Artificial, Photoinduced Activation of Nitrogenase Using Directed and Mediated Electron Transfer Processes

Nitrogenase, a bacteria-based enzyme, is the sole enzyme that is able to generate ammonia by atmospheric nitrogen fixation. Thus, improved understanding of its utilization and developing methods to artificially activate it may contribute to basic research, as well as to the design of future artificial systems. Here, we present methods to artificially activate nitrogenase using photoinduced reactions. Two nitrogenase variants originating from Azotobacter vinelandii were examined using photoactivated CdS nanoparticles (NPs) capped with thioglycolic acid (TGA) or 2-mercaptoethanol (ME) ligands. The effect of methyl viologen (MV) as a redox mediator of hydrogen and ammonia generation was tested and analyzed. We further determined the NPs conductive band edges and their effect on the nitrogenase photoactivation. The nano-biohybrid systems comprising CdS NPs and nitrogenase were further imaged by transmission electron microscopy, visualizing their formation for the first time. Our results show that the ME-capped CdS NPs–nitrogenase enzyme biohybrid system with added MV as a redox mediator leads to a five-fold increase in the production of ammonia compared with the non-mediated biohybrid system; nevertheless, it stills lag behind the natural process rate. On the contrary, a maximal hydrogen generation amount was achieved by the αL158C MoFe-P and the ME-capped CdS NPs.


Introduction
Electrical communication between enzymes and their surroundings is tightly controlled by natural diffusional redox mediators. [1][2][3][4][5] Alternatively, protein-protein interactions at specific docking sites can force short electron transfer distances that, in turn, allow efficient electron transfer. The outer shells of redox enzymes are, by design, insulators and so prevent short circuits with undesired electron carriers present in surrounding proteins. In the last several decades, the theoretical and experimental foundations of these processes were developed, improving our understanding thereof. Establishing electrical communication between redox proteins and electrodes or inorganic nanomaterials is a major challenge that has been at the forefront of research for the last two decades. Using redox mediators or directed electron transfer configurations has led to a variety of sensing or energy generation devices.
For example, glucose oxidase (GOx) and glucose dehydrogenase (GDH) were each conjugated with electrodes and nanoparticles (NPs) for the development of amperometric glucose biosensors. [6][7][8][9] Photosystem I and Photosystem II proteins were each coupled to either electrodes to construct photobioelectrochemical cells or to Pt and Pd nanoclusters to generate hydrogen. [10][11][12][13] Hydrogenase was coupled to NPs or to an electrode, facilitating high turnover rates in both hydrogen oxidation and proton reduction reactions. [14][15][16] In all these examples, tailored chemistries for direct electron transfer processes or integration of redox mediators enabled enzyme activation. Several parameters should be taken into account when establishing electrical communication between biotic and abiotic materials: (i) short electron transfer distance, (ii) orientation of the proteins' active site with respect to the electron acceptor or donor, and (iii) adjusting the redox potential to allow the directional electron transfer process.
Nitrogenase, a bacteria-based enzyme with a α2β2δ4 heterooctameric or α2β2 heterotetramer composition (Azotobacter vinelandii, VFe-P, and MoFe-P, respectively), is the sole enzyme able to generate ammonia by atmospheric nitrogen fixation 17,18 Although these exceptional enzymes have been studied for several decades, their activation mechanism has yet to be fully resolved.
For example, nitrogenase was coupled to an electrochemical mediated system in order to reduce azide or CO2 into NH3 or H2 / formate under highly biased electrodes. 19,20 By integrating a nitrogenase-based cathode with a hydrogenase-based anode, a biofuel cell mimicking the Haber-Bosch process was developed. 21,22 In most cases, cobaltocene was used as a redox mediator, hence a bias of -1.1V vs. Ag/AgCl was used for activation.
The nitrogenase αβ subunits consist of two metal centers: (i) The P-cluster, an iron-sulfur center (8Fe:7S), which acts as an electron mediator accepting the shuttled electrons from the Fe-protein; and (ii) the M-cluster, a unique complex with a 7Fe:9S:M:1C center, which acts as the catalytic center and allows N2 fixation to occur. M represents molybdenum, vanadium or iron metal ions, and C represents carbide.
Activation of nitrogenase requires Mg-ATP and reduced Fe-protein. The ATP allows a structural conformation change in the Fe-protein, leading to efficient electron transfer toward the P-cluster. 23 ATP does not act as an electron donor, and only supports the binding of the Fe-protein to the nitrogenase Pcluster. The mechanism of the process is not yet fully understood and methods to overcome the gated process should be developed. It is known that 8eare involved in the process, which produces two moles of NH3 and one mole of H2 as a by-product (see Eq. 1).

Equation 1: Nitrogen fixation process.
It has been shown that nitrogenase can be activated by using CdS NRs, anchored Ru(bpy)3, or pyrene moieties without the addition of ATPs and Fe-protein. [24][25][26] Although NH3 production by MoFe-P is very attractive, several challenges limit further progress: (i) The MoFe-P redox-active sites, the P-cluster, and the M-cluster, are all extremely sensitive to the presence of oxygen; (ii) The MoFe-P electron transfer mechanism for NH3 generation requires eight successive moles of electrons to generate two moles of NH3.
Furthermore, the reduction mechanism or the way in which the N2 molecule coordinates with the M-cluster complex is not fully understood and is still under debate; 27,28 and (iii) In nature, electron transfers occur in eight consecutive steps of a single electron transfer toward the P-cluster, which thereafter transfers these electrons to the M-cluster. In this process, the reduced Fe-protein acts as an electron donor, while an ATP molecule functions as a kind of key that forces a structural change in the Fe-protein, thus allowing the electron transfer process. This rate-limiting process, which occurs in nature, may be improved and methods to bypass this regulated mechanism should be developed to allow artificial activation of nitrogenase enzyme by external stimuli or an electron source other than the Fe-protein. Several basic questions remaine unanswered: How does the interface between nitrogenase and the artificial system (electrode, NPs, etc.) influence the electron transfer process? How does the nitrogenase bind to the electrode or semiconductor?
As mentioned, artificial activation of nitrogenase has been recently realized, although the role of the Feprotein or its absence in artificial systems, the electron transfer process, and the conditions affecting the ammonia and hydrogen generation remained unresolved and warrant further investigation.
Here, we show a direct and diffusional approach to the artificial activation of MoFe-P by CdS NPs 10 nm in diameter (see Fig. 1). We examined the role of different ligands and their effect on the direct electron transfer process, and then examined the mediated electron transfer process using methyl viologen (MV) as a redox mediator. The generated NH3 was determined by an updated analytical method that minimizes the false-positive response that may occur when using the commonly used fluorescence-based probe.
Two variants of MoFe-P, which express a His-tag, were isolated from Azotobacter vinnelenlii mutated strains DJ1194 and DJ995 as previously described. 29,30 The mutated L158C MoFe-P, isolated from DJ1194, consists of cysteine at the 158 instead of lysine as expressed in the wild type (WT MoFe-P). The close proximity of Cys 158 to the iron-sulfur P-cluster at the nitrogenase's electron acceptor site could potentially allow a short electron transfer distance between a bonded external electron donor and the enzyme's P-cluster. Indeed, Tzecan and co-workers showed methods of integrating Ru(bpy)3-based photosensitizers with nitrogenase for independent Fe-protein activation. The artificial system was limited by photosensitizer degradation, preventing continuous long-term activation. 24 For the artificial activation of nitrogenase, 10-nm CdS NPs were synthesized using a hydrothermal method. 31 The CdS particle size was determined using transmission electron microscopy (TEM) imaging and UV/Vis spectroscopy (Figs. 2a and 2b). Using ligand exchange steps, we altered the CdS NPs surface by replacing the oliel amine/oleic acid with shorter thioglycolic acid (TGA) or 2-mercaptoethanol (ME) ligands, (see Fig. S2). The CdS NPs were dissolved in aqueous solution and their charges were varified using gel electrophoresis and zeta potential measurements (Fig. 2c). As expected, the TGA-modified CdS NPs exhibited dominant-negative charges and traveled through the 0.6% agarose gel to the positive pole.
On the other hand, the ME-modified CdS NPs showed only a minute effect and did not migrate from the well, although small negative charges were measured by the zeta potential setup. Ligands have an important effect on the energy levels or traps on the NPs' surface. [32][33][34] Since the position of the energy levels in the CdS NPs could greatly affect the photoinduced electron transfer process, we determined the band edges of CdS-TGA and CdS-ME NPs using capacitance measurements (Mott-Schottky technique). 35,36 The TGA form exhibited a Conductive band (CB) value of -0.9V vs. Ag/AgCl, pH 7.2 while the ME form exhibited a higher negative potential of -1.2V (Fig. 2d). As shown previously, thiol-Cd interactions form electrons or hole traps on the surface of NPs [32][33][34] , which subsequently slow down recombination reactions and allow improved electron transfer processes. The TGA-and ME-modified CdS NPs were further utilized for photoinduced nitrogenase activation. In each experiment performed, oxygen-free CdS NPs (stored in an Ar-filled glove box, O2<0.2 ppm) were mixed with phosphate buffer (PB) pH 7.2 and ascorbic acid, which was used as an electron donor. The sealed vial was then purged in the dark with N2 for 1 hr followed by injection of the MoFe-P solution, in the glove box, using an insulin syringe (Pic solution, 0.5mL, 31G).
Then, the sealed vails were irradiated using a 465-nm LED light. The H2 generated was monitored using a GC-TCD and the NH3 generation was determined using an o-phthaladehyde (OPA) fluorescence probe, which was later extracted to chloroform as depicted in Fig. S6. The NH3 detection method was validated using NH3 samples of known concentrations, as described in the supporting information. With this method, a detection limit of 5µM can be reached, (see Fig. S5). Using the developed method and GC-TCD chromatography, we examined how the charges on the NPs' ligands affect the biohybrid system activation. Figure 3 depicts the amounts of H2 and NH3 obtained using CdS NPs-MoFe-P mixtures. We examined two variants of the nitrogenase protein: WT MoFe-P (His-Tag, DJ995) and L158C mutated nitrogenase (His-Tag, DJ1094). The cysteine-mutated nitrogenase potentially has two Cys positions that may directly bond the NPs through the strong thiol-Cd interaction. The mixtures of CdS-ME NPs with both of the nitrogenase variants generated double amounts of H2 compared with the TGA NPs mixtures. We also observed that using L158C MoFe-P, generated up to ~2000nmol of H2 after only two hours of blue light irradiation.
MV is a commonly used redox relay unit for both biotic and abiotic electron transfer processes [37][38][39] . The pH-independent one-electron reaction has a strong reducing potential of -0.67V vs. Ag/AgCl in aqueous solutions, which can be used for MoFe-P P-cluster reduction (~-0.5V). MV was therefore introduced into the CdS-ME/TGA NPs-MoFe-P mixture, and H2 and NH3 generation was monitored. Interestingly, when MV was added to the mixture as an electron shuttle mediator, H2 generation was completely inhibited. By measuring the extracted NH3-OPA conjugates without the presence of MV, both MoFe-P variants produced similar amounts of NH3 regardless of the capping ligands. On the other hand, adding MV to the CdS-TGA NPs led to a two-fold increase in the amounts of NH3 generated compared with the MV-free reactions.
Moreover, when WT MoFe-P and L158C MoFe-P were activated with CdS-ME NPs, a five-fold increase in NH3 production was observed. We further tracked the rate of NH3 generation (Fig. S8) and recorded similar rates of 0.1 nmol/min for both variants activated by CdS-ME NPs in the presence of MV. In our design, we assume electrostatic interactions between CdS NPs and MoFe-P. To examine this hypothesis, we examined the nano-biohybrid system using TEM imaging. Samples were stained with uranyl acetate to allow nitrogenase imaging, which demonstrated the formation of nano-biohybrids of nitrogenase and the CdS-ME NPs (red arrows in Figs. 4 and S10).

Discussion
The results obtained indicate that the addition of MV can enhance the amount of NH3 generated compared with nitrogenase that is activated solely by CdS NPs. Moreover, the results show that ME-capped NPs lead to better catalytic performance compared with the charged TGA ligand. To better understand this phenomenon, we first examined the CB edges of the capped NPs (Fig. 2d). The CB edge of CdS-ME NPs is ~300mV more negative than that of CdS-TGA NPs. As the data reveals, we can exclude the option of direct N2 reduction by the CdS NPs surface with or without MV as redox mediator or catalyst. We also examined whether chemically reduced MV can contribute directly to nitrogenase activation, and found that NH3 generation was low (Fig. S9). The data obtained suggest that the CdS surface is essential for the reduction process, and that MV dramatically increases process efficiency. This might be as a result of a conformational change in the nitrogenase proteins that allows better electron injection into the nitrogenase P-cluster. The ME ligand has two different roles in the catalytic process: (i) it shifts the CB potential to -1.2V (vs. Ag/AgCl), and (ii) it weakens the electrostatic interaction with MV and/or the MoFe-P. Analysis of the result obtained without MV leads us to conclude that the CB potential has only a minute effect and we can, therefore, suggest that the less prominent charges of the ME ligands enable improved release of mediator leading to higher turnover rates. The L158C MoFe-P consists of two Cys sites that can potentially bond directly to NPs. If indeed the conjugation occurs at the Cys 158 site, an improved electron transfer rate should be expected. By comparing the results obtained with L158C MoFe-P and with WT MoFe-P, we can conclude that no direct integration occurs or that it may shift the catalytic process to prefer a hydrogenation mechanism that minimizes N2 fixation. The enhanced amount of H2 obtained with L158C MoFe-P could indicate the latter. It should be noted that when MV is introduced into the system, H2 cannot be measured; Although MV is a well-known electron transfer carrier, it does not transfer protons. It is possible that the protein's natural activity is prohibited and that the fast electron injection caused by MV eliminates H2 generation and drives solely to ammonia production. It has been shown that the Fe protein binds to nitrogenase via a positively charged docking site. [40][41][42] We hypothesized that TGA-CdS NPs with negative charges on their surface will bind to the docking site and force improved electron transfer process.
By examining the obtained data (Fig. 3), we conclude that such an effect did not occur. While surface charges can enable electrostatic interactions between NPs and MoFe-P, they cannot force the gated conformational changes required to allow the electron transfer process; this key process must be controlled for improved activity. Examining the TEM imaging (Figs. 4 and S10), it is clear that at least one L158C nitrogenase is bonded to each ME-CdS NP. These results confirm that although conjugation is indeed achieved, it is not sufficient for the directed electron transfer process we were aiming for. Further investigation of artificial photoactivation using smaller sized NPs might link directly to the C158 site and subsequentially provide sufficient electron flux to ungate the nitrogenase electron transfer mechanism. If indeed will be realized, MV will become otiose as a result of the direct electron injection to the P cluster.
A great example for that was already shown, while photoreduction of acetylene was performed using Ru(bpy)3 photosensitizer bonded to L158C MoFe-P. 24 In conclusion, we examined the photoactivation of nitrogenase by CdS NPs. NPs were modified with TGA acid or ME ligands and their aptness to bind and activate nitrogenase was tested. We show for the first time TEM imaging of the hybrids. We further examined the systems using electrochemical, spectroscopic, and analytical tools. The effect of MV, as a redox mediator, on the generation of ammonia and hydrogen was examined. Our results show that ME-modified NPs have an advantage in hydrogen generation compared with TGA ligands, while a similar amount of ammonia was detected for both configurations. Adding MV led to complete elimination of hydrogen generation, while ammonia amounts increased five-fold. These results improve our understanding of the interface between the nitrogenase enzyme and abiotic materials used for its artificial activation.

Materials and equipment:
All chemicals and solvents used were at least analytical grade and were used as received. UV-Vis spectra were acquired using an Evolution 201 -Thermo Scientific spectrophotometer. Gas samples were analyzed via gas chromatography (GC) using an HP/Agilent 6890 (G1530A) system equipped with a 5-Å column and a thermal conductivity detector. Fluorescence intensity was measured by a plate reader (Synergy H1, BioTek) using a Hellma black quartz microplate. All electrochemical measurements were performed using a Biologic SP200 Potentiostat. TEM measurements were performed using a BioTWIN T12 and zeta potential was measured using a Malvern Zetasizer Ultra setup.
CdS nanoparticle synthesis (10 nm): CdS NPs were synthesized using a previously reported method with some modifications (see TEM images in Figs 1a and S1). 43  L of ME (>99.0%, Sigma-Aldrich), and 120 L of 1 M NaOH was added to the CdS/toluene solution.
This solution was then vigorously stirred overnight in the dark until the ligand-modified NPs were transferred to the aqueous phase. Upon completion, the colorless toluene phase was discarded and the watersoluble CdS-ME NPs were centrifuged at 200 x g for 2 min to yield a clear, yellow solution (see Fig. S2).
This clean solution was then filtered using a 30 kDa centrifuge filter (Amicon Ultra-15, Mercury) at 4,200 rpm to remove any excess ligand. CdS-ME NPs were then re-dissolved in H2O and their concentration was determined by UV-Vis spectroscopy (excitation coefficient of 2.56x10 6 M -1 ·cm -1 at 472 nm).
CdS ligand exchange to thioglycolic acid (CdS-TGA NPs): TGA (98%, Sigma-Aldrich) was added to a stock solution of CdS-ME NPs until the solution turned hazy (5 µL TGA per mL of CdS-ME). The solution was then stirred in the dark for 5 hr following which 1 M NaOH was slowly added until the solution became clear. Upon completion, the CdS-TGA NPs were cleaned using a 30 K centrifuge filter (Pall Corporation) at 4,200 rpm. The CdS-TGA NPs were then re-dissolved in H2O and the concentration was determined by UV-Vis spectroscopy. CdS-ME to CdS-TGA ligand exchange was also confirmed via agarose gel electrophoresis and zeta potential measurement.
Light-driven NH3 and H2 production from nitrogenase: Reaction mixtures were prepared in 1. The lower phase (chloroform) was then collected and transferred to a new tube. The extracted NH3/OPA conjugates was diluted five-fold with chloroform and loaded into a 96-well black quartz plate (Hellma).
Fluorescence intensity of the different samples was measured using a plate reader (Synergy H1, BioTek; λ excitation=410 nm, λ emission=472 nm). NH3 was quantified using the calibration curve shown in Fig. S5.    where C is the molar concentration (mole·L -1 ), A is the absorbance, L is the path length (cm), and 472nm is the molar extinction coefficient at 472 nm (2.56x10 6 M -1 cm -1 ).
Equation S2 used for the determination of the Ebg, when λ is the wavelength of the first peak and h represents Planck's constant (4.136x10 -15 ev·sec), c is the velocity of light (4.136x10 17 nm/sec)