Formate metabolism in the acetogenic bacterium Acetobacterium woodii

in the model


Introduction
Acetogenic bacteria are a specialized group of strictly anaerobic bacteria that convert 2 mol of carbon dioxide into acetate via the Wood-Ljungdahl Pathway (WLP) (Drake, 1994;Müller, 2003;Ragsdale and Pierce, 2008). The WLP consists of two linear branches, the methylbranch in which CO 2 is reduced to the methyl group of acetate and the carbonyl branch in which a second molecule of CO 2 is reduced to CO, the precursor of the carbonyl group of acetate. Both branches convey into acetyl-CoA, the precursor of biomass as well as of acetate. The electron donor for the WLP can be hydrogen (Wood et al., 1986;Schuchmann and Müller, 2013), allowing for lithotrophic growth, or sugars (Fontaine et al., 1942;Balch et al., 1977;Leigh et al., 1981), alcohols (Buschhorn et al., 1989;Bertsch et al., 2016) or aldehydes (Daniel et al., 1991). In addition, acetogens grow by conversion of other C1 substrates that are fed directly into the WLP such as carbon monoxide (Genthner and Bryant, 1982;Lynd et al., 1982;Lorowitz and Bryant, 1984;Savage et al., 1987;Daniel et al., 1990;Tanner et al., 1993;Bertsch and Müller, 2015;Weghoff and Müller, 2016), formaldehyde (Kallen and Jencks, 1966;Schink, 1994), or methyl groups (Kreft and Schink, 1997;Poehlein et al., 2012;Kremp et al., 2018;Lechtenfeld et al., 2018;Kremp and Müller, 2020). To provide the electrons for CO 2 reduction to CO, part of formaldehyde or the methyl groups have to be oxidized to CO 2 by a reversal of the methyl branch of the WLP. Formate is also an intermediate in the WLP and known as a substrate for some acetogens (Balch et al., 1977;Leigh et al., 1981;Breznak and Switzer, 1986;Traunecker et al., 1991;Mechichi et al., 1998;Küsel et al., 2000;Simankova et al., 2000).
Growth of acetogens on C1 compounds is of high biotechnological interest. Some acetogens are already used as biocatalysts for ethanol production from H 2 + CO 2 + CO (synthesis gas). Methanol is another promising feedstock for acetogenic biorefineries, as is formate. Formate can be produced from carbon dioxide and hydrogen and the acetogens Acetobacterium woodii and Thermoanaerobacter kivui have highest rates of formate production (Schuchmann and Müller, 2013;Schwarz et al., 2018). Formate can also be produced chemically, for example, via electrochemical reduction of CO 2 with electricity generated by fossil carbon, biomass or renewable energy (Yishai et al., 2016). Formate can then be used as feedstock for various aerobic and anaerobic bacteria and archaea. Surprisingly, little is known about the enzymology of acetogenesis from formate in acetogenic bacteria. To fill this gap is the more important since A. woodii has a gene cluster that encodes a second formyl-THF synthetase along with a potential formate transporter. Here, we have addressed the pathway for acetogenesis from formate and the enzymes involved in A. woodii.

Growth of A. woodii on formate
In order to optimize growth of A. woodii on formate, various concentrations of sodium formate were tested as sole carbon source (Fig. 1A). Cells grown under fructose limitation (2.5 mM) were transferred to media containing sodium formate as carbon and energy source. Growth did not start immediately but only after a lag phase of ≈ 6 h. The final OD 600 increased with increasing formate concentrations and a maximum was reached at 200-300 mM formate. Formate concentrations higher than 300 mM did not lead to higher ODs. The highest growth rate occurred with cells grown on 100 and 200 mM formate, reaching 0.12 h À1 (Fig. 1B). Above 300 mM formate, the growth rate decreased drastically, reaching the lowest rate of 0.06 h À1 with 500 mM formate. The following experiments were performed with 200 mM formate.
Formate conversion during growth led to acetate formation (Fig. 2). 205.5 mM formate was consumed in 35 h as the growth reached its stationary phase (OD 0.6) and 46.7 mM acetate was produced in 40 h. The ratio of formate:acetate was 4.4:1 with an electron recovery of 91%. The pH of the medium increased from 7.4 to 9.3. To analyse the effect of the initial pH on growth and final yields, growth experiments were performed in medium with low initial pH of 4.5, 5.2, 5.8 and 6.4. Cells did not grow at pH 4.5, 5.2 and 5.8 and at pH 6.4 growth rate and final yield was the same as at pH 7.2 (data not shown). Growth as well as final yield was also not stimulated when the bicarbonate buffer was changed to Tris, glycylglycine, imidazole or PIPES. Therefore, the bicarbonate buffer was used in the following for cultivation.

Transcriptional organization of genes involved in formate uptake and initial metabolism
In the genome of A. woodii, there is one formate transporter annotated (Poehlein et al., 2012). The putative formate/nitrite transporter gene, fdhC (Awo_c08050) is located next to a formyl-THF synthetase encoding gene, fhs2 (Awo_c08040), in the same direction (Fig. 3A). Formyl-THF synthetase catalyses the conversion of formate to formyl-THF and this has been studied in the thermophilic acetogen Moorella thermoacetica (Lovell et al., 1990;Radfar et al., 2000a,b). Upstream of these A. Dependence of growth of A. woodii on the formate concentration. The growth experiments were performed in 5 ml complex medium in the presence of 0 (Â), 50 (▼), 100 (■), 200 (▲), 300 (♦), 400 (•), 500 (□) and 1000 mM (Δ) sodium formate as substrate. The pre-culture for inoculation was grown on 100 mM sodium formate. The experiments were performed in biological duplicates and one representative growth curve is shown. B. Dependence of the growth rate of A. woodii on the formate concentration. Data were taken from the experiments depicted in panel A. Each data point indicates a mean with standard error of the mean (SEM). two genes is an acyl-acyl carrier protein thioesterase encoding gene (Awo_c08030) with the same direction as fhs2/fdhC. Downstream is a gene encoding a fructose-1,-6-bisphosphatase (fpb, Awo_c08060) in the opposite direction. In addition to fhs2, there is another gene, fhs1 (Awo_c09260) also potentially encoding a formyl-THF synthetase. fhs1 is part of the gene cluster that encodes methenyl-THF cyclohydrolase (fchA, Awo_c09270), methylene-THF dehydrogenase (folD, Awo_c09280) and methyl-THF reductase (RnfC2, Awo_c09290; MetV, Awo_c09300; MetF, Awo_c09310), which catalyse the next steps after formyl-THF formation in the methylbranch of the WLP (Schuchmann and Müller, 2014). Fhs1 and Fhs2 are 98% identical on the amino acid level.
To identify the transcriptional organization of the genes fhs2 and fdhC, mRNA was prepared from cells grown on 200 mM formate in minimal media and harvested in the exponential growth phase (OD 600 of 0.1). cDNA was synthesized and used as template in a PCR with primers that link the intergenic regions between Awo_c08030 and fhs2 (Awo_c08040), fhs2 (Awo_c08040) and fdhC (Awo_c08050), and fdhC (Awo_c08050) and fpb (Awo_c08060) (Fig. 3B). This bridging PCR revealed that fhs2 and fdhC are on one transcript while Awo_c08030 and fpb genes are not part of this operon.

Transcriptomic analyses for genes involved in formate metabolism
To identify genes involved in formate metabolism, transcriptome analyses were performed. As observed before, the presence of yeast extract in the growth medium led to the expression of genes involved in the utilization of glycine betaine (GB) (Lechtenfeld et al., 2018) or alanine (Dönig and Müller, 2018) as carbon and energy source (Table S1).
In addition, genes involved in flagella formation were upregulated only compared to fructose-grown cells (Table S1). On the other hand, genes involved in riboflavin synthesis, ribD (Awo_c00550), ribE (Awo_c00560), ribAB (Awo_c00570) and ribH (Awo_c00580) were the most downregulated genes in fructose-grown cells (Table S2). Moreover, some genes encoding proteins involved in glycolysis such as 1-phosphofructokinase (Awo_c03330; fruK), fructose-specific PTS system II components (Awo_c03340), 6-phosphofructokinase (Awo_c12790; pfkA), pyruvate kinase (Awo_c12800; pyk), triosephosphate isomerase (Awo_c24510; tpiA2) and a transcriptional regulator (Awo_c24540) were also downregulated. Twenty genes encoding ribosomal  A. Genomic organization of fhs2/fdhC genes. B. Transcriptional organization of fhs2/fdhC genes. For the analysis, total RNA from A. woodii was prepared and the contaminating DNA was removed. cDNA was synthesized by reverse transcriptase. To analyse the transcriptional organization, cDNA was used as template for PCR to bridge the intergenic regions of the fhs2/fdhC genes (lane 1). Genomic DNA and RNA were used as positive and negative controls respectively (lanes 2, 3). Primers used in this study are described in Materials and Methods and their location is indicated by arrows in panel A. proteins were downregulated as well. In comparison to H 2 + CO 2 -grown cells, genes involved in utilization of lactate (Weghoff et al., 2015;Schoelmerich et al., 2018) were downregulated as well as several genes involved in sugar metabolism (Table S3). Besides, one gene cluster encoding an ABC transport system was highly downregulated during formate metabolism. Overall, the changes in transcript levels were low in general when formate-and fructose-or H 2 + CO 2 -grown cells were compared.
The transcript counts of fhs1 (Awo_c09260) and fhs2 (Awo_c08040) were higher in cells grown on formate than those grown on fructose, with a fold change of 2.0 and 1.2, while counts were slightly lower compared to H 2 + CO 2 -grown cells (Table 1). fdhC (Awo_c08050) levels in formate-and fructose-grown cells compared to fructose were similar, however, fdhC was downregulated compared to H 2 + CO 2 , with a fold change of À1.5. The abundance of HDCR (Awo_c08190 -Awo_c08260) transcripts was higher in cells grown on formate compared to fructose and H 2 + CO 2 , except fdhF1 and HycB1. On the other hand, other genes encoding WLP enzymes in the methyl branch such as fchA, folD, rnfC2, metV and metF were upregulated compared to fructose-grown cells, while these were downregulated compared to H 2 + CO 2 -grown cells. The levels of genes encoding CODH/ACS in formate-grown cells was higher than in fructose-grown cells, but similar to H 2 + CO 2 -grown cells.
The genes encoding enzymes involved in acetate formation from acetyl-CoA such as phosphotransacetylase (Awo_c19620; pta) and acetate kinase (Awo_c21260; ackA) were downregulated during growth on formate. Gene clusters encoding the energy-conserving Rnf complex, the ATP synthase and the electron bifurcating hydrogenase (Schuchmann and Müller, 2014) were upregulated in formate-grown cells compared to fructosegrown cells, but were downregulated compared to H 2 + CO 2 -grown cells. The sodium salt of formate was used as a substrate, but the genes potentially involved in Na + homeostasis were downregulated. The gene encoding carbonic anhydrase (Awo_c05870; cynT) was similarly expressed with all substrates. In addition to the WLP, glycine synthase pathway is a possible CO 2 -fixing metabolic pathway. Among acetogens, it has been reported that Clostridium drakei has a functional glycine synthase as well as glycine reductase pathway (Song et al., 2020). A. woodii possesses genes encoding all four proteins involved in the glycine synthase pathway, but not in one cluster: three genes encoding H protein (gcvH1, Awo_c09330; gcvH2, Awo_c22560; gcvH3, Awo_c32810), two genes encoding dihydrolipoamide dehydrogenase (lpdA1, Awo_c09320; lpdA2, Awo_c12540), one gene encoding aminomethyltransferase (gcvT, Awo_c32800) and two genes encoding glycine dehydrogenase (gcvPB, Awo_c32780; gcvPA, Awo_c32790). Except lpdA1 and gcvH1, all genes for the glycine synthase pathway were slightly upregulated during formate metabolism, especially a potential gene cluster, Awo_c32780 -Awo_32810. A. woodii does not have genes encoding a glycine reductase. During growth on formate, pyruvate must be generated as it serves as a precursor for various biosynthetic reactions. A. woodii possesses genes encoding pyruvate:ferredoxin oxidoreductase (PFOR) as well as pyruvate formate lyase (PFL). PFOR catalyses the oxidation of pyruvate to acetyl-CoA and CO 2 with the concomitant reduction of ferredoxin (Chabriere et al., 1999). Awo_c24330 (nifJ) was strongly transcribed during growth on formate compared to cells grown on fructose or H 2 + CO 2 . Awo_c06200 (porB) and Awo_c06210 (porA) were upregulated, but transcript counts were much lower compared to those of nifJ. PFL catalyses the conversion of pyruvate to acetyl-CoA and formate and vice versa (Zelcbuch et al., 2016). All genes encoding for PFL were expressed stronger in formate grown cells, but total transcript counts were extremely low.

Generation and characterization of deletion mutants
To characterize their role in formate metabolism, the genes fhs2, fdhC and fhs2/fdhC were deleted using the genetic system and the ΔpyrE deletion strain of A. woodii described before (Westphal et al., 2018). For generation of the mutants Δfhs2, ΔfdhC and Δfhs2/fdhC, suicide plasmids pMTL84151_JM_dfhs2, pMTL_84151_SK_dfdhC and pMTL84151_SK_dfhs2/fdhC were constructed, carrying each 500 to 1000 bp of upstream and downstream flanking regions (UFR and DFR) of the respective gene leaving 3 bp behind the start codon and 3 bp in front of stop codon. The promoter region aimed to be remained in the chromosome. Moreover, the plasmids contained pyrE gene from Clostridium acetobutylicum (Westphal et al., 2018) and the chloramphenicol/thiamphenicol resistance cassette (catP) from Clostridium perfringens (Werner et al., 1977) together with its own promoter. For the first selection, the plasmids were integrated into the chromosome at one flanking region under antibiotic pressure with thiamphenicol. Subsequently, disintegration was forced by counter-selection with 5-fluoroorotate since the presence of pyrE cassette in the integrated plasmid enabled production of 5-fluorouracil which is toxic to the cells.
The ΔfdhC mutant had a slightly reduced growth rate compared to the wild type but the final OD 600 was reduced by 55% (Fig. 4A). At the same time, formate consumption as well as acetate formation was reduced (Fig. 4B). The Δfhs2 mutant as well as Δfhs2/fdhC double mutant had a more severe growth effect with an extended lag phase; at the same time, formate consumption as well as acetate formation had a lag phase.
In order to study how the deletion of the formate transporter affects growth on different concentration of formate, cells were grown with 20 to 500 mM of formate as substrate. With 20 mM formate, the wild type and the ΔfdhC mutant showed similar maximum OD 600 (0.15) and growth rate (0.07 h À1 ) (Fig. S2). However, with increasing formate concentrations, the growth rate and the final OD 600 of the ΔfdhC mutant decreased and the difference to the wild type stayed almost constant up to 500 mM formate tested.

Cell suspension experiments
To compare H 2 or acetate production by the mutants, resting cells of the wild type and the mutants were prepared, as described in Material and Methods. When high concentration of sodium formate (350 mM) was given to the wild type and the Δfhs2/fdhC mutant, both strains metabolized formate with production of 19 mM acetate and 100 mM H 2 in carbonate-free imidazole buffer or 35 mM acetate and 18 mM H 2 in carbonate-containing imidazole buffer, respectively. In the presence of low concentrations of sodium formate (4 mM) in carbonatecontaining imidazole buffer, the Δfhs2/fdhC mutant was impaired in formate utilization and acetate production. H 2 production was very low and more or less identical in both strains (Fig. 5). When the experiments were performed in the absence of bicarbonate and CO 2 , the wild type as well as the mutant produced much more hydrogen, up to ≈ 3 mM, and less acetate (Fig. S3). Again, formate utilization and acetate production was reduced in the Δfhs2/fdhC mutant.
Since the formyl-THF synthetase is also involved in acetogenesis from H 2 + CO 2 , we compared acetate formation from H 2 + CO 2 in resting cells of the wild type and the Δfhs2/fdhC mutant. The Δfhs2/fdhC mutant produced less acetate from H 2 + CO 2 (Fig. 6). Acetate formation from H 2 + CO 2 was described to be accompanied with a transient production of formate (Peters et al., 1999;Oswald et al., 2018). This was also observed here. However, the double mutant accumulated around thrice the amount and accumulation was not transient but formate remained at a high level. This phenomenon was also seen with different concentrations of H 2 .

Discussion
Here, we have investigated growth of A. woodii on formate and acetogenesis from formate as well as the genes encoding enzymes involved in formate metabolism. Formate is the first intermediate of the methyl branch of the WLP and was reported to be used as carbon source in A. woodii already in 1977 (Balch et al., 1977) as well as other acetogens such as Eubacterium limosum (Genthner and Bryant, 1987), Acetobacterium carbolinicus (Eichler and Schink, 1984) or Clostridium ljungdahlii (Tanner et al., 1993).
The first step in formate metabolism is its uptake. At the optimal concentration of 200 mM and pH 7.0 (pKa formic acid = 3.75), 0.1 mM of the substrate was in the form of formic acid, 199.9 mM as formate. Formic acid can diffuse across membranes (Falke et al., 2010) and thus, there is no essential need for a transport system. The hallmark of active transport is not only to increase the uptake rate but, more important, to accumulate the substrate inside the cell, at the expense of ATP hydrolysis or the electrochemical ion gradient across the membrane. However, FdhC of A. woodii is similar to FocA of E. coli, a presumptive channel that facilitates diffusion but does not accumulate but only equilibrate formate concentration outside and inside the cell (Hunger et al., 2014;Wiechert and Beitz, 2017). The high concentration of formate required for optimal growth may reflect its uptake by diffusion or facilitated diffusion of formic acid. However, with an alkalinization of the external pH (at a constant internal pH), formic acid will be dragged out of the cells, arguing for active transport of formate. Anyway, the experiments described here demonstrate that FdhC is not essential for growth on the formate concentration tested; however, the presence of FdhC leads to higher final optical densities and lower growth rates. This gives a clear fitness advantage to the cells.
The pathway of acetogenesis from formate has been postulated on theoretical grounds. Four mol of formate enter the WLP, then 3 mol of formate are oxidized to CO 2 to provide the electrons for the reduction of another mol of formate plus 1 mol of CO 2 to acetate via the WLP. A key enzyme in formate metabolism is the formate dehydrogenase (FDH) that has been studied in several species. The thermophilic acetogen M. thermoacetica has a tungsten-selenium-containing, NADP-specific FDH (Andreesen and Ljungdahl, 1973;Yamamoto et al., 1983;Deaton et al., 1987). In Clostridium aciduriri, the reduction of CO 2 to formate is catalysed by an electron bifurcating FDH (Wang et al., 2013). A. woodii (Schuchmann and Müller, 2013) and the thermophilic acetogen Thermoanaerobacter kivui (Schwarz et al., 2018) have a hydrogen-dependent CO 2 reductase (HDCR) in which a formate dehydrogenase module is connected via two electron-transferring subunits to a Fe-Fe hydrogenase module. Overall, the genes coding for enzymes of the WLP during growth on formate are upregulated compared to growth on fructose, but slightly downregulated compared to growth on H 2 + CO 2 . Interestingly, A. woodii has a second gene (fhs2) encoding a formyl-THF synthetase. fhs2 builds an operon with fdhC encoding the putative formate transporter. A similar genetic arrangement was not found in other acetogens. T. kivui has fhs1 (TKV_c19930) and fdhC (TKV_19940) in the same order with an intergenic region of 842 bp (Hess et al., 2014), however, it is located in a cluster with other genes for the methyl branch of the WLP. During formate metabolism fhs2 and fdhC as well as fhs1 together with the genes involved in the methyl branch of the WLP were upregulated compared to fructose-grown cells but slightly downregulated compared to H 2 + CO 2 -grown cells. Overall, the fold change was low. As fhs2 and fdhC are transcribed in a single transcriptional unit apart from the Fig. 5. Formate uptake and conversion into acetate and H 2 by resting cells of A. woodii wild type and Δfhs2/fdhC. Cells were grown on 20 mM fructose and harvested in late exponential phase. After washing twice, the cells were resuspended in 10 ml imidazole buffer (50 mM imidazole, 20 mM NaCl, 20 mM KCl, 20 mM MgSO 4 , 60 mM KHCO 3 , pH 7.0) in 120 ml serum flasks under a N 2 /CO 2 atmosphere. 4 mM of sodium formate was given to cell suspension of the wild type (■) and the Δfhs2/fdhC mutant (•). Formate (A) and acetate (B) were determined by HPLC. (C) H 2 was determined by gas chromatography and presented as mmol l À1 cell suspension. Each data point indicates a mean AE SEM; n = 2 independent experiments.
genes for the methyl branch of the WLP, we expected that expression of this transcriptional unit enhances the formate metabolism in A. woodii. Indeed, the deletion mutants Δfhs2 and Δfhs2/fdhC showed impaired growth and reduced formate consumption compared to the wild type.
When H 2 + CO 2 was given as substrate, formate was transiently accumulated and immediately reutilized in the wild type under carbonate-free condition. The same was observed before (Peters et al., 1999;Oswald et al., 2018). However, the Δfhs2/fdhC mutant could not reutilize formate, instead, formate accumulated to 2.6 mM. From all these observations, we conclude that Fhs2 and FdhC play an important role for initial formate uptake and conversion to formyl-THF especially in the presence of low formate concentrations.
From the physiological, genetic and biochemical data presented we propose that formate is metabolized according to the following equation: as depicted in Fig. 7. This is also supported by an observed Na + dependence of acetogenesis from formate and the essential requirement of the Rnf complex for acetate formation from formate (data not shown). A possible bypass of carbon through a potential glycine synthase as in C. drakei (Song et al., 2020) is not supported by the data. How biomass is built during growth on formate would be a remaining question. Pyruvate, a precursor for many biosynthetic processes, can be generated by carboxylation of acetyl-CoA, generated via the WLP. Pyruvate synthesis via PFL would be advantageous during formate metabolism since formate is an educt in the reaction and the reaction is thermodynamically and kinetically independent of external CO 2 concentration in the environment. We observed in the transcriptome analysis that genes encoding PFOR, porB, porA and nifJ are slightly upregulated during growth on formate. The genes pflA, pflB1, pflB2, pflB3 and pflC were upregulated but the overall transcript level was low. Thus, we assume that PFOR is mainly responsible for pyruvate synthesis during formate metabolism. However, which PFOR is responsible for the reaction and why A. woodii possesses three pfl genes are the remaining questions for further research.

Experimental procedures
Organisms and cultivation A. woodii strains were routinely cultivated under anaerobic condition at 30 C in the complex medium described previously (Heise et al., 1989). The A. woodii strains used in this study are listed in Table S4. As substrates for growth, 20 mM fructose, 60 mM methanol, 1 bar of H 2 + CO 2 (80:20, vol/vol) and various concentrations of sodium formate were used. E. coli DH5α and BL21 (DE3) strains were cultivated aerobically at 37 C in LB medium either in presence of chloramphenicol (30 ng μl À1 ) or ampicillin (100 ng μl À1 ). Growth was monitored by measuring the optical density at 600 nm (OD 600 ).

Transcriptome analyses
For transcriptome analyses, cells were grown on 100 mM sodium formate and harvested in the early exponential growth phase with OD 600 of 0.1. Harvested cells were resuspended in 800 μl of RLT buffer of the RNeasy Mini Kit (Qiagen, Hilden, Germany) with β-mercaptoethanol (10 μl ml À1 ), and cell lysis was performed using a Fig. 7. Acetogenesis from formate in A. woodii. Formate is imported via the formate transporter (FdhC) or by diffusion. Three mol formate are oxidized to 3 mol CO 2 and 3 mol H 2 by HDCR and 1 mol formate is converted to formyl-THF via formyl-THF synthetase (Fhs1/ Fhs2). Formyl-THF is further reduced to methyl-THF. One mol CO 2 is reduced to CO and combined with the methyl group by the CODH/ACS complex, generating acetyl-CoA. Acetyl-CoA is further converted to acetate. Balancing of reducing equivalents is achieved by action of the bifurcating hydrogenase (HydABC) and the Rnf complex. The ATP synthase has a Na + /ATP stoichiometry of 3.66 (Matthies et al., 2014), the Na + /e À stoichiometry of the Rnf complex is assumed to be 1 Na + /e À , based on theoretical consideration (Schuchmann and Müller, 2014). laboratory ball mill. Subsequently, 400 μl of RLT buffer (RNeasy Mini Kit) with β-mercaptoethanol (10 μl ml À1 ) and 1200 μl of 96% [v/v] ethanol were added. For RNA isolation, the RNeasy Mini Kit was used as recommended by the manufacturer, but instead of RW1 buffer RWT buffer (Qiagen, Hilden, Germany) was used to also isolate RNAs smaller than 200 nt. To determine the RNA integrity number (RIN), the isolated RNA was run on an Agilent Bioanalyzer 2100 using an Agilent RNA 6000 Nano Kit as recommended by the manufacturer (Agilent Technologies, Waldbronn, Germany). Remaining genomic DNA was removed by treatment of the samples with TURBO DNase (ThermoFisher Scientific, Waltham, MA, USA). The RiboZero kit (Illumina Inc., San Diego, CA, USA) was used to reduce the amount of rRNAderived sequences. For sequencing, the strand-specific cDNA libraries were constructed with a NEBNext Ultra directional RNA library preparation kit for Illumina (New England BioLabs, Frankfurt am Main, Germany) using 100 ng rRNA depleted RNA and 12 PCR cycles. To assess the quality and size of the libraries, samples were run on an Agilent Bioanalyzer 2100 using an Agilent High Sensitivity DNA kit as recommended by the manufacturer (Agilent Technologies, Waldbronn, Germany). The concentration of the libraries was determined using the Qubit ® dsDNA HS Assay kit as recommended by the manufacturer (Life Technologies GmbH, Darmstadt, Germany). Sequencing was performed by using the MiSeq instrument (Illumina Inc., San Diego, CA, USA) using the MiSeq reagent kit v3 (150 cycles) for sequencing in the paired-end mode and running 2 Â 75 cycles. Fructose-grown cells were used as reference dataset . Normalization of the reads was done with DeSeq2 (v 1.28.1) (Love et al., 2014) and foldchange-shrinkages were calculated with DeSeq2 and the apeglm package (v 1.10.0) (Zhu et al., 2019). For analysis based on KEGG annotation, the clusterProfiler (v 3.16.0) (Yu et al., 2012) and pathview (v 1.28.0) (Luo and Brouwer, 2013) packages were used. Genes with a log 2 -fold change (FC) of +2/À2 and a p-adjust value < 0.05 were considered differentially expressed. The sequence data have been submitted to the SRA database.

Preparation of resting cells
Cells of A. woodii wild type, Δfhs2, ΔfdhC and Δfhs2/fdhC were grown on 20 mM fructose in 500 ml complex medium to late exponential growth phase (OD 600 of 1.2 to 1.5) and then harvested by centrifugation (Avanti™J-25 and JA-10 Fixed-Angle Rotor; Beckman Coulter, Brea, CA, United States) at 7000 Â g and 4 C for 10 min. The harvested cells were washed twice with 30 ml of buffer containing 50 mM imidazole (pH 7.0), 20 mM KCl, 20 mM MgSO 4 , 4 mM DTE and 4 μM resazurin by centrifugation at 8500 rpm (5948 Â g) and 4 C for 10 min (Avanti™J-25 and JA-25.50 Fixed-Angle Rotor; Beckman Coulter, Brea, CA, United States). Then, the cells were resuspended in 5 ml of buffer and kept in 16 ml Hungate tubes. All the steps were performed under strictly oxygen free conditions in an anoxic chamber (Coy Laboratory Products, Grass Lake, MI, United States) filled with N 2 /H 2 (96%-98%/2%-4%; v/v). After taking out of the anoxic chamber, the headspace of Hungate tubes filled with resting cells was changed to 100% N 2 . The total protein concentration in the resting cells was measured using the method by (Schmidt et al., 1963).

Cell suspension experiments
The resting cells were filled into 115 ml serum flasks to a volume of 10 ml of imidazole buffer (50 mM imidazole, 20 mM KCl, 20 mM NaCl, 20 mM MgSO 4 , 4 mM DTE, 4 μM resazurin, pH 7.0) and to a total protein concentration of 1 mg ml À1 . As substrate, either formate (4 mM or 350 mM) or 1 bar overpressure of H 2 + CO 2 (80%/20%; v/v) was added to the resting cells. The experiments were started with incubation at 30 C in water bath with shaking (230 rpm). 0.5 ml samples were taken at each time point for determination of formate and acetate.
Determination of H 2 , formate and acetate H 2 was determined by gas chromatography as described previously (Weghoff and Müller, 2016). The concentrations of formate and acetate were determined by high performance liquid chromatography (P680 HPLC Pump, ASI-100 Automated Sample Injector and Thermostatted Column Compartment TCC-100, Dionex, Sunnyvale, California, USA) with a HyperREZ XP Carbohydrate H + ion exchange column for sugar and acid separation (Thermo Fisher Scientific, Waltham, Massachusetts, USA) as described previously (Moon et al., 2019).