To elucidate the genetic background of lactate metabolism and fermentation pathways leading to the formation of n-caproate, n-butyrate and iso-butyrate, we manually curated the functional annotation of genes involved in the following bioprocesses: acetyl-CoA formation from lactate and ethanol, reverse β-oxidation cycle, energy conservation and hydrogen formation. Besides our newly isolated strains, we also included the other eleven chain elongators in this analysis. Especially for those strains reported to use lactate as electron donor, corresponding genes of lactate oxidation were also considered in the manual curation.
Lactate oxidation to acetyl-CoA
Lactate can serve as carbon and energy source for chain-elongating bacteria. As shown in Fig. 5, first lactate needs to be transported into the cell, which is facilitated by lactate permease (LacP). Genomes of BL-3 and BL-6 were predicted to harbor the corresponding CDSs, which are located in a gene cluster encoding lactate racemase (LacR) (Fig. 6a and 6c). The gene cluster encoding LacP and LacR was also found in all other lactate-based chain elongators (Fig. 6d-6 h). The fermentation starts with the oxidation of lactate via pyruvate to acetyl-CoA catalyzed by an NAD-dependent lactate dehydrogenase (LDH) and a pyruvate ferredoxin oxidoreductase (PFOR). All three genomes encode predicted LDH proteins, which are highly similar to each other. Specifically, the BL-3 genome was predicted to have four LDH genes, one of which is located in a gene cluster (Fig. 6a, CDS labels: 11486–11488) comprising also genes for the electron transfer flavoprotein (EtfAB). The BL-4 genome harbors four LDH genes with one located in the gene cluster (Fig. 6b, CDS labels: 2199–2205) encoding membrane-associated energy-converting NADH:ferredoxin oxidoreductase (RnfABCDEG). The BL-6 genome has three LDH genes with one found in a cluster (Fig. 6c, CDS labels: 3216–3223) including genes for butyryl-CoA dehydrogenase (BCD), EtfAB, LacR and LacP. A similar gene cluster (Fig. 6e, CDS labels: 01775–01795) containing genes for LacR, LDH, EtfAB and BCD was found in the genome of Ruminococcaceae bacterium CPB6. As for the enzyme PFOR or its synonym pyruvate synthase, all three genomes contain the corresponding genes, enabling the oxidation of pyruvate to acetyl-CoA. Acetyl-CoA then enters the reverse β-oxidation cycles. CDSs for LDH and PFOR were found in all other lactate-based chain-elongating species (Fig. 6d-6 h).
Ethanol oxidation to acetyl-CoA
The ethanol-based chain elongation pathway is well elucidated in C. kluyveri , and is of particular importance in several biotechnology studies [23–25]. Genome data of BL-3 and BL-6 suggest that they may have the capacity to utilize ethanol as additional or alternative substrate. Small, uncharged molecules like ethanol diffuse through the cytoplasmic membrane and can be oxidized via acetaldehyde to acetyl-CoA. NAD-dependent alcohol dehydrogenase (ADH) and NAD(P)-dependent acetaldehyde dehydrogenase (ADA) catalyze this conversion (Fig. 5). The corresponding CDSs were found in the genomes of BL-3 and BL-6, while BL-4 lacks the gene encoding ADA.
n-Butyrate and n-caproate formation
Transformation of acetyl-CoA to butyryl-CoA includes three intermediates: acetoacetyl-CoA, 3-hydroxybutyryl-CoA and crotonyl-CoA. The involved enzymes are acetyl-CoA acetyltransferase (ACAT), NAD- and NADP-dependent 3-hydroxyacyl-CoA dehydrogenase (HAD), enoyl-CoA hydratase (ECH) and NAD-dependent butyryl-CoA dehydrogenase complex (BCD/EtfAB) (Fig. 5). The formation of n-butyrate further requires butyryl-CoA:acetate CoA transferase (CoAT) to catalyze the reaction of butyryl-CoA and acetate to yield acetyl-CoA and the corresponding fatty acid. Transformation of butyryl-CoA to caproyl-CoA probably happens with the same set of enzymes (ACAT, HAD, ECH and BCD/EtfAB) and a CoAT to remove the CoA from caproyl-CoA, resulting in the formation of n-caproate. We came up with the same assumption as described for the ethanol-based chain elongation mechanism of C. kluyveri  – caproyl-CoA can be a further elongated acyl-CoA when a second analogous cycle proceeds, and CoAT was reported to have a broad substrate specificity [26, 27]. All three genomes contain the genes encoding ACAT, HAD, ECH, BCD, EtfAB and CoAT (Additional file 4 including the summary of all related CDSs). As for BL-3, three sets of ACAT, HAD, ECH, BCD and EtfAB genes are present in the genome, with one cluster encoding CoAT, ACAT, ECH and HAD (Fig. 6a, CDS labels: 13110–13113) as well as one cluster encoding ECH, BCD, EtfAB and HAD (Fig. 6a, CDS labels: 20308–20313), other CDSs are scattered in the genome. As for BL-4, one gene cluster encoding all six enzymes is present in the genome (Fig. 6b, CDS labels: 1867–1873). Two similar clusters were found in the genomes of Eubacterium limosum (Fig. 6k, CDS labels: 21760–21785) and Eubacterium pyruvativorans (Fig. 6i, CDS labels: 280031–280037). Another set of HAD, ACAT, ECH, CoAT genes clusters together with acetyl-CoA:oxalate CoA-transferase (ACOCT) and (R)-2-hydroxyisocaproyl-CoA dehydratase (HadABC) genes (Fig. 6b, CDS labels: 1158–1165). The genome of BL-6 harbors two sets of the ACAT, HAD, ECH, BCD and EtfAB genes separated into several sub-clusters, with one comprising genes for HAD, ACAT, ECH, CoAT and HadABC (Fig. 6c, CDS labels: 0555–0562) and two sub-clusters of genes encoding the BCD/EtfAB complex. One set of genes encoding the BCD/EtfAB complex is located in the same cluster with genes for LDH, LacR and LacP (Fig. 6c, CDS labels: 3216–3223) as mentioned above. We found that the genes encoding BCD are in close vicinity to the genes of EtfAB in the genomes of all three isolates (Fig. 6a-6c), which is commonly conserved as a key feature among all genomes of other chain-elongating bacteria (Fig. 6d-6n). Besides CoAT, the acyl-CoA thioesterase (ACT) may also catalyze the formation of n-butyrate and n-caproate from the terminal acyl-CoA (Fig. 5). Our data suggest that the genome of BL-3 may encode the predicted proteins annotated as thioesterase superfamily proteins. We further compared their protein sequences in all the databases used (see the results in Additional file 5) and confirmed that these thioesterase proteins were not involved in the terminal step of reverse β-oxidation (see CDS labels and final annotations in Additional file 4, sheets BL-3). Genomes of BL-4 and BL-6 both contain the CDSs for ACT (see CDS labels in Additional file 4, sheets BL-4 and BL-6), but presenting a low identity (≤ 40%) to proteins in the databases (see alignment details in Additional files 6 and 7).
Another possible pathway for the n-butyrate formation from n-butyryl-CoA was identified in the genome of BL-3. As shown in Fig. 5, the generation of butyrate phosphate is catalyzed by phosphate butyryltransferase (PTB). Thereafter, it is converted to butyrate via butyrate kinase (BUK), producing one ATP in this step.
The formation of iso-butyrate as a product of lactate-based chain elongation was experimentally proven in all three isolates. However, the genome analysis did not reveal hints on the assumed pathway, i.e. reversible n-butyrate/iso-butyrate isomerization [28, 29]. As described by Matthies and Schink , the conversion of n-butyrate to iso-butyrate first requires activation to n-butyryl-CoA. Next, the isomerization of n-butyryl-CoA via iso-butyryl-CoA to iso-butyrate is catalyzed by a butyryl-CoA:isobutyryl-CoA mutase (BM) and an isobutyryl-CoA:acetate CoA transferase (CoAT) as shown in Fig. 5. All three genomes lack potential CDSs for BM. The only homologue was found in the genome of BL-3 annotated as methylmalonyl-CoA mutase (see the CDS labels and final annotations in Additional file 4, sheets BL-3), which is known to isomerize (R)-methylmalonyl-CoA to succinyl-CoA, a step involved in the propionate fermentation pathway. These mutases catalyze the rearrangement of carboxyl groups represented as migration to the adjacent carbon atom, in which enzyme activities depend on coenzyme B12 . One possible reason for the conversion of n-butyrate to iso-butyrate is that bacteria can maintain the pool of iso-butyrate for synthesizing valine during growth in amino acid-deficient medium . As this isomerization step does not release any free energy, another possible explanation is that bacteria try to overcome inhibition effects of the accumulated nbutyrate, because the corresponding fatty acid of the unbranched form is more toxic than the branched form. As suggested for a methanol-based chain elongation system [3, 12], the formation of iso-butyrate may facilitate bacteria to further obtain energy from the chain elongation process.
Energy conservation and hydrogen formation
As shown in Fig. 5, the cytoplasmic BCD/EtfAB complex catalyzes the transformation of crotonyl-CoA (hexenoyl-CoA) to butyryl-CoA (caproyl-CoA) and simultaneously transfers electrons from NADH to ferredoxin, a mechanism that has been described as flavin-based electron bifurcation . ATP can be produced by the ATP synthase using the ion motive force that is generated by a membrane-associated, proton-translocating ferredoxin:NAD+ oxidoreductase (Rnf complex) in the oxidation of ferredoxin . The genomes of BL-3 and BL-4 contain the operon arranged as rnfCDGEAB encoding the six subunits of the Rnf complex as shown in Fig. 6a and 6b. This gene organization (shown as rnfBAEGDC in the other DNA strand) was also found in other genomes of chain-elongating bacteria (Fig. 6d-n). For the genome of BL-6, we could only find four genes for subunits of the Rnf complex during the functional annotation (see CDS labels in the Additional file 4, sheet BL-6), but it contains the CDSs encoding the analogous membrane-associated energy-converting hydrogenase (Ech complex), which was proposed to generate hydrogen for maintaining the cytoplasmic redox balance caused by the oxidation of ferredoxin [34, 35]. As shown in Fig. 6c, CDS labels 2699–2708, a cluster encoding six subunits of the Ech complex and CDSs for the hydrogenase maturation were found. The Ech complex was also identified in the MAG of Candidatus Weimeria bifida (Fig. 6m). Additional hydrogenases include hydrogen:ferredoxin oxidoreductase (H2ase), which was found in the genomes of all three isolates, and the bifurcating [Fe-Fe]-hydrogenase (HydABC) using electrons from NADH and reduced ferredoxin, of which no homologous genes were detected (see CDS labels in Additional file 4, sheets BL-3, BL-4 and BL-6).
Apart from the BCD/EtfAB complex, the predicted EtfAB-containing complexes for energy coupling may also include the LDH/EtfAB complex. The redox potential of the pyruvate/lactate pair (E0’ = -190 mV) is much higher than that of the NAD+/NADH pair (E0’ = 320 mV), which introduces a thermodynamic bottleneck of the lactate oxidation coupled to NAD+ reduction. Our annotation results show that strains BL-3, BL-6 and Ruminococcaceae bacterium CPB6 have LDH genes next to EtfAB genes (Fig. 6a, CDS labels: 11486–11488; Fig. 6c, CDS labels: 3217–3220; Fig. 6e, CDS labels: 01780–01790). Therefore, similar like the mode of lactate metabolism in the strict anaerobic acetogen Acetobacterium woodii, we assume that the LDH/EtfAB complex of these species can also use flavin-based electron confurcation to solve the energetic enigma: driving electron flow from lactate to NAD+ at the cost of exergonic electron flow from reduced ferredoxin to NAD+ [33, 36].
The manually curated annotation of all above-mentioned CDSs in the genomes of other lactate-based chain-elongating strains are provided in Additional files 8–12 (CDSs predicted with all methods integrated in the MicroScope platform).