Structural basis for bacterial energy extraction from atmospheric hydrogen – Nature

Statistics and reproducibility

Huc purification and PAGE analysis

Huc was purified from M. smegmatis cells on eight separate occasions throughout the course of the study. The yield of purified Huc (mg per litre of culture) varied by a factor of five between purifications. The relative abundance of the Huc oligomer and low molecular weight species (Extended Data Fig. 1a) varied approximately twofold between purifications. SDS–PAGE and activity staining was performed on Huc from each purification (n = 8). The abundance of HucS, HucL and HucM subunits in the Huc oligomer (Extended Data Fig. 1b) and its activity (Extended Data Fig. 1c) was consistent between purifications, so gels from a single purification were shown. Huc oligomer from different purifications was utilized for downstream enzymatic, electrochemical and spectroscopic analysis. Statistical methods were not used to predetermine sample size for this study. Blinding and randomization were not used as they are not appropriate for this kind of study.

HucM mutant analysis

Native PAGE analysis of Huc activity in wild-type and ΔhucM cell lysates was performed three times using cell lysates derived from independently grown cultures (n = 3 biological replicates). All replicates displayed comparable relative activity and band sizes. A single representative gel is shown (Extended Data Fig. 1e).

Huc differential scanning fluorimetry

Differential scanning fluorimetry experiments to determine the Huc melting temperature were performed three times on a single batch of purified Huc (n = 3 technical replicates) with consistent results between samples. Given the consistent purity and activity of the Huc oligomer, analysis of a single sample was sufficient to determine the melting temperature of the enzyme. Representative curves from a single experiment are shown (Extended Data Fig. 1d).

Gas chromatography

Huc H₂ consumption experiments were performed on 4 different Huc preparations with each electron acceptor (NBT or menadione) (n = 4, biological replicates) with 3 replicate vials used in each case (n = 3 technical replicates). The data obtained were consistent between both biological and technical replicates. E. coli Hyd1 H₂ consumption experiments were performed on a single enzyme preparation (n = 1 biological replicates), with 3 replicate vials used (n = 3 technical replicates). Figure 1a,b and Extended Data Fig. 1f show data from a single enzyme preparation with data collected concurrently for all samples presented. Data are presented as a mean of the 3 technical replicates (n = 3) with error bars showing s.d. Control vials containing no enzyme showed no H₂ consumption or loss and were included in each experiment with 2 or 3 technical replicates (n = 2 or 3).

Hydrogen electrode experiments

Huc oxygen tolerance and electron acceptor dependence experiments were performed 3 times on separate Huc purifications (n = 3, biological replicates) with 3 separate electrode traces collected for each O₂ concentration or electron acceptor for each Huc sample (n = 3, technical replicates). For menadione activity experiments, 5 technical replicates were performed due to increased instrument noise when using this compound (n = 5, technical replicates). The data obtained were consistent between both biological and technical replicates. The mean of the three replicates for each condition (five for menadione) from a single Huc sample is shown with error bars showing s.d. (Fig. 1c,d).

Protein film electrochemistry

Cyclic voltammograms of Huc in 100% H₂ (Fig. 1e) were recorded using a single enzyme preparation (n = 1 biological replicates). Two immobilized Huc films were recorded (n = 2 technical replicates), with each film scanned at least two times. The data obtained were consistent between technical replicates. Two control films with no immobilized enzyme were included in each experiment (n = 2 technical replicates), each scanned two times. Cyclic voltammetry for Huc, Hyd1 and Hyd2 in 10 ppm to 5% H₂ were performed on a single preparation of each enzyme (n = 1, biological replicates) with 3 different protein films of each (n = 3 technical replicates) (Fig. 1f and Extended Data Fig. 2). The Huc used in these experiments was from a different purification to that in the 100% H₂ experiments. It is common in protein film electrochemistry to find variation in absolute enzyme coverage between different films arising from minor differences in film preparation; therefore it was not reasonable to compare the absolute current from electrochemistry between different protein films. However, the trends in current in response to changes in H₂ partial pressure were consistent between technical replicates. For determination of onset potentials for Huc, the mean of three objective readings of onset potential for two Huc protein films was determined, as shown with error bars showing s.d. (Extended Data Fig. 2a).

Cryo-EM imaging

Cryo-EM imaging was performed on Huc from 2 separate purifications (n = 2 biological replicates). In total, 2,226 and 3,113 images were collected from each of the purifications with single representative images of free and membrane-associated Huc down in Extended Data Fig. 3a. Two-dimensional and three-dimensional reconstructions of were performed independently on each dataset.

EPR experiments

EPR experiments were performed on a single enzyme preparation (n = 1 biological replicates), incubated under either H₂ or Ar before flash freezing. Each individual power and temperature setting was recorded once (n = 1 technical replicate), and each spectrum shown represents the average of 4 scans (Fig. 3c, Extended Data Fig. 5b–d). Given the consistency of the power and temperature dependence when cross-correlated, this was sufficient for the conclusions drawn in the manuscript.

FTIR experiments

FTIR experiments were performed on a single enzyme preparation (n = 1 biological replicates). The population of reduced states upon H₂ exposure and the formation of Ni-B upon exposure to air was performed with at least three individual attenuated total reflectance (ATR)-FTIR spectroscopy sample preparations (n = 3, technical replicates) leading to similar results (Fig. 3d and Extended Data Fig. 8a–d,h). The kinetics transferring between these redox state populations were recorded on one sample preparation (n = 1 technical replicates) (Extended Data Fig. 8e).

Mass spectrometry

The identification of dihydromenaquinone-9 in the purified Huc protein by LC–MS (Fig. 4e) was performed once (n = 1 technical replicate) due to the qualitative nature of the analysis and sample quantity constraints. While the estimate of occupancy (Supplementary Fig. 5) using these data is quantitative, the purpose of that analysis was predominantly to demonstrate that the signal was consistent with that anticipated for the ligand, which only required a single replicate.

Molecular dynamics experiments

Molecular dynamics simulations were performed 3 times for each condition simulated (n = 3). Cumulative gas occupancy plots are presented as the mean of these simulations with dotted lines showing s.d. (Extended Data Figs. 6b and 7b).

Publicly available data

Publicly available structural coordinates utilized in this study can be accessed via the following PDB accession codes, 5AA5, 2FRV, 4U9H, 6EHQ, 4C3O, 3AYX, 5MDK.

Bacterial strains and growth conditions

M. smegmatis mc²155 and its derivatives were grown on lysogeny broth (LB) agar plates supplemented with 0.05% (w/v) Tween 80 (LBT). In broth culture, M. smegmatis was grown in either LBT, or Hartmans de Bont minimal medium supplemented with 0.2% (w/v) glycerol, 0.05% (w/v) tyloxapol, and 10 mm NiSO₄ (HdB). E. coli was maintained on LB agar plates and grown in LB broth. Selective LB or LBT medium used for cloning experiments contained gentamycin at 5 μg ml⁻¹ for M. smegmatis and 20 μg ml⁻¹ for E. coli. Cultures were incubated at 37 °C with rotary shaking at 150 rpm for liquid cultures unless otherwise specified. The strains of M. smegmatis and its derivatives and E. coli are listed in Supplementary Table 4.

Genetic manipulation of M. smegmatis

For Huc production, protein expression was performed using the M. smegmatis mc²155 strain PRC1, which lacks the glycerol response regulator gylR (gylR Leu154→frameshift) and the Hhy hydrogenase (hhyS inactivation)^49,50. To facilitate Huc purification, a 2×StrepII tag was inserted at the N-terminal end of the small subunit (MSMEG_2262, HucS) through allelic exchange mutagenesis as described previously⁵¹. An allelic exchange construct, hucS-2×StrepII (2656 bp), was synthesized by Genewiz and cloned into the SpeI site of the mycobacterial shuttle plasmid pX33 to yield the construct pHuc-2×StrepII (Supplementary Table 4). pHuc-2×StrepII was propagated in E. coli DH5α and transformed into wild-type M. smegmatis mc²155 PRC1 cells by electroporation. Gentamycin was used in selective solid and liquid medium to propagate pX33. To allow permissive temperature-sensitive vector replication, transformants were incubated on LBT gentamycin plates at 28 °C until colonies were visible (5–7 days). The resultant catechol-positive colonies were subcultured in LBT gentamycin broth at 40 °C for 3–5 days and then diluted from broth onto fresh LBT gentamycin plates and incubated at 37 °C for 3–5 days to facilitate the integration of the recombinant plasmid, via flanking regions, into the chromosome. The second recombination event was facilitated by subculturing catechol-reactive and gentamycin-resistant colonies in LBT supplemented with 10% sucrose (w/v) for 3–5 days, followed by dilution onto LBT agar plates supplemented with 10% sucrose (w/v) and incubating at 37 °C for 3–5 days. Gentamycin-sensitive and catechol-unreactive colonies were subsequently screened by PCR to distinguish wild-type revertants from HucS-2×StrII mutants. A colony that was PCR positive for the HucS-2×StrII insertion was designated M. smegmatis mc²155 PRC1 HucS-2×StrII and used for subsequent experiments. The primers used for screening are listed in Supplementary Table 4. HucM was deleted using allelic exchange mutagenesis using the same PRC1 parent strain and procedure for the 2×StrepII tagging of HucS. An allelic exchange construct consisting of ~1,000 bp of the genomic sequence on either side of HucM was used.

Huc purification

M. smegmatis mc²155 PRC1 HucS-2×StrII cells were propagated in large volumes (10–20 l) of HdB until 24 h into stationary phase. Cells were harvested by centrifugation and resuspended in lysis buffer (50 mM Tris, 150 mM NaCl pH 8.0) supplemented with 0.1 mg ml⁻¹ lysozyme, 0.1 mg ml⁻¹ DNase, 1 mM MgCl₂ and protease inhibitor cocktail tablets. Cells were lysed by cell disruptor (Emulsiflex C-5), and cell debris were removed by centrifugation at 30,000g for 15 min. Biotin blocking buffer (IBA) was added to clarified cell lysates (1 ml per 25 ml of lysate) before loading onto a 1 ml StrepTrap column (Cytiva), and the column was washed extensively with lysis buffer before bound Huc was eluted with lysis buffer + 2.5 mM desthiobiotin. Huc-containing fractions from StrepTrap were determined by SDS–PAGE, pooled, and concentrated using a 100 kDa MWCO centrifugal concentrator (Amicon, Millipore). Concentrated Huc was loaded on a Superose 6 10/300 column (Cytiva), fractions containing oligomeric Huc were confirmed by SDS- and native PAGE, pooled, and concentrated to ~6 mg ml⁻¹ before flash freezing in liquid N₂ and storage at −80 °C. Typically purification yielded 10–20 μg l⁻¹ of the Huc complex.

PAGE analysis, activity staining and western blotting

For SDS–PAGE and native PAGE, samples were run on Bolt 4–12% SDS–PAGE and native PAGE 4-16% gels (Invitrogen) respectively, according to the manufacturer’s instructions. Gels were stained for total protein by AcquaStain Protein Gel Stain (Bulldog) and hydrogenase activity using the colorimetric electron acceptor NBT hydrogenase. For NBT activity staining, gels were incubated in 50 mM Tris, 150 mM NaCl pH 8.0 supplemented with 500 μM NBT in an anaerobic jar amended with 7% H₂ anaerobic mix. Incubation was conducted for 30 mins to 12 h, depending on the level of activity. Bands exhibiting hydrogenase activity were identified based on the purple colour of reduced NBT. For western blotting, proteins were transferred to nitrocellulose membrane from SDS–PAGE or native PAGE gels using the Transblot Turbo semidry transfer apparatus (BioRad) at 25 V for 30 min. Following transfer, the protein-containing nitrocellulose membrane was blocked with 3% (w/v) BSA in PBS (pH 7.4) with 0.1% (v/v) Tween 20 (PBST). The nitrocellulose membrane was washed three times in 20 ml of PBST and finally resuspended in 10 ml of the same buffer. Strep-Tactin HRP conjugate was then added at a 1:100,000 dilution. Bound Strep-Tactin HRP was detected by chemiluminescence using the ECL Prime detection kit (Cytiva) following the manufacturer’s specifications, blots were visualized using a ChemiDoc (BioRad).

Amperometric measurement of Huc activity

H₂ consumption by purified Huc was measured amperometrically with a Unisense H₂ microsensor polarized to +100 mV for 1 h. Initially, the electrode was calibrated with known H₂ standards of 0%, 1%, and 10% H₂ (v/v) in gas saturated buffer (50 mM Tris, 150 mM NaCl, pH 8.0). This buffer was prepared by firstly bubbling all buffers with 100% nitrogen gas for 1 h to remove any traces of oxygen, then bubbling 100% H₂ (v/v) through the de-oxygenated buffer for 10 min. Before degassing and regassing, all buffers contained electron acceptors to a final concentration of 200 μM (menadione, NBT, or benzyl viologen). For each reading, 1 ml of 1% (8 µM) H₂-infused buffer containing the electron acceptor was added to the microrespiration assay chamber, to a final concentration of 1% H₂. The electrode was subsequently placed into the chamber to equilibrate. Once equilibrated (approximately 10 min), BSA (0.3 mg ml⁻¹) and Huc (1–3 nM in 0.3 mg ml⁻¹ BSA) were added to the chamber via a needle so as not to disrupt the gas balance in the solution. The changes in H₂ concentration were measured using the SensorTrace Suite v3.4.00 and linear rates of H₂ consumption were measured from the addition of Huc until complete H₂ consumption. The linear rate of H₂ consumption by Huc was calculated between a concentration of approximately 2.5 μM H₂ to 0.0125 μM H₂ over eight time points. These rates of H₂ consumption were used to plot a Michaelis-Menton curve and calculate Huc-specific activity in the presence of different electron acceptors, as well as the maximal velocity of Huc.

Mass spectrometric menaquinone detection

Samples were prepared for LC–MS analysis using a modified Folch extraction. In brief, a 100 µl solution of purified protein (~57 µg) was treated with 2000 µl of 2:1 chloroform:methanol (v/v) after which the mixture was shaken for 10 min and allowed to stand for a further 50 min. Four-hundred microlitres of water was added and the mixture was shaken for 10 min after which the sample was allowed to stand until the two phases had completely separated. The lower chloroform-rich phase was then transferred to a 2 ml sample vial and the solvent was removed under a stream of nitrogen gas. The resulting residue was reconstituted in 100 µl of 2:1 chloroform:methanol v/v and transferred to a 200 µl sample insert, the solvent was again removed and the sample reconstituted in a 7:3 mixture of LC solvent A:LC solvent B (v/v). Samples were analysed using a Dionex RSLC3000 UHPLC (Thermo) coupled to a Q-Exactive Plus Orbitrap MS (Thermo) using a C18 column (Zorbax Eclipse Plus C18 Rapid Resolution HD 2.1 × 100 mm × 1.8 μm, Agilent) with a binary solvent system; solvent A = 40% isopropanol and solvent B = 98% isopropanol both containing 2 mM formic acid and 8 mM ammonium formate. Linear gradient time, %B as follows: 0 min, 0%; 8 min, 35%; 16 min, 50%; 19 min, 80%; 23 min, 100%; 28 min, 100%; 30 min, 0%; 32 min, 0%. The flow rate was 250 µl min⁻¹, the column temperature 50 °C, and the sample injection volume was 10 µl. The mass spectrometer operated at a resolution of 70,000 in polarity switching mode with the following electrospray ionization source conditions: spray voltage 3.5 kV; capillary temperature 300 °C; sheath gas 34; Aux gas 13; sweep gas 1 and probe temp 120 °C.

The occupancy of MQ9(II-H₂) associated with Huc was calculated by comparison against a single sample containing the standard, MQ9. MQ9 was the closest structural analogue to MQ9(II-H₂) (differing only in the presence of an additional double bond in the second isoprenyl unit) that was commercially available and was purchased from Toronto Research Chemicals. The peak area for the ammonium adduct of MQ9 in the standard sample containing 452.64 ng of MQ9 was 7.40 × 10⁹ and the peak area for the ammonium adduct of MQ9(II-H₂) from the protein sample was 6.77 × 10⁹ (Supplementary Fig. 5). Assuming that the response of MQ9(II-H₂) and MQ9 standard are equivalent, this provides an estimate of the amount of MQ9(II-H₂) in the protein sample of 414 ng. The protein sample contained 57 µg of Huc which corresponds to 68.4 pmoles (MW of Huc is 833 kDa). As there are eight MQ9(II-H₂) per Huc complex at full occupancy, this corresponds to 547.2 pmoles or 430.4 ng (MW of MQ9(II-H₂) = 786.63 g mol⁻¹) of MQ9(II-H₂). Consequently, the occupancy of Huc in MQ9(II-H₂) is estimated to be 96.2% (414 ng / 430.4 ng). If MQ9(II-H₂) is bound outside the observed binding sites in the Huc structure, then the actual occupancy in these sites will be lower. However, given the high quality of the density of the Huc head group present, we expect that the MQ9(II-H₂) present at these sites accounts for the majority in the sample.

Mass spectrometric identification of HucM

The 18-kDa band on SDS–PAGE originating from the Huc complex was excised and extracted from the gel matrix. Upon destaining, the proteins were reduced with tris(2-carboxyethyl)phosphine (Pierce), alkylated with iodoacetamide (Sigma), and digested with mass spectrometry grade trypsin (Promega). The extracted peptides were analysed by LC–MS/MS on an Ultimate 3000 RSLCnano System (Dionex) coupled to an Orbitrap Fusion Tribrid (ThermoFisher Scientific) mass spectrometer equipped with a nanospray source. The peptides were first loaded and desalted on an Acclaim PepMap trap column (0.1 mm id × 20 mm, 5 µm) and then separated on an Acclaim PepMap analytical column (75 µm id × 50 cm, 2 µm) over a 30 min linear gradient of 4–36% acetonitrile/0.1% formic acid. The Orbitrap Fusion Tribrid was operated in data-dependent acquisition mode with a fixed cycle time of 2 s. The Orbitrap full ms1 scan was set to survey a mass range of 375–1,800 m/z with a resolution of 120,000 at m/z 400, an AGC target of 1 × 10⁶, and a maximum injection time of 110 ms. Individual precursor ions were selected for HCD fragmentation (collision energy 32%) and subsequent fragment ms2 scan were acquired in the Orbitrap using a resolution of 60,000 at m/z 400, an AGC target of 5 × 10^5, and a maximum injection time of 118 ms. The dynamic exclusion was set at ±10 ppm for 10 s after one occurrence. Raw data were processed using Byonic (ProteinMetrics v4.3) against a protein database covering M. smegmatis mc²155. The precursor mass tolerance was set at 20 ppm, and fragment ions were detected at 0.6 Da. Oxidation (M) was set as dynamic modification, carbamidomethyl (C) as fixed modification. Only peptides and proteins falling below a false discovery rate of 0.01 were reported.

Gas chromatography analysis

Gas chromatography experiments were used to measure the consumption of H₂ gas by both pure Huc protein and cultures of M. smegmatis expressing Huc. For pure protein experiments, to assess the rate of H₂ oxidation of Huc with menadione as the electron acceptor, 5 ml of buffer (50 mM Tris, 150 mM NaCl, pH 8.0, 0.3 mg ml⁻¹ BSA) with 200 µM menadione was contained in a 120 ml sealed serum vial, in triplicate. The headspace of the vial was flushed with 3 ppm H₂ in an atmospheric gas mix for 10 mins and then 100 nM Huc was added via a syringe to begin the reaction. For experiments comparing the H₂ oxidation rate of Huc and Hyd1, the same reaction buffer contained 200 µM NBT in place of menadione and a headspace of either 10 or 100 ppm H₂ in N₂ was used, with 50 nM Huc or Hyd1 added to start the reaction. Before the experiment, Hyd1 was activated by incubation in an atmosphere of 7% H₂ in N₂ for 18 h, before being added anaerobically to the serum vials. Vials were incubated at room temperature, stirring, and hydrogen gas concentration was monitored using a pulsed discharge helium ionization detector (model TGA-6791-W-4U-2, Valco Instruments Company) over specific time intervals, until significant H₂ oxidation was no longer observed, or the limit of detection had been reached.

Differential scanning fluorimetry

To determine the stability of the Huc complex, thermal melting was performed from 20 to 90 °C using a PrometheusNT.48 DSF (Nanotemper) using high sensitivity capillaries. The ratio of change in fluorescence at 330 and 350 nm was monitored to determine protein unfolding. Melting was performed with a Huc concentration of 0.2 mg ml⁻¹ in buffer containing 50 mM Tris, 150 mM NaCl pH 8.0.

Cryo-EM imaging

Samples (3 µl) were applied onto a glow-discharged UltrAuFoil grid (Quantifoil GmbH) and were flash-frozen in liquid ethane using the Vitrobot mark IV (ThermoFisher Scientific) set at 100% humidity and 4 °C for the prep chamber. Data were collected on a Titan Krios microscope (ThermoFisher Scientific) operated at an accelerating voltage of 300 kV with a 50 μm C2 aperture at an indicated magnification of 105 kx in nanoprobe EFTEM mode. Gatan K3 direct electron detector positioned post a Gatan Quantum energy filter, operated in a zero-energy-loss mode with a slit width of 10 eV was used to acquire dose fractionated images of the Huc complex without an objective aperture. Two initial Huc datasets were collected, one low concentration dataset (0.4 mg ml⁻¹) composed of 2,226 movies, and a medium concentration dataset (1 mg ml⁻¹) composed of 3,113 movies. Movies were recorded in hardware-binned mode (previously called counted mode on the K3 camera) yielding a physical pixel size of 0.82 Å pixel⁻¹ with an exposure time of 6 s amounting to a total dose of 66.0 e⁻ Å⁻² at a dose rate of ~7.5 e⁻ pixel⁻¹ s⁻¹, which was fractionated into 60 subframes. A further high concentration dataset (4 mg ml⁻¹) of 9,868 micrographs was also recorded using the same microscope but at a higher magnification of 165 kx on K3 detector, the resulting pixel size was 0.5 Å. The slit width for zero-loss filtering was set to 10 eV and images were collected at a dose rate of 4.37 e⁻ pixel s⁻¹. Movies were collected with an exposure time of 4 s amounting to a total dose of 60.4 e⁻ Å⁻², which were further fractionated into 60 subframes. The defocus range was set between −1.3 and −0.3 μm.

Cryo-EM data processing and analysis

Micrographs from all datasets were motion-corrected using UCSF Motioncor 1.0.4 and dose-weighted averages had their contrast transfer function (CTF) parameters estimated using CTFFIND 4.1.8, implemented using Relion 3.1.2 (ref. ⁵²).

Huc oligomer reconstruction from dataset 1

Particle coordinates were determined by crYOLO 1.7.6 using a model trained on manually particles picked from 20 micrographs⁵³. Unbinned particles were extracted from micrographs using Relion 3.1.2, before being imported into cryoSPARC 3.3.1 for initial 2D classification to remove bad particles, followed by ab initio model generation and 3D refinement⁵⁴. Refined particles were reimported into Relion 3.1.2 and CTF refinement was performed, followed by Bayesian polishing⁵². Particles were reimported into cryoSPARC 3.3.1 for final 2D classification to remove residual bad particles, followed by non-uniform 3D consensus refinement to generate final maps at 2.19 Å (Fourier shell correlation (FSC) = 0.143, gold standard).

HucS₂L₂ subunit dataset 1

HucS₂L₂ particles were picked using gautomatch V 0.53 (developed by K. Zhang) using a diameter of 80 Å and a minimum distance between picked particles set to 20 Å. The resultant particles were extracted and binned 4 times to a box size of 56 pixels. A small subset of unbinned particles with 224 pixel box size were subjected to 2D classification to exclude noisy particles and the remainder were selected and subjected to 3D ab initio model generation with 4 classes using cryoSPARC v3.0.1 (ref. ⁵⁴). A class corresponding to a population of HucS₂L₂ subunits showed clear C₂ symmetry and this particle set was subjected to homogeneous refinement with C₂ symmetry applied resulting in a 3.25 Å resolution reconstruction. The resulting volume was used as the initial model for further classification and refinement of the full dataset as shown in the workflow Supplementary Fig. 3. Particles corresponding to the subunits were selected from binned dataset following 2D classification and heterogeneous refinement. The cryoSPARC 3D alignments from heterogeneous refinement were then exported to Relion using the csparc2star.py script in the pyem v0.5 package. In Relion, particles were re-extracted with no binning using the refined coordinates from the imported particles. These were re-imported into cryoSPARC and subjected to heterogeneous refinement followed by homogeneous refinement with C₂ symmetry applied. The FSC from the resultant reconstruction showed particle duplication as a result of over-picking using gautomatch. The refined particle coordinates were imported into Relion as described above and were cleaned to remove duplicated particles. The cleaned re-extracted particles were subjected to a round of CTF refinement followed by auto-refinement and Bayesian polishing. The resultant polished particles were imported back into cryoSPARC and CTF refinement, and non-uniform refinement with CTF parameter optimization was performed to reach a final global resolution of 1.67 Å (FSC = 0.143, gold standard), which was close to the Nyquist limit for the dataset (1.64 Å).

HucS₂L₂ subunit dataset 2

A second dataset at higher magnification was collected to overcome the Nyquist limit imposed on the map resolution of dataset 1. The particles were picked, extracted, and binned as described above and in Supplementary Fig. 2. Multiple rounds of 2D classification, ab initio classification, and heterogeneous refinement was performed to retain only particles containing the full Huc oligomer. Homogeneous refinement was performed on the binned particles with the HucS₂L₂ subunit map as an initial model in order to centre the particles on the HucS₂L₂ subunit. These particles were then re-extracted in Relion based on the refined coordinates to a box size of 380 pixels. Homogeneous refinement with C₂ symmetry resulted in a subunit reconstruction map of 2.15 Å. This step ensured the retention of all HucS₂L₂ particles with a good signal to noise ratio. In order to ensure that all symmetry-related HucS₂L₂ particles corresponding to the Huc oligomer are retained, the above-selected particles were re-extracted with a larger box size of 512 pixels and rescaled to 128 pixels (4× binned). The resultant particle set was subjected ab initio model generation followed by homogeneous refinement using C₄ symmetry to yield a 4.10 Å map of the Huc oligomer. The particles were re-extracted based on the refined coordinates from the previous step at a box size of 576 pixels and then rescaled to 288 pixels (2× binned). These particles were then subjected to homogeneous refinement followed by CTF refinement and Non-Uniform refinement in cryoSPARC to yield a 2.05 Å resolution map⁵⁴. The refined particle set was then re-imported into Relion, and Bayesian polishing was performed. The resultant ‘shiny’ particles were extracted to a final box size of 576 pixels with no binning and were re-imported into cryoSPARC to perform symmetry expansion and final refinement. Duplicate particles were excluded to retain 153,359 Huc oligomers. Resultant particles were subjected to one round of refinement imposing C₄ symmetry, which yielded a 2.19 Å map. The resultant refined particles were symmetry expanded and subjected to local refinement within cryoSPARC 3.3.2 with no symmetry to yield a 2.16 Å map for HucS₂L₂. The resultant vol and particles were aligned to C₂ symmetry axis to exploit the C₂ symmetry within the HucS₂L₂. The symmetry aligned particles were then refined with C₂ symmetry followed by iterative Global CTF refinement, per particle defocus refinement followed by Ewald Sphere correction and a final round of per particle defocus refinement to reach a final global resolution of 1.52 Å (FSC = 0.143, gold standard).

HucM stalk reconstruction

The particle set resulting in a 2.15 Å map during the HucS₂L₂ subunit reconstruction from dataset 2 was re-extracted at a box size of 512 pixels and rescaled to 128 pixels (4× binned). The resultant particle set was subjected to 2D classification followed by ab initio model generation and homogeneous refinement using C₄ symmetry to yield a 4.10 Å map of Huc oligomer. The particles were re-extracted based on the refined coordinates from the previous step at a box size of 576 pixels and then rescaled to 288 pixels (2× binned). These particles were then subjected to homogeneous refinement followed by CTF refinement and non-uniform refinement in cryoSPARC to yield 2.05 Å resolution map⁵⁴. The refined particles set was then re-imported into Relion and Bayesian polishing was performed, the resultant ‘shiny’ particles were extracted to a final box size of 576 pixels with no binning. These particles were subjected to ab initio model generation with three classes to retain particles that showed clear density for the HucM tail region. The selected particles were then ctf refined and Non-Uniform refined to yield a 2.09 Å map. The stalk density was isolated using volume tools in ChimeraX 1.3 and low pass filtered to 20 Å to derive the initial model for local refinement in cryosparc⁵⁵. A dilated cosine padded softmask was generated using the initial model. The coordinates at the attachment of the stalk to the main body of the subunit were determined using the volume tracer tool in Chimera and used to define new fulcrum point for localized refinement (Supplementary Fig. 4). Two rounds of masked local refinement resulted in improving the density of the stalk region associated with the membrane, as shown in the figure to a resolution of 5.7 Å (no mask). The resultant FSC calculates a map resolution of 5.21 Å (FSC = 0.143, gold standard) but the map features are more consistent with a resolution of 6–8 Å on visual inspection.

Huc model building and visualization

A model of the HucS and HucL subunit dimers was generated using AlphaFold and docked into one-half of the high-resolution Huc Dimer maps using ChimeraX 1.3 (refs. ^56,57). The model was refined and rebuilt into map density using Coot 0.92 (ref. ⁵⁸). [NiFe], [3Fe–4S], and menaquinone cofactors associated with Huc were downloaded from the PDB and customized restraints generated using Elbow within the PHENIX package before they were fitted and refined into maps using Coot^59,60. The model was then refined using real-space refinement within PHENIX⁶¹. Once model building was complete the model was symmetry expanded using the Map symmetry tool and waters were added using DOUSE within the PHENIX package⁶⁰. The refined dimer model was docked into the Huc oligomer map using ChimeraX⁵⁶, with HucM manually built into the available density for one subunit using Coot⁵⁸, followed by iterative real-space refinement within PHENIX and further model building. Once model building was complete the model was symmetry expanded using the Map symmetry tool within the PHENIX package⁶⁰. Model quality was validated using MolProbity⁶². Images and movies were generated in Coot, ChimeraX, and Pymol^56,58. The location and diameter of the gas channels of Huc and other [NiFe] hydrogenases was determined using the CAVER code⁶³.

AlphaFold structural modelling

AlphaFold modelling was performed using AlphaFold version 2.1.1 implemented on the MASSIVE M3 computing cluster^57,64. The sequence of the HucM C-terminal region (amino acids 80-189) was provided and modelling was run in multimer mode, with four molecules of HucM requested. The five ranked models produced by AlphaFold were compared for consistency with the top-ranked model used for further analysis and figure generation.

Molecular dynamics parameterization and simulation

The 1.52 Å cryo-EM structure of Huc hydrogenase was used to construct the simulation system using the Charmm36m forcefield^65,66 with the TIP3P⁶⁷ water model. Sodium counterions were added to the solution to neutralize the system charge. Three sets of simulations were set up containing no gas molecules or containing 250 hydrogen or 250 oxygen molecules enriched in solution. Gas molecules were randomly inserted into the water phase. Each simulation was initiated with different starting velocities and gas positions and three independent simulations of each system were performed. Simulations were run for 50 ns (wild-type) or 100 ns (mutants). Molecular visualization and analyses were performed using VMD software⁶⁸. The topologies of the metal cofactors were constructed based on the coordinates from the cryo-EM structure. Previous work on hydrogenase cofactors⁶⁹ has demonstrated the transferability of structurally identical metallic cofactors between hydrogenases. Thus, atomic partial charges and force constants for the bond lengths of the metal cofactors were taken from previous studies of the structurally similar Desulfovibrio fructosovorans [NiFe] hydrogenase⁶⁹. Equilibrium values of bond lengths and angles within the metal cofactor were taken directly from the cryo-EM structure. Molecular dynamics simulations were performed in a single redox state. The ready active Ni-SI form of the active site and oxidized form of the iron–sulfur clusters were chosen based on previous similar analyses⁷⁰. The [NiFe] cofactor had bond constraints placed across the CN and CO ligands as well as between the iron atom and two bound sulfur atoms. Lennard–Jones parameters were taken from the Charmm36m^65,66 forcefield where appropriate, with exception of the cyanuric carbon⁷¹ with the iron parameters taken from physical data to have a non-zero epsilon value⁷². The QL molecular oxygen model developed by Javanainen et al.⁷³ was chosen to better define the quadrupole within the molecule. The oxygen atoms are given mass and non-zero Lennard–Jones parameters as well as a partial charge. A massless virtual site, OD1, exists at the midpoint of two oxygen atoms to carry the neutralizing charge, simulating the charge distribution within the oxygen quadrupole. Likewise, the hydrogen model was given mass and charge distribution coherent with the work of Hunter et al.⁷⁰. Molecular dynamics simulations were performed with the GROMACS v 2021.3 simulations suite. Energy minimization was performed using the steepest descent algorithm followed by a second minimization step using the conjugate gradient approach, both set to a maximum force of 250 kJ mol⁻¹ nm⁻¹. Equilibration simulations were performed within the NVT ensemble at 310 K. Position restraints were applied to the protein, cofactors and magnesium ion, with a 10³ kJ mol⁻¹ nm⁻² force constant for 100 ps. Further equilibration was carried out within the NPT ensemble with pressure maintained at 1 bar using the Parrinello–Rahman barostat⁷⁴, for 100 ps. Production simulations in the NPT ensemble were each of duration 50 ns, at a temperature of 310 K. Long-range electrostatics were treated with the Particle Mesh Ewald method⁷⁵, with values for the real-space long-range cut-off for these and Van der Waals forces being set to 10 Å. Covalent bonds containing hydrogen atoms were restrained using the LINCS algorithm^76,77 allowing a 2 fs integration step.

EPR spectroscopy

The composition of the iron–sulfur clusters of Huc was analysed by EPR spectroscopy. X-band EPR measurements were performed on a Bruker ELEXYS E500 spectrometer equipped with a SuperX EPR049 microwave bridge and a cylindrical TE₀₁₁ ER 4122SHQE cavity in connection with an Oxford Instruments continuous flow cryostat. Measuring temperatures were achieved using liquid helium flow through an ITC 503 temperature controller (Oxford Instruments). Samples of Huc isolated in air were prepared under a neat argon atmosphere or flash-frozen following incubation under 1 atm H₂, and X-band EPR spectra were collected in the temperature range 40–7 K, at varying microwave powers. Preliminary simulations of the Ni signals in the EPR spectra were carried out using EasySpin v.6.0.0-dev47 (ref. ⁷⁸).

FTIR spectroscopy

A 4 µl volume of 5 mg ml⁻¹ Huc enzyme solution in a buffer containing 50 mM Tris, 150 mM NaCl pH 8 was deposited on an ATR crystal surface. The sample was applied under laboratory atmosphere (air), dried under 100% nitrogen gas, and rehydrated with a humidified aerosol (100 mM Tris-HCl (pH 8)) as described previously^79,80. A custom-built PEEK cell (inspired by Stripp et al.⁸¹) that allows for gas exchange sealed the ATR unit (BioRadII from Harrick) mounted in the FTIR spectrometer (Vertex V70v, Bruker). Spectra were recorded with 2 cm⁻¹ resolution, 80 Hz scanner velocity, and averaged over a varying number of scans (at least 100 scans). All experiments were performed at ambient conditions (room temperature and pressure, hydrated enzyme films). Data were analysed using OPUS 8 and Origin 2021 software. To assess the response of the Huc [NiFe] cluster to ambient air, N_2, and H₂, the PEEK cell was flushed sequentially with each gas in the following sequence: N₂ (ambient air isolated enzyme), ²H₂, N₂, H₂, ambient air. Changes in the Huc spectra were allowed to stabilize before the spectra in each gas were collected.

Protein film electrochemistry

Huc in 100% H₂ experiment

Protein film electrochemistry experiments were carried out under anaerobic conditions. The 3-electrode system was made up of (1) Ag/AgCl (4 M KCl) as reference electrode, (2) 5 mm diameter pyrolytic graphite edge (PGE) encapsulated in epoxy as the rotating disk working electrode, and (3) graphite rod as the counter electrode. The glass cell featured a water jacket for temperature control and a gas inlet/outlet for H₂ flow. The buffer used was composed of 5 mM 2-[N′-morpholino]-ethane sulfonic acid (MES), 5 mM 2-[N′-cyclohexylamino]ethane sulfonic acid (CHES), 5 mM N′-[2-hydroxyethyl]piperazine-N′−2- ethane sulfonic acid (HEPES), 5 mM N′-tris[hydroxymethyl]methyl-3-amino propane sulfonic acid (TAPS), 5 mM sodium acetate, with 0.1 M Na₂SO₄ as carrying electrolyte, titrated with H₂SO₄ to pH 7.0 at 20 °C, and purged with N₂ for 3 to 4 h. To remove residual O₂ in the PGE electrode, cyclic voltammograms were run at 100 mV s⁻¹ from 0 to −800 mV (versus SHE, standard hydrogen electrode) for 10 scans. To adhere Huc to the deaerated PGE electrode the surface was abraded with P1200 sandpaper and rinsed with purified water. An aliquot of 5 mg ml⁻¹ Huc enzyme solution in a buffer containing 50 mM Tris, 150 mM NaCl pH 8 was mixed with polymyxin B sulfate and transferred to the electrode surface. The cell solution was then saturated with H₂ (1 bar, 10 min) before the cyclic voltammogram of the system with the immobilized enzyme was recorded at 10 mV s⁻¹. The cyclic voltammogram of the blank electrode was recorded with no enzyme immobilized. Electrochemical data were acquired using a PGSTAT10 and GPES 4.9 software (Metrohm/Autolab). Data were analysed using Origin 8 software. All potential values are referenced versus SHE. Experiments were conducted on two independent Huc films, with each film scanned at least two times.

Huc, Hyd1 and Hyd2 in 0 to 5% H₂ experiments

Protein film electrochemistry experiments were conducted in a N₂-filled anaerobic glove box (Glove Box Technology Limited, O₂ < 2 ppm). Experiments were carried out using a glass electrochemical cell with gas inlet and outlet valve⁸². The main cell compartment was surrounded by a water jacket to control temperature. All experiments were performed at 25 °C. A saturated calomel electrode (SCE) was used as the reference, and was held in an isolated glass arm containing 0.10 M NaCl and connected to the main cell compartment by a Luggin capillary. The SCE was calibrated with ferrocene-methanol as +244 mV versus SHE at 25 °C, and potentials are corrected back to V versus SHE in all figures. The working electrode was a PGE rotating disk electrode, 2 mm diameter, and was polished with silicon carbide paper (first P1200 and then P4000). Electrode rotation was controlled by a Metrohm Autolab IME663 rotator, and the electrode was rotated at 2,000 rpm in all experiments shown here in order to achieve effective mass transport to/from the immobilized enzyme film. A platinum wire was used as the counter electrode. Cyclic voltammetry was controlled by an Autolab PGSTAT 10 potentiostat using NOVA software version 1.10. All experiments were conducted with gas flushing through the headspace of the electrochemical cell. Gas mixtures were prepared from cylinders of pure Ar, pure H₂, 5% H₂ in N₂, 1,000 ppm H₂ in N₂ (all from BOC), using gas mass flow controllers (Smart-Trak2, Sierra Instruments) to obtain the desired partial pressure of H₂ within an inert carrier gas (Ar or N₂). A mixed buffer system was used in all experiments; this consisted of 15 mM in each of MES, HEPES, TAPS, CHES (all from Melford) and sodium acetate (Sigma), with 0.1 M NaCl (Fisher) as supporting electrolyte. All solutions were prepared using purified water (Millipore: resistivity 18.2 MΩ cm) and titrated with HCl to pH 7.0 at 25 °C. Buffers were flushed with N₂ overnight to remove O₂ before being taken into the glove box. For preparation of films of Huc, 2 µl enzyme solution (containing 20 mg ml⁻¹ polymyxin B sulfate as coadsorbate) was spotted onto a freshly polished PGE electrode surface and left for 3 min; the electrode was then rinsed with purified water to remove unabsorbed enzyme. The electrode was placed in the electrochemical cell containing 8 ml of pristine mixed buffer. To minimize protein film loss during measurements, 0.025 mg ml⁻¹ polymyxin B sulfate was added to the mixed buffer. The cell headspace was flushed with gas for at least 15 min to fully equilibrate with the cell solution before applying a potential. Cyclic voltammograms were recorded with constant gas flow through the headspace of the cell at 1 bar. E. coli Hyd1 and Hyd2 were prepared as described previously^83,84. These hydrogenases were first activated under 100% H₂ at room temperature for at least 5 h. Films of Hyd1 and Hyd2 were prepared as for Huc, but without use of polymyxin B sulfate.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.