Introduction
Between 12-29% of the organic carbon that is delivered to the sea floor is mineralized by biological sulfate (SO42-)/bisulfite (HSO3-) reduction (1). As such, SO42-/HSO3-reducing organisms (SRO) play substantial roles in the global sulfur and carbon cycles, both today (2, 3) and in the geologic past (4–6). Dissimilatory reduction of SO3-/HSO3-to hydrogen sulfide (H2S) in most SRO is catalyzed by the enzyme dissimilatory sulfite reductase (Dsr) in a reaction that requires six electrons: SO32- + 6e- + 8H+ H2S + 3 H2O (7). Based on fractionation of sulfur isotopes preserved in sulfide and sulfate minerals in rocks, dissimilatory SO42-/HSO3-reduction (presumably via a Dsr-like enzyme) is thought to have evolved as early as 3.47 billion years ago (6). The earliest evolving SRO may have reduced SO3-/HSO3-(8–12), that would have been readily produced through solvation of sulfur dioxide (SO2) released into the biosphere via widespread volcanism on early Earth (13). Subsequently, the ability for SRO to use SO42- as an additional electron acceptor likely occurred in response to the gradual oxidation of Earth, culminating in the Great Oxidation Event (GOE) that occurred ~ 2.3 billion years ago (14). Preceding the GOE for several hundred million years, sustained production of O2 allowed for oxidative weathering of continental sulfides that led to the release of SO42- to oceans (15). At the same time, sustained production of O2 would have led to a decrease in the availability of HSO3-, since it is unstable in the presence of strong oxidants like O2 and Fe(III) (16), both of which became more abundant on an oxygenated Earth. The combination of decreased availability of HSO3- and increased availability of SO42- may have represented the selective pressure to recruit ATP sulfurylase (Sat) that catalyzes the ATP-dependent activation of SO42- to adenosine 5’-phosphosulfate (APS), and APS reductase (AprAB) that reduces APS to HSO3-, thereby allowing for the use of SO42- as an oxidant. In potential support of this model, the reduction of SO42- to HSO3- is an endergonic process (requires ATP), whereas HSO3-reduction is exergonic and is the major energy conserving step during SO42- reduction (17). Further evidence in support of HSO3- reduction preceding SO42- reduction comes from physiological studies that reveal higher growth yields in model SRO when grown with HSO3- relative to those grown with SO42- (18, 19).. As such, the use of SO42-, which imparts an additional energetic burden on SRO, appears to be an adaptation to allow for respiration of an oxidant, SO42-, that was much more widely available later in Earth history.
Extant Dsr enzymes are hetero-tetrameric and composed of two highly homologous A and B subunits thought to have evolved from gene duplication (20). Electrons for HSO3-reduction derived from small organic molecules (i.e., lactate) or H2 are thought to be transferred to DsrAB via a third labile subunit, DsrC, through two C-terminal cysteine residues (7). Thousands of DsrAB sequences have been generated from cultivars and from environmental amplicon-based or metagenomic surveys that have been used to characterize the ecology of SRO and/or to reconstruct their evolutionary histories (21–24). These studies have revealed that SRO inhabit a broad range of habitat types, including subsurface, hydrothermal, soil, and freshwater/marine sediment environments. Moreover, DsrAB are much more widespread throughout archaeal and bacterial lineages than suggested from cultivars only a decade ago (23, 24). In particular, metagenomic surveys have significantly expanded the known taxonomic and genomic backgrounds where DsrAB are found, although many of the taxa harboring DsrAB are uncultured and thus, the function of DsrAB in these organisms is inferred from closely related cultivars or based on phylogenetic clustering among defined DsrAB groups (23, 24).
The recovery of SRO and their corresponding Dsr sequences from environments that are subject to extremes of pressure, temperature, salt concentration, and pH indicate that the organisms harboring Dsr have diversified to function under diverse physiological conditions. For example, a novel DsrAB-encoding euryarchaeote (within the Diaforarchaea/Thermoplasmatota group) was recently discovered in moderately acidic (pH range of ~3.0 to 5.4), high temperature (~50°C to 75°C) springs in Yellowstone National Park (YNP) through metagenomic sequencing (11). Several nearly complete genomes were recovered across multiple springs and years that were representative of these organisms, and none encoded Sat or AprAB, consistent with the ability of these organisms to grow with HSO3-, but not SO42- (11). Phylogenetic analyses suggest these euryarchaeote DsrABs to belong to the earliest evolving lineage that also includes sequences belonging to other thermophilic Archaea largely found in hydrothermal vents or hot springs, and that which are generally inferred to respire HSO3- (but not necessarily SO42- ). These results add credence to the notion that Dsr evolved to allow for the reduction of HSO3- and later diversified to allow for the reduction of SO42- in habitats characterized by more modest environmental conditions. However, while it is clear that DsrAB has substantively diverged at the primary sequence level, it is unclear if this divergence translates to structural variation and whether conserved structural features of DsrAB exist that are invariant to change, irrespective of environmental conditions.
To begin to assess whether structural changes have occurred throughout the evolutionary history of Dsr, we modeled the structural characteristics of early-branching archaeal Dsr (i.e., from the newly characterized euryarchaotes described above) and compared these to those from an early-diverging group of Dsr (i.e., from the Archaeoglobales) and from a later-evolving group of Dsr (i.e., from Desulfovibrioand other Deltaproteobacteria). Sequence co-evolution (co-variance) analysis can provide information about residue-residue contacts and on functional coupling between distant sites (25–28). Sequence conservation and co-variance analysis of DsrAB sequences in the context of representative three-dimensional structures and models across the family was used to identify functionally-critical direct interactions, as well as longer-range potential functional couplings consistent with an intersubunit allosteric network, reminiscent of the negative cooperativity in the Mo-Fe nitrogenase (29). Observations of structural evolution are discussed in the context of environmental- and taxonomic-level adaptations that are likely to have taken place during the evolution of Dsr and the organisms that encode these enzymes.