISspec.83 Calculated (stick) IR spectra were broadened with fwhm values derived from fits of experimental spectra using Voigt functions (50% Gaussian and Lorentzian characters) and calculated and experimental IR intensities were normalized to a sum of 100% over all CO/CN− ligands;36 vibrational modes in PDOS spectra were broadened by Lorentzians (fwhm 15 cm−1) and the energy axis was scaled by a factor of 0.99 for comparison with experimental spectra. The root-mean-square deviation (rmsd) between experimental and calculated vibrational frequencies (from analysis of NRVS or IR spectra) was calculated using eq 3 (n = number of included vibrational modes):38,47 rmsd =
∑ (Fcal − Fexp)2 /n n
(3)
IR frequencies (Fcor), which were corrected for theory-inherent deviations from ideal correlation with experimental data using the offset and slope values from a linear fit to data in a Fcal vs Fexp plot (not shown), were derived using eq 4,36,47 Fcor = (Fcal − offset)/slope
(4)
and used alternatively for rmsd calculation. In-house software and functionalized EXCEL-sheets were used to process Gaussian09 and NISspec output files. NRVS vibrational modes were visualized with ChemCraft.
■
RESULTS H-Cluster State Populations from Infrared Spectroscopy. The H-cluster in purified [FeFe]-hydrogenase HYDA1 (wild-type protein) was labeled with 57Fe using our earlier established maturation procedures, including reconstitution with a synthetic diiron compound (Fe2(μ-adt)(CO)4(CN)2, adt = (SCH2)2NH), to yield fully functional enzyme.32,51,52,55 This approach provided HYDA1 with 57Fe only in [4Fe-4S] (57/56) or [FeFe] (56/57) (and 56Fe in the corresponding subcomplex) as well as in both subcomplexes (57/57). Selected H-cluster states, namely Hox, HoxH, or Hred, and Hox-CO serving as a reference, were accumulated in the 57Fe C
DOI: 10.1021/acs.inorgchem.9b00100 Inorg. Chem. XXXX, XXX, XXX−XXX
Article
Inorganic Chemistry labeled proteins under N2, H2, or CO gas atmosphere. Hcluster state populations in samples for NRS were determined using infrared spectroscopy and quantitative evaluation of the absorption band patterns due to the CO/CN− ligand stretching vibrations, as established earlier32,36−38,47 (Supporting Information; Figure S1, Table S1). The following state populations were obtained (±5%): Hox, 78%; HoxH, 82%; Hred, 63%; and Hox-CO, 87%. About 10% Hox-CO was found also in the samples not exposed to CO gas, besides of minor populations (
