Control of carbon monoxide dehydrogenase orientation by site-specific immobilization enables direct electrical contact between enzyme cofactor and solid surface

Design of native and synthetic CODH-Ls with single and dual gbp-fusion sites for surface immobilization

Based on the LOMETS analysis of distance and the contact map for selecting the gbp-fusion site on CODH-L (see Fig. 1a, b, respectively), residue 1 (N-terminus) and residue 803 (C-terminus) were considered to be ideal for gbp-fusion because they have no contact with the catalytic domain and are solvent-exposed. Thus, native CODH-L was overexpressed, purified, confirmed by sodium dodecyl-sulfate polyacrylamide gel electrophoresis (SDS-PAGE), and tested for its CO oxidation activities. Subsequently, the obtained recombinant CODH-L was used as a template for further genetic modification, shown in Fig. 1c. The SDS-PAGE results presented in Fig. 1d and Supplementary Fig. 1 show similar molecular masses for all variants at ~90 kDa, which suggests uniform glycosylation of the native and synthetic CODH-L. These results indicate that this technique does not modify the intrinsic enzyme activities, since the biocatalytic activities of the synthetic CODH-Ls were highly preserved (Fig. 1e), thereby rendering the gbp-fusion site(s) (i.e., the immobilization site(s)) the only variable to consider. The predicted structures of the synthetic CODH-Ls were then generated by iterative threading assembly refinement (I-TASSER)^52,53,54, and the structures with the highest confidence values (C-score) are shown in Fig. 2. Using various tools, the root mean square deviation (RMSD) of the synthetic enzymes against the native CODH-L (for all atoms) was calculated to be <2 Å, as tabulated in Supplementary Table 1, which indicates that no significant structural changes took place following the addition of gbp to the enzyme.

Estimation of the distance between the enzyme active site and the electrode surface

In a DET-based system, it is known that the enzyme bioactivity^16,17, the surface and solution chemistry, the substrate diffusion¹², and the ET distance at the enzyme–electrode interface^11,12,13,14 are among the factors that can affect the ET rate. In this study, CODH-L was designed and engineered to fuse the gbp at the N- or C- terminus, or at both termini, thereby generating various immobilization sites on the CODH-L. More specifically, the orientation of the enzyme molecules can be controlled depending on the immobilization site. As a result, the distance between the enzyme cofactor and the electrode surface, d_ET, is the only variable parameter to consider in this study, assuming that all other factors, such as the enzyme bioactivity and the substrate diffusion, are negligible (or constant). It was therefore hypothesized that d_ET is the main limiting factor of DET (considering that all other factors are constant); therefore, we initially estimated d_ET. More specifically, the estimated d_ET based on the crystal structure of the native CODH-L is shown in Fig. 3 and Table 1. This estimation was carried out carefully considering the following points: 1) binding takes place at the terminus because the gbp is fused at the terminus, and 2) the size of the gbp is 1.3 kDa, which is only 1.49% of the size of CODH-L, and so becomes negligible in the estimation. Thus, to estimate d_ET when the gbp is fused at the N- or C-terminus, the distance is taken under the condition that the longer enzyme axis is perpendicular (i.e., at 90°) to the electrode surface; therefore, the actual distance may be shorter, considering that the single tethering site used to bind the enzyme to the electrode surface may induce orientational flexibility (see Supplementary Movie).

Using PyMol, d_ET was therefore estimated to be 29.56 ± 3.29 and 39.97 ± 1.16 Å for CODH-L_gbpN and CODH-L_gbpC, respectively (Table 1). In the case, where the gbp was fused at both termini, it was hypothesized that the enzyme molecule would be positioned “lying down”, i.e., with the longest axis parallel to the electrode surface. As the N- and C-termini of CODH-L are freely exposed, located away from the main domain, and extending in the same direction, a fusion of the gbp at both termini would potentially stabilize the enzyme further and limit the orientations that the enzyme could adopt, as indicated in Fig. 3. Based on these considerations, the distance to reach the cofactor would be ~10.27 ± 1.38 Å, which is <14 Å from the nearest residue facing the electrode, thereby rendering it a promising candidate for DET.

Assembly properties of the native and synthetic CODH-Ls on the planar gold surface

Atomic force microscopy (AFM) provides information regarding the morphological and topological features of the electrode surface, which in this case is modified with the enzymes. The bare gold substrate (treated only with buffer) measured ~2 nm in height and possessed −2 nm voids, indicating an atomically flat surface, thereby allowing the enzyme orientation on the gold surface to be examined (Supplementary Fig. 2). Considering that the CODH-L possesses dimensions of ~85 × 75 × 70 Å⁵⁵ and that the bare gold substrate contains nanometer-scale voids, it was assumed that a height (thickness) >5 nm indicates that the enzyme was immobilized on the gold substrate (Fig. 4a–d, Supplementary Fig. 2a–e). This is because the shortest dimension of the enzyme is 70 Å (7 nm) and the voids are −2 nm; therefore, the measured height (thickness of enzyme on the gold surface) would be ~5 nm if the enzyme were immobilized such that its shortest axis was perpendicular to the gold surface. The cross-sectional height profile at a lateral distance of 5 µm of the selected line indicates that the native CODH-L was not well immobilized (Fig. 4a, lower panel). This was expected since the success of the immobilization process is highly dependent on the formation of weak non-specific bonds between the enzyme and the gold surface.

It was found that the heights varied between 6 and 18 nm (±12 nm height difference), 6 and 9 nm (±3 nm), 6 and 12 nm (±6 nm), and 7 and 8 nm (±1 nm) for the native CODH-L, CODH-L_gbpN, CODH-L_gbpC, and CODH-L_gbpNC, respectively. These notable differences in the measured enzyme heights obtained by cross-sectional analysis imply differences in the orientations of the bound molecules, plausibly due to the binding site being only at one end in the cases of CODH-L_gbpN and CODH-L_gbpC.

More specifically, the presence of the gbp at a single site renders the longest protein axis perpendicular to the gold substrate; i.e., it appears to be “standing” on the substrate surface. This results in relatively unstable enzyme molecules, which tend to tilt to find the lowest energy orientation, thereby leading to the observed differences in their heights. In contrast, the height was highly uniform in the case of CODH-L_gbpNC, indicating that the enzymes are immobilized in a near-uniform orientation when the gbp is fused to both termini. The corresponding height histogram, based on a more detailed analysis of the enzyme height on the gold substrate, is shown in Supplementary Fig. 3. The native and synthetic CODH-L, with the exception of CODH-L_gbpNC, demonstrated a wide distribution of heights, ranging from 6 to 18.9 nm, indicating the presence of more than one enzyme layer on the surface.

Notably, the height distribution of CODH-L_gbpNC ranged from 6 to 14 nm, and the histogram shows that >45% of the immobilized protein had a height of 6–6.9 nm, indicating a uniform adsorption geometry for the CODH-L_gbpNC molecules. It was, therefore, speculated that the presence of two freely exposed immobilization sites on CODH-L_gbpNC allowed the enzyme molecules to be positioned stably in an ideal orientation, i.e., “lying down” on the gold surface with the longer protein axis parallel to the gold surface, ultimately resulting in a uniform adsorption geometry. In addition, following modification, the surface roughness was found to have increased compared to that of the bare substrate, again reinforcing the hypothesis that the enzymes were successfully immobilized on the substrate (Supplementary Table 2), since the roughness is related to the protein concentration on the substrate surface.

AFM analysis following the immobilization of native and synthetic CODH-L on a screen-printed gold electrode

Supplementary Fig. 4 shows the AFM images and the 3D-surface profiles of the bare screen-printed gold electrode (SPGE), DRP-250 AT, and the enzyme-immobilized SPGE. As can be seen in the image of the bare SPGE, the surface is not perfectly flat, since significant voids and raised regions are present on the micrometer scale. For the native CODH-L immobilized on the electrode, there were no notable differences in the 3D image or the surface height compared to those of the bare SPGE. This was attributed to the reliance of the native CODH-L to take part in weak non-specific binding with the gold electrode surface, and so it can easily desorb from the electrode surface.

In contrast, significant differences were observed for the synthetic CODH-Ls bearing the gbp at the N- and/or C-terminus, with the surface now appearing to be flatter and the voids becoming less apparent. This can be attributed to the fact that the immobilized enzyme fills the deep voids, so that the peak-to-valley height decreases. This is especially prominent in the case of CODH-L_gbpNC, in which the surface becomes flattened, indicating a high enzyme coverage on the SPGE surface.

Because of the presence of significant voids and raised regions on the micrometer scale, it was not possible to observe how the enzyme molecules were oriented on the SPGE. However, in the AFM study into the SPGE with a structured topography and an atomically flat gold surface, it was demonstrated that gbp-fused enzymes are highly specific for gold surfaces, regardless of their surface properties, as all three fusion enzymes were successfully immobilized on the gold surface.

Gold-binding properties and kinetics of the native and synthetic CODH-Ls as determined using quartz crystal microbalance and surface plasmon resonance analyses

The gold-binding activities of the native CODH-L and the synthetic CODH-Ls were examined using quartz crystal microbalance (QCM) analysis, which allows “dynamic” information to be obtained regarding an enzyme adsorption process; this contrasts with the “static” information obtained by AFM. As shown in Fig. 5a, b, the native CODH-L showed negligible binding to the gold surface. This was attributed to the fact that adsorption of the native CODH-L onto the gold surface relies on weak non-specific interactions, which are influenced by the polarity and wettability properties of the surface, in addition to the presence of surface-exposed cysteine residues^56,57.

The CODH-L_gbpC system exhibited the strongest gold-binding activity (Δf = 670 Hz), and this was followed by CODH-L_gbpN (Δf = 616 Hz) and CODH-L_gbpNC (Δf = 483 Hz). In contrast to our initial expectation that the presence of gbp at both ends would give a higher gold-binding activity, the obtained results indicate that improved binding activities were achieved when the synthetic CODH-L contained gbp at one of its two ends. This unexpected observation can be plausibly explained by considering the immobilization of these enzymes on the electrode. More specifically, in the case of gbp at either the N- or the C-terminus, the enzyme molecules are “standing”; i.e., their longest protein axis is perpendicular to the gold electrode surface because of the site of gbp fusion at the exposed terminus, as confirmed by AFM imaging. Since the enzyme molecules are standing, the calculated enzyme footprint is 52.5 nm². In contrast, in the case of CODH-L_gbpNC, the presence of two gbp units results in the enzyme molecules “lying down” on the gold surface, i.e., with the longest protein axis parallel to the gold electrode surface. As a result, less space is available because of the larger enzyme footprint, i.e., 63.75 nm². With the experimental value, Δf, and the molecular weight of the enzymes being known, the surface coverages of the enzymes can be estimated by applying the Sauerbrey equation. The surface coverages of CODH-L, CODH-L_gbpN, CODH-L_gbpC, and CODH-L_gbpNC were, therefore, determined to be 0.14, 8.19, 8.91, and 6.3 pmol cm⁻², respectively.

Subsequently, the enzymes were prepared at concentrations of 12.5, 25, 50, 100, and 200 nM; their respective SPR sensorgrams are shown in Fig. 5c, d, while the kinetic adsorption and desorption parameters for all constructs are listed in Table 2. As expected, the results shown in Fig. 5c, d indicate that the native CODH-L exhibits the lowest adsorption rate because of its random non-specific interactions with the gold surface, which render it difficult to undergo stable immobilization, and the enzyme molecules can be easily washed off the surface. The introduction of the gbp at either the N- or the C-terminus increased the gold-binding ability almost 4-fold over that of the native CODH-L in terms of the association rate constant, k_a. However, the dissociation rate constants, k_d, of CODH-L_gbpN and CODH-L_gbpC differed from one another, which was possibly related to the different nearby residues surrounding the gbp-fusion site. Hence, the equilibrium dissociation constant, K_D, of CODH-L_gbpN was 3-fold smaller than that of CODH-L_gbpC.

The effect of gbp-fusion at both termini of CODH-L was found to be significant, as evidenced by the dramatic increase in k_a (over 3-fold compared to a single gbp fused at either terminus) and a reduction in the K_D value. Overall, the SPR binding studies demonstrated that fusion of the gbp to the CODH-L did indeed enhance the gold-binding affinity, with the presence of two gbp molecules resulting in a higher adsorption rate and thereby facilitating stable immobilization of the enzyme molecules on the gold surface.

ET properties at the enzyme–electrode interface

SPGEs modified with the native and synthetic CODH-Ls were then tested to determine their ET properties using cyclic voltammetry (CV) in phosphate buffer (PB) at pH 7.2. It should be noted that the selection of an appropriate electrolyte for the electrochemical measurements should be based on the assumption that the electrolyte has no effect on the electrochemical characteristics of the enzymes being studied^58,59. Thus, 100 mM PB buffer, at pH 7.2 was used to evaluate the bioelectrochemical properties at the enzyme–electrode interface, since this is the optimal buffer composition and pH for this CODH⁴⁹. Fig. 6 shows representative CV profiles for the native and synthetic CODH-Ls in the absence and presence of a CO-saturated PB solution. As a control, CV data were recorded for the bare SPGE both in the presence and absence of CO, and for the enzyme-modified SPGE in the absence of CO. It was found that in the CO-saturated PB solution, all CODH-Ls, with the exception of the CODH-L containing gbp at both termini, exhibited a very weak, almost negligible DET current, indicating that direct electrical contact at the enzyme–electrode interface was not effectively achieved.

A significant DET current was only observed in the case of CODH-L_gbpNC, for which a pronounced oxidation current onset was observed at −0.62 V vs. Ag/Ag⁺. This onset value agrees with the reported literature, considering that the redox potential of CO oxidation is −0. 76 V vs. Ag/Ag⁺⁶⁰. In addition, the CV data show a peak at −0.5 V, reflecting the redox potential of Mo–Cu, i.e., −0.57 V vs. Ag/Ag⁺⁶⁰. Although the surface coverage (Γ, mol cm⁻²) of CODH-L_gbpNC was calculated to be 5.38 nmol cm⁻², indicating the presence of a monolayer on the electrode surface, we observed a cyclic voltammogram with a drawn-out shape. This observation can be explained in terms of the disorder among the immobilized enzyme molecules participating in DET, which results in a spread of the ET rates, as reflected by the trailing edge in the CV data⁴⁰. More specifically, Leger et al.^40,61 attributed this linear change in current against the driving force to enzyme molecules immobilized on the electrode surface not being orientated in identical configurations, and to this difference in orientation leading to the spread of the interfacial ET rate constants. Thus, the enzyme molecules that are in an ideal orientation contribute to the catalytic signal at low driving forces, while the disorientated enzyme molecules that have poor electrical contact with the electrode can only contribute to the catalytic signal above a certain driving force^40,61. It should be noted here that CO is the only redox-active species present because the PB does not contain any molecules that produce prominent redox peaks. Supplementary Fig. 5a, c show the dependence of the DET current on the scan rate, wherein the linear dependence of the peak current on the square root of the scan rate indicates that this is a surface-controlled process. In addition, Supplementary Fig. 5c, d show five consecutive scans and chronoamperometry at −0.1 V, respectively. These results reinforce the claim that the oxidation current does originate from the oxidation of CO by CODH at the electrode surface, and that the immobilized enzymes are highly stable.

By introducing the gbp at the N- or the C-terminus of the CODH-L, or at both termini, it was possible to immobilize the different variants at distinct immobilization sites, resulting in a range of ET distances between the active sites and the electrode surfaces. Our results show that efficient DET occurs exclusively when the gbp is fused at both ends of the CODH-L with the kinetic parameter, ET rate (k_ET), at 14.44 s⁻¹, as calculated by plotting the peak potential, E_p against the logarithm of the scan rate, log v (Supplementary Fig. 6) based on the Laviron method⁶². This agrees with the estimated distance between the redox center and the electrode surface, i.e., 10.27 ± 1.38 Å (< 14 Å), which theoretically facilitates DET.

To investigate whether the DET catalytic current observed for CODH-L_gbpNC was caused by the shorter d_ET and to verify whether the difference in the ET kinetics was due to the difference in binding strength, a soluble redox mediator was added to examine the immobilized enzymes for all variants. In this system, the mediator (i.e., methylene blue, MB) should react with all enzyme molecules present at the electrode surface. Thus, the enzyme–electrodes that had been previously tested for DET were tested for MET after the addition of MB (final concentration = 50 µM) to the CO-saturated PB solution. A pronounced mediated oxidation current was observed for all constructs, indicating that the native CODH-L, CODH-L_gbpN, and CODH-L_gbpC were successfully immobilized on the electrode with catalytic activity retention, but that they were not within the necessary DET distance (Supplementary Fig. 7). The oxidation current peak was observed at approximately −0.19 V vs. Ag/Ag⁺, which is consistent with the standard redox potential of MB at −0.188 V vs. Ag/Ag⁺. Upon the addition of MB, an ~70 µA increase in the oxidation current was observed for CODH-L_gbpNC compared to that observed in the DET current (Table 3). This implies that a fraction of the immobilized enzymes adopted orientations that brought the active sites further from the electrode surface, or that the enzymes were immobilized in a non-monolayer fashion. As a result, only the first-layer enzyme molecules were able to transfer electrons directly to the electrode surface, and so upon the addition of the MB redox mediator, which can react with all enzymes on the surface, the MET current became significantly higher than the DET current. From these CV results, it was verified that CODH-L_gbpNC possessed a direct electrical contact with the electrode surface because there was a suitable ET distance between the two components.

These results rule out the hypothesis that the differences in the ET kinetics observed for the four CODH-L variants are due to differences in the binding strength, as verified from the catalytic current caused by MET, where all constructs were successfully immobilized and retained their biocatalytic activity, but only the synthetic CODH-L_gbpNC positioned the active site close to the electrode surface.