A novel experimental method to determine substrate uptake kinetics of gaseous substrates applied to the carbon monoxide‐fermenting Clostridium autoethanogenum

Syngas fermentation has gained momentum over the last decades. The cost‐efficient design of industrial‐scale bioprocesses is highly dependent on quantitative microbial growth data. Kinetic and stoichiometric models for syngas‐converting microbes exist, but accurate experimental validation of the derived parameters is lacking. Here, we describe a novel experimental approach for measuring substrate uptake kinetics of gas‐fermenting microbes using the model microorganism Clostridium autoethanogenum. One‐hour disturbances of a steady‐state chemostat bioreactor with increased CO partial pressures (up to 1.2 bar) allowed for measurement of biomass‐specific CO uptake‐ and CO2 production rates ( q CO ${q}_{{CO}}$ , q CO 2 ${q}_{{{CO}}_{2}}$ ) using off‐gas analysis. At a pCO of 1.2 bar, a q CO ${q}_{{CO}}$ of −119 ± 1 mmol g−1X h−1 was measured. This value is 1.8–3.5‐fold higher than previously reported experimental and kinetic modeling results for syngas fermenters. Analysis of the catabolic flux distribution reveals a metabolic shift towards ethanol production at the expense of acetate at pCO  ≥ $\ge $  0.6 atm, likely to be mediated by acetate availability and cellular redox state. We characterized this metabolic shift as acetogenic overflow metabolism. These results provide key mechanistic understanding of the factors steering the product spectrum of CO fermentation in C. autoethanogenum and emphasize the importance of dedicated experimental validation of kinetic parameters.

processing (Molitor et al., 2016).This gas is currently detoxified by combustion, leading to approximately 3.9 billion tonnes of CO 2 emissions per year (IEA, 2020;World Steel Organization, 2022).Thus, for the transition to a bio-based society, a robust technology for the conversion of CO into chemical building blocks is needed.

Purely chemical gas conversion processes such as
Fischer-Tropsch synthesis are sensitive to gas composition and impurities (Dry, 2002).Therefore, novel technologies for syngas and flue gas conversion must be developed.Anaerobic acetogenic microorganisms are known to convert syngas constituents into products such as acetate and ethanol (Abrini et al., 1994;Arantes et al., 2020;Lee et al., 2019;Levy et al., 1981).These organisms are less sensitive to changes in gas composition than metal catalysts and can withstand the presence of impurities like NH 3 and H 2 S (Ahmed & Lewis, 2007;Klask et al., 2020;Rückel et al., 2021).Additionally, processes with microbial catalysts are less energy-demanding because they can be operated at mild temperatures and pressures.
It is crucial to have a thorough understanding of the kinetics of microbial conversions for the design of large-scale CO fermentation processes (Wett et al., 2007).However, there is currently a lack of kinetic information related specifically to syngas fermentation (Heffernan et al., 2022).This knowledge gap is likely related to the difficulty of accurately measuring substrate concentrations, as both gas-liquid mass transfer of poorly soluble CO and microbial uptake rates govern the substrate concentration in the broth.Consequently, classical batch bottle incubations are inadequate for the kinetic characterization of gas fermenting microbes (Sipkema et al., 1998).
Therefore, previously reported kinetic parameters obtained in batch bottle experiments (Mohammadi et al., 2014) are untrustworthy, as it is highly challenging to determine whether the observed rates were limited by mass transfer, biological capacity, or inhibited by carbon monoxide from the available data.Furthermore, the pH, biomass-, substrate-, and product concentrations are not constant over time in batch bottle incubations.All these factors affect gas-liquid mass transfer properties, which in turn influence CO availability (Cotter et al., 2009;Puiman et al., 2022).Some of these limitations are overcome when using chemostat cultivation in bioreactors.Mathematical modeling studies using steady-state chemostat data for kinetic parameter estimation have also been conducted.Such studies estimated maximum biomassspecific CO uptake rates q ( ) CO max between −37.5 and −46.3 mmol g X -1 h -1 for different closely related Clostridium strains (de Medeiros et al., 2019).More recently, however, experimental data was published in which a much higher q CO of approximately −70 mmol g X -1 h -1 was sustained during chemostat operation (de Lima et al., 2022).These variations between model predictions and new experimental observations highlight the significant influence of input data on kinetic parameter estimation.Therefore, gas-fermenting microbes should not only be cultivated in chemostats, but also in dynamic environments to improve the verification of kinetic model parameters.
Bioreactors with controllable gas-liquid mass transfer properties and online off-gas analysis are highly suitable for assessing the substrate uptake capacity of gas fermenting organisms (Novak et al., 2021).
However, parameter estimation through dynamic cultivation approaches still requires careful experimental design.Useful kinetic models often include many parameters, which must be identified in different experiments.Given the significant disparity between modeling and experimental results for CO uptake rates in recent literature, this study focuses on one of the essential kinetic parameters, q CO max , of Clostridium autoethanogenum in dynamic bioreactor experiments.
Bioreactor cultivation was done using a chemically defined medium without yeast extract as described by Valgepea et al. (2017).
The mineral medium contained per liter: 0. were turned on to feed the reactor with a chemically defined medium and control the dilution rate at 0.01 h −1 .The gas composition and sparging rate were controlled using separate mass flow controllers for CO and N 2 (Brooks Instruments).The total gas flow rate was kept at 100 mL min −1 at 1.0 atm, and the reactor was fed with a 90%/10% mixture of N 2 and CO, respectively, during steady-state operation.
The reactor was continuously agitated at a speed of 600 rpm using two equally spaced six-bladed mechanical stirrers, and pH was maintained at 5.5 ± 0.05 using 2 mol L −1 NaOH (Applikon InControl).
The reactor temperature was maintained at 37°C using a water jacket and a thermostat bath (E300; Lauda).The off-gas was cooled using a cryostat set to 4.5°C to prevent culture broth evaporation.

| k L a determination
The gas-liquid mass transfer capacity of the bioreactor was assessed at different stirring speeds using a dynamic method.Briefly, this method is based on measuring the concentration of dissolved oxygen in the medium using a dissolved oxygen (DO) sensor (Mettler Toledo).
Initially, oxygen was stripped out from the reactor by sparging with N 2 until DO = 0%.Then, air was supplied until oxygen saturation in the medium was reached.The saturation curve was linearized by taking its natural logarithm, and the slope of the obtained curve represented the k L a.After determining the k L a for oxygen, the respective parameter was recalculated for carbon monoxide using the diffusion coefficients of the two gases (2.03 × 10 −5 cm 2 s −1 for CO and 2.0 × 10 −5 cm 2 s −1 for O 2 .This procedure was performed in duplicates for five different stirring speeds (300, 500, 600, 700, and 800 rpm) in a 3 L bioreactor at 37°C and a gas flow rate of 100 mL min −1 .The working volume was 2 L. During the 800 rpm test, excessive foaming occurred.Therefore, a working of 1.6 L was used during chemostat operation.C. autoethanogenum was subjected to increased p CO for 1 h by increasing the CO concentration in the inlet gas without changing the overall gas supply rate.pCO was increased with 0.05 atm increments up to pCO = 1.0 atm and with 0.1 atm increments up to pCO = 1.2 atm (Table 1).Pulse experiments up to pCO = 0.85 atm were carried out during Steady State I, and of the four pulses at pCO = 0.9 atm.

| CO pulse experiments
The rest of the pulse experiments were carried out after reaching Steady State II.After 1 h the inlet gas concentration was returned to the original settings (p CO = 0.1 atm).The duration of the pulses was set to 1 h to prevent significant differences in biomass concentration between the start and the end of the pulse.Shorter pulses of 30 min were not sufficient to obtain the stable off-gas profiles required for the calculation of q CO .Limiting biomass growth during the pulses allowed us to use the steady-state biomass concentration to calculate the biomass-specific CO uptake rates during the pulses.pH control was switched off during the pulse experiments to prevent influencing the off-gas measurements.Pulse experiments were carried out in biological duplicates, triplicates, or quadruplicates.Liquid samples were taken either right after termination of the pulse experiment (pulses of 80% CO and above) or when the off-gas measurements returned to the steady-state values.The optical density of the broth was measured to assess whether significant growth had taken place during the pulse experiment.If the optical density increased more than 5% compared with the steady-state value during a pulse experiment, the chemostat was left undisturbed for at least a day before administering the next pulse.The OD never increased more than 10% during a single pulse experiment, validating the use of the steady-state biomass concentration for the calculation of the biomass-specific substrate uptake rate.

| Analytical methods
The off-gas composition of gas fermentation experiments was followed using an X-stream Enhanced XEGP Off-Gas Analyzer equipped with an infrared detector for CO and CO 2 (6 mm and 30 mm, respectively) measurement and a thermal conductivity detector for H 2 measurement (Emerson).During pulse experiments with an inlet gas fraction of CO above 60%, the off-gas was diluted with 100 mL min −1 N 2 gas to stay within the measuring range of the off-gas analyzer.
Biomass concentrations were monitored by measuring optical density (OD 660 ) and the amount of VSS in the broth using 100 mL effluent (APHA/AWWA/WEF, 1999).Both OD and VSS were always determined in duplicate.OD and VSS correlated linearly (VSS = 0.2935 × OD 660 ).Acetate, lactate, formate, 2, 3 − butanediol, and ethanol concentrations were determined using a Vanquish (Thermo Fisher Scientific) high-performance liquid chromatography machine (HPLC) with an Aminex HPX-87H column (BioRad) at 50°C coupled to a refractive index-and ultraviolet detector (Waters).1.5 mmol L -1 phosphoric acid was used as eluent at 0.75 mL min -1 .Biomass was removed from the reactor samples before storage or measuring fermentation products by filtration using a 0.22 µm membrane filter (Millipore, Millex-GV).
2.6 | Determination of biomass-specific CO uptake and CO 2 production rates The calculation of q CO and q CO 2 was conducted from the data gathered in the last 10 min of the CO pulse.During this time frame, the gas profile was stable and reflected the biological consumption of CO.The first 50 min of the experiment were excluded from calculations, as the gradual increase in CO measured in the off-gas in this period is a result of incomplete gas mixing in the headspace of the reactor (Supporting Information: Figure S1).
CO consumption leads to the production of acetate, ethanol, and CO 2 via the following two stoichiometries: (1) (2) CO uptake and CO 2 production were calculated using gas-and liquid mass balances.First, the volumetric outflow of the reactor was calculated using the inert fraction (N 2 ) of the inlet gas as reference, assuming the gas phase was in steady state in the last 10 min of the pulse experiment: (3) For CO and CO 2 in the gas phase, mass transfer to and from the liquid, respectively, must be considered: The general mass balance equation for a species in the liquid is as follows: For CO, this equation can be simplified under the assumption that the in-and outflowing liquids contain negligible concentrations of CO: When stable off-gas values were measured during a pulse experiment, pseudo-steady state was assumed for CO, due to which the dM dt COterm becomes zero.As CO is consumed, the value of q CO will become negative.
CO 2 has a higher solubility than CO, so the concentration in the outflowing liquid must be taken into account.Note: To maintain a constant inlet gas flow rate (F G, in ) of 100 mL min −1 a supply of N 2 was introduced.Additionally, an N 2 flow in the off-gas was used to dilute the bioreactor off-gas to values within the measuring range of the off-gas analyzer.
CO 2 speciation is strongly affected by the increasing CO 2 partial pressure and the increasing pH during a CO pulse.Therefore, dissolved concentrations of CO 2 species (H 2 CO 3 , HCO 3 -, CO 3 2− ) were calculated at each collected data point (1-min intervals).It was assumed that the gas and liquid phase were in equilibrium at each data point.The dissolved inorganic carbon concentration was calculated using the equilibrium constants for the different CO 2 species at 37°C as a function of the off-gas partial pressure and pH.
The equilibrium assumption was verified by estimating the mass transfer rate in the bioreactor k a ( ) from the pulse data and the dissolved CO 2 concentration required for the observed transfer rate.
This concentration differed less than 10% from the maximum equilibrium concentration of CO 2 , confirming that assuming equilibrium for the inorganic carbon species in the system was valid.The increase in total CO 2 -equivalents in the liquid was calculated through integration over the timespan of the pulse and used to correct the observed q CO2 from the off-gas measurements.

| Flux distribution analysis
Assuming the carbon flux to biomass is negligible, the product spectrum of syngas fermentation can be inferred directly from the off-gas measurements.All carbon and electrons that enter the cell must be recovered in the products.The compounds considered were CO, H 2 , CO 2 , acetate, and ethanol.2,3-BDO production was neglected as this was never produced in quantifiable amounts in this study.Then, two balance equations describe the microbial conversion (Equations 7, 8).Three of the stoichiometric coefficients are needed as input, and the other two can be calculated.Here, CO, H 2 , and CO 2 are used as inputs, as they are continuously measured, to calculate ethanol and acetate.All fluxes are defined as positive, considering CO and H 2 as substrates and CO 2 , acetate, and ethanol as products.
Since there is a stoichiometric coupling between substrates and products, biomass-specific rates can also be used as input to calculate the production rates.In that case, v q = i i .The MATLAB script for flux distribution analysis can be found here: https://data.4tu.nl/private_datasets/6gzFKW_ nNWmOXB0eypwsUGa6YTm1LNpIYaV7BFrx5XQ The balance equations for carbon and electrons (with inorganic carbon as reference oxidation state) are: In addition to the distribution of excreted products, we calculated the flux distribution over key catabolic reactions to assess the effect of different feeding regimes on the energy conservation of the microorganism.With the biochemistry represented in Supporting Information S1: Figure S6, a script of intracellular balance equations was written in MATLAB R2020a to calculate the intracellular distribution of carbon and electrons, given a specific stoichiometry of CO, CO 2 , and H 2 .This was done using all steps of the catabolism that use redox carriers (NAD(H), NADP(H), Fd( 2− )).For a feasible overall stoichiometry, each electron carrier's oxidation and reduction must be balanced.This led to the derivation of the redox carrier balances, with the enzyme abbreviations as used in Supporting Information: Figure S6: In this model, we assumed that all CO entering the cell is oxidized upon generation of reduced ferredoxin (v v = CO CODH ).To be able to solve the system, it was constrained so that no metabolites could accumulate intracellularly and that all ATP invested in the methyl branch of the Wood-Ljungdahl pathway (WLP) was recuperated in the acetate kinase reaction.Lastly, an additional flux for acetate excretion (with no additional costs) was added as acetate can either be excreted or further reduced to acetaldehyde and ethanol.
The constraints were formulated as: This system of equations was solved using a least-squares nonlinear solver.The system proved to be robust: it was not sensitive to changes in initial values, and the solution space did not have to be constrained for a correct solution to be obtained.
The biomass-specific ATP production rate (q ATP ) was calculated in steady-state by multiplying q Rnf with the assumed H + /ATP stoichi- ometry of C. autoethanogenum (Supporting Information: Figure S6).
Dividing the imposed steady-state growth rate in the chemostat (µ = 0.01 h −1 ) by q ATP gives the molar growth yield on ATP (Y ATP ) (Stouthamer & Bettenhaussen, 1973).Assuming that Y ATP stays constant over the range of tested p CO , the growth rate during a CO pulse was calculated by multiplying q ATP with Y ATP .

| Steady-state chemostat operation
C. autoethanogenum was cultivated in a chemostat with a working volume of 1.6 L at a dilution rate of 0.01 h −1 and sparged with 100 mL min −1 of 10% CO and 90% N 2 gas.Steady state was assumed when stable off-gas, biomass, and product concentrations were measured for at least two liquid volume exchanges.During CO pulse experiments, steady state was disturbed once due to technical problems.After restarting the system, a new steady state was reached, though with a statistically different (p < 0.05) biomass concentration (0.52 ± 0.02 and 0.59 ± 0.03 g L −1 for Steady Stages 1 and 2, respectively).Slight differences in broth composition can affect the mass transfer rates in bioreactor systems (Puiman et al., 2022b), which could have caused the slight differences in biomass concentrations in the different steady states.As the biomass concentration is used to calculate q CO , the results for the two different steady states are reported separately (Table 2).Both steady states were characterized by a q CO of approximately −10 mmol g x −1 h −1 and a q CO 2 of approximately 6.2 mmol g x −1 h −1 .The liquid product spectrum consisted of acetate and small amounts of ethanol.

| Biological CO uptake increases linearly with increasing CO partial pressures
To assess the CO uptake capacity of C. autoethanogenum, the CO concentration in the inlet gas was increased for 1 h.The maximum rate of CO uptake depends on the mass transfer rate, which was manipulated by adjusting the inlet partial pressure of CO and the headspace pressure (Table 1).
Figure 1 shows a clear linear correlation between p CO in the inlet gas and q CO and q CO 2 .It was observed that q CO decreased up to −119 ± 1 mmol g x −1 h −1 with increasing p CO .q CO 2 increases linearly up to a p CO of 1.0 atm, after which it plateaus at approximately 90 mmol g x −1 h −1 .This might indicate that the microorganism metabolized CO at near-maximal rates at a pCO > 1.0 atm.Up to a p CO of 0.45, trace amounts of hydrogen were measured in the T A B L E 2 Biological CO consumption and product formation rates, and corresponding biomass concentrations in steady state chemostats.F I G U R E 1 Biomass-specific fluxes during CO pulses.Left panel: CO uptake rate (q CO ) and CO 2 production rate (q CO 2 ) as a function of the fraction of CO in the inlet gas.The biomass-specific rates were determined from two, three, or four individual biological replicates, and the error bars depict the standard deviation.The shaded area shows the error range of the calculated rates based on the standard deviation in the steadystate biomass concentration and the standard deviations.Right panel: Flux distribution analysis illustrating the relationship between CO partial pressure, the rate through central carbon metabolism (q WLP ) and the production rates of acetate (q Ac ) and ethanol (q EtOH ).The rates were calculated using q co and q co 2 , considering that all carbon and electrons consumed by C. autoethanogenum must be recovered in the products and that oxidation and reduction of redox carriers in the metabolism must be balanced.Negative values for a biomass-specific rate indicate net consumption of a metabolite.
reactor off-gas, but the biomass-specific hydrogen production rate decreased with increasing p CO (Supporting Information: Figure S3).
The differences between the q CO and q CO 2 profiles could be accounted for by changes in metabolic strategy (Figure 1).

| Acetate consumption for sustaining increased q CO leads to shifts in product spectrum
The effect of increasing CO uptake rates on the metabolism of C. autoethanogenum was assessed with a stoichiometric model of the central carbon metabolism.The measured q CO , q CO 2 , and q H 2 were used as input values to calculate the fluxes of product formation and the flux through the WLP, as shown in Figure 1 (right panel).A shift was observed from the formation of mainly acetate at a p CO of 0.1-0.35atm to the formation of mainly ethanol until a p CO of 0.6 atm.Above pCO of 0.6 atm a net consumption of acetate for additional ethanol production was observed.The flux distribution analysis based on off-gas data was cross-verified with measurements of pH and metabolites in the culture broth (Supporting Information: Figure S4), confirming that shifts in the catabolic strategy of C. autoethanogenum can be identified based solely on off-gas measurements.Notably, the overall rate through the WLP did not increase proportionally with increasing q CO .The highest observed q WLP was 13.6 mmol g x −1 h −1 .
Estimated growth rates, q Rnf , and q CO during the pulse experi- ments were analyzed to investigate the effect of shifts in the product spectrum on the energy conservation of C. autoethanogenum.The biomass yield on ATP (Y ATP ) was calculated from chemostat data (Table 2) and equaled 2.6 g x per mol ATP.The biomass-specific ATP production rate (q ATP ) during the pulse experiments was calculated by multiplying q Rnf , following from the flux distribution analysis, with the assumed H + /ATP stoichiometry of C. autoethanogenum (Supporting Information: Figure S6).The growth rate (µ) was calculated assuming that Y ATP is equal in chemostat and pulse conditions, following • ATP ATP , and was found to increase linearly with increasing p CO (Figure 2, left panel).The activity of the Rnf complex and the membrane-bound ATPase is the only source for net ATP production in C. autoethanogenum when growing on CO (as the WLP is ATP neutral) (Supporting Information: Figure S6).Therefore, the rate of the Rnf complex relative to the CO uptake rate is a proxy for the ATP yield on CO during the pulse experiments and decreased with an increasing flux towards ethanol (Figure 2).

| DISCUSSION
This study aimed to identify q CO max of C. autoethanogenum by using dynamic CO-feeding experiments.1-h pulses were administered to a steady-state chemostat culture, gradually increasing the p CO up to 1.2 atm.By increasing the fraction of CO in the inlet gas and the headspace pressure while maintaining a constant overall inlet gas flow rate, the mass transfer rate in the system was effectively enhanced without affecting the k L a.A similar approach was employed by Novak et al. (2021) to determine q H 2 of Acetobacterium woodii, though in that case, higher stirring speeds were used to obtain a new steady state at an increased mass transfer capacity in a H 2 /CO 2 -fed system.The advantage of the method used in our study is that the k L a remains constant, and the pulse experiments have a short duration, allowing the system to return to steady state rapidly.These characteristics facilitate the performance of multiple biological replicates and save time compared to obtaining different steady-state datasets.
In a gas-fed bioreactor, the mass transfer or biological conversion rate limits the substrate conversion rate.When the latter is the case, F I G U R E 2 Estimated biomass-specific growth rate (µ, left) and the flux of the Rnf complex (q Rnf ) divided by the observed q CO (right).The biomass-specific growth rate was estimated by using q Rnf and the stoichiometries of the Rnf and ATPase complexes shown in Figure 1 to calculate the biomass-specific ATP production rate.The ATP production rate was multiplied by the biomass yield on ATP (Y ATP ) derived from steady-state data to obtain the biomass-specific growth rate.q Rnf /q CO gives a proxy for the ATP yield per mol of CO consumed.
the substrate concentration in the liquid will build up.So far, it has been widely accepted that the build-up of CO in the liquid can lead to metabolic arrest, with CO inhibition constants around 0.1 mM (Ruggiero et al., 2022).This occurs because CO binds to the active site of metalloenzymes, rendering them inactive (Menon & Ragsdale, 1996;Schuchmann et al., 2018).Various enzymes involved in the WLP rely on metal clusters.Consequently, a decrease in metabolic activity due to CO toxicity would result in reduced CO uptake rates, resulting in noticeable changes in the off-gas profile.These observations would indicate that the limit of the biomass-specific CO uptake rate (q CO max ) has been reached.Interestingly, in our experiments, q CO continued to increase over the entire range of CO partial pressures, up to −119 ± 1 mmol g x −1 h −1 .This high biological capacity for CO conversion by C. autoethanogenum was not exhibited at lower p CO , which could either be due to low affinity for CO or due to mass transfer limitation in the bioreactor system.The study by Mann et al. (2021) showed high CO affinity by C. ljungdahlii, which is able to bring the dissolved CO tension close to zero during continuous cultivation.Given the >99% genetic resemblance (Utturkar et al., 2015) and highly similar physiological characteristics of C. autoethanogenum and C. ljungdahlii, it is likely that C. autoethanogenum also has high CO affinity.Furthermore, the actual CO concentration in the liquid must have been below the inhibition constant for CO, as we observed no apparent decrease in the metabolic capacity in our experiments.Therefore, the linear relation between the measured q CO and the CO supply rate suggests that the liquid CO concentration increased only marginally (i.e., sufficiently to sustain a higher rate).Our data, therefore, suggests that the biological conversion of CO was mass transfer limited over the entire range of tested p CO .Future dynamic bioreactor work combining dissolved CO measurements with off-gas analysis could further resolve the inhibition kinetics of CO.
In this study, we could not identify q CO max of C. autoethanogenum in our bioreactor system.This contrasts with previous studies with C. ljungdahlii, where a metabolic collapse occurred due to CO accumulation in the liquid at a q CO of −42 mmol g x −1 h −1 (Mann et al., 2021).In their study, high mass transfer rates were maintained longer than in the experiments performed here.This shows that the time of exposure to elevated p CO could also affect microbial survival rates.
Conducting CO pulse experiments of varying durations could provide valuable insights into the CO consumption capacity of C. autoethanogenum and warrant an exciting avenue for further investigation.
Our study proves the need for dedicated and well-controlled experimental validation of critical kinetic parameters.It shows that the CO uptake capacity of C. autoethanogenum has been severely underestimated to date.The highest observed q CO in this study is 2.5-3.5-foldhigher than the q CO max reported in kinetic modeling studies that used only steady-state chemostat or batch bottle data for parameter estimation and 1.8-fold higher than the highest q CO reported in the literature for C. autoethanogenum (de Lima et al., 2022;de Medeiros et al., 2019;Mohammadi et al., 2014).
Furthermore, verifying whether mass transfer or biological capacity limited the CO conversion rate in previous studies is impossible.Our study shows that mass transfer limitation in the model systems likely leads to the prediction of lower values for q CO max than the q CO of −119 ± 1 mmol g x −1 h −1 observed here.To improve the predictive power of kinetic models further, obtaining the actual q CO max of C. autoethanogenum is pivotal.Bioreactor systems with a different geometry or impeller type could be helpful, as increased stirring rates also increase the mass transfer capacity of a bioreactor system (de Lima et al., 2022;Hermann et al., 2020;Mann et al., 2021;Novak et al., 2021;Valgepea et al., 2017).Higher stirring speeds in the bioreactor system used here led to vortex formation around the impellers, which was detrimental to the mass transfer rate (Supporting Information: Figure S5).Alternatively, a pressurizable bioreactor system could be employed to increase p CO further and investigate its impact on q CO , or higher q CO could be achieved in a system operated at a lower biomass concentration.
When interpreting the results of this study, it should be considered that the bioreactor has been operated for five months (including a one-month recovery after a technical disturbance) under the pulse-feeding regime.Therefore, it cannot be completely excluded that the culture has adapted to fluctuating conditions.For example, enzyme levels could be regulated to increase the cellular CO uptake capacity, or alterations could be made to the cell wall structure to reduce the permeability of CO.
However, adjustments in the proteome are unlikely, as previous work has shown that protein expression did not change even in chemostat experiments where periodically highly fluctuating CO and H 2 uptake rates were observed (Mahamkali et al., 2020).Still, phenotype-altering mutations might accumulate in the genome during long-term exposure to fluctuating conditions (Ingelman et al., 2023).It can, however, be argued that strains with such mutations resemble industrial workhorses since microorganisms continuously experience fluctuating conditions in industrial-scale systems (Puiman et al., 2023).
The combination of online off-gas measurements and flux distribution analysis provides a powerful means to monitor product distribution during CO fermentation without liquid measurements.This approach highlights the effectiveness of dynamic bioreactor experiments with online data analysis for studying microbial physiology, as also demonstrated earlier by Allaart et al. (2021) and Stouten et al. (2021).However, careful experimental design is pivotal to extracting the desired information from the experimental data.In our study, we observed that acetate was the main catabolic product for pulses up to a p CO of 0.35.However, as the p CO increased, more ethanol was formed at the expense of acetate in the fermentation broth.This shift from net acetate production to net acetate consumption is noteworthy.
Similar observations have been made in C. autoethanogenum and other closely related Clostridium species (de Lima et al., 2022;Hurst & Lewis, 2010;Mann et al., 2021;Phillips et al., 2015).The observed changes in physiology align with the overflow hypothesis of Allaart et al. (2023), stating that acetate reduction to ethanol is exploited to mitigate CO toxicity.
In the experiments performed here, we controlled the rate at which CO was supplied via the mass transfer rate.Through metabolic flux analysis, we observed that the flux through the WLP did not increase significantly beyond a p CO of 0.6 atm.However, the increased CO uptake rate was compensated with higher rates of acetate reduction to ethanol.This is likely to be related to the redox state of the electron carriers and the fact that acetate was available in the cultivation broth (Diender et al., 2019;Mahamkali et al., 2020;Richter et al., 2016;Schulz et al., 2023;Xu et al., 2020).Effectively, the product spectrum changed as a function of the mass transfer rate.In other words, the product spectrum of CO fermentation can be controlled by controlling the CO uptake rate through the operational conditions of the bioreactor.This provides an appealing alternative to tedious genetic engineering strategies for improving ethanol production from CO-containing waste.
CO pulse experiments were carried out after a steady state was reached in the chemostat.Due to a technical disturbance, experiments were conducted starting from two highly similar steady states (Steady State I and II).A two-sample t-test was performed to determine whether the difference between the average biomass concentrations in the two stages was statistically significant.The means and standard deviations of the biomass concentrations are presented in Table2.Steady State I was maintained for a longer duration than Steady State II, resulting in sample sizes of 89 and 6, respectively.