Unexpected metabolic rewiring of CO₂ fixation in H₂-mediated materials–biology hybrids

Significance

Employing renewable electric energy to power material–microbe hybrids for chemical synthesis has emerged as a promising approach for a sustainable society. When the materials and microbes are not in physical contact with each other, it is commonly assumed that the materials component that facilitates an electron transfer mediated by redox molecules like H₂, does not serve to perturb microbial metabolism significantly. However, this study revealed that the electrochemical system can induce a fortuitous metabolic rewiring in planktonic S. ovata cells and an increased efficiency of utilizing provided reducing equivalents for CO₂ fixation. This observation underscores the importance of revisiting existing assumptions of materials–biology interaction and adopting a more holistic approach to understand the underlying mechanisms at material-microbe interfaces.

Abstract

A hybrid approach combining water-splitting electrochemistry and H₂-oxidizing, CO₂-fixing microorganisms offers a viable solution for producing value-added chemicals from sunlight, water, and air. The classic wisdom without thorough examination to date assumes that the electrochemistry in such a H₂-mediated process is innocent of altering microbial behavior. Here, we report unexpected metabolic rewiring induced by water-splitting electrochemistry in H₂-oxidizing acetogenic bacterium Sporomusa ovata that challenges such a classic view. We found that the planktonic S. ovata is more efficient in utilizing reducing equivalent for ATP generation in the materials–biology hybrids than cells grown with H₂ supply, supported by our metabolomic and proteomic studies. The efficiency of utilizing reducing equivalents and fixing CO₂ into acetate has increased from less than 80% of chemoautotrophy to more than 95% under electroautotrophic conditions. These observations unravel previously underappreciated materials’ impact on microbial metabolism in seemingly simply H₂-mediated charge transfer between biotic and abiotic components. Such a deeper understanding of the materials–biology interface will foster advanced design of hybrid systems for sustainable chemical transformation.

The hybrid approach that integrates microbial catalysts with synthetic electrocatalysis has gained momentum in recent years for a variety of chemical transformations in the context of a sustainable future (1–6). Potentially powered by renewable energy such as solar electricity, such bioelectrochemical materials–biology hybrids have been deployed for the sustainable production of valuable commodity chemicals (7–9) and efficient remediation of environmental pollutants (10–12). In those systems, the material components serve as the sole electron donor and transduce electricity to chemical driving forces powering microbial metabolisms for desired applications under the electroautotrophic conditions (10, 13). The materials’ efficient charge transfer and the microbes’ synthetic versatility and selectivity synergistically coalesce at the materials–biology interactions, leading to high efficiencies and reaction throughput for challenging chemical transformations such as the fixation of CO₂ and N₂ (5, 14–19). Hence, a detailed understanding of the mutual interactions between the biotic and abiotic components (20–22) is of critical importance for the development of sustainable techniques addressing the challenges that our society is facing.

The conventional wisdom of the materials–biology hybrids assumes that while a direct physical contact of the biotic and abiotic components may affect beneficial synergies on both components (3, 6, 13), an indirect charge transfer via mediated redox couple may lead to minute if any changes on individual components’ functionality. Indeed, in the research of bioelectrochemical fixation of CO₂ when H₂ generated by electrochemical proton reduction is the mediating redox species (23, 24), H₂-oxidizing chemolithotrophic Ralstonia eutropha (7, 25, 26), diazotrophic Xanthobacter autotrophicus (8, 14, 27), methanogenic Methanosarcina barkeri (28, 29), and acetogenic Sporomusa ovata (15, 30–34) were readily plugged into water-splitting electrocatalytic systems assuming no metabolomic responses and gene regulations due to the perceived indirect charge transfer in the hybrids. However, the hypothesized invariance of microbial metabolism in those H₂-mediated electroautotrophic systems has not been tested. As recent studies at the physical interface between light-absorbing semiconductors and CO₂-fixing organisms uncover unexpected material-induced metabolomic modulation (16, 35), the potential risk of understating metabolic rewiring in materials–biology hybrids with indirect charge transfer looms large, due to the absence of a thorough examination of microbial metabolism in those systems. A detailed, appropriate understanding of how microbial metabolism responds to materials’ presence in H₂-mediated electroautotrophic systems will facilitate the design and development of advanced hybrids in the future.

We aim to rigorously test the hypothesized invariance of microbial metabolism in H₂-mediated materials-biology hybrids and hence shed more lights on the fundamental interactions between the abiotic and biotic components during the mediated indirect charge transfer (Fig. 1A). Built upon our previous experience (7, 14, 15, 31), the model H₂-mediated hybrid system that we aim to study includes H₂-oxidizing S. ovata for the fixation of CO₂ into acetic acid via the Wood–Ljungdahl pathway (36), as well as biocompatible water-splitting electrocatalysts made of cobalt–phosphorous (Co–P) alloy for hydrogen evolution reaction (HER) (Fig. 1B) (7). Previous studies have established a H₂-mediated delivery of reducing equivalents from the electrode to planktonic S. ovata under sufficiently high water-splitting current densities (31, 37). Yet, it remains unknown whether the electrochemical process(es), seemingly innocent and only responsible for H₂ generation, will perturb microbial metabolism beyond the delivery of H₂.

Fig. 1.

Enhanced CO₂ fixation in H₂-mediated materials-biology hybrid. (A) Schematic comparison of S. ovata growth and electron efficiency (e^– efficiency) under electroautotrophic and chemoautotrophic conditions. (B) Scanning electron microscope image of Co–P alloy cathode for the electrocatalytic hydrogen evolution reaction (HER). (Scale bar, 10 µm.) (C) Current density–Voltage (j–V) characteristics of Co–P cathode before and after a 48-h electrolysis with a scan rate of 50 mV/s in electroautotrophic culture. SHE, standard hydrogen electrode; iR compensated. (D and E) The amounts of electrons consumption from respective electron donors and recovered from CO₂ fixation in S. ovata cultures under electroautotrophic and chemoautotrophic conditions in 72 h, before correcting for the carryover extracellular nutrients (D and E, are Entry 1 and 3 in the SI Appendix, Table S1, respectively). The displayed e^– efficiencies in (A) are the corrected ones after subtracting the carryover extracellular nutrients (Materials and Methods). Data represent biological triplicates (n = 3, same below). Error bars represent the SEM, the same below.

In this work, we report the unexpected finding that the seemingly innocent water-splitting electrocatalysis induced observable and significant metabolomic rewiring of planktonic S. ovata in the materials-biology hybrids of CO₂ fixation. We found a higher efficiency for microbes to utilize the generated H₂ for CO₂ fixation in the hybrid system (large than 95%), than the values observed (less than 80%) when S. ovata was cultured chemoautotrophically with externally supplied H₂ under otherwise the same experimental conditions (Fig. 1A). Metabolomic and proteomic studies comparing the electroautotrophic and chemoautotrophic cultures unraveled fortuitous metabolomic rewiring including membrane electrochemical potential regulation and enhanced proton transmembrane transport in the hybrid systems, leading to an increased adenosine triphosphate (ATP) generation and an accelerated rate of CO₂ fixation. Extracellular metabolomic data hint a process of metabolic rewiring potentially mediated by extracellular redox-active metabolites (RAMs) such as riboflavin. Our observations challenged the conventional simple picture of H₂-mediated indirect charge transfer and suggested the presence of previously underappreciated metabolic impact of inorganic electrodes on microbial catalysts. Such an advance in fundamental understanding at the materials–biology interactions heralds additional design considerations for a sustainable chemical transformation that maximizes the benefits of materials–biology hybrids.

Results

Enhanced CO₂ Fixation in H₂-Mediated Electroautotrophic Hybrid.

H₂-mediated bioelectrochemical fixation of CO₂ was established in a two-chamber setup under a flowing gas environment of CO₂:N₂ (20:80, v:v, 6 mL min^–1) at 30 °C (SI Appendix, Fig. S1). In the cathode chamber of such setup, a biocompatible Co–P alloy HER electrocatalysts developed in our previous studies (Fig. 1B and SI Appendix, Fig. S2 A and B) (7, 14, 31, 38), composed of a P:Co weight ratio of 1:4.4 as characterized by energy-dispersive X-ray spectroscopy (SI Appendix, Fig. S2C), was applied to couple with acetogenic bacterium S. ovata strain H1 (DSM 2662) (39) for the reduction of CO₂ into acetic acid. Reported to yield H₂ with 100% Faradaic efficiency (38), Co–P alloy is electrocatalytically active for HER at least over 48 h based on the measured cyclic voltammograms (Fig. 1C) and is known to not inhibit the growth or CO₂ fixation of S. ovata (31). Over 48 h, a minute change of less than 0.04 V in terms of onset potentials, measured with –0.5 mA/cm² as the cutoff current density, suggest robust HER electrocatalysis in biological medium. The minimal medium deployed in the process does not contain any carbon substrates, except the dissolved bicarbonate (NaHCO₃), trace vitamin additives, and the minute amounts of cysteine as the reducing reagent (0.2 mM) for an anaerobic solution (SI Appendix, Supplementary Note 1). Two-week control experiment inoculated with S. ovata in CO₂:N₂ (20:80, v:v) but without the provision of either H₂ or electricity did not yield observable bacterial growth or acetate accumulation from CO₂ fixation (SI Appendix, Fig. S3). Hence, any observation of bacterial growth or acetate accumulation originates from the provision of H₂ or electricity for CO₂ fixation.

We report that S. ovata under electroautotrophic conditions in the bioelectrochemical setup is nearly 100% efficient in utilizing reducing equivalents and fixing CO₂ into acetic acid by repeated independent tests (reported as different entries in SI Appendix, Table S1). Chronoamperometry measurements for the materials–biology hybrid at –1.34 ± 0.10 mA/cm² (Entry 1, n = 3, same below) equivalent to about –0.77 to –0.96 V vs. standard hydrogen electrode (SHE). In such a flowing CO₂/N₂ condition, the H₂ supply rate is 2.40 ± 0.19 millimole per day (mmole d^–1), equivalent to an e^– supply rate of 4.80 ± 0.37 mmole d^–1, while the acetate is the almost exclusive soluble organic end product and accumulation rate of acetate was determined as 0.67 ± 0.08 mmole d^–1 based on the measurement of ¹H NMR spectroscopy (Entry 1 in SI Appendix, Table S1). Such an observed level of acetate production rate is at least comparable to our previous report (31) and other reported bioelectrochemical systems deploying acetogens for CO₂ fixation (SI Appendix, Table S2). In two independent tests, the apparent electron efficiency η_app,electro, or the apparent Faradaic efficiency that describes the percentage of electrons contributing to CO₂ reduction into acetic acid, was calculated to be 112 ± 6% and 110 ± 9% (Entry 1 and 2 in SI Appendix, Table S1, respectively), based on the trajectory of acetate accumulation sampled every 24 h (Fig. 1D). The higher-than-100% η_app,electro is proposed to originate from the residual extracellular oxidizable organics carried over during bacterial inoculation (40), since the chemical oxygen demand (COD) of the extracellular component of the S. ovata inoculum at the beginning of experiments was measured to be 1.99 ± 0.85 mmole of electron equivalents. Accounting the contribution of residual extracellular oxidizable organics (SI Appendix, Fig. S4), the corrected e⁻ efficiency of electroautotrophic CO₂ fixation η_electro = 99 ± 4% and 95 ± 10% in two independent tests (Entry 1 and 2 in SI Appendix, Table S1, respectively). After the three-day electroautotrophy, the visual inspection did not observe any biofilms on the Co–P cathode’s surface, nor did images of scanning electron microscopy. There was 30 ± 14 µg of protein from S. ovata, a proxy of biomass hence bacterial cell numbers (41), on the electrode, in comparison to the 3,851 ± 147 µg of protein in total from the planktonic cells in the 100-mL reactor. Image analysis of the scanning electron microscopy estimated that a small percentage of the electrode surface, on the order of 0.03 cell μm⁻², included attached bacterial cells (SI Appendix, Fig. S5). We estimated that the bacterial cells with direct materials contact, necessary but not sufficiently indicative of a direct electron transfer (42), constituted no more than 1% of the predominantly planktonic bacterial cell cultures. Thus, our observations support a near-unity of e⁻ efficiency for H₂-mediated materials–biology hybrids of CO₂ fixation.

Surprisingly, the e⁻ efficiency is lower for the microbes grown under chemoautotrophic conditions, even though both electroautotrophic and chemoautotrophic conditions utilize H₂ as the direct reductant for the CO₂-fixing through the Wood–Ljungdahl pathway (36). The headspace of chemoautotrophic S. ovata culture was intermittently dosed with mixed gas (H₂:CO₂, 80:20, v:v) every 12 h including 2 mmole H₂ substrate (i.e., e^– supply rate of 8 mmole d^–1) whose amount was higher than the ones in bioelectrochemical experiments (4.80 ± 0.37 mmole e^– d^–1). However, based on temporal trajectory of acetate accumulation, the acetate production rate in chemoautotrophic S. ovata culture is 0.48 ± 0.02 mmole d^–1 (Fig. 1E and entry 3 in SI Appendix, Table S1), which was lower than 0.67 ± 0.08 mmole d^–1 in electroautotrophic cultures (Entry 1 in SI Appendix, Table S1). Quantifying H₂ consumption in the headspace via gas chromatography established the mass balance of reducing equivalents and led to the averaged apparent e⁻ efficiency η_app,chemo of 81 ± 4% (Fig. 1E). The corrected e⁻ efficiency η_chemo of 71 ± 4% after subtracting the contribution of extracellular carried-over nutrients (see above) is significantly lower than the η_electro of 99 ± 4% in the bioelectrochemical case (Entry 1 and 3 in SI Appendix, Table S1). Meanwhile, a lower protein yield, a good surrogate of biomass accumulation (16, 43), was observed for chemoautotrophic S. ovata (14.9 ± 1.0 μg protein mL⁻¹) than the electroautotrophic ones (20.9 ± 2.1 μg protein mL⁻¹) (SI Appendix, Table S1). A small amount of ethanol, ~0.1 mM and ~0.5 mM at 72 h and 120 h, respectively, was observed as another electron sink in chemoautotrophic cultures but not electroautotrophic ones (SI Appendix, Fig. S6). Overall, our results suggest that the electroautotrophic S. ovata is more efficient in utilizing reducing equivalent H₂ for the acetate production and biomass accumulations.

Additional experiments, originally meant as controls to study the impacts of H₂ delivery conditions, allude to different metabolic conditions between chemoautotrophy and H₂-mediated electroautotrophy. Instead of intermittently dosing S. ovata with 80% H₂ at 1 bar pressure that strongly favors CO₂ fixation thermodynamically (44), we aimed to mimic the mass transport and delivery amount of H₂ in electroautotrophy and conducted chemoautotrophic experiments in the same bioelectrochemical setup, with a continuous gas flow of 1% H₂ (H₂:CO₂:N₂, 1:20:79, v:v:v) whose H₂ mass transport and delivery amount resemble if not replicate the ones in electroautotrophic conditions (Entry 5 and 6). Unexpectedly, minimal bacterial CO₂ fixation hence biomass accumulation was observed after multiple attempts (SI Appendix, Fig. S7, Entry 5 and 6 in SI Appendix, Table S1), different from the chemoautotrophic growth under 80% H₂ (Entry 3 and 4 in SI Appendix, Table S1). Multiple attempts of chemoautotrophic culturing under 1% H₂ in the batch mode, which observed the active bacteria growth under 80% H₂, also failed to yield detectable bacterial CO₂ fixation (Entry 7 in SI Appendix, Table S1). These observations suggest that the low H₂ partial pressure (1% H₂) disfavoring CO₂ fixation thermodynamically (44) has a significant metabolic impedance of chemoautotrophy, despite the fact that robust microbial CO₂ fixation with high e⁻ efficiency was observed in electroautotrophy. Our observations lead us to hypothesize that metabolic stimulation and modulation of S. ovata are present in the seemingly simple H₂-mediated electrochemical system, warranting detailed comparative, integrated metabolomic, and proteomic studies for a deeper fundamental understanding of the hybrid system.

Observed Metabolomic Shift in Planktonic S. ovata Cells Under Electroautotrophic Conditions.

There are significant metabolic profile differences between the electroautotrophic and chemoautotrophic planktonic S. ovata. To understand the difference, we extracted and measured intracellular metabolites with triplicates (n = 3) from the electroautotrophic and chemoautotrophic cultures. Since chemoautotrophic S. ovata under 1% H₂ in the bioelectrochemical setup were not active in acetate production, only S. ovata active of CO₂ fixation under electroautotrophic and chemoautotrophic conditions with 80% H₂ were collected for subsequent metabolomic and proteomic analyses (Entry 1 and 4 in SI Appendix, Table S1, respectively). Out of more than 400 intracellular metabolites involving the majority of metabolic pathways that we searched for, a total of 118 metabolites were measured in both electroautotrophic and chemoautotrophic cultures and 11 metabolites were exclusively detected in electroautotrophic one (Dataset S1). The visualization of the principal component analysis (PCA) with the first two components (Fig. 2A), accounting for a combined 77.7% of variance, shows that replicate cultures grown under the same conditions were grouped together and distant from cultures grown under different conditions, indicating the significant metabolomic difference between electroautotrophic and chemoautotrophic cultures.

Fig. 2.

Metabolomic shift in planktonic S. ovata cells induced by electroautotrophy. (A) The principal component analysis (PCA) of intracellular metabolomes extracted from S. ovata cells grown under both electroautotrophic (gray circle) and chemoautotrophic (gray triangle) conditions. Each symbol represents one of three biological replicates. (B and C) Heatmaps visualizing abundance changes of metabolites involved in the Wood–Ljungdahl pathway including redox cofactors, energy cofactors, and CO₂ fixation intermediates. (D and E) Abundances and tandem mass spectra of intracellular acetyl-glutamate (D) and intracellular riboflavin (E). Riboflavin was also detected extracellularly (Dataset S1). Panels (B, D, and E) share the same color legend which is shown in panel (B). Numerical results for metabolite abundance are shown in Dataset S1. *, significant difference between conditions (fold change ≥ 2 and P < 0.05); n.d., not detected; NAD⁺ and NADH, Nicotinamide adenine dinucleotide; NADP⁺ and NADPH, nicotinamide adenine dinucleotide phosphate; AMP, adenosine monophosphate; GMP, guanosine monophosphate; ADP, adenosine diphosphate; GDP, guanosine diphosphate; ATP, adenosine triphosphate (ATP); CH-THF, methenyl-tetrahydrofolate; CH₃-THF, methyl-tetrahydrofolate; acetyl-CoA, acetyl coenzyme A; FMN, flavin mononucleotide; FAD, flavin adenine dinucleotide.

Our metabolomic study suggests increased redox cofactor pools in the materials–biology hybrids for CO₂ fixation. The redox cofactors including nicotinamide adenine dinucleotide in both oxidized (NAD⁺) and reduced (NADH) forms, and oxidized nicotinamide adenine dinucleotide phosphate (NADP⁺) all display higher abundance in S. ovata cells grown bioelectrochemically, while the concentration ratio between NADH and NAD⁺ is 1.5-fold higher in electroautotrophic cells than chemoautotrophic ones (Fig. 2B). Such an increase of redox cofactor pool and a more reduced, electron-rich intracellular status may promote a faster turnover of redox-based metabolic activities (45) and a larger driving force for CO₂ reduction (46).

We also analyzed nucleotides as energy cofactors and found out that abundances of adenosine diphosphate (ADP) and adenosine monophosphate (AMP) are threefold higher in electroautotrophic cells than chemoautotrophic cells. Meanwhile, ATP, guanosine diphosphate (GDP), and guanosine monophosphate (GMP) were exclusively detected in electroautotrophic cells (Fig. 2B) possibly owing to their lower abundances in chemoautotrophic conditions, molecule instability, and the extended static period required for LC–MS-based metabolomic analysis. Luminescence-based ADP/ATP ratio analysis was conducted and ADP/ATP ratios in electroautotrophic and chemoautotrophic cells are 7.7 ± 0.3 and 9.3 ± 0.5, respectively. The generally high value of ADP/ATP ratio, a lower state of adenylate energy charge, is consistent to the values observed in acetogenic Acetobacterium woodii in chemoautotrophy (47), indicative of scarce ATP generation and the thermodynamic limitation in the generally slow-growing acetogens. Yet, the increased pool of nucleotide cofactors and lower ADP/ATP ratio observed in electroautotrophic cells than the chemoautotrophic ones imply a more active cellular metabolism under electricity supply (48).

Moreover, the increased cofactors abundance in H₂-mediated electroautotrophic S. ovata is consistent with the abundance changes of key intermediates in the Wood–Ljungdahl pathway (36). Both electroautotrophic and chemoautotrophic samples exhibit detectable amounts of methenyl-tetrahydrofolate (CH-THF) and methyl-tetrahydrofolate (CH₃-THF), hence affording the expression of Wood–Ljungdahl pathway (Fig. 2 B and C). Yet, acetyl coenzyme A (acetyl-CoA), the key intermediate before the generation of ATP and acetic acid via substrate-level phosphorylation, was only detectable in chemoautotrophic cells. The undesired accumulation of acetyl-CoA in chemoautotrophic cells hints that bacterial cells in the materials–biology hybrid are more efficient in converting acetyl-CoA to acetate for ATP production, a limiting factor of Wood–Ljungdahl pathway (36), consistent with the lower ADP/ATP ratio in electroautotrophy than chemoautotrophy (7.7 ± 0.3 vs. 9.3 ± 0.5) and a faster kinetic of ATP production. Those results corroborate a CO₂-fixing mechanism in electroautotrophy not only more active but also more energy efficient in the context of ATP generation.

There are two more noteworthy metabolomic differences between electroautotrophic and chemoautotrophic cells. First, one of the largest abundance changes of intracellular metabolites is the 150-fold to 500-fold increase of N-acetyl-amino acids, such as acetyl-glutamate, acetyl-ornithine, and acetyl-lysine, in the electroautotrophic cells (Fig. 2D). The magnitude of such abundance change, in conjunction of their possible implication towards the osmotic protection (49), offers some intriguing clues toward a hypothesized shift of membrane electrochemical potential (Δμ_proton) in the materials–biology hybrid (see below). Second, we attempted extracting the extracellular metabolites in both scenarios and searching for additional extracellular redox mediators. Intracellularly, RAMs including riboflavin (Fig. 2E), flavin adenine dinucleotide (FAD), and flavin mononucleotide (FMN) (SI Appendix, Fig. S8) were detected under both electroautotrophic and chemoautotrophic conditions. Yet, only riboflavin was also detected extracellularly under both conditions, albeit at a much diluted concentration on the order of nmol/L. Comparison of riboflavin abundance suggests that there is no statistically significant difference for extracellular riboflavin between electroautotrophy and chemoautotrophy (Dataset S1), while chemoautotrophy witnessed a higher abundance of intracellular riboflavin (Fig. 2E). The extracellular concentration of riboflavin is too low to contribute significantly if anything to the delivery of reducing equivalent and CO₂ fixation but may serve as a signal molecule toward an altered microbial metabolism in the H₂-mediated materials–biology hybrids.

Materials-Induced Differential Gene Expression during Electroautotrophic Conditions.

A total of 1,993 proteins of S. ovata were detected in chemoautotrophic and electroautotrophic cultures (detailed annotations and abundances are available in Dataset S2), among the annotated S. ovata DSM2662 genome encoding 5,103 proteins (Uniprot Proteome ID: UP000015521) (50). In addition, 1,697 proteins were stably expressed (fold change < 1.5) under both conditions (Fig. 3A). Among the five hydrogenases that were annotated in the genome of S. ovata (51), one [NiFe] hydrogenase and two [Fe] hydrogenases were detected with similar level of expression. Such results support H₂ as the primary energy source for microbial CO₂ fixation under both chemoautotrophic and H₂-mediated electroautotrophic conditions and that the rate of H₂ delivery should not contribute significantly to the observed difference in e⁻ efficiency. All the enzymes involved in the Wood–Ljungdahl pathway including formate dehydrogenase (FdhA), formyl-THF synthase (Fhs), methenyl-THF cyclohydrolase (FchA), methylene-THF dehydrogenase (FolD), methylene-THF reductase (MetFV), CO dehydrogenase/Acetyl-CoA synthase (AcsABCDEFV and CooC), and acetate kinase (AckA) were all stably expressed in both electroautotrophic and chemoautotrophic cells (Fig. 3 B and C). The lack of statistically significant changes in abundance indicates that microbial capability of CO₂ fixation into acetic acid remained unperturbed in the materials–biology hybrid. Such result also suggests that other factors connected to but not within the Wood–Ljungdahl pathway, such as ATP generation as one key limiting factor for acetogenic organisms (see above) (52), are potentially responsible for the observed difference in e⁻ efficiency (Fig. 1 D and E).

Fig. 3.

Electrochemical system–induced differential gene expression in S. ovata cells. (A) Volcano plot demonstrating detected proteins in electroautotrophic and chemoautotrophic cells. The differentially expressed proteins between electroautotroph/chemoautotroph and chemoautotroph/electroautotroph were determined by fold change ≥ 1.5 and P < 0.05. The number in color represents the number of proteins determined in the associated category. (B) Expression of gene clusters of interest. The Wood–Ljungdahl cluster includes genes for methenyl-THF cyclohydrolase (fchA), methylene-THF dehydrogenase (fold), CO dehydrogenase/acetyl-CoA synthase (cooC and acsABCDEFV), formyl-THF synthetase (fhs), heterodisulfide reductase (hdrAC), and methylene-THF reductase (metFV) involved in the Wood–Ljungdahl pathway. The ATPase cluster contains genes for ATP synthase (atpADGH). The Rnf cluster includes genes encoding ferredoxin:NAD⁺ oxidoreductase (rnfABCDEG). The Cyt cluster contains genes for cytochrome bd (cydAB). (C) The involvement of the expressed proteins in the Wood–Ljungdahl pathway and the speculative ATP production pathway. Numerical results for protein abundance are shown in Dataset S2. Ferredoxin (Fd).

We also observed significant differences (fold change ≥ 1.5 and P < 0.05) of gene expressions in some key proteins that hints a favored membrane-based proton translocation and ATP generation in S. ovata grown electroautotrophically. Differential gene expression was observed that 102 proteins were up-regulated in S. ovata cells grown with electrochemical system while 194 proteins were up-regulated in cells grown under H₂-containing headspace (Fig. 3A). Notably, the subunit of ATP synthase (AtpH), the enzyme responsible for ATP generation based on proton translocation (53), is up-regulated in the materials–biology hybrid (Fig. 3B). This corroborates an enhanced ATP production under the bioelectrochemical condition as observed in metabolomic analysis. Meanwhile, ferredoxin:NAD⁺ oxidoreductase (Rnf) and heterodisulfide reductase subunit C (HdrC), the two enzymes that are reported to sequentially translocate H⁺ and generate transmembrane proton gradient by a series of membrane-bound electron transfer (54–56), are either stably expressed for Rnf or up-regulated for HdrC. Such information suggest that the establishment of transmembrane H⁺ gradient is favored in the electroautotrophic condition. Such an argument is corroborated by the observed upregulation of cytochrome bd ubiquinol oxidase (Cyt bd) subunit A (Cyd A) and several proteins neighboring the same gene cluster (Fig. 3B), which are the putative components of the electron transport chain that connects HdrC via ubiquinol pool (Fig. 3C) (56). Meanwhile, the Na⁺/H⁺ antiporter (Nha) involved in maintaining cytosol pH homeostasis (57) was found significantly upregulated under electroautotrophic conditions. This observation suggests a shift and possible regulation of Δμ_proton in electroautotrophic S. ovata cells as discussed in-depth below.

Integrated Omic-Analysis Unraveled Metabolic Rewiring.

Our experimental observations collectively support a more active and efficient metabolism in the materials–biology hybrid, for the generation of energy carriers and the reduction of CO₂ into acetic acid. We demonstrated a H₂-mediated electroautotrophy, based on the absence of noticeable biofilm formation on Co–P alloy cathode (SI Appendix, Fig. S5), the estimated percentage of more than 99% for planktonic cells, and the inability of detecting other known redox mediators at appreciable concentrations (>nM). We observed an accelerated production of acetic acid and increased values of e⁻ efficiency (Fig. 1 D and E and SI Appendix, Table S1), even though the hybrid system consumed similar amounts of reducing equivalent namely H₂ as the chemoautotrophic ones do under the similar levels of hydrogenase expression in both conditions. Additional experiments under different methods of H₂ delivery suggest that mass transport of H₂ does not contribute to such observed differences of e⁻ efficiency. Instead, it is the significant difference in thermodynamic Gibbs free energies of CO₂ reduction (ΔG) under different H₂ partial pressures that leads to drastic observed difference of chemoautotrophic CO₂-fixing rates between 1% and 80% H₂ partial pressure. ΔG is calculated to be –16.7 and –27.8 kJ/mol per one equivalent of H₂ for 1% and 80% H₂ partial pressures at 1 bar, respectively, under otherwise the same physiological conditions (SI Appendix, Fig. S9). It is remarkable that the more efficient electroautotrophy (Entry 1 and 2 in SI Appendix, Table S1) occurs at a much lower H₂ partial pressure, i.e., against a less favored ΔG, than the compared chemoautotrophic ones with 80% H₂ at 1 bar in the headspace (Entry 3 and 4 in SI Appendix, Table S1).

Meanwhile, the abundance of redox-active and more importantly non-redox-based energy cofactors, quantitatively assayed via measuring the NADH/NAD⁺ and ADP/ATP ratios, indicated that intracellular conditions in the hybrid system were more reduced and energy-rich (Fig. 2B), while the metabolic activities and protein expressions in the Wood–Ljungdahl pathway remained largely unperturbed (Figs. 2B and 3B). Such an enhanced metabolic energy supply was consistent with the up-regulated expression of ATP synthase and other proteins related to the generation of transmembrane proton gradient (Fig. 3 B and C). In the Wood–Ljungdahl pathway (36), acetogenic organisms are quite ATP-limited under chemoautotrophic conditions due to the net zero ATP synthesis by substate-level phosphorylation. The organisms are predicted to yield at most 0.5 to 1.5 ATP through the transmembrane H⁺ gradient for each acetic acid molecule after an 8-electron reduction of CO₂ (58, 59). Hence, our observations suggest that the efficiency of ATP generation in S. ovata was increased in the materials–biology hybrid.

We hypothesize that the up-regulated expression of Hdr and cytochromes contribute to an increased H⁺ translocation hence microbial generation of ATP in electroautotrophic conditions (Fig. 4). S. ovata is reported to generate transmembrane proton gradient hence ATP through a confurcation mechanism (56), in which ferredoxin generated by electron-bifurcating hydrogenase (Bif) first transforms to NADH via proton-pumping Rnf and subsequently to NAD⁺ via a H⁺-pumping electron transport chain including ubiquinol, Hdr, and speculatively cytochromes (Cyt) (Fig. 4). Our observed upregulation of Hdr and cytochromes will potentially increase the extent of the second H⁺-pumping step by the confurcation mechanism. Such metabolic shift in the materials-biology hybrid can alleviate the limited ATP yield with faster microbial CO₂ fixation and a higher e⁻ efficiency.

Fig. 4.

Proposed metabolic rewiring in electroautotrophic S. ovata. The enzymes in black are stably expressed under both electroautotrophic and chemoautotrophic conditions, while the enzymes in red were up-regulated under electroautotrophic conditions. The metabolites in red were more abundant in electroautotrophic cultures. Blue arrows represent electron flow, and green arrows represent carbon flow in S. ovata metabolism. The proton translocation and ATP synthesis are highlighted by yellow arrows. RAM_ox and RAM_red are the oxidized and reduced redox-active metabolites, respectively; H₂ase, hydrogenase; Bif, electron-bifurcating hydrogenase; Nha, Na⁺/H⁺ antiporter; Acetyl-AA, acetyl-amino acids.

A relevant side evidence supporting the proposed increase of proton translocation resides in the observed significant upregulation of Na⁺/H⁺ antiporter (Nha) expression and the increased accumulation of N-acetyl-glutamate in the H₂-mediated materials-biology hybrids (Fig. 4). While there is a sufficient thermodynamic driving force of CO₂ fixation at 80% H₂, at low H₂ partial pressure as the case of electroautotrophy, it is critical to maintain a homeostasis of suitable membrane proton electrochemical potential (Δμ_proton), typically −140 to −200 mV as commonly reported (57), in order to satisfy two contradicting factors for sustained CO₂ fixation (SI Appendix, Fig. S10, more detailed discussion in SI Appendix, Supplementary Note 2). First, a negative enough Δμ_proton must be maintained in order to have sufficient energy to drive ATP formation owing to the structural H⁺/ATP stoichiometry of ATP synthase (60); second, Δμ_proton cannot be too negative in order to thermodynamically enable H⁺-pumping in aforementioned coupling sites driven by exergonic redox reactions (57). Therefore, ideally a suitable Δμ_proton level tailored to the specific H₂ partial pressure must be maintained to enable CO₂ fixation in acetogens, and a less negative Δμ_proton better suits a lower H₂ partial pressure owing to change of ΔG values for CO₂ fixation. As Na⁺/H⁺ antiporter (Nha) is known to modulate pH homeostasis by exchanging extracellular H⁺ with intracellular Na⁺ hence increasing the value of Δμ_proton (SI Appendix, Supplementary Note 2 and Fig. S10), our observed significant upregulation of Nha during electroautorophy is consistent with the expected shift of Δμ_proton for sustained CO₂ fixation under low H₂ partial pressure. Similarly, the accumulation of acetyl-glutamate, a common osmolyte (49), could be explained as a response toward the increased osmotic pressure due to the extensive loss of intracellular Na⁺. The electroautotrophic condition is hypothesized to rewire the metabolism by modulating Δμ_proton and achieve a balance between proton-pumping at coupling sites and ATP synthesis, which is not present under chemoautotrophic conditions under similarly low H₂ partial pressure.

Metabolic Rewiring Hypothetically Based on Extracellular Redox-Active Metabolites (RAMs).

We hypothesize that it is the trace amount of extracellular RAMs such as riboflavin detected in our experiments that putatively signal the metabolic rewiring. The stable expression level of hydrogenase, as well as the slow if any chemoautotrophic CO₂ fixation under similarly low H₂ pressures, preclude H₂ as the signaling pathway in electroautotrophy. Therefore, other signaling pathways, probably redox-based, must be present in addition to the H₂-mediated materials-biology hybrids. We detected intracellular FAD, FMN, and riboflavin in all tested conditions (Fig. 2E and SI Appendix, Fig. S8). Small amount of extracellular riboflavin was also detected at most on the order of nM (Dataset S1), which is too low to contribute significant delivery of reducing equivalents for CO₂ fixation. Yet, such extracellular riboflavin exhibits reversible redox peaks in cyclic voltammograms with the half-wave potential (E_1/2) of –0.209 V vs. SHE (SI Appendix, Fig. S11), in accordance with the reported value of –0.210 V vs. SHE (61). Because riboflavin is electrochemically redox active (62), riboflavin will be almost exclusively under its reduced form under the electroautotrophic conditions of electrochemical H₂ generation (–0.77 to –0.96 V vs. SHE). Such a predominantly reduced population of extracellular riboflavin should differ appreciably from the one under chemoautotrophic condition when direct reduction of riboflavin by H₂ is kinetically slow in the absence of catalyst (63).

Soluble redox-active electron shuttles, commonly derivatives of flavins, quinones, phenazines, and cytochromes, are known to mediate long-distance electron transfer at a redox potential close to HER (64, 65) and trigger metabolic rewiring in microorganisms (66, 67). Therefore, our observed presence and electroactivity of extracellular riboflavin lead us to propose that the extracellular riboflavin, along with other possible extracellular RAMs, should serve as redox-sensing signal molecules. Because extracellular riboflavin’s abundance does not change significantly between electroautotrophic and chemoautotrophic samples, we deem the redox states of extracellular riboflavin, instead of the absolute concentration, that contributes to the observed metabolic rewiring. The proposed ratiometric signaling, based on the concentration ratio between reduced and oxidized riboflavin, is consistent with the regulation of microbial activities by altering the redox status of the extracellular RAMs (Fig. 4), which has been exemplified in the case of pyocyanin for bacterium Pseudomonas aeruginosa (68). This ratiometric figure-of-merit is also in line with the fact in bioenergetics that it is the ratiometric values including [NAD(P)H]/[NAD(P)⁺] and adenylate energy charge that fundamentally govern microbial metabolism (69). We hypothesize that this previously underappreciated redox signaling in S. ovata system triggers the observed metabolic rewiring including a tailored Δμ_proton level, up-regulated Hdr and Cyt for proton translocation, more efficient ATP generation, and a more reduced intracellular redox state for H₂-mediated microbial CO₂ fixation (Fig. 4).

Our reported metabolic rewiring shares some similarities yet differs from previous literature precedencies related to electrofermentation (12, 70). Acetogens under heterotrophic fermentation, such as Clostridium acetobutylicum (71), are known to shift the fermentation product distribution toward more reduced chemicals under the presence of exogenous mediators such as viologens, commonly under a reductive electrochemical driving force (72). It is similar that both our case and the ones of electrofermentation (72) witness a more reduced cellular redox state. Yet, the reported autotrophic metabolism differs greatly from the heterotrophic fermentation because the limiting factor here is the limited ATP generation owing to the small thermodynamic energy gain of acetogenesis and the zero net ATP yield from substrate-level phosphorylation in the Wood–Ljungdahl pathway (36), while ATP generation in heterotrophic fermentation is not a major concern. Hence, the major impacts from the varied redox states of extracellular riboflavin reported here reside on the proton translocation and ATP synthesis that leads to the observation of a smaller ADP/ATP ratio during electroautotrophy, which is significantly different from the previous reports in electro-fermentation (70, 72). Indeed, the fact that only chemoautotrophic cultures yield ethanol (SI Appendix, Fig. S6), the more reduced product than acetate, corroborates the aforementioned difference between our case and the ones of electro-fermentation, when more reduced products were promoted under additional electrochemically reductive energy input (70, 72).

Discussion

In this work, we showcase the previously undetected metabolic rewiring in S. ovata from a H₂-mediated materials-biology hybrid. Such results illustrate the metabolic versatility and complexity of S. ovata, thanks to its high e⁻ efficiency of CO₂ fixation (73), robustness toward extragenic interferences (74), and the plentiful redox-based coupling sites involving ubiquinol and cytochromes (55). Future studies about the detailed signaling pathway between the microbes and the water-splitting electrodes, currently proposed via the mediation of extracellular riboflavin, will be welcomed and should yield useful insights about proton translocation and ATP generation that are critical in the CO₂-fixing Wood–Ljungdahl pathway. Such insights will be beneficial for the design of hybrid systems for sustainable CO₂ fixation with S. ovata as the biotic component.

Moreover, our unexpected observation of the fortuitous metabolic rewiring highlights the necessity of thorough examination about the interactions in materials–biology hybrids. We posit that the observed metabolic rewiring may exist in a wide range of seemingly simple H₂-mediated hybrid systems with potential benefits that so far have been under-appreciated and under-utilized. In addition to the extensive research efforts being devoted toward materials design and synthetic biology for the efficient production of targeted chemicals in the hybrids (1, 2, 4), we advocate the deployment of multiomic studies for a comprehensive understanding of microbial metabolism in search of unexpected synergies. Such research efforts will enable us to maximize the efficiency and throughput of CO₂ fixation using the materials-biology hybrids for a sustainable future.

Materials and Methods

Materials and Chemicals.

The chemicals and materials were purchased from Sigma-Aldrich, unless otherwise specified. The carbon cloth was purchased from Fuel Cell Earth, and the stainless-steel mesh was purchased from Alfa Aesar. The ¹H NMR standard sodium 3-(trimethylsilyl) propionate 2,2,3,3-D4 (TSP-D4) and deuterium oxide (purity > 99.9%) were obtained from Cambridge Isotope Laboratories, Inc. Methanol, acetonitrile, ammonium acetate, ammonium hydroxide, and acetic acid were purchased from Fisher Scientific.

Electrode Fabrication and Characterization.

The cathode and anode were coated with Co–P alloy and cobalt phosphate (CoPi) catalysts, respectively, using electrochemical deposition following established protocols (7, 14, 31). A single-chamber electrochemical setup installed with a working electrode of deposition substrate, a Pt counter electrode, and a Ag/AgCl (3M KCl) (CH Instruments) reference electrode was connected to Gamry Interface 1000 potentiostat (Gamry Instruments). The Co–P alloy cathode was prepared by a 15-min chronoamperometry in an electrolyte consisted of 0.15 M H₃BO₃, 0.1 M NaCl, 0.33 M NaH₂PO₂·H₂O (sodium hypophosphite), and 0.2 M CoCl₂·6H₂O, during which the stainless-steel mesh was poised at –1.5 V vs. reference electrode. The CoPi anode was deposited on the carbon cloth in an electrolyte consisted of 10 mM Co(NO₃)₂ and 0.1 M methylphosphonate with a pH adjustment to 8. Chronocoulometry was performed with an applied potential of 0.85 V vs. reference electrode until 500 mC cm^–2 of charge was transferred.

The morphologies and element compositions of electrocatalysts (SI Appendix, Fig. S2) were characterized by JEOL JSM-6700F field emission scanning electron microscope (JOEL) equipped with Ametek energy-dispersive X-ray spectroscopy (EDS) device (EDAX-AMETEK) or ZEISS Supra 40VP field emission scanning electron microscope (ZEISS), after Au sputtering with Hummer 6.2 sputter coater (Anatech). A pretreatment followed an established protocol (33) was applied for the electron microscopy characterization of Co–P cathode with potentially adhered cells. The cathode was fixed in 2.5% glutaraldehyde overnight and sequentially immersed in ethanol (25%, 50%, 75%, 90%, 100%, 10 min each) and then transferred into hexamethyldisilazane (HMDS) for 1 h. The cathode was left to air dry overnight prior to characterization.

Chemoautotrophic Culturing of S. ovata.

The S. ovata were routinely maintained at 30 °C in 160 mL serum bottles containing 100 mL of mineral salts medium (75, 76) (SI Appendix, Supplementary Note 1) closed with butyl rubber stoppers (Chemglass Life Sciences LLC) held in place with aluminum crimps. The headspace was filled with 1-bar H₂:CO₂ (80:20, v:v) with sufficient agitation. For the growth comparison under different conditions, the inoculum of H₂-grown S. ovata cells was transferred in biological triplicates.

The chemoautotrophic cultures were grown at 30 °C in the 160-mL serum bottle containing 50 mL mineral salts medium (SI Appendix, Supplementary Note 1), 50 mL inoculum, and 60 mL headspace, which were maintained under similar if not the same experimental conditions as the electroautotrophic cultures except for the electron donor (see below). H₂ was intermittently supplied to the chemoautotrophic cultures by replacing the headspace (60 mL) with H₂:CO₂ (80:20, v:v) every 12 h at 1 bar with sufficient agitation. Under such conditions, S. ovata utilized H₂ as the sole electron donor and CO₂ and NaHCO₃ as the carbon sources. All cultures were grown under strictly anoxic conditions in the presence of reducing reagents of 0.2 mM Na₂S and 0.2 mM cysteine. At least two independent batch experiments, each with three replicates, were performed for the sake of reproducibility (Entry 3 and 4 in SI Appendix, Table S1). A negative control was performed on the same medium for S. ovata growth in the absence of electricity or H₂, and the headspace was replaced by N₂:CO₂ (80:20, v:v) every 12 h at 1 bar (SI Appendix, Fig. S3). To mimic the mass transport in electroautotrophic cultures (see below), chemoautotrophic growth of S. ovata was also tested in the same bioelectrochemical setup (see below) yet in the absence of electricity (Entry 5 and 6 SI Appendix, Table S1). Instead, a continuous gas flow of H₂:CO₂:N₂ (1:20:79, v:v:v) was bubbled to the setup from the gas tube submerged around 2 cm below the liquid surface, mimicking if not replicating the mass transport and the delivery of H₂ in electroautotrophy, under similar electron supply rate and agitation rate (i.e., 150 rpm) as the electroautotrophic cultures except for the electron donor. Two different gas flow rates of 4 mL min^–1 and 6 mL min^–1, controlled by the mass flow controller (Omega Engineering), were tested with almost the same results (Entry 5 and 6 in SI Appendix, Table S1). Aliquot samples were taken from the cathodic chamber periodically for further analysis.

Electroautotrophic Materials–Biology Hybrids of S. ovata.

The electroautotrophic cultures were established with triplicates following previously reported procedure with slight modifications (31). Customized three-electrode, two-chamber (150 mL each) electrochemical cells (H cells), separated by Nafion 117 membrane (Chemours), were applied for S. ovata incubation to allow the proton exchange (SI Appendix, Fig. S1). The gas-tight cathodic chamber had been flushed with N₂:CO₂ (80:20, v:v) to maintain the anoxic environment, followed by the transfer of the mineral salts medium (50 mL) and S. ovata inoculum (50 mL) under strict anoxic conditions. The cathode chamber includes Co–P cathode (2 × 2 cm) and leak-free Ag/AgCl reference electrode (+0.209 V vs. standard hydrogen potential, SHE, Innovative Instruments, Inc.) under a constant gas supply of N₂:CO₂ (80:20, v:v) at the rate of 6 mL min^–1 controlled by the mass flow controller. In order to maintain similar if not the same gas transport as the case of chemoautotrophic growth (see above), the gas supply is purged over the headspace of cathodic chamber. The anodic chamber includes CoPi anode and the headspace was continuously flushed with N₂:CO₂ (80:20, v:v). Both chambers were stirred at the constant rate of 150 rpm and placed in a water bath maintained at 30 °C during the three-day experiment. Aliquot samples were taken from the cathodic chamber periodically for further analysis.

The electrochemical settings of electroautotrophic cultures were achieved with a Gamry Interface 1000E potentiostat interfaced with a Gamry 8-channel ECMB multiplexer. The electrochemical potentials (E) have been iR compensated with the solution series resistance determined daily by electrochemical impedance spectroscopy and are all referenced to SHE using the following Eq. 1:

$$E \mathrm{v}\mathrm{s}. \mathrm{SHE}=E \mathrm{vs}. \mathrm{Ag}/\mathrm{AgCl}+0.209 \mathrm{V}.$$ [1]

Cyclic voltammetry (CV) between –0.39 V to –1.39 V vs. SHE at a scan rate of 50 mV s^–1 was conducted daily to check the system’s stability. Multiplexed chronoamperometry, whose E values are adjusted daily to maintain a desired level of current intensity, was performed to ensure electroautotrophic systems.

Residual Extracellular Oxidizable Organics Carried Over during Bacterial Inoculation.

The chemical oxygen demand (COD) in the extracellular components of the culture after S. ovata inoculation was analyzed in order to more accurately quantify the e⁻ efficiency of CO₂ fixation. The extracellular components were collected by filtering culture through sterile polyethersulfone membrane (0.22 µm, Avantor) to exclude planktonic S. ovata cells. Colorimetric COD analysis was performed using TNT82206 kit (Hach) following the manufacturer’s protocol, after establishing a standard curve prepared in the same mineral salts medium to avoid any matrix effects interfering with the quantification. The potential electron storage in the extracellular oxidizable component in carryover was calculated based on Eq. 2.

$$N_{\mathrm{e},\mathrm{carryover}}\left(\mathrm{mmole}\right)=\mathrm{COD}\left(\mathrm{mmole}\right)\times 4-N_{\mathrm{acetate}}\left(\mathrm{mmole}\right)\times 8.$$ [2]

Here, COD represents the amount of oxygen required to completely oxidize the extracellular components, $N_{\mathrm{acetate}}$ the observed generation of extracellular acetate (see below), and N_e,carryover presents the amount of reducing equivalents, in the unit of millimoles (mmole) of electrons, that can contribute to bacterial CO₂ fixation. Here, we assume that 1 COD equivalent offers 4 electron equivalents and the formation of one equivalent acetate requires 8 electron equivalents via CO₂ fixation.

Product Quantification and Electron Efficiency Analysis.

The liquid products in the sampled aliquots were quantified by the 400 MHz NMR spectrometer (Bruker) equipped with Bruker AV400 Sample Changer. Representative ¹H-NMR spectra under chemoautotrophic conditions are provided in SI Appendix, Fig. S6. While the electroautotrophic cultures almost exclusively yield acetate, chemoautotrophic cultures yields small amount of ethanol and occasionally formic acid as an ephemeral product. We note the technical difficulties of accurate quantification of low-level ethanol concentrations as our Hungate anaerobic technique demands gas purging despite ethanol’s high volatility. The acetate concentration was quantified using the chemical shift of 1.92 ppm in ¹H NMR spectroscopy on a liquid sample containing culture suspension and internal standard of TSP-D4 (31), after the establishment of a standard calibration curve. The total amount of acetate accumulation was calculated based on the measured acetate concentration and liquid volume. The H₂ partial pressure in the culture headspace was measured by an 8610C gas chromatography system (SRI Instruments) equipped with Hayesep D and MS 5A packed columns, and a thermal conductivity detector (TCD) (16). The total amount of H₂ in the liquid culture was calculate every 12 h based on measured H₂ partial pressure, by considering both the gaseous and dissolved H₂ with additional inputs of the monitored headspace gas pressure (McMaster-Carr) and the known H₂ solubility with a Henry solubility coefficient of 7.7 × 10⁻⁶ mole m⁻³ Pa⁻¹ (77). Last, the protein concentrations in S. ovata cultures were determined using the Bradford assay following an established protocol (43). Lysed culture suspension was prepared by treating 0.5 mL culture suspension in a boiling water bath and then mixed with Pierce Coomassie reagent (Pierce Coomassie Protein Assay Kit, Thermo Scientific) for a colorimetric analysis of 595-nm absorbance using Cary 60 UV-Vis spectrophotometer (Agilent Technologies), following the manufacturer’s protocol after establishing a calibration curve with albumin as the standard sample. To quantify proteins in cells attached to Co–P cathode, the Co–P cathode (2 × 2 cm) was collected after incubation and immediately immersed in 1 mL cell-free mineral salts medium followed by the same extraction procedures. At the same time, the extraction control including the Co–P cathode but without electroautorophic incubation was conducted and returned no protein detection, suggesting that the extraction procedures will not lead to false positive observations in the Bradford assay.

The apparent electron efficiencies (e^– efficiencies) η_app under electroautotrophic and chemoautotrophic conditions (η_app,electro and η_app,chemo, respectively), without considering potential contribution from the extracellular organics carried over from inoculation, were calculated using Eqs. 3 and 4, respectively.

$$\begin{array}{c}{\eta }_{\mathrm{app},\mathrm{electro}}=\frac{∆C_{\mathrm{acetate}}\times V_{\mathrm{medium}}\times 8 }{{\mathit{\int }}_{t_0}^{t}I\mathrm{d}t\times F}\times 100\%,\hfill \end{array}$$ [3]

$${\eta }_{\mathrm{app},\mathrm{chemo}}=\frac{∆C_{\mathrm{acetate}}\times V_{\mathrm{medium}}\times 8 }{∆N_{{\mathrm{H}}_2}\times 2}\times 100\%.$$ [4]

Here, ΔC_acetate represents concentration change of acetate that each requires 8 electrons from CO₂ reduction, V_medium the total volume of liquid culture, ${\int }_{t_0}^{\mathrm{t}}I\mathrm{d}t$ the total amount of charge transfer during electroautotrophy, F the Faraday constant (96,485 C mole^–1), and $∆N_{{\mathrm{H}}_2}$ the amount of H₂ consumed during chemoautotrophy that each contributes two electrons for CO₂ reduction.

Moreover, because we found that the extracellular organics carried over from bacterial inoculation contribute appreciable amount of reducing equivalents toward bacterial metabolism (see above), it is more appropriate to calculate the corrected e^– efficiencies η by accounting the reducing equivalents in extracellular organics (N_e,carryover) for both electroautotrophic (η_electro) and chemoautotrophic (η_chemo) conditions, respectively:

$${\eta }_{\mathrm{electro}}=\frac{∆C_{\mathrm{acetate}}\times V_{\mathrm{medium}}\times 8 }{{\mathit{\int }}_{t_0}^{t}I\mathrm{d}t\times F+N_{\mathrm{e},\mathrm{carryover}}}\times 100\%,$$ [5]

$${\eta }_{\mathrm{chemo}}=\frac{∆C_{\mathrm{acetate}}\times V_{\mathrm{medium}}\times 8 }{∆N_{{\mathrm{H}}_2}\times 2+N_{\mathrm{e},\mathrm{carryover}}}\times 100\%.$$ [6]

Both apparent and corrected e^– efficiencies under each experimental condition were reported in SI Appendix, Table S1.

Luminescence-Based ADP/ATP Ratio Analysis.

The two-step D-luciferin-based ADP/ATP ratio analysis was performed using MAK15 kit (Sigma-Aldrich) following the manufacturer’s protocol. Then, 10 µL bacterial culture under both electroautotrophic and chemoautotrophic conditions was collected immediately after the three-day experiments and instantly mixed with reaction reagent on a 96-well plate with a clear bottom. The luminescence intensities were recorded by microplate reader SpectraMax iD3 (Molecular Devices) and the ADP/ATP ratio was calculated following the manufacture’s protocol.

Metabolome Extraction and Sample Preparation.

After a three-day experiment, 1-mL bacterial suspension was vacuum-filtered on a nylon membrane (25 mm diameter, 0.2 µm pore size, Millipore). The membrane was immediately transferred to a 6-well plate containing 1 mL of cold extraction solvent for nonfolate species (–20 °C, 40:40:20, methanol/acetonitrile/water) to quench metabolism. For intracellular folate species, the extraction solvent was 80:20 acetonitrile/water with 2.5 mM sodium ascorbate and 25 mM ammonium acetate stored at 4 °C for stability. After 20 min of metabolite extraction at the respective cold temperatures, the cell extract was collected, transferred to a microcentrifuge tube, and centrifuged at 17,000 × g for 10 min at 4 °C. The supernatant was transferred to a new centrifuge tube, dried under nitrogen flow, and reconstituted in LC-MS (liquid chromatography-mass spectrometry) grade water before transferring to vials for LC-MS analysis (78). Additionally, to determine extracellular metabolites, the filtrate of bacterial suspension was diluted with LC-MS grade methanol at a 20:80 ratio. After centrifuging at 17,000 × g for 10 min at 4 °C, the supernatants were transferred to vials for later LC-MS analysis.

Metabolomics Analysis.

LC separation was achieved in a Vanquish UHPLC system (Thermo Fisher Scientific) equipped with a hydrophilic interaction chromatography column (Xbridge BEH Amide XP column, 130Å, 2.5 µm, 2.1 mm × 150 mm, Waters, Milford, MA). Mobile phases were 20 mM ammonium acetate and 20 mM ammonium hydroxide in 95:5 water:acetonitrile (pH 9.4) as eluent A and acetonitrile as eluent B. Column temperature set at 25 °C, and a flow rate was 150 µL min^–1. The gradient was as follows: 90% B at 0 min, 90% B at 0.5 min, 75% B at 1.5 min, 75% B at 5.5 min, 70% B at 6.5 min, 70% B at 7.5 min, 50% B at 8.5 min, 50% B at 10.5 min, 25% B at 11.5 min, 25% B at 12.5 min, %0 B at 14.5 min, 0% B at 20 min, 90% B at 20.5 min, and 90% B at 26 min. The total run time was 26 min at a flow rate of 150 µL min^–1 (79). The injection volumes were 2 and 10 µL for extracellular and intracellular metabolite samples, respectively. LC separation of folates species was on a reversed-phase column (Agilent InfinityLab Poroshell 120 Bonus-RP 2.7 µm, 2.1 × 150Â mm) with 10 mM ammonium acetate and 0.1% vol of acetic acid in 95:5 water:acetonitrile as eluent A and acetonitrile as eluent B. Gradient conditions were 2% B at 0 min, 98% B at 20 min, 98% B at 25min, 2% B at 26 min, 2% at 35 min. The flow rate was 200 µL min^–1, and the column temperature was 25 °C. The injection volume was 20 μL.

For mass spectrometry, the above-described LC systems were interfaced with Q-Exactive Plus Hybrid Quadrupole-Orbitrap Mass Spectrometer (Thermo Fisher Scientific). The mass spectra were collected in positive mode scanning mass-to-charge (m/z) ratio range between 60 to 900, and negative mode scanning m/z 60 to 200 and m/z 200 to 2,000 at a resolution of 140,000 at m/z 200, with automatic gain control (AGC) target 3e6, and maximum injection time (IT) of 500 ms. For folate species, mass spectra were collected only in positive scanning m/z 442 to 476 at a resolution of 35,000 at m/z 200 and a maximum IT of 200 ms. The authenticated standards including more than 400 metabolites were applied to verify the retention time (80). For MS/MS analysis, full MS/dd-MS2 mode was used. Full MS was in positive scanning m/z 100 to 1,000 at a resolution of 140,000 at m/z 200, and a maximum IT of 120 ms. MS/MS scanning was with a resolution of 70,000 at m/z 200, a maximum IT of 100 ms, an isolation window of 2 m/z, minimum AGC 1.0e3, and normalized collision energy (NCE) of 30 with an inclusion list of targeted metabolites.

Data were collected using Thermo Scientific Xcalibur data acquisition and interpretation software (version 4.2.27) and Q-Exactive MS series Tune interface (version 2.11). Data analysis was processed using the Maven software package (81). Metabolite identification was based on the exact m/z with mass error less than 10 ppm and a retention time difference smaller than 1 min to the authenticated standards (79). The abundance of each metabolite was reported after integrating the top peak area in extracted ion chromatogram and normalized to the protein concentration in culture.

Proteome Extraction and Sample Preparation.

After a three-day experiment, cell pellets from 40-mL culture suspensions were collected by centrifuging at 7,547 × g for 20 min using Sorvall ST 8 centrifuge (Thermo Fisher Scientific). The proteome was extracted using the established protocol (16). Cell pellets were resuspended in 0.5-mL lysis buffer consisted of 100 mM Tris, 8 M urea, 35 mM sodium dodecyl sulfate, protease inhibitor (Thermo Fisher Scientific), and benzonase nuclease (MilliporeSigma), and sonicated using the point sonicator for 10 s on the ice bath followed by rotation at 4 °C for 30 min. The proteins present in the supernatant were collected by centrifuging at 16,000 × g for 15 min at 4 °C. The proteins were reduced, alkylated, and digested following the established procedures (16). Following the digestion, the peptides eluted from carboxylate-modified magnetic beads were dried by vacuum centrifugation and reconstituted in 5% formic acid before analysis by LC-MS/MS (Liquid chromatography–tandem mass spectrometry).

Proteomics Analysis.

Peptide samples were separated on a 75 μm ID, 25 cm C18 column packed with 1.9 μm C18 particles (Dr. Maisch GmbH HPLC) using a 140-min gradient of increasing acetonitrile concentration and injected into a Thermo Orbitrap-Fusion Lumos Tribrid mass spectrometer. MS/MS spectra were acquired using Data Dependent Acquisition (DDA) mode.

MS/MS database searching was performed using MaxQuant (1.6.10.43) against the S. ovata DSM 2662 reference proteome from Uniprot (UP000015521, 5045 entries). The domain annotation and functional prediction of proteins were obtained from UniParc. The KEGG orthology (KO) assignment based on the amino acid sequence of the protein was performed using BlastKOALA which was used for pathway analysis against KEGG database (82).

Statistical analysis of MaxQuant label-free quantitation data was performed with the artMS Bioconductor package which performs the relative quantification of protein abundance using the MSstats Bioconductor package (default parameters). Intensities were normalized across samples by median-centering the log2-transformed MS1 intensity distributions. The abundance of proteins missing from one condition but found in more than two biological replicates of the other condition for any given comparison was estimated by imputing intensity values from the lowest observed MS1-intensity across samples.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Data, Materials, and Software Availability

The metabolomic and proteomic data reported in this study are available in the Datasets S1 and S2. All other data are included in the manuscript and/or supporting information.