Structural basis for bacterial energy extraction from atmospheric hydrogen

Diverse aerobic bacteria use atmospheric H2 as an energy source for growth and survival1. This globally significant process regulates the composition of the atmosphere, enhances soil biodiversity and drives primary production in extreme environments2,3. Atmospheric H2 oxidation is attributed to uncharacterized members of the [NiFe] hydrogenase superfamily4,5. However, it remains unresolved how these enzymes overcome the extraordinary catalytic challenge of oxidizing picomolar levels of H2 amid ambient levels of the catalytic poison O2 and how the derived electrons are transferred to the respiratory chain1. Here we determined the cryo-electron microscopy structure of the Mycobacterium smegmatis hydrogenase Huc and investigated its mechanism. Huc is a highly efficient oxygen-insensitive enzyme that couples oxidation of atmospheric H2 to the hydrogenation of the respiratory electron carrier menaquinone. Huc uses narrow hydrophobic gas channels to selectively bind atmospheric H2 at the expense of O2, and 3 [3Fe–4S] clusters modulate the properties of the enzyme so that atmospheric H2 oxidation is energetically feasible. The Huc catalytic subunits form an octameric 833 kDa complex around a membrane-associated stalk, which transports and reduces menaquinone 94 Å from the membrane. These findings provide a mechanistic basis for the biogeochemically and ecologically important process of atmospheric H2 oxidation, uncover a mode of energy coupling dependent on long-range quinone transport, and pave the way for the development of catalysts that oxidize H2 in ambient air.

Based on the MQ9 standard, this peak corresponds to 525.9 pmoles of MQ9(II-H2), which with eight binding sites per Huc molecule, gives an occupancy of 96.1%.

Supplementary Note 1. Analysis of Huc by EPR spectroscopy
We used EPR spectroscopy to assess the configuration of the HucS iron-sulfur clusters.At 30 K, the EPR spectrum of the "as-isolated" Huc was dominated by a featureless, isotropic signal centered at g = 2.03.This signal lacks the sharp, low-field feature and anisotropic lineshape that is typical for an isolated, oxidized (S = ½) [3Fe-4S] + cluster 3,4 , likely due to spin-spin interactions between the clusters that broaden the spectral features (Fig 3c) [5][6][7][8] .At a lower temperature

Supplementary Note 2. Putative mechanisms of O2 tolerance and insensitivity
In previously studied respiratory O2-tolerant hydrogenases, it is proposed that O2 binds to the [NiFe] active site and is rapidly reduced by four electrons to the innocuous oxidation state of water.Three of these electrons derive from normal redox transitions of the medial and proximal iron-sulfur clusters of the small subunit along with the active site itself [9][10][11][12] .The fourth electron originates from an unusual on the HucSL dimer in the presence of an excess of either H2 or O2.Simulations were run for 50 nanoseconds (ns) and repeated three times.Huc was stable and retained its original secondary structure throughout the simulations.H2 penetrated the Huc gas channels, with the molecule reaching within 8 Å of the active site within the first 10 ns of the simulations (Extended Data Fig 6b).O2 entered the gas channels more slowly but did reach within 8 Å of the active site in the second half of the simulations (Extended Data Fig 6b).By the end of the simulations, comparatively similar numbers of H2 and O2 molecules reached within 12-14 Å of the active site (approximately half the length of the Huc gas channel) (Supplementary Table 2).However, a much smaller number of O2 molecules reached within 8-10 Å of the active site compared to H2 (Extended Data Fig 6b).This indicates that bottlenecks along the length of the tunnel selectively limit O2 access to the interior of Huc, which is consistent with the varying diameter of the gas channels calculated using the CAVER3 code (Fig. 3f) 19 .As discussed in the main text, an additional bottleneck in the Huc gas channel prevented O2 from reaching a proximity of closer than 5 Å to the Huc catalytic cluster in our simulations (Extended Data Fig 6c,d).Conversely, in a number of simulation frames, H2 molecules are observed within 3 Å of the catalytic cluster and, in several instances, in a location analogous to the ligand-bound state of the cluster (Extended Data Fig 6e,f).This indicates that the ultimate point selection against O2 is a bottleneck after the convergence of the three gas tunnels immediately preceding the active site entrance.Interestingly, the conformation of amino acids in this region appears responsive to whether Huc was simulated in the presence of H2 or O2.In the simulations in the presence of O2, arginine 443 from HucL largely adopted a conformation where it formed a close hydrogen bond with serine 104 from HucL (Extended Data Fig 6h,i).This state was never fully realized in the simulations with H2 and may be indicative of long-range structural changes that prime Huc for O2 exclusion from the catalytic site (Extended Data Fig. 6g,i).
To assess the role of the bottleneck immediately preceding the Huc active site in excluding O2, we simulated Huc models in which key residues that form it (glutamate 15, isoleucine 64 and leucine 122) were mutated to the less bulky sidechain alanine, either in pairs (E15A + I64A; I64A + L122A) or a triple mutant (E15A + I64A + L122A).Simulations were run on these mutants and wildtype Huc (WT) for 100 ns and repeated three times.As for the previous simulations, for WT and the I64A + L122A, O2 was unable to reach closer than 4.7 Å to the catalytic cluster (Extended Data Fig. 7a,b).However, for the E15A + I64A and E15A + I64A + L122A mutants, O2 molecules reached within bonding distance of the catalytic cluster during the simulations (Extended Data Fig. 7a,b).In both mutants, the O2 molecules were able to access a region below the active site in close proximity to the Fe(CN)2CO cofactor, while in the E64A + I64A mutant O2 molecules reached the active site nickel ion (Extended Data Fig. 7c).
Based on these simulations we hypothesize that both these mutants would be sensitive to O2 to only 1% O2 (Extended Data Fig. 8f) 18 .These data indicated that the Huc [NiFe] cluster is oxidised much more slowly than oxygen-sensitive hydrogenases, likely due to the restricted access of O2 to its active site and potentially because of the altered properties of its redox chain.
Fig 5c).Again, a complex spectrum is observed, suggesting the presence of multiple species with slight electronic distinctions or splitting due to interactions with the clusters, which, even in the EPR-silent reduced state, likely remain paramagnetic.Work towards elucidating the active-site electronic structure through EPR spectroscopy is underway but beyond the scope of this report.

FTIR
spectra after transferring Huc in a mixed Ni-SI, Ni-R, and Ni-C state from 100% N2 into 20%:80% N2 to O2.Initially, the Ni-SI state was further populated at the expense of Ni-R and Ni-C, before the Ni-SI state was replaced by Ni-B over a timescale of minutes (Extended Data Fig.8e).These data suggest that the hydrogen-bound Ni-R and Ni-C states must first convert to the empty reduced Ni-SI state before the hydroxide-bound oxidised Ni-B state is formed (Extended Data Fig.8g,h).This is in contrast to the oxidation profile observed for the oxygen-sensitive hydrogenase Hyd2 from E. coli which was previously shown to rapidly adopt Ni-A and Ni-B states at the expense of all other states when exposed