ORP control for boosting ethanol productivity in gas fermentation systems and dynamics of redox cofactor NADH/NAD+ under oxidative stress

Abstract Gas fermentation processes have attracted considerable attention in recent years as they hold high potential for capturing and converting C1 waste gases into a range of biofuels and commodity chemicals. The production of solvents in gas fermentation is typically achieved by exploiting the solventogenic metabolism of acetogenic cultures, which is generally triggered upon exposure to stressful conditions, e.g. low pH. Although the oxidoreduction potential (ORP) is a well-known trigger of the cellular stress response, it has been scarcely investigated as a process control parameter in gas fermentation. Thus, this study focused on evaluating the potential of ORP control strategies for boosting the productivity of ethanol by exploiting the metabolic response to oxidative stress of acetogenic cultures. The dynamics of the redox cofactor pool and ratio as a function of the extracellular ORP and other operational parameters were also studied by monitoring the intracellular levels of the redox cofactor NADH/NAD+. The results showed that increasing the ORP to oxidizing conditions using dilute H2O2 triggered a 3.7-fold increase in the specific ethanol productivity, from 0.63 ± 0.04 mmol∙gCDW−1 h−1 at an ORP of -210 mV to 2.32 ± 0.19 mmol∙gCDW−1 h−1 at 160 mV. Additionally, the concentration and product selectivity towards ethanol also increased considerably due to the partial inhibition of the chain elongation under oxidative stress. Boost in ethanol productivity and inhibition of the chain elongation were both found to be driven by the presence of H2O2 rather than by the ORP per se. Studying the profile of the redox cofactors revealed a highly dynamic nature in the pool and ratio of NADH/NAD+ as a function of the specific uptake rate and the ratio of acetate-to-ethanol, respectively. The latter was explained by analyzing the thermodynamics of the aldehyde:ferredoxin oxidoreductase (AOR) pathway, which showed that the intrinsic thermodynamic limitation of this pathway imposes a high Fdred/Fdox ratio (>88 % of reduced ferredoxin) while forcing a highly dynamic NADH/NAD+ ratio in order to maintain the thermodynamic drive in the forward direction. The dynamics of the NADH/NAD+ ratio were also found to be significantly affected by the oxidative stress triggered by dilute H2O2, which confirmed the involvement of the AOR pathway in the detoxification of reactive oxygen species.


Introduction
The threat of global warming and the insufficient CO 2 emission reduction scenarios projected based on the Paris agreement and the national pledges make the implementation of carbon capture and utilization technologies more imperative than ever. Moreover, the need of decoupling transportation fuels from fossil resources along with the fact that carbon-based fuels and chemicals will not be phased out in the future, evidences that the latter must be sourced from renewable feedstocks. Gas fermentation has attracted significant attention as a viable approach to tackle both issues simultaneously, as this technology holds high potential for displacing fossil fuels by capturing and converting C1 waste gases, such as CO and CO 2 , into a range of biofuels and commodity chemicals [1,2]. Additionally, the fact that gas fermentation systems could operate in the future on syngas derived from gasification, if the present challenges related to toxic impurities are overcome, opens the way for the treatment and revalorization of a wide array of organic waste streams and recalcitrant biomasses through a hybrid thermochemical/biological conversion process.
Although significant efforts have been made to expand the product portfolio of gas fermentation systems, so far, ethanol is undoubtedly the most investigated product. The production of ethanol with either high selectivity or high concentration has been reported using several acetogenic cultures, including e.g. Alkalibaculum bacchi, Clostridium aceticum, Clostridium ragsdalei, Clostridium carboxidivorans, Clostridium ljungdahlii, Clostridium autoethanogenum and several open cultures [3][4][5][6][7]. Regardless the biocatalyst used, the production of ethanol from C1 gases takes place through the Wood-Ljungdahl or reductive acetyl-CoA pathway, either through the reduction of acetyl-CoA to acetaldehyde and ethanol by aldehyde/alcohol dehydrogenase (AdhE) using NADH as electron carrier, or through the reduction of acetate by aldehyde:ferredoxin oxidoreductase (AOR) using reduced ferredoxin (Fd red ) and NADH as electron carriers. However, the more favorable bioenergetics of the reductive acetyl-CoA pathway when coupled to the AOR pathway suggest that ethanol is mostly produced through the intermediate synthesis of acetate and further reduction by AOR and AdhE, which has been confirmed experimentally [6,[8][9][10].
A wide range of operational strategies and reactor configurations have been studied in recent years for enhancing ethanol productivity and titer in gas fermentation systems [1,11,12]. Medium composition and nitrogen source are known key parameters determining the performance of acetogenic bacteria and their product specificity. Shen et al. [13] showed that optimizing the trace metal content of the growth medium increased the alcohol-to-product ratio from 47.7%-89.8% in batch experiments. Similarly, using nitrate instead of ammonium as nitrogen source was found to improve growth and ethanol production significantly, but promoted a metabolic crash upon nitrate and nitrite accumulation [14]. Lowering the incubation temperature is yet another option for enhancing the product titer of alcohols, as it was shown to favor the re-assimilation of organic acids and their further reduction to alcohols, and to prevent acid crash under uncontrolled pH due to the lower metabolic rates of the culture [15]. Nevertheless, the most common strategy for boosting the product titer and productivity of ethanol is to increase the level of stress of the culture in order to trigger solventogenesis, which is typically achieved through the modulation of the pH conditions. For example, decreasing the initial pH from 6 to 5 resulted in an increase of ethanol yield from 0.082 to 0.592 e-mol of ethanol/e-mol of syngas in a sequential batch enrichment of an open culture under uncontrolled pH [6]. Similarly, a rather high concentration of ethanol, corresponding to 16.9 g/L, was achieved in a hollow-fiber membrane bioreactor (HFMBR) by operating in batch mode at pH 4.5 and using an open culture [7]. Cyclic pH shifts were also shown to drive a solventogenic metabolic shift in C. autoethanogenum and C. carboxidivorans [16].
The redox potential or oxidoreduction potential (ORP) is another parameter known to be a major trigger of the cellular stress response, involving e.g. the alteration of stress-responsive signaling pathways and the expression of several redox-regulatory proteins such as dehydrogenases [17]. An interesting example is the redox-sensing transcriptional repressor Rex, which is widely distributed in Gram-positive bacteria, and was found to regulate gene expression and modulate the fermentation products distribution in response to changes in the intracellular ORP and the ratio of NADH/NAD + in Clostridium acetobutylicum [18]. The extracellular ORP measures the activity of electrons, and thus, is indirectly sensitive to changes in the metabolic activity and intracellular redox state of the culture. However, as shown by De Graef et al. [19] in glucose-limited chemostat cultures of Escherichia coli, the opposite is also true, i.e. the extracellular ORP also exerts an effect on the intracellular NADH/NAD + ratio. Other studies also showed that controlling the extracellular ORP could drive significant changes in the metabolism of several cultures, e.g. promoting an earlier solventogenic phase in C. acetobutylicum, favoring higher succinic acid yield and productivity in Actinobacillus succinogenes, or boosting the productivity of 1, 3-propanediol by Klebsiella pneumoniae [20][21][22]. In gas fermentation though, research has been focused mostly on the effect of reducing agents addition on the production of ethanol, which showed variable results depending on the reducing agents used [23,24]. Whitham et al. [25] also showed higher ethanol production and increased expression levels in the AOR pathway upon O 2 exposure when C. ljungdahlii was grown mixotrophically. Nevertheless, a systematic investigation on the effect of the extracellular ORP on the solventogenic metabolic shift typically encountered in gas-fermenting Clostridium spp. is yet to be carried out.
The focus of this study is thus to evaluate if the modulation of the extracellular ORP exerts an effect on the solventogenic metabolism and product profile of syngas-converting microbial communities, dominated by Clostridium spp., and growing on syngas as the sole carbon and energy source. In this work, instead of favoring more reducing conditions, the extracellular ORP was controlled through the addition of an oxidizing agent (H 2 O 2 ) in order to trigger the response to oxidative stress and drive a metabolic shift towards solventogenesis in the culture. Additionally, alternative oxidizing agents were studied to determine whether the metabolic changes were driven by the extracellular ORP or the presence of reactive oxygen species (ROS). The intracellular levels of the redox cofactor NADH/NAD + were also monitored in order to (i) track changes in the internal redox state of the cells derived from the modulation of the extracellular ORP, (ii) study the dynamics of the intracellular free cofactor pool (NADH + NAD + ) as a function of the metabolic activity, and (iii) investigate other influencing factors driving changes in the ratio of NADH/ NAD + .
Using ORP control with dilute H 2 O 2 , instead of control of the O 2 saturation level in the broth, allowed for an accurate control and finetuning of the effect of ORP on the activity of the culture while minimizing the risk of generating an explosive atmosphere in the gas phase of the reactor. The results showed a 3.7-fold increase in the maximum specific productivity of ethanol when comparing the lowest (-210 mV) to the highest (160 mV) ORP level studied, which confirms the response to oxidative stress and presence of ROS, and illustrates the potential of ORP control strategies for boosting the production of alcohols in gas fermentation systems. The dynamic behavior of the NADH/NAD + redox cofactor pool was shown to be modulated by the metabolic activity rate, while the dynamics of the NADH/NAD + ratio were found to be tightly constrained by the thermodynamic restrictions of the AOR pathway and the degree of oxidative stress.

Biocatalyst and medium composition
The open culture used corresponded to a microbial community enriched on synthetic syngas (45 % H 2 , 20 % CO, 25 % CO 2 and 10 % N 2 ) in continuous operating mode for 78 days as described previously [26]. The initial composition of this microbial community was dominated by Clostridium spp., with variable presence of several putative species like C. ljungdahlii, C. luticellarii, C. drakei, C. propionicum and C. nitrophenolicum. In order to secure the same microbial composition upon inoculation in all experiments, the microbial community was reactivated directly into the bioreactor using frozen cultures previously stored at -80 • C in 3% w/v polyvinylpyrrolidone (PVP-15 K) [26].
The growth medium used in all experiments corresponded to a modified basal anaerobic (BA) medium, which consisted of several stock solutions. The composition of stock solutions was as follows: salt solution (NH 4 Cl, 100 g L − 1 ; NaCl, 10 g L − 1 ; MgCl 2 •6H 2 O, 10 g L − 1 ; CaCl 2 •2H 2 O, 5 g L − 1 ), sodium sulfate solution (Na 2 SO 4 , 100 g L − 1 ); vitamins solution according to Wolin  , yeast extract (yeast extract, 25 g L − 1 ) and reducing agent solution (Na 2 S•9H 2 O, 25 g L − 1 ). The medium was prepared in deionized water by adding 20 mL L − 1 of salts solution, 20 mL L − 1 of sodium sulfate solution, 10 mL L − 1 of vitamins solution, 10 mL L − 1 of trace metal solution, 10 mL L − 1 of dipotassium hydrogen phosphate solution, 20 mL L − 1 of yeast extract solution, and 4 mL L − 1 of reducing agent solution added in anaerobic conditions.

Experimental set up: bioreactor and ORP control
All experiments were performed in a 2 L stirred tank bioreactor (Biostat B, Sartorius Stedim Biotech GmbH, Germany) with a working volume of 2 L. The bioreactor was equipped with a mass flow controller (Bronkhorst, Netherlands) for adjusting the syngas inflow. The gas outflow was measured with a gas meter (μFlow, Bioprocess Control, Sweden). The pH was measured using a pH/ORP probe (Easyferm Plus VP 225, Hamilton, Switzerland) and was automatically regulated through the addition of 1 M KOH and 0.5 M HCl. The ORP reading was obtained through the same pH/ORP probe, and the ORP level was controlled, within a deadband of 7 mV, using the software MCFS/win 3.0. ORP readings (Ag/AgCl as reference electrode -E Ag/AgCl ) were corrected against the Standard Hydrogen Electrode (SHE) using a E • Ag/ AgCl of 207 mV and a temperature coefficient ). Unless stated otherwise, a dilute solution of H 2 O 2 was used for controlling the ORP level of the fermentation broth.
All experiments were carried out in batch mode in respect to the liquid phase and with continuous gas feeding using a syngas inflow of 10.3 mL min − 1 of a synthetic syngas mixture composed by of 45 % H 2 , 20 % CO, 25 % CO 2 and 10 % N 2 . The temperature was adjusted to 37 • C and the pH to 5 in all experiments. The experiments were started using the same frozen stock cultures with an inoculum size of 0.25 %v/v. Upon inoculation, the agitation was adjusted to 350 rpm and was increased to 400 rpm when the culture became active. Similarly, the ORP control was initiated upon activation of the culture. The objective of the experimental setup was to determine the effect of the ORP level on the product selectivity and productivity of the microbial community. Thus, the ORP was fixed at a constant level throughout the fermentation, and the different ORP levels studied corresponded to -210 mV, -70 mV, 70 mV, 100 mV, 130 mV and 160 mV. The ORP level of these experiments was controlled using a dilute solution of H 2 O 2 , where the concentration of H 2 O 2 was adjusted to a variable range of 0.0025− 0.4% w/w, increasing the H 2 O 2 concentration as the ORP set-point was increased in each experiment. To investigate whether the effect on the metabolism of the microbial community was triggered by the ORP of the fermentation broth or by the specific oxidizing agent used, an additional experiment was carried out, where the ORP level was adjusted to 130 mV using a 15 g L − 1 solution of FeCl 3 as an alternative oxidizing agent.
Monitoring of the performance of the batch fermentations was carried out by sampling from the gaseous and liquid phase of the bioreactor at least five times a week in all experiments. All samples were used for determination of the gas conversion efficiency, concentration of microbial biomass, organic acids and alcohols and the intracellular concentration of NADH and NAD + as described below.

Analytical techniques
The exhaust gas of the bioreactor was analyzed using a gas chromatograph (8610C, SRI Instruments, Germany) equipped with a thermal conductivity detector as described previously [6]. The content of organic acids and alcohols in the fermentation broth was analyzed through a High Performance Liquid Chromatograph (Shimadzu, USA) equipped with a refractive index detector and an Aminex HPX-87H column (Bio-Rad, Denmark). The concentration of microbial biomass was determined indirectly by measuring the optical density at a wavelength of 600 nm (OD 600 ) using a spectrophotometer (DR3900, Hach Lange). The OD 600 measurements were correlated to the Volatile Suspended Solids concentration (cells dry weight -CDW) in the fermentation broth, determined according to standard methods [28].

Redox cofactor NADH/NAD + determination
The intracellular concentration of the redox cofactors NADH and NAD + was determined based on a modified enzymatic cycling assay method described previously [29,30].
In order to secure a homogeneous extraction efficiency across samples, the volume of the samples for NADH and NAD + extraction was adjusted to an absolute content of microbial biomass (CDW) of around 0.3 mg, corresponding to a volume range of around 1.5− 2 mL. Samples for NADH and NAD + determination were rapidly collected separately and were centrifuged at 14,500 rpm for 30 s. After discarding the supernatant, 75 μL of 0.2 M HCl were added for NAD + extraction and 75 μL of 0.2 M NaOH for NADH extraction. All samples were re-suspended and incubated at 60 • C for 10 min in a dry block heater, after which the samples were placed on ice to cool them to 0− 4 • C. The cell extracts were neutralized by adding 75 μL of 0.15 M NaOH for NAD + and 75 μL of 0.15 M HCl for NADH and kept on ice for 10 min for cell debris precipitation. All samples were then centrifuged at 14,500 rpm for 5 min and the supernatant was transferred to new tubes for further analysis.
The concentration of NADH and NAD + of the samples was determined by measuring the reaction rate of the oxidation of ethanol into acetaldehyde catalyzed by alcohol dehydrogenase (ADH) indirectly through an indicator (thiazolyl blue tetrazolium bromide, MTT) using a microplate spectrophotometer (Multiskan sky, ThermoFisher Scientific, Denmark). The assay mixture contained (per sample) 43.3 μL of 1 M Bicine buffer (pH 8.0), 20 μL of absolute ethanol, 20 μL of 4.2 mM MTT, 40 μL of 16.6 mM phenazine ethosulfate (PES), 20 μL of 40 mM EDTA (pH 8.0) and 6.7 μL of 100 U ml − 1 ADH. The assay mixture was preincubated at 30 • C and the reaction was initiated by adding 50 μL of NADH or NAD + cell extract. The change in absorbance at 570 nm was measured for 12 min. The rate of reaction was correlated to the concentration of NADH and NAD + through a 5-point calibration in each run. The standards of NADH and NAD + were subjected to the same extraction protocol (described above) as the cell extracts. All cell extracts were analyzed using four replicates, corresponding to two biological replicates and two technical replicates. The concentration of NADH and NAD + in cell extracts was then normalized per g of microbial biomass (g CDW).

Computational methods
The maximum product yields (e-mol product/ e-mol syngas consumed) were determined according to eq. 1 using the time points with maximum concentration for each product, where n i is the number of moles of compound i and n e − i is the number of e-mol per mol of compound i. A table describing the half-reactions and e-mol-to-mol coefficients used for products and substrates is provided in the supplementary material (table S1 -Supplementary Information file) The specific productivity for each product was calculated for all time points according to eq. 2, where t j represents the time at measurement number j, C i, tj corresponds to the concentration of product i at time point t j , X t j corresponds to the microbial biomass concentration at time point t j and n is the number of measurements. Representative maximum average specific productivities (q p, imax ) were obtained by averaging the maximum q p, i over three consecutive measurements. The calculation of specific activity (or specific substrate uptake rate, q s ) lumped for H 2 and CO was analogous to the q p, i (eq. 2), substituting C i, tj by the moles of H 2 and CO consumed per liter of reactor.
In order to capture the representative volumetric productivities and make them independent of the residual production during the late stationary phase, all volumetric productivities (Q P80 % ) were calculated using the time points when the production started and when 80 % of the maximum concentration for each product was reached in each batch experiment.
Multiple linear regression and correlation tests run on the redox cofactor data were performed using R (version 4.0.2) and ggpubr package (version 0.4.0) [31].
The thermodynamics of the aldehyde:ferredoxin oxidoreductase (AOR) pathway was evaluated through the Gibbs free energy change (Δ r G´) and equilibrium constants (K eq ) based on eq. 3. Standard Gibbs free energies of formation (Δ r G • ) were extracted from Alberty [32] and Flamholz et al. [33]. The Δ r G´was corrected for pH accounting for the concentration of protons and undissociated acetic acid species based on the mass balance of acetate species (Acetate total = Acetate + Acetic acid) and the acetate dissociation constant (k acetate ), resulting in eq. 4, where Q i/j stands for the ratio between compound i and j, and the quotient derives from the correction for pH and the dissociation of acetic acid. See Supplementary Information file for a detailed explanation on the derivation of eq. 3 and 4. Plots of Δ r G´and K eq as a function of ethanol-to-acetate and redox cofactor ratios were made based on eq. 4.

Effect of ORP control on microbial activity
The effect of the ORP on the syngas-fermenting activity of the enriched microbial community was evaluated by adjusting the extracellular ORP at fixed levels, corresponding to -210 mV, -70 mV, 70 mV, 100 mV, 130 mV and 160 mV. The results showed that the modulation of the ORP causes severe alterations on the metabolic activity of the microbial community, with the latter being ultimately translated in significant changes in the solventogenic and chain elongation activity of the culture. Fig. 1 shows a schematic of the pattern of the metabolic activity of a syngas-converting open culture and Fig. 2 shows the fermentation profile for biomass and products in all ORP levels studied. Figs. 3 and 4 present the maximum product yields and averaged maximum specific productivities, respectively, found at each ORP level. The maximum specific uptake rates and biomass concentrations, the volumetric productivities (Q P80 % ) and the final product yields are given in Table 1.
The enriched microbial community exhibited the same pattern of activity in all experiments performed, where acetate was the dominant initial product during the exponential growth phase, and ethanol, butyrate, butanol and caproate (and hexanol under ORP level 160 mV) were synthesized sequentially later on as a result of the reduction of acids into alcohols and a chain elongation reaction (Figs. 1 and 2). This activity pattern is actually common for most syngas-fed acetogenic cultures with the ability to produce solvents and medium chain fatty acids (MCFAs) regardless the type of culture, i.e. whether monocultures, co-cultures or open cultures are used [34][35][36]. However, in this case, the extent and rate of each of these metabolic activities was strongly affected by the modulation of the ORP, especially when transitioning from reducing conditions (ORP level < 0 mV) to oxidizing conditions (ORP level > 0 mV). As a matter of fact, the productivities and maximum product yields obtained under negative ORP levels (-210 mV and -70 mV) (Fig. 3) were consistent with those previously obtained when using the same microbial community under uncontrolled ORP [26], which indicated that the lowest ORP levels studied presented only minor or no effects on the metabolic activity of the culture (table S2-Supplementary  Information file). Additionally, the fact that significant changes in the metabolic activity of the culture were only observed under oxidizing conditions was an early indication that the metabolic changes observed were probably driven by the presence of ROS and the response to oxidative stress rather than by the ORP per se (see Section 3.2 for further details).
The modulation of the extracellular ORP presented a strong effect on the specific productivity and yield of ethanol. The fermentations carried out at negative ORP levels presented a similar performance in terms of concentrations, maximum yields and maximum specific productivities for ethanol, corresponding approximately to 50 mM, 34 % e-mol•e- Figs. 3 and 4). Nevertheless, increasing the ORP to oxidizing conditions triggered a significant increase in the specific ethanol productivity, which underwent a 3.7-fold increase when comparing the lowest ORP level (-210 mV) with the highest ORP level (160 mV) studied, and resulted in a maximum specific ethanol productivity of 2.32 ± 0.19 mmol•gCDW − 1 h − 1 (2.57 ± 0.22 g•gCDW − 1 •d − 1 ) (Fig. 4). Similarly, the volumetric productivity of ethanol at an ORP of 160 mV was also 4-fold higher than that of ORP -210 mV, increasing from 1.51 mmol•L -1 •d -1 to 6.02 mmol•L -1 •d -1 ( Table 1). The ethanol yields at the end of the fermentation followed a trend similar to the volumetric productivity, reaching a maximum of 44.6 % e-mol•e-mol − 1 at an ORP of 130 mV (Table 1). In terms of yields The semitransparent green data points correspond to measurements taken after a major disturbance during the fermentation, as the gas inflow stopped for 3 days, and consequently, these measurements were not taken into account in subsequent analyses. though, the fermentation at an ORP of 160 mV was an exception as its maximum and final ethanol yield was considerably lower than at lower ORP levels due to the exceptionally high chain elongation rate found in this experiment, resulting in higher yields for butyrate, butanol, caproate and hexanol.
As opposed to the trend observed on ethanol production, increasing the ORP to oxidizing conditions was found to be detrimental for the chain elongation (except for 160 mVfurther discussion is given below). As shown in Fig. 4, the averaged maximum specific productivity of butyrate dropped significantly upon increasing the ORP level, decreasing from a maximum of 0.56 ± 0.04 mmol•gCDW − 1 h − 1 at -210 mV to a minimum of 0.09 ± 0.00 mmol•gCDW − 1 h − 1 at 100 mV. Consequently, the specific productivity of butanol was also negatively affected when increasing the ORP level due to the lower availability of butyrate in the fermentation broth and despite the clear boost on the solventogenic activity of the culture (Figs. 2 and 4). The production of caproate followed the same trend, gradually decreasing its final yield as the ORP level was increased. However, the results obtained when operating at 160 mV stand out from the rest, as at these conditions, the production of butyrate, butanol, caproate and hexanol increased significantly due to the higher rate and extent of the chain elongation reaction, reaching a maximum yield of caproate of 15.5 % e-mol•emol − 1 (Table 1).
Increasing the ORP level was also found to have an effect on the maximum biomass concentration, the maximum specific substrate uptake rate and the overall product recovery. As shown in Table 1, the maximum biomass concentration remained between 144 and 169 mgCDW•L − 1 at reducing and mild-oxidizing conditions (-210 mV to 70 mV), whereas further increasing the ORP promoted a significant increase in the maximum biomass concentration, reaching a maximum of 319 ± 2 mgCDW•L − 1 at 160 mV. At the same time, the product recovery remained high (87.0-92.5 % e-mol•e-mol − 1 ) at all ORP levels, with the exception of the fermentation at 160 mV where the product recovery suddenly dropped to 74.5 ± 1.8 % (Table 1). These observations, together with the sudden increase in specific substrate uptake rate observed at an ORP of 160 mV ( Table 1), suggest that the continuous addition of H 2 O 2 led to the spontaneous formation of O 2 in the fermentation broth up to concentrations high enough to allow its consumption by aerobic carboxydotrophs [37], which would explain the simultaneous drop and increase in product recovery and biomass titer, respectively.
The latter could in fact provide a possible explanation for the atypical chain elongation activity found when fixing the ORP at 160 mV. Although acetogenic bacteria are generally referred to as strict anaerobes, it is known that many acetogens possess mechanisms to cope with ROS under oxidative stress, probably derived from the fluctuating ORP conditions found in many of their natural habitats [38]. In fact, several acetogens typically used in gas fermentation systems are well known for their aerotolerance, such as Acetobacterium woodii, Moorella thermoacetica, Sporomusa termitida, C. autoethanogenum and C. ljungdahlii [25,38,39]. Nevertheless, chain-elongating bacteria, with Clostridium kluyveri as model organism, are generally more sensitive to stressful conditions than acetogens, as the reported inhibition of C. kluyveri at low pH or high acid concentration imply [40,41]. Additionally, the authors were unable to find any reference to tolerance to oxidative stress in C. kluyveri in the scientific literature. This species is generally referred to as a strict anaerobe, which could explain the gradual increase in inhibition of the chain elongation found in this study when transitioning to oxic conditions. Thus, a possible explanation for the higher chain-elongating activity found at an ORP of 160 mV could be that the inhibition of the chain elongation, possibly caused by O 2 , was alleviated due to the continuous removal of O 2 by aerobic carboxydotrophs found at these conditions (see Section 3.2 for further discussion).
The specific productivity of acetate and ethanol was affected by the increase in extracellular ORP. The maximum specific acetate productivity was found to decrease by nearly 1.6-fold when comparing the ORP levels -210 mV and 160 mV, while there was a 3.7-fold increase in the maximum specific productivity of ethanol between the same conditions. Nevertheless, the production of acetate and ethanol was completely decoupled (as it took place under different phases of the fermentation, Fig. 2), which indicated that the effect of the ORP on the specific productivity of ethanol was strictly related to the AOR pathway (reduction of acetate into ethanol). The latter is in line with the higher AOR activity and higher AOR (CLJU_c24130) expression found when C. ljungdahlii was exposed to O 2 under mixotrophic conditions, where the need for ROS detoxification triggered a higher AOR activity [25]. Therefore, the higher ethanologenic activity found here is most likely also related to the role played by the AOR pathway in the detoxification of ROS.
Overall, controlling the ORP was found to be a successful operating strategy with high potential to boost the specific solventogenic activity of the culture while improving the selectivity towards ethanol. The inhibition of the chain elongation favored a higher selectivity towards ethanol, as shown by the increase in ethanol yield by 9.4 % e-mol•e-  Table 1). The acetogen Clostridium glycolicum RD-1 had been previously reported to shift its metabolism towards more reduced products as a response to oxidative stress when growing on glucose, which is consistent with the results obtained here [42]. Similarly, Whitham et al. [25] reported an increase in ethanol concentration from 0.7 g L -1 to around 1.5 g L -1 when C. ljungdahlii was exposed to 8% of O 2 under mixotrophic conditions (using H 2 (10 %), CO (20 %) and 5 g L -1 of fructose as substrates in a modified reinforced clostridial medium). In the present study, it was shown that the metabolic shift towards solvents is maintained when the acetogenic culture is grown strictly autotrophically on syngas. However, perhaps more importantly, the effect of the modulation of the ORP (by adding dilute H 2 O 2 ) was shown to trigger a remarkable boost of the specific productivity of ethanol by several-fold, while minimizing the risk of generating an explosive atmosphere (in contrast to using O 2 as oxidizing agent). Additionally, the fact that many commonly used acetogens can cope with oxidative stress suggests that this ORP control strategy could be widely applicable in gas fermentation, not only when using open cultures, but also using defined cultures.
The specific productivities reported here range from 0.63 ± 0.04 mmol•gCDW − 1 h − 1 to 2.32 ± 0.19 mmol•gCDW − 1 h − 1 (0.70-2.57 g gCDW − 1 •d − 1 ) depending on the ORP level. Similar specific productivities of ethanol have been previously reported in other studies using C. ljungdahlii (0.9-1.0 g gCDW − 1 •d − 1 ), C. ragsdalei (0.8-1.15 g gCDW − 1 •d − 1 ), C. carboxidivorans (1.65− 2.31 g gCDW − 1 •d − 1 ) and C. autoethanogenum (1.3-4.2 g gCDW − 1 •d − 1 ), and these are most likely very dependent on the specific reactor configurations and operating conditions used in each case [43][44][45][46][47]. Nevertheless, the authors consider that it is the relative increase in specific ethanol productivity obtained here, rather than its absolute value, what illustrates the potential of ORP control strategies for boosting the solventogenic activity of acetogenic cultures. Future work should assess whether the relative increase in specific productivity is maintained when using a more productive set of reactor systems and operating conditions.

ORP vs. Reactive oxygen species
The fact that the effect on the solventogenic activity of the culture was observed primarily when the ORP was set to oxidizing conditions motivated an additional fermentation using another oxidizing agent to determine whether this effect was triggered by the presence of ROS or by the extracellular ORP level per se. In this case, the ORP of the fermentation broth was controlled at 130 mV by addition of FeCl 3 , which is a mild oxidizing agent (E 0 red = 0.77 V, Fe 3+ + H + + e − ⟶ Fe 2+ ) compared to H 2 O 2 (E 0 red = 1.76 V, H 2 O 2 + 2 H + + 2 e − ⟶ 2 H 2 O) and does not generate ROS upon reduction.
As expected, the control of the ORP at 130 mV by adding FeCl 3 clearly exerted an effect on the metabolic activity of the culture, but it differed significantly from that of adding dilute H 2 O 2 . The results of the fermentation using FeCl 3 for ORP control are shown in Fig. 5, where it can be seen that the solventogenic activity was negatively affected by the change of oxidizing agent. While using H 2 O 2 resulted in a maximum ethanol yield of 44.6 ± 0.1 % e-mol•e-mol − 1 , using FeCl 3 for ORP control yielded a maximum of 5.0 ± 0.0 % e-mol•e-mol − 1 . The volumetric productivity of ethanol also exhibited a significant drop to values lower than in any other ORP level studied using H 2 O 2 (Fig. 5). This indicated that the effects of the ORP on the solventogenic activity Table 1 Volumetric productivity, final product yield, averaged maximum specific activity for H 2 and CO and maximum biomass concentration for each of the ORP levels studied. The averaged maximum specific activity corresponds to the average of the specific uptake rate for the sum of H 2 and CO over three consecutive measurements. The product recovery includes the yields of other minor products such as propionate, propanol, iso-butyrate, valerate and iso-valerate. ORP  observed in Section 3.1 were specifically driven by the presence of ROS in the fermentation broth and the potential involvement of the AOR pathway in the detoxification of ROS. As opposed to the effect on the solventogenic activity, the change of oxidizing agent seemed to alleviate partially the inhibition of the chain elongation found when using dilute H 2 O 2 , as both yield and volumetric productivity of butyrate were around four times higher when using FeCl 3 (Fig. 5). This further supported the hypothesis that the chain elongation reaction was inhibited by the presence of ROS, as discussed above in Section 3.1. Overall, these results clearly showed that the effects of controlling the ORP on the metabolic activity of the culture were determined by the specific oxidizing agent used, in this case dilute H 2 O 2 , rather than by the ORP per se.
The ORP is generally regarded as an important parameter affecting the cellular redox metabolism of microorganisms and several ORP control strategies have been devised in the past based on either metabolic engineering or bioprocess engineering [48]. Studies on bioprocess-engineering-based ORP control strategies tend to focus on the specific ORP values and their effects on the metabolism towards end products, while the importance of the reagents used for ORP control is typically neglected. However, as shown here, changing the oxidizing agent used for ORP control caused drastic differences in the metabolic response of the culture despite the same ORP level was used, as there was no need for ROS detoxification when using FeCl 3 . Wang et al. [21] also reported different acetone-butanol-ethanol (ABE) fermentation profiles when fixing the ORP at -350 mV using different reducing agents (Na 2 S, dithiothreitol and cysteine), which could be explained by e.g. different membrane permeability or toxicity for/of each reducing agent. This shows that the ORP is a mere indicator of the extracellular redox balance instead the ultimate trigger of the metabolic response to stress, and that the specific response mechanism upon exposure to oxidizing/reducing agents should be considered for each case. Therefore, generalizing about the effects of the extracellular ORP on microbial metabolism should be done with caution.

Role of redox cofactor NADH/NAD + under oxidative stress
The previous sections showed that the changes in solventogenic and chain elongation activity observed when increasing the ORP were in fact driven by the ROS detoxification mechanism. The potential involvement of the AOR pathway in the latter was also discussed. Hillmann et al. [49] studied the response to oxidative stress in C. acetobutylicum and showed that several ROS detoxification enzymatic activities, such as oxidase, peroxide reductase and alkylhydroperoxide reductase, used preferentially NADH rather than NADPH as redox cofactor when using O 2 , H 2 O 2 and t-butylhydroperoxide as substrates. Similarly, another study also found that the activity of reverse rubrerythrin, typically mediated by NADH and NADH:rubredoxin oxidoreductase, plays a central role in the reduction of O 2 and H 2 O 2 to H 2 O in C. acetobutylicum [50]. Additionally, the redox-sensitive (NADH/NAD + -sensitive) Rex repressor was observed to control product profile in C. acetobutylicum and was predicted to regulate transcription of the reductive acetyl-CoA pathway in C. ljungdahlii and C. carboxidivorans [18]. The latter indicate that the redox cofactor NADH/NAD + plays a fundamental role in both the detoxification of ROS and potentially the product profile during gas fermentation as well. Thus, in this study, the intracellular concentration of NADH/NAD + was monitored during the fermentations at all ORP levels studied in order to (i) determine the effect of the ORP and other potentially influencing factors on the dynamics of this redox cofactor, and (ii) investigate whether the changes observed in the metabolic activity of the culture could be explained by the dynamics of the redox cofactor pool and ratio.
The redox cofactor pool (NADH + NAD + ) was found to be highly dynamic during the course of the fermentations, with concentrations typically ranging from a maximum of around 9-14 μmol•gCDW − 1 during the exponential growth phase to a minimum of 1 μmol•gCDW − 1 at the stationary phase ( fig. S1-S6, Supplementary Information file).
Analyzing the evolution of the cofactor pool across all datasets revealed a common pattern in all of them, where the concentration of the redox cofactor was rapidly adjusted to the metabolic activity of the culture. This is shown in Fig. 6A for the fermentation at ORP level -210 mV, where it can be seen that the cofactor concentration followed closely the changes in specific substrate uptake rate of the culture (plots for other datasets can be found in fig. S1-S6, Supplementary Information file). The correlation between the redox cofactor pool and the specific substrate uptake rate of the culture was confirmed to be significant for each ORP level and across all ORP levels through the Pearson "R" correlation coefficient (Fig. 6C). This indicated that the intracellular redox cofactor pool was rapidly modulated to (i) ensure a sufficiently high turnover of redox mediators under high metabolic activity and (ii) minimize the energetic cost associated to maintenance of the redox cofactor pool when the metabolic activity rates are low.
The ratio of NADH/NAD + was also found to be highly variable during the fermentations, generally presenting values between 0.01 mol mol − 1 and 0.3 mol mol − 1 ( fig. S1-S6, Supplementary Information file). However, the NADH/NAD + ratio presented a generalized increasing trend along the fermentation across all ORP levels studied, which suggested that the increasing trend was, in principle, independent of the ORP and that there were other factors playing a significant role in determining the NADH/NAD + ratio. Analyzing this common pattern as a function of other fermentation parameters showed that the NADH/ NAD + ratio increased proportionally to the ratio of alcohols-to-acids in solution (Fig. 6B, 6D and fig. S1-S6, Supplementary Information file). The significance of the correlation between these two variables was also confirmed based on the Pearson "R" correlation coefficient across all ORP levels and for each ORP level independently, with the exception of ORP 160 mV (Fig. 6D). In the latter case, the non-significant correlation was most likely related to the rapid changes in product distribution caused by the unusually high chain elongation rate found in this fermentation, which interfered in the linearity of the correlation. In any case, the high significance (p-value, 2.2e-16) and the high correlation coefficient (R, 0.89) of the overall dataset clearly shows that the NADH/ NAD + ratio was highly dependent on the alcohols-to-acids ratio (Fig. 6D).
The following sections elaborate on two key aspects of the dynamics of the NADH/NAD + ratio along the fermentation: (i) on the role of thermodynamics in the generalized increasing trend of the NADH/NAD + ratio found in all experiments (see Section 3.3.1), and (ii) on the effect of the ORP on the dynamics of the NADH/NAD + ratio (see Section 3.3.2).

Thermodynamics as a major driver of the dynamics of NADH/NAD + ratio
As mentioned above, the intracellular ORP (NADH/NAD + ratio), or rather the response to oxidative stress, may be regulated by several transcriptional repressors such as PerR or Rex [18,49]. However, the fact that the increasing trend in NADH/NAD + ratio during the course of the fermentation was observed at all ORP levels, along with the high dependency found between the ratios of NADH/NAD + and alcohols-to-acids, suggested that there were additional processes driving major changes in the profile of NADH/NAD + .
As discussed in Section 3.1, the reduction of acetate to ethanol through the AOR pathway was clearly the dominant pathway during the fermentation, and consequently, was responsible for the bulk of the change in alcohols-to-acids ratio. Nevertheless, this pathway (eq. 3) seems to take place rather close to thermodynamic equilibrium, as shown by its positive standard Gibbs free energy change at pH 7 (5.0 ± 11.9 kJ•mol − 1 , calculated using component contribution [33]), and hence, may be very sensitive to changes in the intracellular concentration of redox cofactors, metabolites, pH, ionic strength and other factors. This indicated that the dynamic behavior of the NADH/NAD + ratio was probably driven by thermodynamics rather than by transcriptional regulation, where the NADH/NAD + ratio played a role in maintaining a constant thermodynamic driving force (Δ r G´) of the forward reaction as A. Grimalt-Alemany et al. the ratio of alcohols-to-acids in solution increased. Based on eq. 3 and eq. 4, it can be seen that there is a linear relationship between the ethanol-to-acetate ratio (Q Ethanol/Acetate ) and the NADH/NAD + ratio (Q NADH/NAD + ). Thus, in the region near to thermodynamic equilibrium, any change in Q Ethanol/Acetate must be compensated by a proportional change in Q NADH/NAD + in order to stay within the thermodynamically feasible region, where the constant of proportionality (or the rate of change of the NADH/NAD + ratio as a function of the ethanol-to-acetate ratio -ΔQ NADH/NAD + /ΔQ Ethanol/Acetate ) is dictated by Δ r G´. This could explain the nearly linear increasing trend found for the NADH/NAD + ratio as the alcohols-to-acids ratio increased and is illustrated in Fig. 7B, where the ΔQ NADH/NAD + /ΔQ Ethanol/Acetate is plotted for different Δ r Gv alues together with the linear regression model of the combined dataset of NADH/NAD + ratios found experimentally. In this case, the Fd red /Fd ox ratio was kept at a constant value of 15, corresponding to 93.75 % of Fd red , as values above 90 % are generally found in living cells [51]. It should be noted that changing the Fd red /Fd ox ratio would only magnify the effect of the NADH/NAD + ratio on the thermodynamic feasibility of the reaction by shifting the K eq without affecting the linear dependency between the NADH/NAD + and ethanol-to-acetate ratios (Fig. 7C).
Accounting for a dynamic Fd red /Fd ox ratio though, the relation between NADH/NAD + and ethanol-to-acetate ratios would deviate from linearity depending on the ΔQ Fd red /Fdox /ΔQ Ethanol/Acetate (Fig. 7D). However, the fact that the increasing trend of NADH/NAD + ratio found experimentally was nearly linear suggested that the Fd red /Fd ox ratio was relatively constant during the fermentation. Additionally, the intrinsic thermodynamic limitation of the AOR pathway imposes a restriction on the minimum Fd red /Fd ox ratio required for the feasibility of the forward reaction, corresponding to ≈ 8 (88.89 % of Fd red ) based on Fig. 7C. This thermodynamic restriction prevents significant fluctuations in the Fd red /Fd ox ratio and explains the dynamic nature of the NADH/NAD + ratio, as the increase in NADH/NAD + ratio was in fact necessary to maintain the driving force of this pathway in the forward direction.

Effect of ORP on the NADH/NAD + ratio
Due to (i) the fact that the production of ethanol through the AOR pathway was the dominant metabolic activity and (ii) the dynamic nature of the redox cofactor data, the potential effects of the extracellular ORP on the NADH/NAD + ratio were evaluated based on the ΔQ NADH/NAD + /ΔQ Ethanol/Acetate using multiple linear regression. Although   Fig. 6. A Evolution of the specific substrate uptake rate and redox cofactor pool (NADH + NAD + ) over time for ORP -210 mV, showing their parallel dynamic behavior during the fermentation. The black line corresponds to a smoothed interpolation of the redox cofactor pool. B Plot of the alcohols-to-acids and NADH/NAD + ratios over time for ORP -70 mV showing a proportional increase between these two variables. The black line corresponds to a smoothed interpolation of the NADH/ NAD + ratios. C Correlation between the redox cofactor pool (NADH + NAD + ) and the specific substrate uptake rate for all ORP levels, and Pearson "R" correlation coefficient and p-values for each ORP level and for all ORP levels together. The lines on the horizontal and vertical axes show the density of data points for each ORP level. D Correlation between the NADH/NAD + ratio and the alcohols-to-acids ratio for all ORP levels, and Pearson "R" correlation coefficient and p-values for each ORP level and for all ORP levels together. The lines on the horizontal and vertical axes show the density of data points for each ORP level. The positive error bars shown in the four plots represent the standard deviation of the redox cofactor data.
it might not necessarily be the case in all datasets, the ΔQ NADH/NAD + /Δ Q Ethanol/Acetate was determined assuming linearity to facilitate the analysis, as in any case, the linear fit would represent the average changes in ΔQ NADH/NAD + /ΔQ Ethanol/Acetate as a function of the ORP level. In order to minimize the interference of other reactions (acetate production and chain elongation) taking place during the fermentations with the reduction of acetate to ethanol through AOR and the NADH/NAD + ratio, a subset of data points corresponding to the time frame used for calculating the volumetric productivity of ethanol (Q P80 % ) was used for linear regression ( fig. S7-S12, Supplementary Information file). The results showed that the regression models of the NADH/NAD + ratio as function of the ethanol-to-acetate ratio were significant across all ORP levels (Fig. 8A), and that the ΔQ NADH/NAD + /ΔQ Ethanol/Acetate and the ORP level were positively correlated (Pearson "R" coefficient of 0.99 and p-value of 0.00017) (Fig. 8B). This indicated that the ORP level indeed had an impact on the NADH/NAD + ratio, where the ΔQ NADH/NAD + /Δ Q Ethanol/Acetate increased by threefold with the increase of ORP following a linear trend (Fig. 8B). Next, to determine whether the increase in the rate of change of the NADH/NAD + ratio was dependent on the productivity of ethanol and confirm the role of the AOR pathway in the detoxification of ROS, the correlation between the ethanol productivity and the rate of change of NADH/NAD + ratio over time (ΔQ NADH/NAD + /Δt) was investigated. The results of the multiple linear regression indicated that the ΔQ NADH/NAD + /Δt and the volumetric productivity of ethanol (Q P80 % ) were highly correlated, with the Pearson "R" coefficient and the p-value being 0.99 and 0.00017, respectively ( Fig. 8C and 8D). Therefore, the fact that the rate of change of NADH/ NAD + ratio is simultaneously correlated with the ORP level and the volumetric productivity of ethanol clearly shows that the activity of the AOR pathway played an important role in the detoxification of the increasing dosage of H 2 O 2 as the ORP level was increased. Based on a thermodynamic interpretation, the increase in ΔQ NADH/NAD + /Δ Q Ethanol/Acetate would also be consistent with the higher thermodynamic driving force needed to allow for a higher AOR activity rate, as shown in Fig. 7B.
Analyzing the dynamics of the NADH/NAD + cofactor showed that redox cofactor pool is closely linked to the metabolic activity of the culture, and that the redox cofactor ratio is strongly influenced by the thermodynamics of the AOR pathway and the extracellular ORP. The correlation between the NADH/NAD + ratio and the productivity of ethanol and the extracellular ORP indicated that the AOR pathway had an active role in the detoxification of ROS. This is in agreement with Whitham et al. [25], who found a significantly higher expression of the RNF complex, AOR and rubrerythrin when C. ljungdahlii was exposed to O 2 under mixotrophic conditions. The increase in NADH/NAD + ratio with the ethanol-to-acetate ratio was consistent with the intrinsic thermodynamic limitation of the AOR pathway, as the NADH/NAD + ratio must increase to maintain the thermodynamic drive in the forward Fig. 7. A Gibbs free energy change (Δ r G´) for the AOR pathway (eq. 3), calculated using eq. 4, as a function of the ethanol-to-acetate ratio and the NADH/NAD + ratio. The Fd red /Fd ox ratio was fixed at 15 in the calculations and the pH was 7. B NADH/NAD + ratio as a function of the ethanol-to-acetate ratio for different Δ r G´values and experimental linear regression model of the combined dataset using all ORP levels (black dashed line). C NADH/NAD + ratio as a function of the ethanol-toacetate ratio for K eq at different Fd red /Fd ox ratios, and experimental linear regression model using all ORP levels (black dashed line). D NADH/NAD + ratio as a function of the ethanol-to-acetate ratio and K eq assuming dynamic Fd red /Fd ox ratios with variable rates of change in ΔQ Fdred /Fdox /ΔQ Ethanol/Acetate . An initial Fd red /Fd ox ratio of 8 was used in these calculations. A rate of change of 0 in ΔQ Fdred/Fdox /ΔQ Ethanol/Acetate corresponds to a constant Fd ox /Fd red ratio.
direction. Richter et al. [9] reported an increased NADH/NAD + ratio when comparing the acidogenic to solventogenic phase of C. ljungdahlii, which is consistent with the results obtained here. On the other hand, the fact that the higher extracellular ORP also contributed to an increase in NADH/NAD + ratio was rather surprising. Generally, the intracellular redox state of the cell would be expected to become more oxidized upon exposure to O 2 or ROS, at least due to the oxidation of NADH during ROS detoxification. The latter is also in line with the findings of Zhang et al. [18] when culturing C. acetobutylicum, where it was found that the NADH/NAD + ratio decreased after exposing the culture to H 2 O 2 . It should be noted though that the NADH/NAD + ratio values reported in their study, even when supplying H 2 O 2 , were significantly higher than those reported here. C. acetobutylicum and C. ljungdahlii have been previously shown to present significant differences in their profile of intracellular redox cofactor ratios, as e.g. C. acetobutylicum was found to have a more oxidized redox state during solventogenesis than during acidogenesis, while C. ljungdahlii presented a considerably more reduced redox state during solventogenesis [9,52]. This could partially explain the unexpected trend in NADH/NAD + ratios obtained here. In this case, the increase in ΔQ NADH/NAD + /ΔQ Ethanol/Acetate in response to oxidative stress was found to be consistent with the higher AOR activity and ethanol productivity, and the higher thermodynamic drive needed for boosting this metabolic activity. However, the specific mechanism triggering the increase in ΔQ NADH/NAD + /ΔQ Ethanol/Acetate in response to oxidative stress could not be explained and is yet to be elucidated. The regulation of the response to oxidative stress through the redox-sensitive transcriptional repressors Rex or PerR could possibly contribute to explain the increase in NADH/NAD + ratio when increasing the ORP. Further research on the interplay of different redox cofactors and key metabolic activities under oxidative stress, through e.g. dynamic metabolic flux analysis, could also shed some light on the trigger of this unexpected behavior and provide a modeling framework for further developing ORP control strategies in gas fermentation.
Overall, this study provided insights on the dynamics of the NADH/ NAD + redox cofactor under oxidative stress and the thermodynamics of the AOR pathway. The total redox cofactor pool (NADH + NAD + ) was found to be rapidly modulated by the cell to ensure high enough turnover of redox mediators to meet the demand as the metabolic activity rate changed during the fermentation. It was also shown that the highly dynamic nature of the NADH/NAD + ratio was in fact necessary in order to compensate for the restrictions imposed by the compromised thermodynamic feasibility of the AOR pathway and the high Fd red /Fd ox required to allow for this metabolic activity. Lastly, the control of the ORP using dilute H 2 O 2 was shown to have an effect on the NADH/NAD + ratio and the productivity of ethanol, which further confirmed that the AOR pathway was involved in the detoxification of ROS upon oxidative stress.

Conclusions
This study evaluated the use of ORP control strategies through the addition of dilute H 2 O 2 for boosting the specific productivity of ethanol in an open-culture-based gas fermentation system. The main drivers of the metabolic response of the culture upon oxidative stress were also Fig. 8. A Linear regression models of the ΔQ NADH/NAD + /ΔQ Ethanol/Acetate for each ORP level studied. B Correlation test between the ΔQ NADH/NAD + /ΔQ Ethanol/Acetate and the extracellular ORP level based on the Pearson "R" correlation coefficient, showing that the ORP exerted an effect of the NADH/NAD + ratio of the culture. C Linear regression models of the ΔQ NADH/NAD + /Δt for each ORP level studied. D Correlation test between the ΔQ NADH/NAD + /Δt and the volumetric productivity of ethanol (Q P80 % ) using the Pearson "R" correlation coefficient.
investigated through the use of alternative oxidizing agents and by studying the dynamics of the redox cofactor NADH/NAD + . The results showed that adding dilute H 2 O 2 to trigger oxidative stress and using the ORP for fine-tuning the degree of stress exerted on the culture is a promising operational strategy for boosting considerably the specific productivity of ethanol and the selectivity of the process. The use of dilute H 2 O 2 may also serve as an alternative to the micro-aerobic fermentation approach in general, and particularly in gas fermentation systems, as this approach minimizes the risk of generating an explosive atmosphere within the reactor. This operational strategy was demonstrated using an acetogenic open culture. However, based on the widespread aerotolerance among acetogenic bacteria, this strategy is expected to be widely applicable within gas fermentation systems targeting solvent production and using e.g. defined mono-or co-cultures. Analyzing the redox cofactor dynamics during the fermentation revealed that the redox cofactor pool is modulated by the metabolic activity rate of the culture, while the dynamics of the NADH/NAD + ratio is tightly constrained by the thermodynamic restrictions of the ethanolproducing pathway and the degree of oxidative stress.

Data availability
Part of the data used in this manuscript is provided in the Supplementary Information file. The complete dataset is available upon request.

Funding
This work was financially supported by the Technical University of Denmark and Innovation Fund Denmark in the frame of SYNFERON project.