Electrochemical characterization of mediators. From a thermodynamic perspective, the formal reduction potential of the mediator must be lower than that of the reaction catalyzed by the [Fe-Fe] hydrogenase, which involves the interconversion of H+/H2 (E0′= -0.636 V vs. Ag/AgCl at pH 6.8), to facilitate efficient transfer of the reducing power towards the desired product. We selected four organic mediators, previously demonstrated by some of us to transfer electrons to [Fe-Fe] hydrogenases inside E. coli cells in a biohybrid system,31 as possible candidates for the construction of the system (Fig. 1A). These mediators were then examined through cyclic voltammetry (CV). As depicted in Fig. 1B, each mediator, starting from its (+ 2) oxidation state, undergoes two quasi-reversible reduction events (Table 1). Thus, each mediator has two redox couples, M2+/M1+ and M1+/M0, with reduction potentials that are thermodynamically favorable for driving H+/H2 interconversion. The more reduced couple of each mediator provides a large driving force for the reaction at the cost of a higher overpotential. Considering only the first reduction step, the reductive power of the mediators follows the order: DQ > DQ-CO2H > DQ-OH > MV.
Table 1
Formal reduction potentials of the investigated mediators.
Mediator | E0′M2+/M1+ (V vs Ag/AgCl) | E0′M1+/M0 (V vs Ag/AgCl) |
DQ | -0.873 | -1.065 |
DQ-CO2H | -0.845 | -1.110 |
DQ-OH | -0.808 | -1.035 |
MV | -0.660 | -1.010 |
Selection of suitable mediator. The capacity of the mediators to drive H2 production by E. coli BL21 DE3 cells containing [Fe-Fe] hydrogenase (E. coli (CrHydA1)) was evaluated by CV analysis. The E. coli (CrHydA1) cells heterologously expressed the CrHydA1 hydrogenase, and the enzyme was activated through artificial maturation to ensure a high and well-defined intracellular concentration.36,38 Fig. 2A shows CVs of 1 pM E. coli (CrHydA1) cells suspended in a deaerated 150 mM phosphate-buffered saline (PBS) at pH 6.8 (orange trace) in comparison to that of the same electrolyte without the cells (black trace). The CVs overlap, indicating (i) the absence of direct electronic interaction between the electrode and the cells; and (ii) the experimental conditions did not compromise cell viability during the measurement period, as cell lysis would result in the release of hydrogenase into the electrolyte leading to an increase in the cathodic current near E0′(H+/H2) arising from DET to the enzyme (vide infra).
Initially, MV was assessed as the electron transfer mediator as the formal potential of its first reduction (E0′ = -0.660 V vs. Ag/AgCl) lies closest to the E0′(H+/H2) among the selected mediators and therefore a successful mediation by MV2+/MV1+ will lead to the lowest overpotential requirement. The voltammogram recorded immediately after the addition of cells to a 2 mM solution of MV in the electrolyte showed a gradual increase in the anodic peak current at -0.620V V vs. Ag/AgCl, assigned to MV1+ → MV2+ transformation with no corresponding changes in the cathodic wave (Fig. S1A). Over time (7–30 min), this change in the voltammograms became more pronounced. This observation contradicts the anticipated response for MET from the electrode to the cytoplasmic hydrogenase. Instead, it suggests that in the presence of cells, MV2+ is converted into MV1+ by some cellular component. We attribute this to the reduction of the mediator by ferredoxin. Ferredoxin is abundant in E. coli and has a formal reduction potential in the range of -0.636 V to -0.734 V vs. Ag/AgCl. 39,40 Thus, ferredoxin having E0′ < -0.636 V can reduce MV2+ (Fig. S1B), leading to a rise in MV1+ concentration near the electrode surface, reflected in the gradual increase in the anodic current over time. No changes in the cathodic peak current corresponding to the second reduction of MV (E0′ = -1.010V vs. Ag/AgCl) were observed either, suggesting that MV1+/MV0 also cannot mediate electron transfer from the electrode to the hydrogenase inside the cells (Fig. S1A) under tested conditions. A sharp anodic peak at -0.973 V vs. Ag/AgCl was observed, which was previously assigned to the one-electron oxidation of the surface-adsorbed MV0 species. 41 Thus, MV0 cannot transfer electrons to the cells as it adsorbs to the electrode surface.
Among the diquat-based molecules investigated in this study, DQ has the lowest formal reduction potential in the first reduction event DQ2+ → DQ1+. It was therefore evaluated as the next mediator. Upon the addition of 2.2 pM of E. coli (CrHydA1) cells to a 200 µM solution of DQ in a deaerated 150 mM PBS at pH 6.8, an increase in the cathodic currents with a concurrent decrease in anodic peak currents compared to CVs of DQ in the absence of cells were observed for both DQ redox couples (black vs orange trace in Fig. 2B and Fig. S2A). With increasing E. coli (CrHydA1) cell concentration, the cathodic current augmented further and the anodic peak almost disappeared (purple trace in Fig. 2B and Fig. S2A). To quantify the current increase, we compared values of the cathodic peak current at -0.925 V vs. Ag/AgCl, corresponding to the DQ2+ → DQ1+ transformation, upon addition of cells to the value recorded without the cells (Fig. 2C). At the highest added concentration of E. coli (CrHydA1) of 8.8 pM, a current enhancement of 39 ± 5% was observed. This current enhancement did not depend on the concentration of the mediator (Fig. S3). It is worth mentioning here that the extent of the peak current increase started to plateau beyond 4.4 pM. This may arise from the precipitation of the cells, which was observed at cell concentrations > 4.4 pM. The current enhancement for the second reduction peak, assessed at -1.126 V vs. Ag/AgCl and assigned to the DQ1+ → DQ0 transformation, was 40 ± 8%, i.e. practically identical to that of the first reduction peak (Fig. S2B). Despite the larger driving force for the more reduced redox couple of DQ for electron transfer to the hydrogenase inside the cells, no additional current enhancement was observed. The observed changes in the CV are consistent with the expected response of a MET system.
To confirm that the changes in current are due to the catalytic activity of hydrogenase and not arising from some other process(es), we substituted the natural cofactor [Fe2(adt)(CO)4(CN)2]2− ([2Fe]adt, adt = azadithiolate) with [Fe2(pdt)(CO)4(CN)2]2− ([2Fe]pdt, pdt = propanedithiolate) during the enzyme maturation stage, (yielding E. coli (CrHydA1-PDT)). 42 In [2Fe]pdt, the amine bridgehead of the adt ligand is replaced with a methylene group that prevents [2Fe] subsite protonation and impedes catalytic turnover. 35,43 Upon performing electrochemical measurements with E. coli (CrHydA1-PDT), both cathodic and anodic currents of DQ were found to slightly decrease with increasing cell concentration (Fig. 2D), most probably due to blocking of the active electrode surface by cells. This observation underscores that the active form of the cytoplasmic CrHydA1 hydrogenase is necessary to manifest the changes illustrated in Fig. 2B and confirm that the observed changes in the CVs upon E. coli (CrHydA1-ADT) addition are indeed appearing from MET from the electrode to the cytoplasmic hydrogenase.
Following successful mediation by DQ, the reductive capability of DQ-CO2H was assessed. Upon the addition of E. coli (CrHydA1), cathodic currents emerged from − 0.636 V vs. Ag/AgCl, i.e. from E0′(H+/H2) (Fig. S4A). The cathodic current also increased with increasing concentration of E. coli (CrHydA1). However, this was approximately 98 mV higher compared to the onset of the reductive current corresponding to DQ-CO2H2+ → DQ-CO2H+ 1 reduction in DQ-CO2H (-0.734 V vs Ag/AgCl, black trace). Such a response is unexpected in a MET system, where changes are typically constrained to the CV features of the mediator. 44 Instead, we attribute this behaviour to direct electron transfer from the electrode to the hydrogenase, resulting in currents starting from the formal potential of the H+/H2 couple. 45 Therefore, the observed current response suggests that DQ-CO2H is adversely impacting the cellular integrity leading to the release of the cytoplasmic hydrogenase into the electrolyte. To test this, we first measured the CV of DQ-CO2H in the presence of 4.4 pM E. coli (CrHydA1) (Fig. S4B, red trace) and thereafter exposed the electrolyte to air for 1 minute. Any hydrogenase present in the electrolyte will become irreversibly inhibited due to its extreme sensitivity to oxygen. 46,47 The electrolyte was subsequently deoxygenated before being transferred to the glovebox and subjected to CV measurements again. The previously observed current increase completely disappeared (Fig. S4B, pink vs red trace), confirming that the changes indeed stemmed from the hydrogenases released into the electrolyte (contrary to the behavior observed in the presence of DQ, as discussed below). Consequently, DQ-CO2H was excluded from further consideration in this study.
Finally, DQ-OH was evaluated as the mediator. As for DQ, an increase in the cathodic current with a concurrent decrease in the anodic current was again observed in CVs with the addition of E. coli (CrHydA1) to the 200 µM DQ-OH solution for both DQ-OH redox couples (Fig. S5A). Similar to DQ, the changes in the currents were confined to the CV features of DQ-OH, suggesting that, unlike DQ-CO2H, DQ-OH does not adversely impact cellular integrity. Fig. S5 shows the current increase at both the first (-0.855 V vs. Ag/AgCl) and second (-1.105 V vs. Ag/AgCl) reduction peak potentials upon the addition of cells in comparison to the value recorded without the cells. The maximum increase of 27% was observed at the first peak potential upon the addition of 8.8 pM of cells. As in the case of DQ, the current enhancement for the first and second reduction peaks of DQ-OH was similar. Overall, the current enhancements observed with DQ were consistently higher compared to DQ-OH. Therefore, further stability studies of the system were performed with DQ as the mediator.
Cell integrity analysis. As DQ-CO2H prompted the release of hydrogenase into the electrolyte, we opted to verify cellular integrity following electrochemical studies with DQ to ensure that the observed responses originated from cytoplasmic hydrogenase. After recording CVs of DQ in the presence of 8.8 pM E. coli (CrHydA1), we centrifuged the electrolyte to precipitate the cells. The resulting supernatant is devoid of E. coli (CrHydA1) but will still contain hydrogenase released from the cells in the event of cell lysis. The supernatant was subjected to CV (Fig. S6). The CV traces of the supernatant overlapped with that of DQ in isolation (light purple vs. black trace in Fig. S6) suggesting that (i) all current changes were arising from the interaction between DQ and cytoplasmic hydrogenase present in E. coli (CrHydA1), and (ii) no hydrogenase is present in the electrolyte. Further, SDS-PAGE analysis on both the supernatant and the centrifuged cells was performed to check the presence of hydrogenase. While the centrifuged cells showed a strong band close to 50 kDa corresponding to hydrogenase, no corresponding band was present in the supernatant (Fig. S7). Indeed, the SDS-PAGE analysis showed no indication of proteins in the electrolyte. This further confirms that in measurements using DQ as a mediator, the cells remain intact under our experimental conditions and the observed current enhancements are entirely due to MET between the electrode and cytoplasmic hydrogenases.
To assess the impact of DQ on the cellular integrity and bioelectrochemical performance of E. coli (CrHydA1) over an extended duration, we incubated 8.8 pM E. coli (CrHydA1) cells with 200 µM DQ in a sealed glass vial at 4 ˚C under strictly anaerobic conditions for 24 hours. We then compared the CVs of the solution before and after 24-hour storage (Fig. S8A purple vs light purple traces). After 24 hours, the system retained ~ 95% of its initial activity with no significant changes in the CV features (Fig. S8B). The latter observation further confirms that the cells remained intact as the release of cytoplasmic hydrogenase into the electrolyte is expected to change CV features (Fig. S4A), as observed with DQ-CO2H. The minimal change in activity, coupled with the preservation of cellular integrity, suggests that DQ can serve as an effective and safe mediator for electromicrobial production with suspended E. coli cells engineered with functional [Fe-Fe] hydrogenases.
Oxygen tolerance of the electromicrobial production system. It has been previously observed that E. coli provides protection to cytoplasmic hydrogenase from the external environment, particularly oxygen. 29,48 CrHydA1 is an extremely oxygen-sensitive enzyme and in our measurements using DQ-CO2H as the mediator, we observed that exposing the electrolyte to air for one minute leads to the complete disappearance of the current enhancements originating from hydrogenase present in the electrolyte (Fig. S4B). To check the oxygen sensitivity of the electromicrobial production system based on DQ-mediated E. coli (CrHydA1), we exposed the electrolyte to air for 10 or 20 minutes. We then compared CVs of 200 µM DQ in the presence of 8.8 pM E. coli (CrHydA1) before and after oxygen exposure (Fig. S9, purple vs light purple traces). As shown in Fig. S9B, the activity drops to 74% and 70% of the initial enhancement after 10 and 20 minutes, respectively. Such retention of activity after 20 minutes of air exposure suggests that the cytoplasm of E. coli is able to partially protect CrHydA1 from oxygen, most likely due to the high respiration rate of the bacteria resulting a low intracellular concentration.
Modeling of electrochemical data. To unravel the kinetics of MET to cytoplasmic hydrogenase and pinpoint the rate-determining step of the electromicrobial synthesis platform at the level of cells, we constructed a kinetic model allowing us to simulate the experimental cyclic voltammograms. We considered two different approaches to model microbial cells: (I) by modeling the activity of individual cells 49–52; or (II) by using a continuum approach, where the behavior of single cells is neglected, and the average concentration of components is used instead (Scheme 1). 33 The difference between these two approaches lies in what is considered the catalyst and how the concentration and activity of the catalyst are calculated. In the first approach, E. coli cells are the catalyst. This approach is based on the adaptation of early reported modeling frameworks for modeling CVs of biofilm anodes.49–52 In the second approach hydrogenase is the catalyst and it is assumed that it is homogenously distributed in the reaction volume. Both approaches are based on molecular electrochemistry methods commonly used to model enzymatic kinetics.53,54 Further, they incorporate steps reported for the modeling of biofilm-modified anodes but adapted to model the cathodic process.49 For both approaches the same steps with slight modifications were used (Scheme 1).
Step 1 represents the reversible reduction of the mediator at the electrode surface characterized by the formal potential and standard heterogeneous electron transfer rate constant (k0), both of which can be obtained by simulating CVs of the mediator without the addition of cells (Fig. 1, Table S1). Mediator oxidation and reduction rates at the electrode surface depend on the potential applied relative to the formal potential of the mediator as described by the Butler-Volmer equation. 55 Initially, only the DQ response was modeled and only one redox couple of the mediator, DQ2+/DQ1+, was considered. The model was subsequently further extended to include the second redox couple, M+ 1/M0 and DQ-OH.
Step 2 describes the reduction of the biocatalyst (either the E. coli cells or hydrogenase) by electron transfer from the mediator. The biocatalyst in its native state was designated as BioCat(Ox), while its states following successive reductions were labeled as BioCat (Red) and BioCat (Red2). The electron transfer from the mediator to hydrogenase could occur directly or via an intermediate intracellular species, which undergoes initial reduction before transferring the electron to hydrogenase. Thus, in the context of [Fe-Fe] hydrogenase, BioCat(Ox), BioCat (Red), and BioCat (Red2) can be interpreted as the well-known Hox, HredH+ and HHyd states, all of which have been shown to accumulate in vivo.38 However, the model does not a priori make any assumption on the exact nature of the three redox states. For simplicity, the rates associated with the electron transfer steps (kET), regardless of their actual complexity, were denoted as kET1 and kET2, respectively, where kET1 and kET2 were assumed to represent the rate-limiting step of the entire electron transfer process from the mediator to the biocatalyst for both approaches. In the modeling, the magnitudes of kET1 and kET2 were assumed to be equal (kET1= kET2= kET)
Step 3 is the catalytic turnover of the biocatalyst in which protons are converted to hydrogen. In models of biofilm-modified anodes, this step is usually assumed to follow Michaelis-Menten kinetics. In our modeling for both approaches it was assumed that there is excess of the substrate and this step was modeled as a single catalytic step with the rate constant kcat. This assumption should be valid as the pKa of the proton-donating Cys169 in the active site of CrHydA1 is 12,56,57 so it should always be protonated at the pH 6.8 used in the experiments. While the internal pH of the cell might differ from the pH of the buffer used, and it was reported to be approximately 0.5 pH units higher in E. coli cytoplasm, 58,59 it is still sufficiently low to maintain effective protonation of the Cys169.
Mass transport of the mediator, cells, and hydrogen occurs only by diffusion in the model and is described by Fick’s law of diffusion, 55 parametrized by the corresponding values of the diffusion coefficients (D) taken from the literature (Table S1). Further, experimental conditions are selected in such a way that mass transport of the mediator does not limit the catalytic response as the current enhancement does not depend on the concentration of the mediator (Fig. S3) and this condition is reproduced in the model. To simplify the model, the reactions of Steps 2 and 3 were considered as unidirectional occurring only in the direction of H2 generation. Considering the reduction potentials, this assumption should hold for all diquat derivates studied here. In short, current enhancement in the presence of E. coli cells results from the reduction of the excess of the oxidized mediator at the electrode surface (Step 1). This excess is generated by oxidation of the reduced mediator by the biocatalyst via electron transfer between the reduced mediator and the biocatalyst (Step 2), which in turn is sustained by the catalytic activity of the biocatalyst (Step 3).
Approach I. To model the activity of individual E. coli cells we first estimated the kcat of individual cells using the earlier calculated concentration of hydrogenase inside the cell (43 000 molecules or 70 µM) and the estimated catalytic activity of a single hydrogenase enzyme, 400 s− 1.32 Assuming a linear correlation between the total enzyme amount and rate, a calculated value of kcat of 1.72×107 s− 1 is obtained. Using the calculated kcat, 8.8 pM concentration of the cells, and values of other parameters specified in Table S1, simulated CV traces reproduced the experimentally observed current response (Fig. 3A vs Fig. 2B, purple trace) for kET ~ 400–500 kcat. Higher values of kET relative to the kcat resulted in a larger enhancement of the cathodic current. To simulate the gradual addition of E. coli (CrHydA1) during the measurements (Fig. 2B), we varied the concentration of the biocatalyst from 2.2 to 8.8 pM for kET = 500 kcat (Fig. 3B). A gradual increase in the cathodic current with a subsequent decrease in the anodic current was observed in the simulated CVs (Fig. 3B), reproducing the experimental data.
We further performed simulations to mimic control experiments (Fig. 2D). Hydrogenases artificially matured with [2Fe]pdt can undergo a single electron reduction 60 and this was used in the simulations. In this case, both kET2 and kcat were set to 0, while kET1 was set to 6.88×109 s− 1 (i.e. 500 × 1.72×107 s− 1). The simulated catalytic CV does not show any current enhancements compared to the mediator CV (black vs. purple curves in Fig. S10A), in agreement with the experimental data.
Approach II. The average concentration of hydrogenase in solution at the added concentration of 8.8 pM cells, when neglecting the existence of individual E. coli cells but considering that each cell on average contains 43000 CrHydA1, 32 was calculated to be ~ 0.4 µM. Using the kcat of 400 s− 1 32 and values of other parameters shown in Table S1, simulated catalytic CV also reproduced the experimentally observed current enhancement (Fig. 3C vs Fig. 2B, purple trace) when kET ~ 300–400 kcat. Similar to the results obtained with Approach I, increasing the value of kET relative to the kcat value led to an increase in the catalytic current. Figure 3D shows the dependence of the simulated current responses on the concentration of the hydrogenase corresponding to experimental changes in E. coli (CrHydA1) cells concertation (Fig. 2B). As with Approach I the simulation reproduced the experimental data. The control experiment using E. coli (CrHydA1-PDT) (Fig. 2D) was also successfully reproduced using Approach II (Fig. S10B).
Thus, both Approach I and II allowed to reproduce the experimental data. In both cases, the current enhancement increased with increasing the value of kET relative to kcat (up to very large values of kET, at which point kcat became limiting, as shown in Fig. S11), demonstrating that this is the rate-limiting step of the process. While values of kET that reproduced the experimental data were somewhat different for Approach I and II (kET ~ 400–500 kcat vs. kET ~ 300–400 kcat, respectively), when used to calculate the initial rate of the electron transfer reaction (Step 2), they gave the same value of ~ 10− 5 M/s (Note S1).
We further investigated the effect of the driving force on the rate-limiting step of the process. For this, simulations of experimental current enhancements for both redox couples of DQ and DQ-OH were performed (Note S2, Fig. S12). Since Approach I and II gave the same results, only Approach I was used for these simulations. Figure 4C shows simulated CVs reproducing experimental data (Fig. S2A) and current enhancements for two cathodic peaks (Fig. 4A) obtained with DQ. In order to reproduce experimentally observed shapes of catalytic CVs for both redox couple of DQ (equivalent current increase at both cathodic peaks) for all concentrations of the cells, the electron transfer rate from the fully reduced DQ0 to the E. coli (CrHydA1) should be at least an order of magnitude lower than the electron transfer rate from DQ1+ (Fig. S13A). This effectively means that mediation occurs primarily by the DQ2+/DQ1+ redox couple. The contribution of the DQ1+/DQ0 redox couple to the mediation process would result in a large current enhancement for the second redox peak (Fig S13A, light purple traces), which was not observed experimentally. Figure 4D depicts simulated CVs reproducing experimental data (Fig. S5A) and current enhancements for two cathodic peaks of DQ-OH (Fig. 4C). Cathodic current enhancements in the case of DQ-OH were significantly lower than for DQ for both cathodic peaks and all concentrations of the E. coli (CrHydA1) cells. To reproduce these experimental data kET~200 kcat is required. Similar to DQ, mediation occurs primarily by the DQ-OH2+/DQ-OH1+ redox couple (Fig. S13B). The driving force for electron transfer by the M2+/M1+ redox couple for DQ-OH is 0.065V lower compared to DQ (Table 1), while the rate of the electron transfer to the E. coli (CrHydA1) is ~ 2.5 times lower for DQ-OH than for DQ.