Since 2013, there has been an explosion in the number of research articles published on Clostridium autoethanogenum, an acetogen capable of producing platform chemicals such as ethanol and 2,3-butanediol from greenhouse gases. However, no review focusing solely on C. autoethanogenum has appeared in the literature. This review outlines the research conducted into this organism in three broad categories (Enzymology, Genetics, and Systems Biology) and suggestions for future research are offered.
Clostridium autoethanogenum is a strictly anaerobic bacterium belonging to a group of microorganisms known as acetogens [1,2]. These bacteria typically produce acetate by fixing carbon monoxide (CO) or carbon dioxide (CO2) via the Wood–Ljungdahl Pathway (WLP, Figure 1) [3,4]. C. autoethanogenum is so named because it can grow autotrophically, producing ethanol from CO as its sole source of carbon and energy. While acetate and ethanol are the main components of the organism's native product profile, lactate and the valuable platform chemical 2,3-butanediol (2,3-BD) are also formed in smaller amounts via pathways involving the reduction of acetyl-CoA to pyruvate (see Figure 1).
Autotrophic metabolic pathways in C. autoethanogenum.
The industrial application of C. autoethanogenum as a producer of ethanol and 2,3-BD from greenhouse gas is not just an exciting prospect; it is becoming a reality. In fact, the Chicago-based company LanzaTech  already produce ethanol and 2,3-BD from C. autoethanogenum at commercial scale using steel-mill off-gas (a mixture of CO, CO2, and H2). LanzaTech have also secured intellectual property rights on much of the relevant industrial and biological technology [6–10]. Nevertheless, the need for improved efficiency and diversification in product profiles has been recognised  and represents an important mission for the scientific community.
The impressive advancements and challenges in developing a C. autoethanogenum-based bioprocess have caused an explosion in the number of related research articles (see Figure 2). The publication of multiple genome sequencing efforts [12–14], including a manual genome annotation , has been fundamental to this increased activity. However, even as we approach the 25th anniversary of the discovery of C. autoethanogenum, no review has appeared which treats the study of this organism separately from its close relative, Clostridium ljungdahlii, an acetogen with important phenotypic and genomic differences [15–17]. For example, C. autoethanogenum possesses a native CRISPR (Clustered Regularly Interspaced Short Palindromic Repeats) system , a potential advantage given the forecast dominance of CRISPR as a tool for genetic manipulation in biotechnology . Other genotypic differences between the organisms include an additional sixth hydrogenase and a putative formate transporter  (both absent in C. ljungdahlii). Taking such differences into account, this review aims to provide an overview of progress in recent scientific research focusing solely on C. autoethanogenum. The published work fits into three categories: Enzymology, Genetics, or Systems Biology, which are examined in turn, followed by a discussion that includes critical analysis of research conclusions and provides an assessment of the communication between disciplines. Directions for future study are also outlined.
Research articles focusing solely on C. autoethanogenum since its discovery in 1994.
The chemical conversions carried out by an industrial microorganism depend fundamentally on the enzymes encoded in its genome. Any effort to engineer the improved performance of these conversions must therefore include a proper characterisation of this enzymatic equipment — all the way from substrate to product. While the WLP [3,4] and glycolytic pathways, respectively, enabling CO2 fixation and degradation of fructose, had both been identified in C. autoethanogenum , enzymes enabling energy conservation [19–22] at what has been termed the ‘thermodynamic limit of life’  required characterisation. In this section, we show how progress made in the study of individual enzymes has enabled answers to questions regarding the broader picture of metabolism, such as energy conservation.
In the case of C. autoethanogenum, the first such enzymological study focused on the set of oxidoreductases involved in product formation from CO . The outstanding discovery was an enzyme complex consisting of an electron-bifurcating NADP-dependent [FeFe]-hydrogenase (HytA-E) and formate dehydrogenase (FdhA), which exhibited reversible hydrogenase, formate dehydrogenase, and formate hydrogen lyase activities in vitro. In this instance, the term ‘electron bifurcation’ [25–28] refers to the enzyme's harnessing of energy released in the transfer of electrons from hydrogen to NADP+, enabling the endergonic transfer of electrons from hydrogen to ferredoxin. This complex thus catalyses the first step in the methyl branch of the WLP (from CO2 to formate), while also providing the means for the utilisation of hydrogen as an energy source.
Certain aspects of the functional characterisation of this complex were problematic. Its NADP-dependence would make hydrogen utilisation unfavourable for thermodynamic reasons and was thought to necessitate the existence of an NAD-dependent version of the enzyme. Furthermore, the observed inhibition of the enzyme's formate-producing activity by CO — a property usually confined to hydrogenases — led to the hypothesis that the internal steady-state CO concentration is lower than the apparent inhibition constant (Ki) of CO . Following this, the identification of a CO dehydrogenase (CODH) resulted in further speculation about the dynamic properties of C. autoethanogenum metabolism — for instance, that the build-up of cytoplasmic CO caused by a limitation of oxidised ferredoxin could lead to a complete deregulation of metabolism (a ‘culture crash’) . Given new information about individual enzyme components, this form of speculation on system properties is a recurring theme in the C. autoethanogenum literature.
Further characterisation of the oxidoreductases  focused on cells grown on CO2 + H2. The set of oxidoreductases was refined by analysing gene expression levels during ethanol production. Subsequent inference of enzyme abundance revealed that five (out of six) hydrogenases encoded in the C. autoethanogenum genome were potentially functional under the subject conditions. Of these, only the electron-bifurcating NADP-dependent [FeFe]-hydrogenase (HytA-E) characterised in ref.  was expressed to significant levels, even during growth on CO2 + H2. This refutes the hypothesis made in the recent study  that an NAD-dependent version would be expressed for growth under such conditions.
In addition to the previously identified Acs (acetyl-CoA synthase)/CODH complex, two further CODH genes were identified, with potential roles in ensuring functional robustness. Another important result was the isolation of catalytically active acetaldehyde : ferredoxin oxidoreductase (AFOR), an enzyme understood to provide an alternative route for ethanol production via direct reduction of acetic acid [29,30]. Steady-state flux distributions were determined manually for the subsequent metabolic scheme, enabling the calculation of theoretical ATP yields. Importantly, ATP yields per mole ethanol were higher than those per mole acetate, which some authors associate with the preference for ethanol . ATP yields per mole substrate uptake were, however, higher for acetate, which is likely to benefit the cell under carbon-limited growth regimes. Furthermore, physiological conditions render ethanol production sub-optimal in terms of exergonicity relative to acetate production (on CO2 and H2); thus, it was concluded that the avoidance of acidification is the main advantage of ethanol formation . Intriguingly, the cofactor combinations required for the purportedly electron-bifurcating methylene-THF reductase (MetFV) could not be determined, providing a tantalising open question for the later study.
A picture of pathway-level behaviour can be constructed using the enzymological results reported above, noting that the functioning of biochemical pathways in C. autoethanogenum is constrained by an important feature of anaerobic metabolism: redox balance, i.e. the oxidation of one compound must be accompanied by the concomitant reduction of another, and fermentation products must always exist in a ratio such that the ‘degree of oxidation of the system as a whole remains constant’ . With this in mind, the question of how energy is conserved in the fixation of CO to acetate (in the absence of H2) can be explored, starting from the fundamental principle that for any chemical reaction to proceed, energy must be released in the conversion of substrates to products, i.e. the change in Gibbs-free energy of reaction must be negative (ΔrG < 0). The term ‘energy conservation’ describes the processes enabling the conversion of this energy into a form (usually ATP) that can be used to support biomass production and other essential cellular tasks. For example, C. autoethanogenum catalyses the exergonic conversion of CO into acetate with the following stoichiometric weightings:
In this case, the question of energy conservation is twofold: (1) How is CO (the energy substrate) used to provide energy for the fixation of CO2 to acetyl-CoA? (This conversion follows an uphill energetic gradient, see Figure 3) and (2) How is CO used to generate a positive ATP yield? Answers to both these questions can be built up using the enzyme characterisations reviewed above. First, following eqn (1), 4 mol of reduced ferredoxin are generated during the oxidation of CO to CO2 by CODH. Since CO2 cannot drive further reduction of electron carriers, all reducing power must be derived from these reduced ferredoxins, including the NAD(P)H required in the methyl branch of the WLP. The allocation of ferredoxin proceeds as follows (see Figure 4): one out of 4 mol of ferredoxin are consumed in the (re-)reduction of CO2 to CO by the ACS/CODH complex, while 0.5 mol are consumed in the reduction of CO2 to formate by FdhA/HytA-E. The NADH requirement of MetFV is met by electron transfer from ferredoxin to NAD+ as carried out by the Rnf complex [20,21,29,33]. This transfer of electrons from the highly electronegative Fd/Fd2− redox pair (E˚′ = −500 mV) to NAD+/NADH (E˚′ = 320 mV) enables a proton motive force, which in turn drives ATP synthesis via the membrane-bound F1Fo ATP synthase [19,22,29].
Formation energies of compounds in the methyl branch of the WLP.
Energy conservation in C. autoethanogenum: acetate production from CO showing the allocation of reduced ferredoxin.
If MetFV is bifurcating, an additional transfer of electrons from Fd2− to NADH is carried out by the Rnf complex to balance the production of an additional Fd2− by MetFV, which serves to increase the ATP yield through Rnf/ATP synthase-driven ATP synthesis. Both Fd2− and NADH are required in the production of NADPH by the Nfn complex. In this reaction (catalysed by Nfn), 2 mol of NADPH are produced from 1 mol of Fd2− and 1 mol of NADH, and thus, 0.75 mol of NADH and 0.75 mol of Fd2− are required to meet the NADPH demand of both FdhA and methylene-THF dehydrogenase (0.5 and 1.0 mol, respectively). The ATP requirement of FtfL is met by substrate-level phosphorylation in the conversion of acetyl-CoA to acetate via acetate kinase (Ack). By assuming that the rotor of the ATP synthase consists of 11 c-subunits (see Discussion) and that MetFV is bifurcating, the ATP yield per mole CO uptake (YATP) in this case would be 0.375. Figures 5 and 6 show similar schemes for ethanol production from CO, demonstrating how the enzymes summarised above conserve energy. An in-depth description of energy conservation in the case of CO2 + H2 metabolism is given in ref. .
Energy conservation in C. autoethanogenum: ethanol production from CO (acetyl-CoA reduction to acetaldehyde) showing the allocation of reduced ferredoxin.
Energy conservation in C. autoethanogenum: ethanol production from CO (acetate reduction to acetaldehyde) showing the allocation of reduced ferredoxin.
The enzymatic information summarised above can also be used to form predictions about unreported modes of metabolism in C. autoethanogenum, for example, formate growth. Formate has previously been trialled as a carbon source for C. autoethanogenum , but no growth was observed. It is possible, however, that the 100 mM concentration used to grow the cells was toxic and a lower concentration may support growth. Indeed, 100 mM and 30 mM concentrations of formate have been shown to be toxic for Escherichia coli  and C. ljungdahlii , respectively. The possibility thus remains that the formate transporter, if genuine, could be involved in formate uptake — though it is questionable if a transporter would be necessary given that such a small molecule can probably diffuse passively across the cytosolic membrane.
Furthermore, the reported growth of C. ljungdahlii on formate  would suggest that the transporter is not essential for growth on formate (if, indeed, this genetic difference between C. ljungdahlii and C. autoethanogenum is genuine and not a result of mis-annotation). One possible consequence of this hypothetical mode of metabolism is in the ongoing development of a ‘formate bioeconomy’ , in which electrical energy is stored as either CO or formate before incorporation into commodity chemicals or biomass via a microbe-mediated bioprocess. Thus, if C. autoethanogenum can support growth on formate, it might be well suited as a chassis for the formate-bioeconomic programme . A hypothetical scheme for formate metabolism is set out in Figure 7 using the enzymes described above [24,29]. In this scheme, YATP = 0.24.
Energy conservation in C. autoethanogenum: acetate production from formate showing the allocation of reduced ferredoxin and NADPH. Boxes, arrows, and numbers are defined in the legend of Figure 4.
The hypothesis regarding functionality of the CO dehydrogenase isozymes, i.e. that the two monofunctional CODHs are non-essential for CO autotrophy , was tested by means of generating and characterising gene knockout (KO) mutants for the corresponding enzyme-encoding genes. In total, three genes encoding separate copies of CODH are present in the C. autoethanogenum genome: cooS1, a monofunctional carbon monoxide dehydrogenase active in the CO-oxidising direction; cooS2, an isozyme of cooS1; and acs/codh, encoding a complex formed by CODH and Acs in which the reduction of CO2 is coupled to the production of acetyl-CoA. The absence of growth in CO-fed cultures of the acs/codh KO mutant was consistent with its proposed central function in the WLP. As expected, fructose-grown acs/codh KO mutant cells exhibited growth; however, the apparent abolition of acetate was less intuitive. This observation suggests that the inactivation of the Acs-dependent WLP, which provides a route for the utilisation of reducing equivalents and CO2 released during the degradation of fructose to pyruvate, causes a redistribution of flux towards ethanol, lactate, and 2,3-BD.
The unchanged behaviour of cooS1/S2 KO mutants when growing on CO was consistent with the hypothesis that these enzymes provide functional redundancy. Curiously, cooS1 KO mutants grown on CO2 + H2 reached approximately double the optical density (OD) of cooS2 KO mutants, with no apparent lag phase. The authors speculated that this phenotype could be explained if CooS2 acted as a competitor for CO2, its removal thus enhancing Acs/CODH efficiency and subsequent biomass production.
To assess acetate-/aldehyde-involving oxidoreductases, mutants were engineered using allelic coupled exchange , including adhE KO, aor1 single KO, and aor1 + 2 double KO (where aor genes encode AFOR) . These mutants were characterised under autotrophic and heterotrophic conditions, with a focus on the resulting product and growth profiles. One hypothesis, stated in ref. , was that the metabolic route to ethanol via Ack and AFOR provides a more energetically favourable form of ethanol production for the cell. While this qualitative prediction was consistent with mutant behaviour on CO, it conflicted with the product profiles of fructose-grown cells, for which the aor2 KO mutant yielded most ethanol. The explanation offered for this result was that the assumed objective of maximising ATP generation through Ack-catalysed substrate-level phosphorylation may not hold under heterotrophic conditions, during which sufficient ATP is produced in glycolysis .
Glycolysis and gluconeogenesis are somewhat peripheral areas of metabolism in the field of gas fermentation in acetogens. Nevertheless, an investigation was conducted into the two glyceraldehyde-3-phosphate dehydrogenase (GAPDH) isozymes separately encoded by CAETHG_1760 and CAETHG_3424. CAETHG_1760 is located in a gene cluster with glycolysis-related genes (CAETHG_1756–1760) . KO mutants of the two GAPDH genes exhibited interesting phenotypes . The analysis of cell extracts  showed that both NAD and NADP cofactors enabled GAPDH activity in the anabolic direction on gas-fed growth and catabolic direction during fructose-fed growth. It was hypothesised that one gene encoded an NAD version and the other an NADP version .
The CAETHG_1760 KO could not be obtained, while the CAETHG_3424 mutant could only support fructose growth. This result suggests that both enzyme products of the two genes must be present to allow growth on gas. However, failure to achieve a CAETHG_1760 mutant by ClosTron mutagenesis  could be due to reasons (perhaps technical) other than its essentiality. Furthermore, flux distributions from in silico analysis given in ref.  (Supplementary Materials) predict growth on gas with zero flux carried by the CAETHG_1760-encoded reaction (see the Systems Biology section).
While clear progress has been made in characterising the relatively small set of enzymes apparently involved in ethanol production, the behaviour of the complete metabolic network, including its various possible layers of regulation, remained unexplored. Moreover, a better understanding of the system's response to bioprocess parameters was needed; previous studies report changes in fermentation performance in relation to variations in growth medium and bioreactor/culture set-up [1,42–44], no study had yet attempted to understand the inner workings of the system orchestrating these changes.
Marcellin et al.  provided a system-level description of the metabolism of C. autoethanogenum using a GSM, incorporating transcriptomic, proteomic, and metabolomic data. The authors contrasted metabolic behaviour under gas (CO, CO2 + H2) and fructose growth, building on the knowledge derived thus far from genetic and enzymological studies. The analysis of metabolomics data revealed that during autotrophic growth, the intracellular NAD+ concentration far exceeded that of heterotrophic growth, while NADH concentrations remained constant between growth conditions . The Rnf complex , which transfers electrons from reduced ferredoxin to NAD+, was highly expressed under autotrophic conditions but completely absent under heterotrophic conditions, consistent with its proposed role in ATP generation by the maintenance of a proton gradient (ΔμH+) [23,29]. Furthermore, the high NAD+/NADH concentration ratio reported in ref.  may help to drive the forward reaction of Rnf under autotrophy, when its functioning is essential. The organism's use of alcohol dehydrogenases also differed between heterotrophic and autotrophic conditions, agreeing with observations made by Mock et al. . Gene KO targets (executed with ClosTron technology ) included GAPDH and Nfn encoding genes, both of which were among the oxidoreductases characterised in ref. .
GSM simulations showed that the reaction catalysed by the CAETHG_1760-encoded GAPDH is essential only under high rates of glycolytic flux and without co-expression of CAETHG_3424, suggesting synthetic lethality. The hypothesis tested was that CAETHG_1760 encodes conventional GAPDH, while CAETHG_3424 encodes a non-phosphorylating GAPDH (or ‘GapN’, as described in ref. ) reversibly catalysing glyceraldehyde-3-phosphate (GAP) to 3-phosphoglycerate (3PG) and avoiding the ATP requirement in the conversion of 3PG to 1,3-bisphosphoglycerate during gluconeogenesis. Simulations of this scheme, while demonstrating the potential existence of an energetically draining futile cycling, ultimately failed to explain the observed phenotypes . The Nfn KO mutant showed stunted growth on gas, which could not be improved with increased H2 supply. The function of Nfn is to interconvert NADH and NADPH (see Figure 1), and thus, the KO mutant's behaviour suggests that Nfn provides the main, or perhaps the only route for NADH production from the NAPDH directly generated from H2 by hydrogenase HytA-E. Unfortunately, flux balance analysis (FBA) of the Nfn KO was not reported in ref. . Such analysis may help to understand the deleterious effect of the Nfn KO when hydrogen is the system's source of energy.
A refined version of the GSM (iCLAU786) established by Marcellin et al.  was used to generate optimal flux distributions and shadow prices for the identification of amino acid growth supplements . The flux distributions, computed using FBA, were compared with growth rates and uptake rates of batch-grown cells. Subsequently, results were assessed in terms of their consistency with metabolomics and transcriptomic data. Shadow price analysis and FBA of the GSM returned nine amino acids which improved theoretical growth yields when supplied as additional nutrients. Model simulations and experimental tests showed that the provision of arginine abolished acetate production under otherwise autotrophic conditions, and gene expression data were consistent with the proposed use of the arginine deiminase pathway for the generation of ATP, leading the authors to advocate its use as a growth supplement for ethanol production in C. autoethanogenum.
Further system-level study provided a detailed description of the relationship between carbon, energy, and redox metabolism in C. autoethanogenum during continuous growth . Highlights included an attempt to answer the question of MetFV electron bifurcation posed by Mock et al. , and an explanation for the causes of failure in continuous culture. The key experimental variable in this study was mass transfer rate, i.e. the rate at which gaseous CO becomes aqueous CO. This variable was increased by agitation of the culture medium until biomass concentrations no longer increased in response. A major result of this study was that carbon flow could be shifted towards ethanol; the acetate/ethanol ratio shifted from 6 to 1 with increasing biomass concentrations. No clear transcriptomic explanation for this redistribution of carbon was given. In fact, repression of the conventional ethanol-producing enzymes (ADHE and AFOR) was observed.
Furthermore, the authors reported that no change in the expression of the gene encoding the redox-sensitive gene derepressor, Rex , was detected, despite changes in the ratio of concentrations of NAD(P)H and NAD(P)+. In silico flux distributions computed with FBA showed that the only favourable MetFV reaction stoichiometry, in terms of achieving positive ATP yield and supporting biomass production, was bifurcation with ferredoxin and NADH, a seemingly conclusive answer to the question of MetFV functionality posed in ref. . Unfortunately, other factors which may undermine this conclusion cannot be overlooked, including the subject model's assumed ATP synthase reaction stoichiometry and the absence of alternative sites of electron bifurcation, e.g. LDH .
Following this result, estimates for the total cost of ATP maintenance (rATP) could be generated with FBA, in which the maximisation of ATP dissipation was the objective function. Results from this analysis demonstrated that cells growing at higher biomass concentrations have higher rATP. Using the GIMM3 algorithm , which integrates expression data into FBA, 2,3-BD production was predicted. This is a major result given that no concrete explanation for 2,3-BD production had been reported prior to the present study. The authors proposed that metabolic modes involved with the production of 2,3-BD might relate to the lowering of intracellular proton concentrations, an objective which reoccurred in later hypotheses relating to culture behaviour at peak biomass concentrations after which cultures entered crash-recovery cycles .
The explanation offered for this phenomenon involved an intricate hypothetical scheme of metabolic events linking culture failure to the ethanol–acetate-shifting phenomenon. The fundamental proposal behind this was that acidification by acetate leads to an increased uncoupling of the proton motive force caused by the increased influx of protons bypassing ATP synthesis [48,52]. Subsequently, it was postulated that the system's compensatory mechanisms detonate the following chain of critical events. Rapid substrate-level phosphorylation via Ack causes a depletion of acetyl-CoA, and this depletion is then exacerbated by the accelerated dissipation of CO to CO2 necessitated by the increased need for reduced ferredoxin in the electron transport chain. This ‘verbal simulation’ of intracellular chaos echoes the hypothesis of hydrogenase inhibition by CO and its potential crash-causing consequences set out by Wang et al. . It is with hypothetical schemes like this that biology exceeds the level of complexity comprehensible to manual deductive reasoning and enters the realm requiring computational analysis, as discussed below.
In summary, recent efforts to understand the metabolism of C. autoethanogenum began with enzymological studies characterising oxidoreductases, establishing the likely stoichiometry of their catalysed reactions and generating hypotheses regarding their place in the wider context of autotrophic metabolism. The metabolic schemes shown in Figures 4–6 show how this information resource can be used to understand metabolic behaviours and to postulate new theoretical modes of metabolism (e.g. formate metabolism, Figure 7), which might now be tested experimentally. Experiments using KO mutants supported two hypotheses regarding their function: first, the proposed functional robustness provided by additional carbon monoxide dehydrogenases and second, the alternative ethanologenic modes of metabolism offered by acetaldehyde : ferredoxin oxidoreductase. In both cases, experimental results were consistent with hypotheses. To characterise the metabolic system in vivo, omics data and genome-scale modelling techniques were used to provide a fuller description of the metabolic system. Subsequent results offered bioprocess strategies for improved ethanol production and led to complicated hypotheses regarding the chain of biochemical events which give rise to cellular behaviour observed in continuous culture.
While enzymology has produced hypotheses testable by the analysis of mutant phenotypes and genome-scale modelling, the mutant phenotype tests have themselves produced hypotheses testable by genome-scale modelling. Closing the systems biology circle, the analyses conducted with GSMs have yielded problems that may be addressed experimentally, or indeed by dynamical modelling. Sequencing efforts combined with enzymological and genetic studies have helped to construct and parameterise GSMs, and will prove important in the construction and parameterisation of kinetic models.
In considering future directions, we note that in C. autoethanogenum, electron-transfer flavoproteins (EtfAB) are not encoded near the MetFV gene, an enzyme with which they are hypothesised to form a complex [12,14,29,30]. This is interesting, because if the genomic location of MetFV and EtfAB is no barrier to their interaction, the possibility of electron-bifurcating (or confurcating) complexes formed by EtfAB and other enzymes must be considered beyond the WLP. In fact, it has already been shown that a complex between LDH and EtfAB enables electron bifurcation and subsequent regeneration of reduced ferredoxin in Acetobacterium woodii . Furthermore, a list of acetogens likely to contain this electron-bifurcating LDH is provided in ref. , to which we would add C. autoethanogenum based on the close proximity of LDHs and EtfAB in the C. autoethanogenum genome sequence [12,14].
Given the potential consequences to an already energetically sensitive system, a comprehensive investigation of potential sites of electron bifurcation should be conducted. The answer to the question of MetFV bifurcation provided in ref. , while elegant may nonetheless be inconclusive; the model in which the inference is made relies on assumptions about the structure of ATP synthase (and thus the efficiency of ATP generation) and the absence of electron-bifurcating modes of energy conservation elsewhere (e.g. LDH). Furthermore, the possibility of reduction of electron carriers other than ferredoxin has not been addressed. These factors conspire to undermine the assertion that NADH + ferredoxin is a mandatory substrate combination for MetFV. One possible future direction is to test the hypothesis, put forward in ref. , that MetFV-mediated bifurcation only takes place in the presence of electron-transfer flavoproteins (EtfAB) and CODH, the latter being required to directly utilise the reduced ferredoxin regained in this mechanism of energy conservation.
Some of the hypotheses covered in this article might be tested effectively with GSMs. For instance, hypothetical proton-dumping by 2,3-BD production could be investigated further with FBA, using carefully chosen objective functions and constraints. While a role for the GapN has been proposed , an amino acid sequence alignment [53,54] reveals that CAETHG_3424 is more similar, in terms of amino acid sequence and amino acid sequence length, to the phosphorylating form (see Supplementary Materials). Moreover, some Gram-positive bacteria [55,56] encode two dedicated phosphorylating GAPDH enzymes, the NAD-specific GapA and the NADP-specific GapB, which are required for glycolysis and gluconeogenesis, respectively. Given the observed induction of the GAPDH encoded by CAETHG_3424 under autotrophic conditions , it appears likely that CAETHG_3424 is functionally equivalent to GapB and its main purpose is gluconeogenesis.
A future development of the analysis from which arginine was identified as a promising growth supplement could be to incorporate actual market cost of arginine in FBA, enabling an assessment of its potential cost-effectiveness in comparison to other medium components. Current prices (sigmaaldrich.com) illustrate this trade-off: 1 kg of arginine costs £403, whereas 1 kg of yeast extract costs £164. Cost analysis of feedstocks with GSMs could be generally beneficial to industrial biotechnology.
Another area requiring attention is the improved parameterisation of Clostridial GSMs  with experimentally determined values. One parameter of particular importance is the composition of the ATP synthase rotor subunit because it is thought to determine the number of protons translocated for the production of 1 ATP (i.e. it determines the stoichiometry of the process). Each c-subunit comprising the ATP synthase rotor translocates 1 proton, and each 120° turn of the rotor enables the production of 1 ATP from ADP and inorganic phosphate . This scheme can be summarised with the following chemical equation:
where nc is the number of c-subunits in the ATP synthase rotor, H+EXT represents external protons, and inorganic phosphate (Pi) has the molecular formula HO4P2− (as given in MetaCyc ). This reaction stoichiometry is simply a lumped version of the reverse ATPase reaction (ADP + Pi + H+ → ATP + H2O) and the proposed rate of proton translocation as determined by the ATP synthase rotor stoichiometry . While the ATP synthases of several bacterial species have been found to contain 11 c-subunits in their rotors, for example Clostridium paradoxum , others have been found to contain different stoichiometries, such as Clostridium thermoautotrophicum for which a rotor consisting of eight to nine c-subunits is likely . This important parameter requires experimental investigation.
Whereas Valgepea et al.  rule out any significant role for the redox-sensitive regulator Rex during a carbon-limited growth experiment based on the constant expression profile of its encoding gene rex (CLAU_1540  and CAETHG_1581 ), we argue that the lack of change in the rate of production of the redox-sensitive regulator Rex would not in itself negate the possibility of its influence because it is an allosterically activated gene regulator . Unchanged expression of rex would indicate its inactivity only if rex genes were autoregulated; this would indicate that no change in Rex-DNA binding had taken place, and that Rex regulation was inactive generally. This could be tested by identifying any Rex binding-sites upstream of Rex-encoding genes/operons. If none exist, Rex regulation of metabolism would remain a strong possibility, especially in a system with high sensitivity to redox poise. A more comprehensive investigation into the role of Rex regulation in C. autoethanogenum will help to resolve this question. Furthermore, an investigation of Rex-regulated metabolism and of general metabolism under non-carbon-limited conditions may be beneficial considering the high rates of gas flow associated with industrial gas sources.
The results reviewed in the present study have inspired a range of hypotheses concerning the dynamics of C. autoethanogenum metabolism. These include possible competition for the substrate of Acs/CODH (CO2) by cooS1 , the deregulation of metabolism caused by the ‘rate-limiting’ nature of ferredoxin oxidation , and dissipation of the PMF (ΔμH+) caused by acidification . Dynamical modelling could be used as an initial test for these hypotheses and to trial competing hypotheses. For example, an alternative to the hypothetical uncoupling of (ΔμH+) by acidification is the simpler ‘overflow’ model proposed by Richter et al.  in which acidification itself induces increased ethanol production by driving the reduction of acetate via AFOR , and that this alone might explain the redistribution of carbon fluxes. The evident need for a formal appraisal of the dynamic behaviours put forward so far in the C. autoethanogenum literature is the key message of this review.
acetaldehyde : ferredoxin oxidoreductase
Clustered Regularly Interspaced Short Palindromic Repeats
flux balance analysis
proton motive force
This work was supported by the Biotechnology and Biological Sciences Research Council [grant numbers BB/K00283X/1, BB/L502030/1, and BB/L013940/1] and the Engineering and Physical Sciences Research Council [grant number BB/L013940/1].
We thank our colleagues at the SBRC for countless helpful discussions. We thank the editors for their patience.
The Authors declare that there are no competing interests associated with the manuscript.