Orientation-Controlled Electrocatalytic Efficiency of an Adsorbed Oxygen-Tolerant Hydrogenase

Protein immobilization on electrodes is a key concept in exploiting enzymatic processes for bioelectronic devices. For optimum performance, an in-depth understanding of the enzyme-surface interactions is required. Here, we introduce an integral approach of experimental and theoretical methods that provides detailed insights into the adsorption of an oxygen-tolerant [NiFe] hydrogenase on a biocompatible gold electrode. Using atomic force microscopy, ellipsometry, surface-enhanced IR spectroscopy, and protein film voltammetry, we explore enzyme coverage, integrity, and activity, thereby probing both structure and catalytic H2 conversion of the enzyme. Electrocatalytic efficiencies can be correlated with the mode of protein adsorption on the electrode as estimated theoretically by molecular dynamics simulations. Our results reveal that pre-activation at low potentials results in increased current densities, which can be rationalized in terms of a potential-induced re-orientation of the immobilized enzyme.

The enzyme coverage was estimated by evaluating AFM images of several independently prepared samples and at least three different spots on each sample. The amount of Re MBH units was counted via the respective periodicity derived from the height profiles as indicated in Figure 1D for all recorded AFM images and was averaged subsequently.

Electrochemical desorption of the amino-1-hexanethiol SAM
Reductive desorption of the amino-1-hexanethiol SAM. The chemically modified Au electrode was subjected to an AC voltammetric sweep from +0.2 to 1.4 V in a buffered solution at pH 5.5 under Ar atmosphere. Scan rate, frequency, and potential amplitude were 50 mV s -1 , 100 Hz, and 4 mV, respectively. In the potential region between 0 and 0.8 V, the amplitude of the capacitive current of the electrical double layer (I) is essentially constant, and its value is about half that of the bare Au electrode, hereby used as a blank (dotted line). At poised potentials lower than 0.6 V, the current increases exponentially, as expected for SAM desorption. [2] The capacitive current measured after SAM desorption (short thick line) is twice that of the Au-SAM electrode and similar to that of the blank (dotted line). Please note that the minimum of the AC trace at 250 mV for the SAM-coated electrodes corresponding to the point of zero charge of the system is very close to 340 mV.  The structure of reduced MBH from Ralstonia eutropha (PDB: 3RGW) (figure E) served as a starting point for the calculations. The structure consists of the small subunit harbouring three FeS clusters and the large subunit carrying the active site and a conserved Mg 2+ ion. The correct ligand configuration of the active site, as determined recently, [2] was taken into account. Furthermore, all crystallographic water molecules were included in the MBH model. The protein matrix was protonated according to pH 7.0 and treated with the CHARMM 27 force field. [3] All histidine residues were modelled as HSD carrying a proton at the δ-nitrogen of the imidazole ring. The FeS clusters and the active site were treated as rigid bodies by constraining their internal motions. Their non-bonding interactions were calculated as reported earlier, [2] where the partial charges were derived by electrostatic potential fits with the Merz-Singh-Kollmann scheme. [4], [5] Moreover, the small subunit of the MBH hetero dimer, HoxK, was elongated by its C-terminal membrane anchor and an associated Strep-tag II peptide used for purification. The construction of the fusion protein is described in Schubert et al. [6] These structural elements are not resolved in the crystallographic structure, but the prediction with the PSIPRED server [7] showed the membrane anchor as a α-helical element, which is in agreement with the recently resolved crystallographic structure of`hydrogenase 1 from E. coli. [8] Thus, the Cterminus was manually generated as an α-helix with VMD 1.8.7. [9] The Strep-tag II peptide was predicted as random coil by PSIPRED [7] and built as an extension to the C-terminal helix as predicted. The final model was energy-minimized, heated to 300 K, and equilibrated for 10 ns with NAMD 2.7 [10] to relax the structure. During these steps performed in vacuo, the experimentally resolved parts were fixed to their positions and only the extension was allowed to move freely.

Surface
The gold electrode was modelled as a perfect three layer Au(111) film with the x-and ydimensions of ca. 120 × 120 Å 2 . The fixed Au(111) film was functionalized with 672 6amino-1-hexanethiol SAM molecules, ca. 8 % of which were protonated and positively charged, which is in line with the approximate pK a (6.0±0.2). [11] In order to neutralize the resulting net charge of the SAM, the bottom layer gold atoms were slightly negatively charged (0.02777 e). All other gold atoms were treated as uncharged. In this way, a neutral surface was generated. The sulfur atoms of the SAM were arranged in a ( 3 × 3 )R30° lattice with a nearest neighbor spacing of 4.98 Å on the perfect Au(111) surface. [12] , [13] The SAM molecules were initially tilted by ca. 30°. Gold and SAM atoms were described with the vdW radii from Bizzarri et al. [14] and the CHARMM force field for lipids, [15] respectively. The functional groups of the SAM were adopted from protonated and deprotonated lysine. Gold and SAM sulfur atoms were fixed to their positions, so that no binding parameters were required.

Surface-MBH system
In order to avoid long re-orientation times of Re MBH on the surface, we searched for energetically favorable starting orientations for the MD simulations. This was done in a systematic scanning of the interaction energy between Re MBH and the surface. The interaction energy of the enzyme (in a minimal separation distance of 5 Å with respect to the SAM) and the surface was evaluated with the NAMD energy plugin in VMD using the parameter set described above. In each step of the scanning, the Re MBH was rotated by the angles Φ and Ψ around the x-and y-axis, respectively (figure F). The resulting interaction energy landscape, shown in figure F, identified two energy minima of Re MBH on the surface. Thus, two models starting from these energetically favorable orientations, shown in figure G, were simulated. In both scenarios, the MBH was placed in a 5 Å separation distance on the surface. Then, the two models were solvated in 120 × 120 × 150 Å 3 (configuration A) and 120 × 120 × 170 Å 3 (configuration B) large TIP3P water [16] boxes. Both systems were neutralized and 15 mM Na + Clwas added.

Molecular dynamics simulations protocol
All simulations were performed with NAMD 2.7 [10] using the CHARMM 27 force field [3] as described above. First, both models were energy-minimized with the conjugated gradient algorithm and heated to 300 K. Then the water was equilibrated for 60 ps. During these steps position restraints of 25 kcal mol 1 Å 2 on all heavy atoms were stepwise decreased until all atoms were allowed to move freely, except those described above. After the preparation, the two models were subjected to 50 ns long production dynamics carried out in an NPaT ensemble (constant number of particles (N), pressure (p = 1 atm), surface area (A), temperature (T = 300 K)) realized by Langevin piston dynamics. [17] Short-ranged electrostatics and vdW interactions in the periodic systems were cut at a distance of 12 Å. Long-ranged electrostatics were calculated by the Particle-Mesh Ewald summation. [18] The time step of 2 fs was enabled by using the Rattle algorithm constraining all bonds containing hydrogen atoms.