A continuous anaerobic filter bioreactor was set up to investigate the effect of CO₂ supply on isobutyrate chain elongation. The reactor system (35 cm height, an internal column diameter of 6.5 cm, a 1 L working liquid volume, and a headspace of 0.15 L) was exactly the same as in previous research on isocaproate formation via ethanol based chain elongation (de Leeuw et al., 2019), except for the addition of a carrier material. This was done to retain microbial biomass and to increase the rate of chain elongation activity (Grootscholten et al., 2013b). After startup (phase I), the reactor was filled with sponge carrier material (0.15 L liquid exclusion volume of 15 by 15 mm polyurethane cubes; Recticel, Belgium) to support additional growth of biomass (phase II). To maintain anaerobic conditions during this procedure the reactor was flushed with N₂ gas. The addition of cubes changed the active liquid volume of the reactor from 1 to 0.85 L. The influent rate was adjusted accordingly (from 22.2 mlh⁻¹–18.9 mlh⁻¹) to maintain a hydraulic retention time (HRT) of around 45 h. The CO₂ supply was doubled in phase III and halved again in phase IV. An overview of the influent carbon sources, the steady state duration for each phase, the HRT, pH and the CO2 supply for the different phases are listed in Table 1.

The reactor was fed with 650 mmol Carbon L⁻¹ (mMC) isobutyrate, 540 mMC ethanol and 1 gL⁻¹ yeast extract [34 mMC (Duboc et al., 1995)] as carbon sources (acetate was omitted from the influent). The reactor and batch experiments were all done using the same macro- and micronutrient composition (g L⁻¹): NH₄H₂PO₄ 3.60; MgCl₂·6H₂O 0.33; MgSO₄·7H₂O 0.20; CaCl₂·2H₂O 0.20; KCl 0.15. In addition, the micronutrients (Pfennig trace metals and B-vitamins) of the designed basal medium described in Phillips j et al. (1993) was used.

The batch experiments were performed in 250 ml serum bottles (150 ml liquid medium). The remaining 100 ml gas headspace was replaced at the start of each batch up to a pressure of 150 kPa (see gas composition per batch in Tables 1–3). The batch bottles were kept in a shaker at 35°C and 150 rpm throughout the whole experiment. The exact step-by-step protocol for the batches is given in the Supplementary Information, including the recipes for the medium preparation stock solutions (Supplementary Tables S1, S4). All batches were carried out in duplicates.

The first experimental series consisted of eight batches (1.A to 1.H) that aimed to investigate if an enriched microbiome that produced i-C₆ could also elongate branched valerates to branched heptanoates. Ethanol (160 or 320 mMC) and acetate (6.5 or 13 mMC) were always added as substrate, whereas the types of branched valerates were varied throughout the series. In batch 1.A and 1.B a racemic mixture of L/D 2-methylbutanoate was added. I-C₅ (i.e., 3-methylbutanoate) was added in batch 1.C and 1.D. In batch 1.E and 1.F a 50:50 mixture of the L/D 2-methylbutanoate racemate and i-C₅ was added to investigate their combined effect on chain elongation. All these batches were performed at two different substrate concentrations (Table 2), because it was unknown to what degree the (potentially toxic) branched valerates could affect the chain elongation process. BES (2-bromoethanesulfanoate) was added at 10 g/L to inhibit methanogenesis (Vogels et al., 1982), except in the control batch 1.G. Additionally a control batch (1.H) was performed without yeast extract to be able to exclude the possible formation of i-C₇ from yeast extract.

In the first batch series ethanol was not completely converted and it remained unclear if this was caused by the drop in pH, a limiting acetate concentration, the increased hydrogen partial pressure or something else [such as product inhibition on the microbiome (Roghair et al., 2018b)]. Therefore, a second and third series were performed to further investigate the effect of increased hydrogen partial pressure in combination with different starting acetate concentrations. In contrast to the first series that contained no hydrogen at the start of the experiment, the second series was performed with hydrogen already present in the headspace at the start of the batch (20% for all batches, except 2.C which contained 80% H₂ at the start). This was done to minimize acetate formation via potential excessive ethanol oxidation which is thermodynamically inhibited at higher H₂ pressures (Roghair et al., 2018a) and to investigate the effect of an elevated H₂ pressure on the chain elongation itself (Li et al., 2008; Angenent et al., 2016). These batches were all started at pH 7 instead of 6.5 to allow for more proton formation due to ethanol oxidation before pH drops down to limiting levels. When the pH drops below 6.3 it could cause limitations in bicarbonate availability (Tomlinson, 1954; Jungermann et al., 1968). When the pH drops even further and approaches pK values of the carboxylates (∼4.9), undissociated fatty acids concentrations rise which causes additional toxicity effects (Infantes et al., 2012). One batch (2.B) was started with an initial acetate concentration ten times higher than the control (2.A). Additionally, to investigate the necessity of acetate during chain elongation, batch 3.B was started with zero added acetate (3.A as control, in the third batch series).

To batch 2.D i-C₄ was added in addition to i-C₅ to get insight into substrate preferences for branched chain elongation. In the third batch series n-valerate was added (batch 1.D) to compare its utilization as electron acceptor with i-C₅ and exclude possible i-C₇ formation via n-C₅. Tables 3, 4 show overviews of the second and third batch series, respectively. The medium was the same as the medium from the first series, except for the indicated changes in the tables.

The continuous reactor as well as the first batch series was inoculated with a mixture of two anaerobic cultures. One volume part was taken from the continuous reactor that elongated i-C₄ to i-C₆ (de Leeuw et al., 2019) and contained a C. kluyveri dominant culture; second equal volume part came from an undefined mixed bovine rumen sample. The bovine rumen liquid from three cows was provided by the Animal Science Department of Wageningen University and Research. Biomass concentration was not measured within the inocula. The inocula were centrifuged at 4,500 rpm for 10 min after which the cell pellets were resuspended in carbon source free medium prior to inoculation as described within the step-by-step protocol in the Supplementary Information section.

The inoculum for the second batch series was taken from batch 1.D of the first series. Its contents were centrifuged in 50 ml tubes at 4,500 rpm and the pellets were subsequently combined and re-suspended with 50 ml carbon source free medium. These re-suspended cells were then used as inoculum for the second batch series as described within the step-by-step protocol. Similarly, the third batch series was inoculated with biomass that originated from batch 2.D. However, before inoculating, batch 2.D was stored for one and a half year at room temperature. Sporulation of bacteria was observed under the microscope, prior to activation. Before starting the third batch series an activation batch was performed using the same conditions as in batch 2.D. The third batch series was then inoculated with this freshly activated biomass.

Samples of the gas phase were taken once per week and analyzed using an established protocol for gas chromatography to determine the fractions of O₂, N₂, CH₄, H₂, and CO₂ (Steinbusch et al., 2011; Chen et al., 2016). Before sampling the pressure was measured using a pressure meter (GMH 3151). At the same time liquid samples (3.5 ml) were taken, centrifuged at 10,000 rpm and stored in a freezer at −20°C. Every 2 weeks these samples were analyzed according an earlier described gas chromatography method (Agilent 7890B, United States, HP-FFAP column, FID detector) (Jourdin et al., 2018) to determine the concentrations of primary alcohols and volatile carboxylic acids (ethanol, propanol, butanol, iso-butanol, pentanol, branched pentanols, n-hexanol, iso-hexanol and acetate, n-butyrate, iso-butyrate, n-valerate, branched valerates, n-caproate, iso-caproate, n-heptanoate, iso-heptanoate and n-caprylate). The branched valerates i-C₅ and L/D 2-methylbutyrate as well as the branched pentanols isopentanol and L/D 2-methylbutanol could not be distinguished with the available equipment because the isomers exhibited the same retention time. Therefore, the batches were designed to investigate their effect on chain elongation separately to analyze which isomers of branched valerates were used for the formation of which branched heptanoate. The expected forms of branched heptanoates, L/D 4-methylhexanoate and i-C₇, could be distinguished as is shown in Supplementary Figure S1 in the Supplementary Information. For the continuous experiment, the data is presented using the averaged values during each phase and a confidence interval (±) using an α of 0.1. For depiction of the batches, the duplicate results are averaged, and the error bars indicate the differences between each measurement within the duplicate. For the carbon and electron balance calculation (see SI page 5 for equations) yeast extract was assumed to be completely consumed.

The continuous reactor experiment was operated without any acetate in the influent with the intention to maximize i-C₄ utilization during chain elongation and to maximize selectivity towards i-C₆ formation. Compared to a previous study on i-C₆ formation (de Leeuw et al., 2019), the current system achieved a 30% higher volumetric i-C₆ (57 ± 4 mMC/day, or 1.1 ± 0.07 g/L/day) formation rate and a 70% higher average i-C₆ broth concentration (125 ± 6.6 mMC, or 2.43 ± 0.13 g/L) in phase III. During the whole operation period the reactor was operating under apparent CO₂ limited conditions (<1 kPa), meaning that the low availability limits chain elongation activity of well-known chain elongators such as C. kluyveri (Tomlinson and Barker, 1954). When the CO₂ load in phase III was increased, overall chain elongation activity increased (n-C₆ formation increase more than i-C₆ formation). There was a higher (branched) i-C₆ productivity, although selectivity towards i-C₆ had dropped (from 27% in phase II to 20% in phase III). Higher in situ acetate formation, both directly via the chain elongation metabolism and via increased excessive ethanol oxidation led to increased straight chain elongation (see Supplementary Table S5). Homoacetogenesis cannot be fully excluded, however the low CO₂ availability limits homoacetogenic activity that requires CO₂ as electron acceptor.

The reactor behavior shows there is a tradeoff to be made when designing the system: 1) selectivity towards i-C₄ elongation is high during acetate and CO₂ limitation (which reduces overall chain elongation activity), or 2) straight chain elongation is stimulated by lifting the CO₂ limitation leading to a decreased selectivity towards alternative carboxylate elongation. In all phases i-C₄ was abundantly available, while acetate was only available via in situ formation. The sensitivity to increases in acetate show that there is a preference towards acetate as electron acceptor over i-C₄ (and i-C_5, in the batches) within the established chain elongation microbiome. In the third batch series it was observed that CO₂ was consumed down to low partial pressures (<1 kPa), while more H₂ was consumed, and overall a higher percentage of ethanol was utilized for chain elongation compared to the previous batches; it suggests the microbiome had developed increased homoacetogenic activity.

The degree by which i-C₅ and i-C₄ are elongated in a batch system varied. Formation of i-C₇ contributed only 4.2% (based on carbon atoms) to the total produced compounds in the batch with both i-C₅ and i-C₄ (2.D). In contrast, i-C₆ formation contributed for 27% to the total product spectrum, even though the molar concentration of i-C₅ was higher than i-C₄. With the L/D 2-methylbutanoate racemate batches no elongation product was observed at all and overall, the chain elongation rate diminished. Moreover, in the batch experiments a higher acetate availability negatively influences the selectivity towards branched chains, like what was observed in the continuous reactor. This is emphasized by the batches performed at 65 mMC and 6.5 mMC initial acetate. A higher initial acetate concentration (batch 2.B) increased total chain elongation activity, but considerably lowered the selectivity towards i-C₇ (1.4%) compared to the control (7.3%) at low initial acetate amounts (batch 2.A). The results suggest the microbiome contains enzymes that can perform branched carboxylate elongation, but only to a certain degree. The varying selectivities can arise from two different scenarios: 1) acetate limitation and 2) no limitation (illustrated in Supplementary Figure S12). Hypothetically, the higher i-C₇ selectivity during acetate limitation (2.A) can arise from kinetic impairment of acetate elongation at low acetate concentrations, while branched carboxylate elongation occurs of maximum speed. In contrast, at higher acetate concentrations (2.B) the alleviated kinetic impairment leads to more acetate elongation relative to branched chain elongation, causing a lower i-C₇ selectivity.

The initially available acetate (6.5 mMC in 2.A versus 65 mMC in 2.B and 13 mMC in 3.A versus 0 mMC in 3.B) in the batch series greatly affected the time it took for chain elongation to occur. These results are in line with earlier studies that show a reduced chain elongation activity during acetate limitation (Spirito et al., 2018). Interestingly, all batches, except for 2.A where 65 mMC acetate was added, showed that first acetate formation started, before any chain elongation commenced (Supplementary Figures S4, S6, S7, S9). Despite presence of sufficient alternative electron acceptors, a minimum amount of acetate seems to be required for chain elongation to occur. The requirement of acetate hints towards a cometabolism for the branched electron acceptors within chain elongation, i.e., branched carboxylates are only elongated during straight chain elongation.

The measured alcohol concentrations during the continuous reactor experiment followed a dependency on the concentrations of ethanol and acetate as well as on the concentration of the corresponding carboxylates (Figure 4A). This finding was in line with the earlier study where i-C₆ and alcohol (i-C₄OH, n-C₆OH and i-C₆OH) formation were found (de Leeuw et al., 2019). It suggests that excessive (hydrogenogenic) ethanol oxidation is coupled to (hydrogenotrophic) carboxylate reduction within the microbiome as shown in Table 6, resulting in a net carboxyl-hydroxyl exchange reaction. A coupling of reactions would imply that the thermodynamics driving force is no longer affected by pH and hydrogen partial pressure (p_H2), in contrast to hydrogenotrophic carboxylate reduction to alcohols that is favored at a lowered pH and an elevated p_H2 (See Supplementary Figure S2 for the p_H2 in the continuous reactor) (Steinbusch et al., 2008). Figure 4B shows that after startup the ΔrG¹ (corrected for measured concentrations) of the combined reactions for each carboxylate—alcohol pair (when correcting for the broth concentrations of the reactants and products) remained between 15–25 kJ per reaction. This value is close to the currently known minimum required energy gain for a catabolic reaction to sustain microbial growth (Jackson and McInerney, 2002), and suggests that this bioconversion could be utilized as energy-providing route by organisms growing in a specific niche. It still needs to be revealed which organism(s) play(s) a role in this alcohol formation.

Possibly chain elongation microorganisms themselves are solely responsible for the formation of the longer alcohols. It is reported that Clostridium kluyveri, a well-known chain elongator, is able to produce small amounts of higher alcohols (Kenealy and Waselefsky, 1985). A batch series performed using pure Clostridium kluyveri with propionate and ethanol under different hydrogen pressures showed that propanol formation increased with an increasing p_H2 (Candry et al., 2020). The alcohol formation during acetate limitation in combination with a high p_H2 could hypothetically be method to get rid of excess electrons when chain elongation-coupled ethanol oxidation is thwarted due to high hydrogen partial pressures. Carboxylate reduction then replaces hydrogen formation as electron sink.

Alternatively, another specialized organism performing the hydroxyl-carboxyl exchange could be present. It would require an organism similar to Clostridium autoethanogenum (Diender et al., 2016) that can harvest the energy from ethanol-derived electrons via an energy-coupled transhydrogenase (Rnf complex) (Westphal et al., 2018) before reducing the carboxylates to alcohols. A second alternative would be syntrophic interaction between ethanol oxidizers and “hydrogenotrophic” carboxylate reducers [via H₂ exchange and/or Direct Interspecies Electron Transfer (DIET) (Nozhevnikova et al., 2020)]. Although the thermodynamic calculations performed with macroscopic data show that the hydrogenotrophic carboxylate reduction is often unfeasible (Figure 4B), a syntrophic coupling of ethanol oxidation and carboxylate reduction would imply that the actual microscopic conditions are such that both (in syntophy-growing) microorganisms are able to proliferate (Smith and McCarty, 1989).

The gradual increase of i-C₄OH formation during phase III and the increased alcohol formation in phase IV compared to phase II indicate that this additional bioconversion capability had slowly become more prominent within microbiome. Consequently, the p_H2 did not recover in phase IV to the earlier values in phase II (9.2 ± 1.3 kPa) after reducing the CO₂ load; it was kept in a lower range (0.6 ± 0.4 kPa in phase IV) by the microbiome, while alcohol formation spiked. The alcohol formation likely acted as an alternative electron sink when methane formation had dropped due to the sudden lower availability of CO₂, as was also observed previously (de Leeuw et al., 2019).

The onset of the alcohol formation implies that the earlier achieved high selectivity towards i-C₆ in phase II could be transient. A low acetate concentration is used as steering parameter in this research to increase the selectivity towards i-C₄ elongation. However, in combination with high ethanol and high other carboxylate amounts, a low acetate concentration leads to a thermodynamic potential that allows an alternative source of in situ acetate formation via hydroxyl-carboxyl exchange.

Chain elongation microbiomes can be engineered to produce various chemicals depending on the supplied feedstock and steered reactor conditions. The higher branched and straight alcohol formations described in this research seem to be thermodynamically dependent on the reactant to product ratio. If the products could be removed in situ this could drive the reaction towards more straight and branched alcohol formation. Increasing the alcohol formation in this way could lead to an interesting biochemical production process in itself; the observed alcohol titers are in a suitable range for in situ extraction via gas stripping (Richter et al., 2016). This method may be used to develop processes that upgrade the ethanol in dilute ethanol-containing residue streams to higher alcohols.

Branched carboxylates such as i-C₄ and apparently also i-C₅ can be used as electron acceptor during chain elongation fermentations with a varying extend of conversion. Operating the reactor under acetate and CO₂ limited conditions increases the selectivity towards branched carboxylate elongation, but as a tradeoff overall chain elongation activity is reduced. The conversions of branched carboxylates to longer chains seem only to occur as a form of co-metabolism during straight chain elongation. It remains to be seen if the co-metabolism, that is expressed as a dependency on straight chain elongation activity, can be lifted. Acetate plays a pivotal role within the chain elongation metabolism as it can both serve as a primer and elongation (acetyl-CoA) unit (Schoberth and Gottschalk, 1969). However, research has already shown that it is possible to increase the affinity of butyrate relative to acetate for an engineered thiolase (Bonk et al., 2018), as well as first efforts to modify thiolases to use branched carboxylates as primers (Clomburg et al., 2018). Further efforts to tailor the thiolase and other involved enzymes via metabolic engineering could offer perspectives where the k_cat and K_m values for branched carboxylates and their conversion intermediates are increased. This metabolic engineering approach would require pure or coculture cultivation of the responsible bacteria, in contrast to the presented research. Aside from bioconversion kinetics, the metabolic fluxes could also have been hampered by uptake and export limitations.

Production of i-C₇ in the observed amounts in this study at this stage are unattractive for direct industrial applications compared to the formation of n-C₆. Still with fractional distillation of the produced broth considerable amounts of i-C₇ may be obtained. In the batch experiments i-C₅ is hardly elongated (∼98% remains unconverted) in the cases where it is supplied in excess and acetate is only present in low amounts. So far, it is remarkable that i-C₄ elongation has different kinetics compared to i-C₅ elongation. It shows that the involved microbiome has not developed fully optimized enzymes for the artificially imposed selective pressure with low amounts of acetate and large amounts of alternative electron acceptors. In addition to metabolic engineering approaches further research on selection pressure and natural adaptation in open culture microbiomes can provide a potential i-C₇ bioprocess development utilizing organic residual streams.