3. Results and discussion

3.1 Characteristics of the SRP core genome and pan-genome

The pan-genomes of three SRP genera are assembled to highlight the genetic and metabolic diversity of SRPs. Based on the sequence similarity analysis, a phylogenetic tree created for all 24 SRPs together shows a clear separation of Archaeoglobus ,Desulfovibrio , and Desulfotomaculum (Figure S1), demonstrating that they belong justifiably to three different genera. Pan-genome of every genus can be subdivided into three subsets: (i) the core genome (genes shared among all species and mostly play a housekeeping role), (ii) the accessory genome (genes present in some but not all species and contribute to species adaptation to different conditions), and (iii) the unique genome (species-specific genes). Pan-genome subset distribution and function profile ofArchaeoglobus , Desulfovibrio , and Desulfotomaculumare shown in Figure 2, and pan-genome and core genome development plot projections are provided in supplementary Figure S2. The core genomes of all three genera display a similar evolution, which rapidly stagnates at close to 800 genes. The proportion of core genome in the pan-genome varies significantly (from 3.7% to 25.8%) due to the differences in pangenome size, but remains relatively low, revealing a rather high inter-species diversity of each genus. In terms of functional annotation according to COGs, almost half of the genes from the core genomes of the three genera are devoted to translation, energy metabolism, and the metabolism of building blocks (carbohydrates, amino acids, nucleotides, coenzymes, and lipids) (Figure 2), which supports Ouzounis and Kyrpides (1996)’s hypothesis that metabolism and translation are conserved and close to the last universal ancestor. Besides, a significant fraction (more than 20%) of core genes of Archaeoglobus andDesulfotomaculum encodes poorly characterized proteins (categories R and S) whose function is hypothetical, while that fraction of Desulfovibrio is only 8.2% because Desulfovibriostrains have been intensively studied (Barton & Fauque, 2009).
Key genes and associated functional proteins involved in central carbon and energy metabolism are provided in the ortholog table generated from the pan-genome analysis (Table S1). Overall, all SRP species contain complete gene sets for glycolysis, while adopting an incomplete pentose phosphate pathway and an incomplete TCA cycle. A complete set of genes for the Wood-Ljungdahl pathway (Schauder et al., 1986; Spormann & Thauer, 1988) was identified in Archaeoglobus . The essential enzymes for DSR are conserved in all studied SRPs, including sulfate transporters, ATP sulfurylase (Sat), APS reductase (AprAB), and dissimilatory sulfite reductase (DsrAB). Two electron-transporting enzyme complexes, dissimilatory sulfite reductase (DsrMKJOP) and quinone-interacting membrane-bound oxidoreductase (QmoABC), are conserved across SRPs, which are assumed to be involved in electron transfer from electron donors to the terminal reductases AprAB and DsrAB respectively. In addition,Desulfovibrio contains several other electron transfer complexes such as Hmc (high-molecular-weight complex), Tmc (transmembrane complex), Rnf (ferredoxin: NAD reductase), Nuo (NADH: quinone oxidoreductase complex), and energy-conserving hydrogenases (Ech or Coo), but few of them are found in Archaeoglobus or Desulfotomaculum , suggesting that electron transfer for sulfate reduction is substantially different among SRPs.

3.2 Model construction and validation

Based on the orthologous relationships resulted from pan-genome analysis, core metabolic models of the 24 SRPs are constructed and used to simulate their metabolic capabilities and growth characteristics under different growth conditions. The resulting core metabolic models consist of 91-110 reactions (Figure 3). They share most of the reactions related to central carbon metabolism but vary significantly in electron transfer pathways. Clustering of the models shows that the phylogenetically closely related SRP species have similar metabolic capabilities. Detailed information of all the reactions and the GPR associations, as well as the model files of 24 SRPs can be found on Github (https://github.com/TANG-Wentao/SRP_MetabolicModel).
To demonstrate the predictive accuracy of the core metabolic model, steady-state growth simulation was performed on the metabolic model under two different growth conditions: lactate-limited and hydrogen-limited sulfate respiration (LS and HS, respectively). The predicted growth yields (YX/S) and specific growth rates (μ) are compared to the experimentally determined values. Multiple literature datasets describing growth yields of DvH are available. The growth yields are dependent on the specific growth rate (μ) and the maintenance coefficient (m) (Russell & Cook, 1995); thus, growth rates and growth yields from both experimental results and model predictions are compiled according to Eq. 1 (Pirt, 1965). As shown in Figure 4, under LS and HS conditions, the predicted growth yields and specific growth rates of DvH are in close agreement with the experimental data.
\begin{equation} \frac{1}{Y}\ =\ \ \frac{m}{\mu}\ +\ \ \frac{1}{Y_{\max}}\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ Eq.1\nonumber \\ \end{equation}
For HS growth, a net synthesis of 1.21 ATP molecule per sulfate reduced is predicted from metabolic modeling, which is close to the theoretical value proposed by a previous study (Badziong & Thauer, 1978). By comparing the growth yield data and the growth energy requirement ofD. vulgaris under HS growth, a net yield of 1 ATP molecule per sulfate reduced was proposed by Badziong and Thauer (1978). The stoichiometry of cell synthesis and energy reactions for growth ofDesulfovibrio under LS growth is provided in Table 2. More than 90% of electron donor is used for energy generation, and the rest is used for cell synthesis, which is in agreement with previous studies (Noguera et al., 1998; Traore et al., 1981). Our results demonstrate that the high-quality core metabolic model equipped with improved annotations in ETC pathways can accurately predict the growth characteristics and energy yield of SRPs, providing a basis for further quantitative analysis of energy metabolism.

3.3 Lactate-sulfate (LS) respiration

Lactate is the most widely used growth substrate by known SRPs and often applied for cultivation purposes. The central carbon metabolism and the predicted metabolic flux of essential reactions associated with energy generation under LS growth are shown in Figure 5 and Figure 6a. Lactate oxidation to acetate is mediated by lactate dehydrogenase (Ldh), and the produced pyruvate from lactate oxidation is then converted to acetyl-CoA via pyruvate: formate lyase (Pfl) and/or pyruvate: ferredoxin oxidoreductase (Por). The predicted pathway of lactate oxidation agrees well with previous studies (Keller & Wall, 2011; P. M. Pereira et al., 2008). Acetyl-CoA is mainly used to produce ATP through substrate-level phosphorylation by Acetyl-CoA synthetase (ADP-forming, Acs), or phosphotransacetylase (Pta)/acetate kinase (Ack) couple. Consistent with previous genome analysis, this pathway is highly conserved in SRPs (Vita et al., 2015). For SRPs without Acs or Pta, such as Desulfovibrio magneticus andDesulfovibrio gigas , acetyl-CoA is predicted to be converted to acetaldehyde by aldehyde:ferredoxin oxidoreductase (Aor) or to citrate by citrate synthase (CIT), which are in turn converted to ethanol and succinate respectively.
For Desulfovibrio and Desulfotomaculum species except forD.magneticus and D. gigas , a theoretical ratio of acetate excretion to lactate utilization of 1:1 is expected in energy metabolism based on the stoichiometry of this pathway, and an average acetate-to-lactate ratio of 0.907:1 is predicted (Figure 6b & Table 2). A difference exists between the theoretical ratio and the predicted value because a small amount of acetyl-CoA is used for the synthesis of the building blocks, and this part of acetyl-CoA plays a non-energetic, assimilatory role in SRPs. For the complete oxidizers, e.g.Archaeoglobus species, acetyl-CoA is predicted to be further utilized by the Wood–Ljungdahl pathway (Schauder et al., 1986; Spormann & Thauer, 1988) for energy generation by coupling the endergonic oxidation of acetate to CO2, thus no acetate is excreted and the ratio of CO2 to lactate is triple that of complete oxidizers (Figure 6b). The ATP yield (ATP/sulfate ratio) varies significantly across different SRPs, with a predicted average of 2.35.
Electrons generated through lactate oxidation are assumed to be used indirectly for sulfate reduction by the hydrogen (H2) cycling pathway (Odom & Jr, 1981). Another study (Lupton et al., 1984) argued that H2 is not an obligatory intermediate, but is produced only as a mechanism to control the redox state of internal electron carriers, and the electrons transport directly through membrane-bound electron carriers to sulfate. Experimental evidence partially supports each hypothesis (Keller & Wall, 2011; Noguera et al., 1998). Based on our modeling work (Figure 6a and Figure 7), these two electron transport mechanisms can complement—rather than compete with—each other for Desulfovibrio . As shown in Figure 7, for the direct pathway, electrons flow from lactate to the membrane-bound electron carrier menaquinone (MQ) and then transfer to sulfate reduction (via Qmo and DsrMK) without H2 involvement. In general, the consumption and evolution of H2 are mediated by hydrogenases (Hase), which are redox metalloenzymes that catalyze the reversible oxidation of H2 (Vignais & Colbeau, 2004). For the H2 cycling pathway, electron equivalents (Fdred and MQH2) and protons generated from Ldh will react with membrane-bound Hase Ech and/or Coo to form H2, which will diffuse to the periplasm and be oxidized back to protons and electrons by periplasmic Hase (Hase_p). The electrons from H2 oxidation are transported to the electron equivalent Type I cytochrome c3(TpIc3), and then transferred back to the cytoplasm via Hmc and Qrc for sulfate reduction, creating a transmembrane proton gradient for ATP synthesis. This predicted phenomenon is supported by a previous study of Desulfovibrio gigas , where only one periplasmic and one cytoplasmic Hase are present. Single deletion of mutants for each of these proteins showed slightly lower growth rates with lactate-sulfate or pyruvate-sulfate than the wild-type strain, suggesting that H2 cycling can be compensated by the direct MQ mediated electron transport (Morais-Silva et al., 2013). Based on our simulation, the so-called H2-cycling mechanism is responsible for around half of the electron transfer inDesulfovibrio under LS growth, except Desulfovibrio alaskensis G20 who contains no Ech or Coo.
Genera Archaeoglobus and Desulfotomaculum do not use the H2 cycling pathway for energy conservation under LS growth due to the lack of a periplasmic space for H2oxidation and the absence of transmembrane electron transfer proteins such as Qrc and Hmc. For Desulfotomaculum species, H2 generated from Ldh is predicted to involve in the flavin-based electron bifurcation mechanism (see section 3.5 for details) for the generation of reduced ferredoxin (Fdred) and NADH via cytoplasmic Hase (Hase_c), thus serving as a redox balancing strategy as suggested by Lupton et al. (1984). NADH and Fdred further drive sulfite reduction via heterodisulfide reductase (Hdr) and/or dissimilatory sulfite reductase (DsrC). For Archaeoglobus species, H2generation is not predicted during LS growth, and a new kind of electron equivalent, reduced coenzyme F420(F420H2), is introduced. F420H2 is generated as a result of complete oxidation of lactate into CO2 via the Wood-Ljungdahl pathway, and is further converted to MQH2by the membrane-bound electron carriers Fqo (F420H2: quinone oxidoreductase). NADH is somewhat a substitute for H2 in Archaeoglobusand Desulfotomaculum . It not only works as the redox currency, but also directly drives energy conservation by Nuo (NADH: quinone oxidoreductase).

3.4 Hydrogen-sulfate (HS) respiration

Molecular hydrogen (H2), an electron sink product of anaerobic fermentation, can be utilized via DSR, methanogenesis, and acetogenesis to enable more efficient energy recovery from organic substrates (Rowland et al., 2018). When H2 is the only available electron donor, most SRPs will grow heterotrophically on acetate, and carbon dioxide can serve as a supplementary carbon source. The average carbon dioxide-to-acetate ratio of SRPs is predicted to be 0.96, which means 67.6% of the carbon is derived from acetate and the other 32.4% comes from carbon dioxide (Figure 8 &Table 2). This result agrees well with experimental findings from previous work (Badziong & Thauer, 1978), which reported that 70% of the carbon is derived from acetate, and the remaining 30% comes from carbon dioxide.Archaeoglobus sulfaticallidus are capable of coupling chemolithoautotrophic growth on H2/CO2to sulfate reduction in addition to heterotrophic growth. ForArchaeoglobus other than A. sulfaticallidus , although the machinery for CO2 fixation via the Wood–Ljungdahl pathway exists, autotrophic growth with H2/CO2 is not feasible. This is because Fdred generated from Hase is completely used up for APS reduction (Qmo), CO2 fixation through the Wood–Ljungdahl pathway cannot be promoted without Fdred.
Metabolic fluxes predicted by FBA for essential reactions related to central carbon metabolism and energy generation are shown in Figure 8a. As expected, a strikingly different flux distribution is observed under HS growth condition as compared to LS growth condition. With H2 serving as the electron donor for growth,Desulfovibrio is predicted to establish a pmf directly by periplasmic H2 oxidation (Hase_p). Besides, the membrane-bound Hase Ech can function in reverse to oxidize H2 and generate Fdred for carbon fixation during growth with H2 for severalDesulfovibrio species. Archaeoglobus andDesulfotomaculum species are predicted to use soluble cytoplasmic Hases (MvH and Hase_c) that are not present in Desulfovibrio , and these Hases are involved in electron bifurcation mechanisms to produce Fdred and balance the redox state of internal electron carriers (see 3.5 for details). Por (pyruvate: ferredoxin oxidoreductase) involves in energy metabolism of all examined SRPs under both LS and HS growth. It can work bidirectionally to oxidize pyruvate into acetyl-CoA and CO2 accompanied with Fdred generation under LS growth, or to synthesize pyruvate via CO2 fixation fueled by Fdred under HS growth. For several SRPs (Figure 8a), formate is predicted to be generated from pyruvate via Pfl and is then oxidized to protons and electrons by periplasmic formate dehydrogenase (Fdh). This formate cycling pathway performs a similar function to the H2 cycling pathway, which is supported by the transcriptome study of P. M. Pereira et al. (2008).

3.5 Energy conservation mechanisms

In sulfate respiration, energy is mainly generated through the conversion of the free energy of a redox reaction into a pmf, which drives ATP production by ATP synthase (ATPase). Despite intensive study, the mechanism of energy conservation during sulfate respiration is still incompletely understood. By simulating the steady-state growth of 24 SRPs under different growth conditions, three mechanisms are proposed for energy conservation via oxidative phosphorylation in this study, as shown in Figure 9.
The MQ-based redox loop is the universal machine used for energy conservation in SRPs. In this scenario, proton translocation is achieved via charge separation performed by coupling MQ reduction with MQH2 oxidation (Simon et al., 2008). A MQ reductase such as Qrc, Hmc, Nuo, and Fqo, together with a MQH2 oxidase such as Qmo, Coo, and DsrMK, constitute the two arms of an energy-conserving redox loop (Figure 9a). The charge separation is established by electron extraction from the positive (P) side and proton uptake on the negative (N) side of the membrane by MQ reductases, paired with electron transfer to the N side and proton release to the P side by MQH2 oxidases (Figure 9a). The second energy conservation mechanism is to build up a pmf via direct proton pumping (Figure 9b). Several electron transfer complexes capable of proton pumping are found in SRPs, including Nuo, Fqo, Rnf, Ech, Coo, and PPi (inorganic pyrophosphate), which provides SRPs with great flexibility in terms of energy metabolism. The Ech and Coo Hases are involved in both direct proton pumping and indirect H2 cycling (Figure 9c). Ech and Coo can convert protons and electron equivalents (Fdred and MQH2) that are generated by the oxidation of organic compounds (e.g., lactate) to H2, simultaneously translocate protons. The H2 is then reoxidized to protons and electrons (TpIc3 pool­) by periplasmic Hase (Hase_p). The H2 cycling has to be operated at very low H2partial pressure to allow H2 generation through Ech and Coo thermodynamically feasible. This is because the H-/H2 couple has lower redox potential (E°’= -414 mV) than MQ/MQH2 (E°’= -74 mV) and Fdred/Fdox (E°’= -398 mV) at standard conditions (Table 3), and the decrease of H2 partial pressure to 1 Pa will increase the lower redox potential of H-/H2 couple to -270 mV. Formate cycling across the membrane works as an alternative energetic pathway of H2 cycling, and carbon monoxide (CO) cycling is also proposed during fermentative growth (Voordouw, 2002).
Flavin-based electron bifurcation (FBEB) is a novel mechanism by which a hydride electron pair from H2, NAD(P)H, F420H2, or formate is split by flavoproteins into one-electron with lower redox potential and one with a higher redox potential than that of the electron pair (Buckel & Thauer, 2018). FBEB is not directly involved in energy conservation; instead, it supports energy conservation through Qmo, Rnf, Coo, and Ech by generating Fdred from relatively high-potential electron donors such as NADH (Figure 9d). The discovery of electron bifurcation also helps to resolve the physiological role of MQ in DSR. MQ pool is widely accepted as a critical mediator in the electron transfer and energy generation in SRPs. However, the redox potential of MQ/MQH2 (E°’= -74 mV) is higher than most of the redox couples found in SRPs (Table 3) and is very similar to APS/HSO3- (E°’= -65 mV). Therefore, MQ can readily receive electrons from other electron donors, whereas donating electrons from MQH2 to other electron acceptors seems thermodynamically unfavorable. Via electron bifurcation, MQH2 oxidation can be achieved by introducing a second electron donor (e.g., Fdred) with lower redox potential, so that electron transfer via Qmo and Coo can proceed (Figure 9 a and c).
Overall, MQ-based redox loop and FBEB are widely used to conserve energy in SRPs. H2 cycling is restricted toDesulfovibrio , and Nuo induced proton pumping plays a crucial role in Archaeoglobus and Desulfotomaculum . A wide variety of proteins are involved in energy conservation of SRPs, and our simulations show that alternative energy-conserving proteins and different energy-conserving mechanisms can be used in response to shifting nutrient availability, which enables SRPs to thrive in a range of different environmental conditions.