A genome-guided analysis of energy conservation in the thermophilic, cytochrome-free acetogenic bacterium Thermoanaerobacter kivui

Background Acetogenic bacteria are able to use CO2 as terminal electron acceptor of an anaerobic respiration, thereby producing acetate with electrons coming from H2. Due to this feature, acetogens came into focus as platforms to produce biocommodities from waste gases such as H2 + CO2 and/or CO. A prerequisite for metabolic engineering is a detailed understanding of the mechanisms of ATP synthesis and electron-transfer reactions to ensure redox homeostasis. Acetogenesis involves the reduction of CO2 to acetate via soluble enzymes and is coupled to energy conservation by a chemiosmotic mechanism. The membrane-bound module, acting as an ion pump, was of special interest for decades and recently, an Rnf complex was shown to couple electron flow from reduced ferredoxin to NAD+ with the export of Na+ in Acetobacterium woodii. However, not all acetogens have rnf genes in their genome. In order to gain further insights into energy conservation of non-Rnf-containing, thermophilic acetogens, we sequenced the genome of Thermoanaerobacter kivui. Results The genome of Thermoanaerobacter kivui comprises 2.9 Mbp with a G + C content of 35% and 2,378 protein encoding orfs. Neither autotrophic growth nor acetate formation from H2 + CO2 was dependent on Na+ and acetate formation was inhibited by a protonophore, indicating that H+ is used as coupling ion for primary bioenergetics. This is consistent with the finding that the c subunit of the F1FO ATP synthase does not have the conserved Na+ binding motif. A search for potential H+-translocating, membrane-bound protein complexes revealed genes potentially encoding two different proton-reducing, energy-conserving hydrogenases (Ech). Conclusions The thermophilic acetogen T. kivui does not use Na+ but H+ for chemiosmotic ATP synthesis. It does not contain cytochromes and the electrochemical proton gradient is most likely established by an energy-conserving hydrogenase (Ech). Its thermophilic nature and the efficient conversion of H2 + CO2 make T. kivui an interesting acetogen to be used for the production of biocommodities in industrial micobiology. Furthermore, our experimental data as well as the increasing number of sequenced genomes of acetogenic bacteria supported the new classification of acetogens into two groups: Rnf- and Ech-containing acetogens.


Background
Acetogenic bacteria represent an ecologically important group of anaerobes that are ubiquitous in nature [1]. They constitute the penultimate limb in the anaerobic food web and convert a number of substrates exclusively to acetate, which is then converted by methanogenic archaea to methane, the final product of anaerobic food webs [2]. Acetogens convert hexoses to three moles of acetate according to: indicating the presence of a pathway that is able to reduce two moles of CO 2 to acetate [3]. This pathway has been elucidated mainly in the thermophilic species Moorella thermoacetica (formerly Clostridum thermoaceticum) and is named according to its discoverers as the Wood-Ljungdahl pathway (WLP) [4]. It involves two branches and each of them contributes to the reduction of one molecule of CO 2 . In the carbonyl branch, CO 2 is reduced to enzymebound CO by the CO dehydrogenase/acetyl-CoA synthase (CODH/ACS). Electrons for this reduction are provided by reduced ferredoxin [5][6][7]. In the methyl branch of the WLP, another molecule CO 2 is reduced first to formate, which is then activated to formyl-THF in an ATP-dependent reaction [8,9]. Subsequently, formyl-THF is converted to methenyl-THF and then reduced stepwise to methyl-THF via methylene-THF. Eventually, the methyl group and the enzyme-bound CO are joined to form acetyl-CoA, catalyzed by the key enzyme of the WLP, the CODH/ACS [5][6][7][10][11][12]. Subsequently, acetyl-CoA is converted to acetyl-phosphate and acetate [13,14]. The latter step involves the production of ATP by an acetate kinase. However, since one mole of ATP is consumed during activation of formate, there is no net ATP gain by substrate level phosphorylation in the WLP. The WLP also enables lithotrophic growth of acetogens on H 2 + CO 2 according to: and, therefore, additional energy must be conserved by a chemiosmotic mechanism. The chemiosmotic process of energy conservation was recently uncovered in Acetobacterium woodii. A soluble, electron bifurcating, ferredoxin-and NAD + -reducing hydrogenase oxidizes molecular hydrogen and generates a reduced electron carrier with a very low redox potential, ferredoxin (E 0 ' ≈ −500 mV; a redox potential in this range has to be assumed since a ferredoxin is the electron donor for the CO 2 /CO couple E 0 ' = −520 mV). Reduced ferredoxin is then oxidized by a membranebound ferredoxin:NAD + -oxidoreductase encoded by the rnf genes and therefore also termed Rnf complex. The Rnf complex is composed of six subunits that harbor (covalently-bound) flavins and iron sulfur centers as electron carriers, and the electrons are transferred to the acceptor NAD + [15]. Electron transfer from reduced ferredoxin to NAD + is exergonic and this electron transfer is used to expel sodium ions from the cytoplasm thus generating an electrochemical sodium ion gradient across the membrane [16]. The electrochemical Na + gradient then drives the synthesis of ATP via a Na + -F 1 F O ATP synthase [17,18]. The energy-conserving module in A. woodii thus comprises only one coupling site, the Rnf complex, and a Na + -F 1 F O ATP synthase. Acetogenesis in A. woodii therefore has a modular appearance: An energyconserving module that is connected to the WLP module by soluble electron carriers such as pyridine nucleotides and ferredoxin. The WLP module is not energy conserving, it serves the function to reoxidize the end products of the anaerobic respiration and to provide acetyl-CoA for biomass synthesis [19]. The use of two different modules for energy conservation and re-oxidation of reduced electron carriers (WLP) opens interesting possibilities for biotechnological applications. In principal, any reductive pathway can be coupled to the energy-conserving module and indeed, acetogens are known that reduce nitrate [20,21], phenylacrylates [22,23] and fumarate [24,25], or acetyl-CoA to ethanol [26] or butyrate [27,28]. Very recently, the genetic coupling of a butanol-production pathway to the energy-conserving module proved successful in Clostridium ljungdahlii [29]. Production of biocommodities at high temperatures has several advantages such as lower costs for cooling and distillation or lower risk of contaminations. The genome of the thermophilic acetogen M. thermoacetica (T opt : 55-60°C) has been sequenced [30] and the way(s) of energy conservation can be predicted from the genome as well as experimental analyses [31]. Unfortunately, this thermophilic model strain grows poorly on H 2 + CO 2 (doubling times of up to 24 h [32]), a substrate used for 3rd generation biotechnology, thus limiting its use in industrial microbiology. In contrast, the acetogen Thermoanaerobacter kivui grows very fast on H 2 + CO 2 with doubling times around 2 h and has an even higher T opt of 66°C [33]. Moreover, its acetate to biomass ratio was about half compared to M. thermoacetica indicating a more efficient way to conserve energy [34]. This is also exemplified by yield measurements: 97% more biomass is produced from one mole of H 2 in T. kivui compared to M. thermoacetica [34]. This prompted us to sequence the genome of T. kivui to predict by bioinformatic analyses followed by experimental analyses how this acetogen couples acetogenesis to chemiosmotic energy conservation.

Growth conditions
T. kivui LKT-1 (DSM 2030) was grown at 65°C in medium that was prepared as described by Leigh et al.   [35]. The contaminating amount of Na + was 160 ± 4 μM, determined using an Orion Star A214 Na + -selective electrode (Thermo Scientific, USA). Growth was determined by measuring the optical density at 660 nm (OD 660 ) with a spectrophotometer.

Preparation of membranes from T. kivui
Cells were grown to an optical density at 600 nm of 1.9 -2.4 as described in 2 × 500 ml complex medium in 1-lflasks (Glasgerätebau Ochs, Bovenden-Lenglern, Germany) with 28 mM glucose as carbon source and harvested anaerobically by centrifugation at 11,500 × g for 10 min at 4°C. Cells were washed once in buffer A (50 mM Tris, 20 mM MgSO 4 , 20% glycerol, 2 mM DTE, 4 μM resazurin, pH 7.5) and, after another centrifugation step, resuspended in 5-10 ml buffer A. Cells were disrupted by a single passage through a French press (110 MPa). Cell debris and whole cells were removed by a centrifugation step for 30 minutes at 23,700 × g and 4°C. The cell extract was separated into the cytoplasmic and membrane fraction by ultracentrifugation (150,000 × g, 2 h, 4°C). The membranes were washed once in buffer A and sedimented again via ultracentrifugation. Finally, the membranes were resuspended in~5 ml buffer A and used immediately for the measurement of Fd red : NAD + oxidoreductase activity. Protein concentrations were determined as described previously [36].

Isolation of chromosomal DNA
Chromosomal DNA of T. kivui was isolated according to the procedure described [39] and modified by Ausubel et al. [40].

Sequencing strategy
The genome of T. kivui was sequenced with a combined approach using the 454 GS-FLX Titanium XL system (Titanium GS70 chemistry, Roche Life Science, Mannheim, Germany) and the MiSeq (Illumina, San Diego, CA). Shotgun libraries were prepared according to the manufacturer's protocols, resulting in 110,391 reads for 454 shotgun sequencing and 5,916,460 150-bp paired-end reads for Illumina sequencing. All of the 454 shotgun reads and 1 Mio 150-bp paired-end Illumina reads were used for the initial hybrid de novo assembly with MIRA 3.4 [41] and Newbler 2.8 (Roche Life Science, Mannheim, Germany). The final assembly contained 42 contigs with an average coverage of 89.20. For scaffolding and contig ordering tasks we used the Move Contigs tool of the Mauve Genome Alignment Software [42] and the genomes of T. mathranii mathranii A3, DSM 11426, T. brockii finii Ako-1, DSM 3389 and T. wiegelii Rt8.B1 as references. Additionally, contigs that could not be ordered with Mauve were examined via Gene Ortholog Neighborhoods based on bidirectional best hits implemented at the IMG-ER (Integrated Microbial Genomes-Expert Review) system [43,44] and with multiplex PCR [45]. Sequence gaps were closed in the Gap4 (v.4.11) software of the Staden Package [46] by PCR-based techniques and primer walking with conventional Sanger sequencing, using BigDye 3.0 chemistry on an ABI3730XL capillary sequencer (Applied Biosystems, Life Technologies GmbH, Darmstadt, Germany).

Gene prediction and annotation and analysis
The software tool prodigal (Prokaryotic Dynamic Programming Genefinding Algorithm) [47] was used for automatic gene prediction, while identification of rRNA and tRNA genes was performed with RNAmmer and tRNAscan, respectively [48,49]. Automatic annotation was carried out with the IMG-ER (Integrated Microbial Genomes-Expert Review) system [43,44], but annotation was afterwards manually curated by employing BLASTP and the Swiss-Prot, TrEMBL, and InterPro databases [50]. Prophage regions were identified using the online tool PHAST [51].

Experiments with resting cells and acetate determination
T. kivui was grown as described in 2 × 500 ml complex medium with 28 mM glucose as carbon source to an OD 600 of 1.9 to 2.4. The culture was centrifuged anaerobically at 11,500 × g and 4°C for 10 min. Cells were washed twice in imidazole buffer (50 mM imidazole, 20 mM MgSO 4 , 20 mM KCl, 4 mM DTE, 4 μM resazurin, pH 7.0). After the last centrifugation step, cells were resuspended in 2 ml imidazole buffer. The protein concentration was determined according to [52], being approx. 100 mg/ml. In order to determine the conversion of H 2 + CO 2 to acetate, 100-ml-serum flasks (Glasgerätebau Ochs GmbH, Bovenden-Lenglern, Germany) were filled with 10 ml imidazole buffer containing 50 mM KHCO 3 . If indicated, NaCl was added to a concentration of 20 mM. If applied, the ionophores ETH2120 and TCS were added to a concentration of 30 μM each. Subsequently, the gas phase of the serum flasks was exchanged to 1 bar H 2 + CO 2 (80:20 [v/v]) and then preheated to a temperature of 65°C in a water bath. The reaction was started by addition of resting cells to a final concentration of 1 mg/ml. Henceforth, 500-μl-samples were taken at time points as indicated. Samples were centrifuged immediately at 18,000 × g for 1 min and the supernatant was stored at −20°C. Determination of the acetate concentration of all samples was carried out using a commercially available kit (Acetic acid, Co. R-Biopharm, Darmstadt, Germany).

Nucleotide sequences and accession number
The sequence data described here have been deposited in GenBank under Acession No. CP009170 [GenBank: CP009170].

General features of the T. kivui genome
The complete genome of T. kivui (accession number CP009170) consists of a circular chromosome with a size of 2.397 Mbp and an overall G + C content of 35.06 mol%. General features of the genome are listed in Table 1. We could identify 2378 putative protein-coding genes, three complete rRNA clusters and 58 tRNA genes. The tRNA necessary for incorporation of selenocystein is also present, but this organism is probably not able to assemble selenoproteins, since the selABC gene cluster coding for essential proteins could not be identified, only a very short fragment of selB is present. 76.05% (1810) of the open reading frames (ORF) could be assigned to a putative function, 568 ORFs (23.86%) were annotated as hypothetical proteins and 35 ORFs as pseudo genes. Approximately 61% (1551 ORFs) of all protein-encoding genes could be assigned to at least one of the 21 functional COGs (Cluster of Orthologous Groups). The two most abundant categories were "general function" and "amino acid transport and metabolism", to which 10.24% and 9.18%, respectively, could be assigned to, followed by "function unknown", "translation, ribosomal structure and biogenesis", "coenzyme transport and metabolism", "replication, recombination and repair" and "energy production and conversion" with 9.06%, 8.30%, 6.83%, 6.65% and 6.00%, respectively.
As mentioned before, "replication, recombination and repair" is one of the most abundant COG category in the genome of T. kivui. Proteins which were assigned to this category are Cas proteins (CRISPR-associated sequences) and the encoding genes are often located adjacent to CRISPR loci (Clustered Regularly Interspaced Short Palindromic Repeats). The CRISPR/Cas system protects the genome against invading mobile elements such as plasmids or phages and consists of contiguous repeats with different length (20 to 47) and different numbers [53][54][55]. The genome of T. kivui harbors 3 CRISPR loci with 35, 48 and 21 repeats, respectively. The first CRISPR locus is flanked by four genes which could be assigned to a III-B/ polymerase-RAMP module CRISPR/Cas system subtype (TKV_c14620-TKV_c14650), the second locus is upstream flanked by a type III-B/polymerase-RAMP module subtype (TKV_c23500-TKV_c23560) and downstream by a hybrid cluster consisting of I-B/Teanap-Hmari and again a III-B/polymerase-RAMP module subtype where the cmr2 gene is interrupted by a transposase (TKV_c23590-TKV_c23770), while a second I-B/Teanap-Hmari subtype CRISPR/Cas system is located adjacent to the third CRISPR locus (TKV_c24261). In addition we could identify a complete subtype III-A CRISPR/Cas system (TKV-c17330-TKV_c17390), which is not associated with a CRISPR locus. All CRISPR/Cas clusters were annotated according to the polythetic classification of CRISPR/Cas systems [56]. There is no complete prophage located on the chromosome of T. kivui.

Substrate utilization
T. kivui can not only grow autotrophically on H 2 + CO 2 (or formate) but also heterotrophically on glucose, fructose, mannose, and pyruvate [33]. Correspondingly, genes coding for all enzymes for glycolysis were found in the T. kivui genome with only the hexokinase and the aldolase being encoded twice (Table 2). Despite the gene cluster TKV_c16300-TKV_c16340 (coding for the conversion of glyceraldehyde-3-P to phosphoenolpyruvate), the genes are spread over the genome. Fructose is taken up and phosphorylated to fructose-1-P by a fructose-specific PTS system (TKV_c23130 and TKV_c23140). Interestingly, the genes encoding component IIB and IIC are fused to one gene, TKV_c23130. Subsequently, fructose-1-P is phosphorylated to fructose-1,6-bisphosphate by FruK (TKV_c2 3150), a 1-phosphofructokinase. On the other hand, mannose is taken up by a second, mannose-specific PTS system (TKV_c06180-TKV_c06200) and thereby activated to mannose-6-P. This substrate is converted to fructose-6-P by a mannose-6-P isomerase (TKV_c16180). Glycolysis results in the formation of pyruvate, which is most probably converted to acetyl-CoA by use of the pyruvate:ferredoxin oxidoreductase. A gene cluster (TKV_c19290-TKV_c19260) encoding this enzyme was found in the genome of T. kivui, which encodes a ferredoxin, as well as the α, β, and γ subunit of the complex. Eventually, the phosphotransacetylase (TKV_c13970) and acetate kinase (TKV_c13960) catalyze the conversion of acetyl-CoA to acetate. Although two potential alcohol deyhdrogenases (TKV_c02600 and TKV_c22590) are encoded in the genome, an aldehyde dehydrogenase was not found. This is in accordance with the finding that ethanol was not detected in cells grown on H 2 + CO 2 , glucose or pyruvate [33].

Genes involved in the Wood-Ljungdahl pathway
The genes encoding the key enzymes in the Wood-Ljungdahl pathway of T. kivui are shown in Figure 1A. The first step in the methyl branch of the WLP is the reduction of CO 2 to formate by a formate dehydrogenase. A corresponding gene cluster very similar to the one of A. woodii was found on the chromosome of T. kivui (TKV_c19950-TKV_c19990). In A. woodii, the cluster consists of 7 genes [57], with fdhF2 being an isogene of fdhF1, encoding a selenium-containing Fdh. A gene encoding an electron transfer protein (hycB) follows each fdh gene. In T. kivui, only a selenium-free Fdh (fdhF/ TKV_c19990) is encoded in the cluster. This gene is followed by hycB3 (TKV_c19980) and hycB4 (TKV_c1 9970), both encoding small FeS proteins. Adjacent to hycB4 is a gene (hydA2/TKV_c19960) encoding a hydrogenase subunit. Both clusters also encode a gene annotated as fdhD. Since the enzyme purified from A. woodii did not contain an FdhD subunit [58], its function remains to be elucidated. In A. woodii, this enzyme complex was recently shown to catalyze the hitherto unknown reduction of CO 2 with electrons coming directly from molecular hydrogen and was therefore named hydrogen-dependent CO 2 reductase (HDCR) [58]. Thus, it is likely that T. kivui reduces CO 2 to formate with H 2 as well. This is in contrast to the close phylogenetic neighbor M. thermoacetica, which was shown to reduce CO 2 with electrons coming from NADPH [59]. Many of the genes encoding functions of the WLP are located in one main gene cluster ( Figure 1B), which lies in close proximity to the genes coding for the hydrogen-dependent CO 2 reductase. The product of the first gene of the operon, fhs (TKV_c19930), probably catalyzes the initial ATP-dependent activation of formate to formyl-THF. Subsequent to fhs is an open reading frame (orf1), whose product is similar to thymidylate synthases. Adjacent to orf1 are genes encoding the methenyl-THF cyclohydrolase (fchA), methylene-THF dehydrogenase (folD) as well as two subunits of the methylene-THF reductase (metV and metF). Those are followed by a dihydrolipoamide dehydrogenase (pdhD), a maturation factor of the CO dehydrogenase (cooC), subunit beta (acsD) and subunit alpha (acsC) of the corrinoid iron-sulfur protein.
The putative operon is completed by genes encoding a methyltransferase (acsE), an acetyl-CoA synthase (acsB) and protein H of a glycine cleavage system (gcvH). A possible function of GcvH within the WLP is unclear, but since this gene is also found within this gene cluster in A. woodii [57] and M. thermoacetica [30], one cannot exclude a so far undiscovered function of this enzyme within the pathway. Thus, as in A. woodii, proteins for the conversion of formate to methyl-THF are encoded by one gene cluster. However, in A. woodii folD and metV do not lie adjacent to each other but are separated by the gene rnfC2, probably coding for a third subunit of the methylene-THF reductase in this organism [57]. In M. thermoacetica, the genes coding for proteins of the methyl branch of the WLP are spread all over the genome. Only recently, MetF and MetV were shown to form a complex with the proteins HdrABC and MvhD [31]. Such genes were not found in T. kivui, implicating that the soluble methylene-THF reductase is composed of only 2 subunits. In A. woodii and M. thermoacetica, a separate gene cluster coding for the methyltransferase AcsE as well as subunits of the CODH/ACS is located somewhere else on the genome. In T. kivui, only two genes encoding a CO Glucose-6-P isomerase TKV_c16780 6-phosphofructokinase TKV_c16900 Aldolase TKV_c01430, TKV_c04080 Triose-P isomerase TKV_c16320 Glycerinaldehyde-3-P deyhdrogenase TKV_c16340 Phosphoglycerate kinase TKV_c16330 Phosphoglycerate mutase TKV_c16310 Enolase TKV_c16300 Pyruvate kinase TKV_c16890 dehydrogenase (acsA/TKV_c20100 and cooS/TKV_c08080), a phosphotransacetylase and an acetate kinase are encoded elsewhere on the genome. Thus the WLP gene cluster is outstanding, since it encodes all proteins (despite AcsA, which is separated by 10 genes) that are required for the conversion of CO 2 to acetyl-CoA.

Transfer of reducing equivalents by soluble enzymes
In all acetogens examined, ferredoxin and NAD(P) serve as common electron carriers in the WLP and in energy conservation. Reducing equivalents required in the WLP of T. kivui may be delivered from molecular hydrogen by a soluble electron bifurcating hydrogenase HydABC complex (TKV_c19580-TKV_c19600) similar to the electron bifurcating hydrogenases of A. woodii and M. thermoacetica, which reduce NAD + and ferredoxin in equimolar amounts [60][61][62]. HydA1 (TKV_c19850) of T. kivui shows 56% and 49% identity to the corresponding subunits of M. thermoacetica and A. woodii, respectively. The identities for HydB (TKV_c19590) to the homologous proteins of M. thermoacetica and A. woodii are 54% and 59%, the ones for HydC (TKV_c19600) amount 42% and 47%, respectively. Apart from HydA2 of the HDCR and the HydABC complex, there is no other soluble hydrogenase encoded in the genome of T. kivui.

F 1 F O ATP synthase of T. kivui
ATP synthases consist of a soluble head domain, that catalyzes ATP synthesis or hydrolysis and a membrane domain that is responsible for ion translocation [63]. Genes encoding all the subunits of a F 1 F O ATP synthase were found in the genome of T. kivui in one gene cluster (TKV_c06410-TKV_c06480). Interestingly, the cluster does not code for the gene atpI, which is the first gene of most bacterial atp operons and was shown to be essential for the assembly of the c ring in A. woodii [64]. The genes flanking the ATP synthase gene cluster of T. kivui apparently do not code for proteins that are required for ATP synthesis.
The membrane-integral c subunit of F 1 F O ATP synthases determines the ion specificity of the enzyme. To check whether the c subunit of T. kivui has a conserved Na + binding site, sequence alignments of several c subunits of Na + -dependent ATP synthases along with the one of T. kivui (atpE/TKV_c06420) were performed. As can be seen in Figure 2, the Na + -dependent F 1 F O ATP synthases from Ilyobacter tartaricus, Propionigenium modestum and A. woodii have the conserved Na + binding motif consisting of the amino acids Q…..ES/T [65,66]. This motif cannot be found in the c subunit of T. kivui ( Figure 2). Here, the glutamine (Q) is changed to an isoleucine (I) and instead of the serine (S) or threonine (T), there is an alanine (A). Since the proton-binding site (E55) of the c subunit is well conserved in T. kivui, this finding suggests that the F 1 F O ATP synthase of T. kivui is H + -dependent. This finding is in line with the hypothesis that the bioenergetics of T. kivui is not based on a sodium ion but proton current across its cytoplasmic membrane. To substantiate this hypothesis, the effect of Na + on growth and acetate formation was analyzed.

Autotrophic growth is not Na + -dependent
In order to determine a Na + dependence of T. kivui during autotrophic growth, cells were transferred four times in Na + -depleted minimal medium, the contaminating amount of Na + was determined to 160 ± 4 μM, Na + -enriched medium contained 71 ± 1 mM NaCl. As evident from Figure 3, the growth rate and the final optical density when growing on H 2 + CO 2 were independent of the Na + concentration. Both cultures grew to an OD 600 of~0.09 within 30 hours. These data strongly support our hypothesis that T. kivui is not Na + -but H + -dependent.

Acetate formation from H 2 + CO 2 is H + -dependent
In order to identify the nature of the coupling ion in acetogenesis of T. kivui irrevocably, cell suspensions of glucosegrown cells were prepared and acetate formation from H 2 + CO 2 was determined. As evident from Figure 4A, the presence of NaCl had almost no effect on the final amount and rate of acetate formation, arguing against Na + as coupling ion. In contrast, the addition of NaCl even led to a slightly decreased acetate production rate. Without NaCl present in the system, the final amount of acetate was about 42 mM, whereas only 35 mM acetate were produced in the presence of 20 mM NaCl. The contaminating amount of Na + was 350 ± 20 μM.
Although the experimentally determined Na + concentration in the assay without added NaCl is well below the K M for Na + during acetate formation in A. woodii, it may well be that the K M for Na + in T. kivui is lower than the contaminating Na + concentration in the assay. To determine whether acetogenesis in T. kivui relies on a transmembrane Na + or H + gradient, the effect of different ionophores on acetate formation was monitored. The Na + ionophore ETH2120 had almost no effect on acetate formation (in the absence or presence of sodium ions) ( Figure 4B), whereas the protonophore TCS led to a complete inhibition of acetate formation. Here, the final amount of acetate was only 5 mM ( Figure 4C). The experiments with resting cells of T. kivui give strong evidence that T. kivui does not belong to the Na + -dependent acetogens, but that H + is the coupling ion.

Membrane-embedded electron transfer coupled to the formation of a transmembrane proton gradient
In A. woodii, the Na + -translocating Rnf complex is the only coupling site [15,16,37]. However, inspection of the genome and the absence of Fd red :NAD + oxidoreductase activity at membranes of T. kivui (data not shown) revealed that T. kivui does neither possess an Rnf complex nor a complex with similar function. Furthermore, no genes encoding for cytochrome synthesis could be detected in the genome. Therefore, the genome of T. kivui was sought for genes encoding potential ion pumping membrane-bound oxidoreductases. As in M. thermoacetica, two gene clusters with Figure 2 Sequence alignment of c subunits of Na + -dependent ATP synthases and the subunit c (TKV_c06420) of T. kivui. The Na + binding motif is highlighted in bold. similarities to Ech-type complexes were found: TKV_c01 230-TKV_c01310 ( Figure 5A) and TKV_c19750-TKV_c19 680 ( Figure 6A). Ech complexes are thought to be the ancestor of complex I [67] and although final proof with a purified enzyme is still pending, strong indications were given that Ech complexes in methanogens couple the electron transfer from reduced ferredoxin to H + with the translocation of H + across the cytoplasmic membrane [68,69]. However, in M. thermoacetica, one Ech-type complex might be coupled to a formate dehydrogenase (Moth_2183) and the second does not possess all residues essential for a catalytic [NiFe] domain in the large hydrogenase subunit EchE (Moth_0980).
The first gene of the first ech gene cluster of T. kivui, ech1A (TKV_c01230), codes for a transmembrane protein with 19 predicted transmembrane helices with a weak similarity (17%) to EchA from Methanosarcina barkeri. The identity to homologous proteins of M. thermoacetica is 21% to Moth_2191, 23% to Moth_2188 and Moth_2187 (encoded by the first ech cluster), as well as 20% to Moth_0985 and 21% to Moth_0986 and Moth_0987 (encoded by the second ech cluster). The product of the following gene, ech1B, has 8 transmembrane helices and shows 21% identity to M. barkeri EchB, the identity to homologous proteins from M. thermoacetica is 22% for Moth_2190 and 26% for Moth_0981. Both subunits are thought to be involved in H + transport [70,71]. ech1C codes for the small hydrogenase subunit, probably involved in electron transfer. The identity to M. barkeri EchC is 38%, to M. thermoacetica Moth_2184 36% and to Moth_0978 25%. The fourth gene of the cluster, ech1F, encodes another subunit of Ech complexes. It has 24% identity to the corresponding subunit of M. barkeri and 26% to the homologous subunits (Moth_2185 and Moth_0982, respectively) from M. thermoacetica. It is predicted to have two 4Fe-4S clusters and therefore might be involved in electron transfer from ferredoxin. The following gene (hycB1/TKV_c01270) encodes a small FeS protein with four 4Fe-4S clusters that shows homology to HycB from Escherichia coli (33% identity) and CooF from Rhodospirillum rubrum (32% identity). Its identity to the homologous protein from M. thermoacetica (Moth_2192) is 36%. The protein products of the next two genes do not have any homologues in known Ech-type complexes, but show sequence similarity to the N-terminus of MetV, the small subunit of the methylene-THF reductase of the Wood-Ljungdahl pathway. The sequence identity to each other is 22%, the identity to MetV of T. kivui is 23% (for TKV_c01280) and 20% (for TKV_c01290), respectively. Both proteins might coordinate one 4Fe-4S cluster. 170 bp downstream of TKV_c01290 is a gene encoding the subunit Ech1D with 27% identity to the corresponding subunit from M. barkeri. It has no predicted cofactors and its function in the complex is unknown so far. The genome of M. thermoacetica does not code for a homologous protein. Adjacent lies the gene ech1E, encoding the catalytic NiFe hydrogenase subunit. The protein sequence shows 40% identity to EchE from M. barkeri and 39% and 33% to Moth_2186 and Moth_0980, respectively, from M. thermoacetica. A model of the protein complex Ech1 encoded by this ech1 cluster is shown in Figure 5B. Electrons coming from reduced ferredoxin may enter the complex at subunit Ech1F and may be transferred via the FeS proteins HycB1, TKV_c01280 and TKV_c01290 onto the catalytic hydrogenase subunit Ech1E that catalyzes H 2 formation. Subunits Ech1A and Ech1B mediate H + transport across the cytoplasmic membrane.
The second gene cluster potentially encoding an Echtype hydrogenase is TKV_c19750-TKV_c19680 ( Figure 6A). The product of the first gene, ech2D, encodes a 14 kDa protein ( Figure 6B) and is 19% identical to EchD from M. barkeri. The sequence identity to its homologue (TKV_c0 1300) in the Ech1 complex is only 8%. The succeeding gene, ech2E, codes for the large catalytic NiFe hydrogenase subunit and the deduced protein product shows 37% identity to M. barkeri EchE, 40% to Moth_2186 and 32% to Figure 3 Effect of Na + on autotrophic growth of T. kivui. Cultures grown on H 2 + CO 2 were transferred into Na + -enriched (•) and Na + -deficient (○) minimal medium. The curves shown are representative for three independent experiments. Precultures were grown for four transfers in the same medium.
Moth_0980 from M. thermoacetica, as well as 45% to TKV_c01310/Ech1E from T. kivui. Thereafter lies a gene encoding a 13 kDa protein (HycB2) with four predicted 4Fe-4S clusters. It has 35% identity to the FeS protein HycB1 (TKV_c01270) of the above described Ech-type complex and is homologous to CooF from R. rubrum (36% identity) and HycB from E. coli (39% identity), as well. The next two genes, TKV_c19720 and TKV_c19710, both encode proteins similar to the membrane-bound subunit EchA (with 22% and 21% identity to M. barkeri EchA, respectively and 18 to 27% identity to the six EchA homologues from M. thermoacetica) and were therefore named ech2A1 and ech2A2. The homologue of Ech2A1 in complex I is NuoL, whereas the homologue of Ech2A2 is NuoM. Ech2C is encoded by the gene TKV_c19700. It has one hydrophobic stalk and represents the small hydrogenase subunit. Its identity to EchC from M. barkeri is 43%, M. thermoacetica Moth_2184 and Moth_0978 are 48% and 37% identical. The gene ech2B encodes another membraneintegral subunit with 7 transmembrane helices but no known cofactors. Its identity to Moth_0981 is 25%, sequence identity to Moth_2190 is only 17%. The last gene encodes another small electron transfer protein with two 4Fe-4S clusters and is similar to EchF, the identity to the corresponding proteins from M. thermoacetica is 28% (Moth_2185) and 24% (Moth_0982). A postulated subunit composition of the Ech2 complex  and electron flow from reduced ferredoxin to H + with concomitant H + export is shown in Figure 6B.
Thus, in contrast to the above-described Ech-type complex, the second has one more membrane-integral subunit that might be involved in H + transport, probably allowing for a more beneficial stoichiometry of the electron:H + ratio. On the other hand, the small FeS proteins with similarity to MetV are missing. Sequence identities of the homologous proteins of the two Ech-type complexes to each other are summarized in Table 3.

Discussion
Model of energy conservation during autotrophic growth of T. kivui The genomic data allow to propose a model for electron and carbon flow during acetogenesis from H 2 + CO 2 in T. kivui. Electrons coming from molecular hydrogen are transferred to the electron carriers NAD + and ferredoxin by the soluble, electron-bifurcating hydrogenase HydABC. In the carbonyl branch of the WLP, the reduced ferredoxin serves as electron donor for the reduction of one molecule of CO 2 to CO, catalyzed by the CODH/ACS. In the methyl branch of the pathway, another CO 2 is reduced to formate with electrons coming directly from molecular hydrogen. This reaction is catalyzed by the hydrogendependent CO 2 reductase (HDCR), as it was also shown for A. woodii. The electron donor for the subsequent reduction of methenyl-THF cannot be predicted, since both NADH and NADPH are used in other acetogens and bioinformatic analyses do not allow to discriminate between the possibilities. However, since NADH is more common, it was used as electron donor for the T. kivui model. The following step, reduction of methylene-THF is still under debate for an indirect role in energy conservation. The redox potential (E 0 ') of the methylene-THF/methyl-THF pair is −200 to −130 mV [72] and electron transfer from NADH (E 0 ' = −320 mV) to methylene-THF is highly exergonic. Therefore, this reaction was assumed already in 1977 to be involved in energy conservation [73]. One scenario is that methylene-THF reduction is coupled to ferredoxin reduction by electron bifurcation, as recently suggested for M. thermoacetica [31]. In T. kivui, MetV and MetF are encoded in the WLP operon (see Figure 1B), however, the genes encoding the putatively bifurcating Hdr subunits of M. thermoacetica are lacking in the T. kivui genome. Therefore, the situation is more similar to A. woodii, which is also lacking Hdr-encoding genes and in which evidence for electron bifurcation was not obtained (Bertsch J, Öppinger C, Hess V, Langer JD, Müller V: A heterotrimeric NADH-oxidizing methylenetetrahydrofolate reductase from the acetogenic bacterium Acetobacterium woodii, in preparation). Actually, MetV and MetF of T. kivui are the minimal subunit composition of an acetogenic, nonelectron-bifurcating methylene-THF reductase. Since cell-free extract of T. kivui does not catalyze NADH-or NADPHdependent methylene-THF reduction (data not shown) we assume an electron carrier in that redox range which may be, for example, a flavodoxin, which is energetically equivalent to NADH. If we take into account a non-electron-bifurcating methylene-THF reductase that uses the electrons derived from a carrier energetically equivalent to NADH we can conclude a quantitative scheme for the bioenergetics of acetogenesis in T. kivui (Figure 7). According to this model, T. kivui can synthesize 0.25 ATP via the F 1 F O ATP synthase per acetate formed.
The ATP:acetate ratio in the phylogenetically close neighbor M. thermoacetica is predicted to be 0.5 [19], thus in comparison, T. kivui conserves only half as much energy from 4 H 2 + 2 CO 2 . Both organisms have an electron bifurcating hydrogenase HydABC that couples initial H 2 oxidation to the reduction of NAD + and ferredoxin. The latter is the electron donor for the energy conserving reaction, the electron flow to H + via the Ech complex that concomitantly translocates protons across the cytoplasmic membrane. The abovementioned difference in ATP production per acetate can be explained by the postulated additional coupling of methylene-THF reduction to the reduction of ferredoxin (or an energetic equivalent) in a second electron bifurcating reaction in M. thermoacetica. Hence this bacterium can transfer four electrons onto the Ech complex, thereby translocating two protons across the cytoplasmic membrane, whilst T. kivui uses the Ech complex only once per acetate formed.
As described in this paper, all data achieved are in accordance with H + based bioenergetics in T. kivui. Based on bioinformatic and experimental data, we could show that the ATP synthase is H + -dependent and the electrochemical ion gradient is probably formed by a H + -translocating Ech type complex. However, these data are in contrast to the results obtained by Yang et al. in 1990 [35], who found a strict Na + dependence when H 2 was the electron donor. In order to assure our diverging results, we prepared our Na + -deficient medium strictly according to the instructions as given in [35]. We measured a contaminating Na + concentration of  TKV_c19710  TKV_c01230  EchA  26   TKV_c19720  TKV_c19710  EchA  18   TKV_c19690  TKV_c01240  EchB  28   TKV_c19700  TKV_c01250  EchC  50   TKV_c19750  TKV_c01300  EchD  8   TKV_c19740  TKV_c01310  EchE  45   TKV_c19680  TKV_c01260  EchF  21   TKV_c19730  TKV_c01270  HycB  35 160 μM, while Yang and Drake reported 200 μM NaCl in their medium. Thus a different amount of Na + in the deficient medium can be excluded as a source of the discrepancy. We reproduced this experiment several times and could never see an effect of Na + on growth. Unfortunately, we were not able to find a reason for the discrepancy, but the additional data obtained in cell suspensions of T. kivui as well as the absence of a sodium ion binding site in the c subunit of the F 1 F O ATP synthase are consistent with the classification of T. kivui as H + -dependent.
In the early history of acetogenic bacteria, M. thermoacetica was used as a model organism to unravel the mechanism of energy conservation in H + -dependent acetogens. The detection of cytochromes [74] led to the assumption that H + translocation in M. thermoacetica is cytochrome-based [75][76][77]. However, despite many decades of intensive research, the participation of cytochromes in H + translocation of M. thermoacetica could never be confirmed and their involvement in energy conservation became questionable. Instead, a role of cytochromes in O 2 reduction was discussed in recent years [32]. With our finding that T. kivui and M. thermoacetica seem to use very similar mechanisms for energy conservation and that the genome of T. kivui does not encode any proteins necessary for cytochrome synthesis, it becomes more and more evident that the function of cytochromes in M. thermoacetica is indeed not in energy conservation coupled to the WLP. Instead, we have recently proposed to classify the acetogens bioenergetically in Rnf-and Ech-containing acetogens [19] and each class has a subclass with Na + -and H + -dependent species. T. kivui thus belongs to the subclass of H + -depending, Echcontaining acetogens.
By further comparing the two acetogens T. kivui and M. thermoacetica, it is remarkable that, despite similar mechanisms for energy conservation, their growth behavior on H 2 + CO 2 is quite different. As mentioned earlier, the doubling time of M. thermoacetica when growing autotrophically is about 24 h [32], while T. kivui grows 10-times faster [33]. But how can this discrepancy be explained in spite of the similar genetic configuration of the two bacteria? Since a weaker coupling of ATP synthesis and autotrophic acetate formation in M. thermoacetica was already excluded, the only explanation left is that at least one reaction of the Wood-Ljungdahl pathway is rate limiting. As M. thermoacetica is postulated to use three reactions that do not seem to be necessary in T. kivui (the transhydrogenase reaction for production of NADPH, reduction of CO 2 with NADPH, and a bifurcating reaction in the course of methylene-THF reduction), one of those might well be the limiting factor. However, in order to irrevocably identify the bottleneck, extensive biochemical studies based on experimental data need to be carried out in future.

Conclusions
Sequencing the genome of T. kivui allowed for a better insight into the reactions enabling the autotrophic reduction of CO 2 to acetate. Since this process is thought to be Figure 7 Model of electron and carbon flow in T. kivui when growing autotrophically. Fd red , reduced ferredoxin; a H + /ATP ratio of 12/3 was assumed for the F 1 F O ATP synthase. *, reduction of methylene-THF might occur using an electron donor with a similar redox potential as NADH. CoFeSP, corrinoid/FeS protein.
one of the first pathways that evolved on the early earth [78], its detailed understanding is of great relevance.
The thermophilic acetogen T. kivui oxidizes the electron donor H 2 by use of an electron bifurcating hydrogenase HydABC. The reduced ferredoxin then fuels the energy conserving membrane module that is coupled to the WLP: the genome of T. kivui harbors two gene clusters, which both encode Ech-type complexes. Although the deduced protein subunit composition of the two complexes alter slightly, both enzymes have all cofactors and catalytic domains that are required to catalyze ferredoxin oxidation as well as H + reduction. The proton gradient formed by Ech can be used by a H + -dependent F 1 F O ATP synthase to drive phosphorylation of ADP.