1. Introduction

Sulfate-reducing prokaryotes (SRPs) are anaerobic bacteria and archaea ubiquitously distributed in nature. They can couple dissimilatory sulfate reduction (DSR) to the oxidation of organic substrates or hydrogen/carbon dioxide in their energy metabolism, and therefore function as crucial mediators of the global sulfur and carbon cycle (Muyzer & Stams, 2008; Rudolf K. Thauer et al., 2007). The corrosive and toxic hydrogen sulfide resulting from DSR is responsible for significant damage to the environment and industrial processes, and is also implicated in human health problems (Beech & Sunner, 2007; Goldstein et al., 2003; Loubinoux et al., 2002; Pikaar et al., 2014; Rückert, 2016; Wang, 2012). On the other hand, the biogenically produced sulfide can be utilized beneficially to remove and recover heavy metals from groundwater and wastewater by capturing the dissolved metals into insoluble metal sulfides (Hao et al., 2014; K. Tang et al., 2009). In addition, the low redox potentials achieved by SRPs enable them to reduce several toxic metals and radionuclides, including divalent mercury [Hg(II)] and the oxyanions of chromium [Cr(VI)], uranium [U(VI)], and arsenic [As(V)] (Bruschi et al., 2007; Lloyd, 2003). SRPs exhibit versatile energy metabolic capacities by using various electron donors and carbon sources under a wide variety of environmental conditions, which also enables them to degrade municipal organic wastes, hydrocarbons, and crude oil in industrial and environmental applications (Hao et al., 2014; Ollivier et al., 2007; Rabus et al., 2013; Rabus et al., 2015). A fundamental understanding of the energy metabolism of SRPs growing with different energy sources is thus crucial for improving their performance in SRP-based biotechnologies, as well as for controlling their activity when they are undesired.
Extensive genomic and molecular studies of SRPs have been conducted in recent years, providing an enormous amount of information on their genetics, physiology, and biochemistry (Muyzer & Stams, 2008; P. M. Pereira et al., 2008; Rabus et al., 2015; Walker et al., 2009). However, a systematic understanding of their energy metabolism remains elusive (Muyzer & Stams, 2008; Rückert, 2016). For instance, DSR has long been recognized to be associated with energy conservation via oxidative phosphorylation, which implies an electron transport chain translocating charges across the cell membrane and generating a proton motive force (pmf) (Fitz & Cypionka, 1989; Kobayashi et al., 1982; Peck, 1960; Wood, 1978). However, the terminal reductase of the DSR pathway, dissimilatory sulfite reductase (DsrAB), is located in the cytoplasm, thus failing to be directly involved in the membrane-associated charge translocation (I. A. C. Pereira et al., 2011; Rabus et al., 2015). The interaction of DsrAB with the membrane-bound electron transport chain to generate a pmf remains a mystery. More importantly, previous analyses indicated that the composition of energy metabolism proteins could vary significantly from one SRP to another (Rabus et al., 2015; Zhou et al., 2011), whereas current biochemical and molecular studies on SRPs are primarily focused on a few model organisms only, such as Desulfovibrio desulfuricans and Desulfovibrio vulgaris (Barton & Fauque, 2009; Rabus et al., 2013). Besides, SRPs can switch from one mode of energy generation to another in response to shifting nutrient availability (P. M. Pereira et al., 2008; Walker et al., 2009). For these reasons, a systematic database describing the energy metabolic properties of SRPs is highly sought after.
Genome-scale metabolic models (GEMs) can systematically compile genomic, biochemical, genetic and high-throughput omics data, thereby forming a mathematically-structured knowledge-base for investigation of metabolic capabilities, generation and testing of hypotheses, and analysis of growth characteristics (King et al., 2016; Thiele & Palsson, 2010). By using GEMs, not only engineers can obtain broader insights into the energy metabolism of SRPs, but also can quantitatively predict the metabolic behaviors of SRPs in engineered ecosystems. Despite their utility, high-quality GEM construction is a time-consuming and labor-intensive process, involving multiple computational and experimental steps, which limits their application to only a few well-understood model organisms (J. Monk et al., 2014; Thiele & Palsson, 2010). To address this challenge, automated GEM reconstruction tools have been developed that can generate an increasing number of GEMs (Gu et al., 2019; Machado et al., 2018). Nevertheless, without manual curation, the quality of an automatically generated draft GEM is always low, due to lower-confidence annotations, extensive gap-filling, and incompleteness and/or inaccuracy of the existing databases involving genes, enzymes, reactions, and metabolites (Edirisinghe et al., 2016; Gu et al., 2019). Many of these problems can be avoided by using a simplified core metabolic model comprised of only the well-studied and biologically critical pathways (Edirisinghe et al., 2016; J. D. Orth et al., 2010a), thus significantly simplifying the laborious construction process while maintaining the utility of GEMs. The core metabolic model aims at a focused understanding of central and energy metabolism. The chosen pathways in the model are often the subjects of textbooks and should be familiar to most readers with basic biochemistry knowledge. To further improve the accuracy and efficiency of metabolic model construction, comparative genomics has been integrated into GEMs by using curated reference GEMs from closely related organisms to derive multiple metabolic models for various strains of a single species (J. M. Monk et al., 2013) or multiple species of a single genus (Seif et al., 2018).
In view of the above, this study aims to develop an efficient approach for constructing metabolic models of multiple organisms, and to apply it to 24 SRP species to study their energy metabolism quantitatively and systematically. To this end, a combination of comparative genomics via pan-genome analysis and the concept of core metabolic model has been applied. The global workflow illustrated in Figure 1 can be divided into four major steps: 1) pan-genome analysis of SRPs to obtain orthologous genes encoded proteins related to central and energy metabolism; 2) pairwise comparison of the gene-protein relations from pan-genome analysis and the reaction-protein relations from the model template to obtain gene-protein-reaction (GPR) associations for construction of 24 SRP metabolic models; 3) validation of the metabolic model of the model SRP Desulfovibrio vulgarisHildenborough (DvH); and 4) Flux balance analysis (FBA) of the developed models to evaluate the energy metabolism of SRPs under different growth conditions.