Gas fermentation has attractive waste carbon valorization properties, for which the need is intensifying (International Panel on Climate Change [IPCC], 2014; Emerson and Stephanopoulos, 2019). Recently, LanzaTech commercialized the first waste gas-to-ethanol process, efficiently incorporating the carbon from steel mill off-gas into fuel quality ethanol via the model acetogen Clostridium autoethanogenum. The key carbon source—carbon monoxide (CO)—accounts for a significant portion of steel mill off-gas and synthesis gas (syngas), which can be generated from multiple high-volume, non-gaseous waste feedstocks (e.g., biomass, municipal solid waste) (Liew et al., 2016). Therefore, LanzaTech's process is significant in that it valorizes waste carbon by fusing two one-carbon gas molecules (C1) into liquid fuel. Furthermore, Handler et al. (2016) found that ethanol produced by LanzaTech's process reduced greenhouse gas emissions by 67 to 98% when compared to petroleum gasoline on an energy content and “cradle-to-grave” basis (feedstock dependent). Carbon dioxide (CO₂) represents a more diverse and plentiful waste stream compared to CO (International Panel on Climate Change [IPCC], 2014), thus embodying a feedstock with greater climate change mitigation and carbon recycling potential.

Increasing acetogenic carbon capture as CO₂ would build on the success of commercial gas fermentation and continue the expansion of the technology as a platform for sustainable chemical production (Redl et al., 2017; Bengelsdorf et al., 2018; Müller, 2019). Compared to other CO₂ valorization methods, acetogens are ideal candidates due to their high metabolic efficiency, ability to handle variable gas compositions, high product specificity, scalability, and low susceptibility to poisoning by sulfur, chlorine, and tars (Liew et al., 2016; Artz et al., 2018). However, metabolism of CO₂ requires an energy source, for which some see an appropriate solution is lacking (Emerson and Stephanopoulos, 2019).

Gas fermenting acetogens harbor the Wood-Ljungdahl pathway (WLP) (Drake et al., 2008), a non-photosynthetic C1-fixation metabolic pathway with the highest-known theoretical thermodynamic efficiency (Fast and Papoutsakis, 2012; Schuchmann and Müller, 2014; Müller, 2019). Various potential energy sources exist for metabolizing CO₂, primarily hydrogen, nitrates, sugars, and arginine. Yet, acetogenic CO₂ valorization, which is actively being developed for industrial implementation (Tizard and Sechrist, 2015), poses challenges along with promise. These include potential adenosine triphosphate (ATP) starvation in autotrophic conditions and carbon catabolite repression in hetero/mixotrophic conditions (Emerson and Stephanopoulos, 2019).

Hydrogen (H₂) is the most recognized energy source for CO₂ utilization—as metabolism of sugars or nitrates cause shifts in metabolism that result in lower CO₂ or H₂ utilization (Liew et al., 2016; Emerson and Stephanopoulos, 2019). H₂ production will also logically transition to renewable sources in the future, whereas production of sugars and nitrates are dependent on less-sustainable methods. Furthermore, levelized cost predictions for solar H₂ indicate a 30% reduction by 2030, potentially becoming competitive with the current levelized cost of fossil fuel derived H₂ by 2035 (Detz et al., 2018; Glenk and Reichelstein, 2019). This is in part due to rapidly decreasing solar electricity costs (IRENA, 2017) and projections of H₂ electrolysis technology development (Detz et al., 2018; Glenk and Reichelstein, 2019). Similarly, atmospheric CO₂ capture via direct air contact showed promising feasibility recently (Keith et al., 2018), which represents an essential development for carbon recycling (Otto et al., 2015). Various power-to-gas technologies are being discussed for mediating fluctuations in renewable power generation (Götz et al., 2016). By extension, gas fermentation to liquid products could couple mediation of renewable power fluctuations to carbon recycling (Redl et al., 2017). This provides an attractive new opportunity for bacterial artificial-photosynthesis, whereby renewable H₂ supplementation facilitates acetogenic CO₂ valorization (Claassens et al., 2016; Haas et al., 2018).

Continuous culture bioprocesses are preferable to batch or fed-batch fermentation bioprocesses (Hoskisson and Hobbs, 2005). Furthermore, systems-level quantification is essential for design-build-test-learn bioprocess optimization by metabolic engineering (Valgepea et al., 2017). Therefore, obtaining quantitative datasets from steady-state chemostat cultures, whose analyses are comparable between experiments, is important for development of these systems (Adamberg et al., 2015). Whilst Bengelsdorf et al. (2018) reviewed autotrophic acetogen growth on CO₂ and H₂ (CO₂/H₂), and Mock et al. (2015) provided notable insight into the CO₂/H₂ metabolism of C. autoethanogenum, the literature lacks a steady-state dataset where carbon flows in a CO₂/H₂ fermentation are quantified. Here we aimed to quantify steady-state CO₂/H₂ fermentation using fully instrumented chemostats and the model acetogen C. autoethanogenum. Subsequently, we showed that CO₂ is a promising feedstock alternative to CO, as more than half of the substrate CO₂ carbon was converted into ethanol. Furthermore, supplementation with CO at low concentrations improved fermentation performance significantly.

Achieving steady-state continuous cultures using CO₂/H₂ mixtures, without cell recycling here, was challenging. Yet, compared to other organisms fermenting CO₂/H₂ with continuous medium exchange, Clostridium autoethanogenum performs well (Table 2). No direct comparisons can be made to other experiments due to variations in conditions, but C. autoethanogenum clearly achieves the highest ethanol production, with comparable quantities of carbonous products also. Acetobacterium woodii, along with Sporomusa ovata, were shown to perform well when compared to a wide range of acetogens under batch CO₂/H₂ conditions (Groher and Weuster-Botz, 2016). Yet, as evidenced by omission of S. ovata from Table 2, few continuous culture characterizations of acetogens are available—an essential step for validation of industrial robustness in gas fermentation. As discussed by Molitor et al. (2019), the lack of yeast extract or C_≥2 substrates is also distinguishing between fermentations.

Notably, CO₂/H₂ cultures displayed higher variability between biological replicates compared to those of CO-containing gas mixtures (Figure 1) (Valgepea et al., 2017). This may indicate variable organism fitness, a trait previously discussed for C. autoethanogenum by Liew et al. (2016), who extensively covered numerous techniques used for enhancing gas fermentation including—coupling to other processes, adaptive laboratory evolution, and metabolic engineering of acetogens using genetic tools. CO-supplementation could be a valuable option for enhancement as it overcomes inherent problems linked to engineering acetogens. Supplementation of low quantities of CO here stabilized the culture, enabled culturing at D = 1 day⁻¹, and achieved higher biomass concentration with a carbon incorporation larger than any other gas type—all without compromising by-product distribution.

While, Valgepea et al. (2018) found that syngas fermentation lead to CO-only fermentation at steady-state, we observed co-utilization of CO and CO₂. Tizard and Sechrist (2015) have also shown co-utilization for C. autoethanogenum continuous cultures, and it seems that co-uptake may also occur for some points of syngas batch fermentation (Infantes et al., 2020). Co-utilization of sugars was found for E. coli in chemostats—where inhibition of consumption, but no change in induction time was observed (Standing et al., 1972). The WLP is most likely no different, in that metabolism of CO is preferential, yet the pathway can co-consume CO₂ under certain conditions.

Various efforts have been made toward enhancing CO₂(+H₂) fermentation to C_≥2 products (Table 2) (Emerson and Stephanopoulos, 2019). Braun and Gottschalk (1981) first discovered the potential for enhancement when Acetobacterium woodii simultaneously consumed fructose and a headspace of CO₂/H₂ during batch cultivation. Growth and acetate production was high but no characterization of the headspace was performed. More recently, continuous glucose-supplemented CO₂/H₂ fermentation of Moorella thermoacetica by Park et al. (2019) did not lead to net uptake of CO₂. Furthermore, Jones et al. (2016) did not show net CO₂ uptake for a wide range of acetogens (not A. woodii) fermenting syngas and fructose. A. woodii generates a sodium ion (Na⁺) gradient (Hess et al., 2013) rather than a proton (H⁺) gradient for membranous ATP generation (Pierce et al., 2008; Poehlein et al., 2015; Bengelsdorf et al., 2018). This may highlight an important metabolic difference from other model acetogens—decoupling the resources of the WLP and membranous ATP generation pathways could facilitate fermentation of sugar and CO₂/H₂ simultaneously.

Other enhancements have also struggled to achieve net CO₂ uptake. Co-culture of C. acetobutylicum and C. ljungdahlii showed syntrophic metabolic coupling when fermenting glucose, fructose, and CO₂/H₂, but no net CO₂ uptake (Charubin and Papoutsakis, 2019). Addition of nitrate to batch CO₂/H₂ fermentation by C. ljungdahlii, increased biomass concentration and subsequently volumetric productivity of acetate (Emerson et al., 2019). However, the specific WLP productivity decreased, meaning lower utilization of CO₂. Other organisms not recognized as gas fermenters can also use mixotrophy to minimize carbon loss, such as Clostridium beijerinckii but have not displayed net CO₂ uptake either (Sandoval-Espinola et al., 2017). To the best of our knowledge, this is the first report where supplementation of a substrate other than H₂, increased productivities of continuous acetogenic CO₂ fermentation while maintaining net CO₂ utilization. Furthermore, the effect of CO supplementation on CO₂ utilization was superlinear, indicating a synergistic mechanism (Park et al., 2019). This is encouraging for development of bioprocesses valorizing CO₂.

Comparisons between fermentation datasets enables us to speculate about the positive effect of CO-supplementation on CO₂/H₂ fermentation. Although, addition of CO led to minimal metabolic shifts (Figure 2—CO₂/H₂ vs. CO/CO₂/H20.5 and Figure S2), FBA showed that CO supplementation caused significant increases to the reduced ferredoxin consumption by AOR and Rnf complex (leq000004 and M002, respectively) compared to CO₂/H₂ (Figure 2). The overflow model proposed by Richter et al. (2016) suggests that high NADH production via Rnf and Nfn complexes (leq000002) is also important for reducing AOR product inhibition. In this way, NADH facilitates fast metabolism of acetaldehyde to ethanol via alcohol dehydrogenase (Adh(E); rxn00543_c0). Decreasing the acetate concentration reduces acidification and the ATP cost for excreting acetate (Valgepea et al., 2018). Including acetaldehyde conversion to ethanol and association to acetic acid, this also leads to consumption of 2 H⁺ (4 here vs. 2 produced via CODH). Therefore, CO consumption decreases the intracellular H⁺ pool, and following Le Chatelier's principle, drives HytABCDE activity. Indeed, the change in specific H₂ uptake relative to specific CO₂ uptake is greater than that of CO (for CO₂/H₂ vs. CO/CO₂/H20.5, Table S7). Subsequently, the relative gain in free energy from H₂ is ~ 4-fold greater than CO. We speculate this is ultimately responsible for the improved fitness of CO-supplemented CO₂/H₂ fermentation by C. autoethanogenum. We propose the following five critical factors to this enhanced metabolism: [1] metabolism of CO increases the intracellular pool of reduced ferredoxin; [2] this stimulates oxidation of ferredoxin, which if consumed by the AOR; [3] reduces ATP costs; and [4] decreases the H⁺ pool/acidification; which therefore [5] drives H₂ uptake for further reduction of ferredoxin. Evidently, additional understanding of acetogenic redox metabolism, from a thermodynamic perspective, is important for developing acetogenic CO₂-valorization as a platform industrial bioprocess (Cueto-Rojas et al., 2015).

Physicochemical properties could also play a key role in CO-supplementation enabling to achieve a stable CO₂/H₂ chemostat culture at D = 1 day⁻¹. Generation of a stable and large non-equilibrium is what drives microbial growth (Qian and Beard, 2005; Igamberdiev and Kleczkowski, 2009; Quéméner and Bouchez, 2014) and gas-liquid mass transfer (Ma et al., 2005). For continuous culture of gas fermenting microbes, an inherent relationship between substrate mass transfer and culture growth exists (Supplementary Materials Note 2.1). An important parameter for these systems is the Gibb's free energy of a system (Cueto-Rojas et al., 2015). This describes the thermodynamic favorability of the reaction system—termed spontaneity. Here, analysis of experimental flux and Gibbs free energy suggests that CO₂/H₂ fermentation is infeasible (ΔG˙OR0= 5.4 kJ/mol/day), whereas CO-supplemented CO₂/H₂ fermentation is feasible (ΔG˙OR0= -12.3 kJ/mol/day; Table S6). Though these calculations use standard conditions, they do indicate how close CO₂/H₂ fermentation is to the thermodynamic limit of metabolism. Theoretically, minute and unobservable changes to chemostat CO₂/H₂ fermentation can disrupt the culture (Henry and Martin, 2016). Thus, increasing the free energy of central metabolism with CO-supplementation appears to keep metabolism in a spontaneous and stable state by increasing reduced ferredoxin production.

The mechanisms for achieving the 2-fold higher specific WLP flux throughput for CO/CO₂/H21 compared to others is less clear but appears to be linked to the difference in primary substrate. CO/CO₂/H21 and CO/H₂ are the most similar CO₂ and CO fermentations, respectively (D ~1 day⁻¹ and carbon to hydrogen feed ratio (~1:3); Table 1), and the maximum carbon incorporation per cell for CO/H₂ was roughly half of that of CO/CO₂/H21 (~450 vs. ~1,000 mmol/gDCW). Theoretically, cells will maximize carbon-to-redox metabolism by minimizing thermodynamic losses. CO supplementation to a CO₂/H₂ culture seems to facilitate this as (H₂/carbon)_feed–(H₂/carbon)_flux was ~0 mol/mol for CO/CO₂/H₂ fermentations only (Table S2)—an indication of the relative magnitude of carbon and redox metabolism. This suggests that high specific fluxes for CO/CO₂/H21 may be a result of (close to) optimal co-factor recycling by C. autoethanogenum's WLP and redox pathway. Thus, the lower energy associated with CO₂ fermentation may, counterintuitively, stimulate specific WLP activity when in the presence of appropriate energy-containing substrates. Further quantifications of CO₂ metabolism and characterizations of enzyme activities are required to confirm these hypotheses (Supplementary Materials Note 2.2), as they assist our ability to engineer the links between redox and carbon metabolisms.

We established a dataset quantifying steady-state of the model acetogen C. autoethanogenum during autotrophic-CO₂/H₂ growth in chemostat cultures. This enabled analysis via FBA, and highlighted CO as a potential supplement. CO supplementation successfully improved metabolic stability and CO₂ utilization. This was the first time that intracellular fluxes for net uptake of CO₂ (with enhancement) where characterized. Industry is actively developing gas fermentation to valorize CO₂ (Tizard and Sechrist, 2015; Haas et al., 2018). Previously, genetic and process engineering of gas fermentation successfully developed the technology for industrial CO valorization (Liew et al., 2016). Therefore, progression to industrial CO₂ valorization is foreseeable, and CO supplementation may play a role in the continuing diversification of industrial gas fermentation.

All datasets generated for this study are included in the article/Supplementary Material.

All authors viewed and approved the manuscript and contributed significantly to the work. KV, EM, and LN conceived the project. JH, KV, and EM designed the experiments and analyzed the results. JH and KV performed experiments, supported by RS, IC, MP, and EM. JH wrote the manuscript with the help of KV, EM, RT, SS, MK, and LN.