Discussion
The diversity of SRO and their DsrAB enzymes has been greatly expanded through recent cultivation-dependent and cultivation-independent approaches (11, 23, 24). Together these studies suggest that Dsr likely emerged to catalyze SO32- reduction and then diversified (through recruitment of APS and Sat) to catalyze SO42- reduction and ultimately HS- oxidation (9–12). The phylogenetic studies conducted herein suggest that model bacterial SROs implicated as major players in contemporary biogeochemical S cycling (e.g., Deltaproteobacteria and Firmicutes) evolved comparatively recently whereas early evolving archaeal SROs (and a few taxonomically patchy bacterial genera) tend to be restricted to hydrothermal or more nutrient-limited extreme environments (11) where oxidant limitation is likely pervasive (50). While the ecological drivers of the evolution of SROs (via Dsr phylogeny) is obscured in the present study by limited corresponding metadata (e.g., cardinal growth parameters, geochemistry) associated with these organisms or the environments from where they were recovered, the broad differences in habitats of early evolving and later evolving SROs suggests that the ecology of Dsr-harboring SROs has evolved over time. Yet, it remains unclear if the structure and thus, functional mechanisms of Dsr have also evolved during its evolutionary history.
Despite the collective abovementioned observations indicating SRO (and thus Dsr) diversification across gradients in temperature, pH, salinity, and pressure, amongst other variables, the phylogenetic studies conducted herein and elsewhere document a general pattern of vertical inheritance and a high degree of overall primary sequence conservation across all DsrAB (11, 23, 24). Consistent with these findings, available structures and our structural models of selected Dsr enzymes that represent much of the known sequence diversity of Dsr generated herein reveal a high degree of structural conservation. Dsr forms a heterotetrameric structure comprising two heterodimers of DsrA and DsrB, the latter of which arose from an ancestral gene duplication (22, 51). The high degree of structural conservation, including at the inferred quaternary level, suggests that all extant lineages of SROs settled on this Dsr structural configuration prior to the radiation of SROs.
The evolutionary co-variance in the extensive inferred conservation in the structure of Dsr revealed a group of three residues in each heterodimer that form a pathway between the two active site sirohemes. Based on the homologous monomeric FPECs from both the A. fulgidusand MV2-Eury A subunit queries, and a second set of monomeric FPECs from the A. fulgidus B subunit query adjacent to two of the three residues, we conclude that the N393A2/1-T351B1/2 (A. fulgidus ) pair is truly coupled evolutionarily. Given the position of N180B1/2 interacting with the active site heme and the Fe-S center, and its relatively close proximity to the N393A2/1 residue, we hypothesize that N180B1/2-N393A2/1 are also evolutionarily coupled. Based on structural link formed by these residues between the two active sites, we hypothesize that, beyond simply stabilizing contacts between heterodimers, the pathway traced by these residues could correspond to an ancient allosteric pathway (i.e., the “heme road”) that could have provided the advantage of allosteric control when the heterodimer to heterotetramer transition occurred. While these heme road residues are not in direct contact, their interactions are mediated by only one or two aromatic residues. Consistent with the existence of an allosteric pathway are the hinge motions predicted by ANM. Inter-domain motions in multi-chain protein complexes are an increasingly appreciated aspect of their dynamic structures and thermodynamics, with the potential to modulate their functions (52–54). Our Gaussian Network normal mode analysis (40, 41, 49) indicates the possibility of intersubunit “rocking” in the structure of the DsrAB complex, like that observed for other dimeric complexes (52). These motions could in principle be frozen out by crystallization of multiple members of the DsrAB family, as also observed across crystal structures of dimeric influenza NS1 protein domains (52). In further support of such interfacial dynamics in DsrAB complexes, which could also relate to the proposed intersubunit allostery, is the lower quality of the electron density maps noted by the authors of the A. fulgidus structure between the A2B2, as opposed to the A1B1, heterodimers (20). This difference in quality could reflect an asymmetry in the structures of the two heterodimers due to allosteric interactions. Re-evaluation of the A. fulgidus DsrAB crystal structure, 3MM5, suggests that the difference between the two heterodimers in the crystal structure comes from distinct packing. This allows for greater flexibility that, according to our TLS or ensemble refinements, appears to be related to rigid-body movements, consistent with our ANM normal mode analyses.
The proposed heme road allosteric pathway could serve to allow for communication between the two active sites, one in each heterodimer, during the delivery of 2 e- from DsrC to one of them. Such an interaction might inhibit DsrC binding and electron injection into one heterodimer, while the other is active. . Interestingly, it has recently been proposed that DsrD, a small protein that is found in late-evolving SROs, acts as an allosteric regulator of DsrAB (55). This is the case for negative cooperativity in the function of the Mo-Fe nitrogenase heterotetramer (29), which exhibits similar rocking normal modes as DsrAB. Examples of statistically detected evolutionarily conserved pathways of energetic coupling within proteins have been reported for several proteins, including PDZ domains, GPCRs, chymotrypsin, lectin and hemoglobin (27, 56–59). We emphasize the importance of experimental validation of putative allosteric networks revealed by statistical analysis of sequence co-evolution. Such validation can be accomplished by combinations of approaches including H/D exchange mass spectrometry (60) and/or NMR (61), and spectroscopic approaches coupled with mutagenesis (62). We note that given the distances, we cannot rule out that the heme road corresponds to a pathway for electron transfer between active sites, although it is difficult to rationalize its utility.
While this proposed pathway warrants experimental scrutiny, the co-evolution at the positions putatively involved also indicates that this pathway was likely established prior to the radiation of all DsrAB, although the precise residues involved in these interactions vary across Dsr enzyme types. Consequently, the presence of a slightly modified allosteric pathway may have allowed fine-tuning of Dsr activity in the context of different physiological backgrounds, including those that operate under chronic energy (dissimilatory e-shuttling) stress imposed by extreme conditions (e.g., temperature, pH and pressure) (50) where the kinetics of HSO3- reduction and thus growth of SRO are likely to be much slower. The combination of the evolutionary coupling analysis and ANM dynamic “normal mode” calculations provide an intriguing set of residues to probe for their involvement in structure-function relationships in future experiments.
In addition to the aforementioned putative allosteric pathway, it is possible that the presence of the W119 residue in A. fulgidus Dsr and the loss of the iron in the sirohydrochlorin moieties in D. vulgaris Dsr represent two separate mechanisms to limit function to the “top” siroheme in each heterodimer, adjacent to the binding site of the DsrC putative electron carrying subunit. Evolution of the gaps and inserts in the region of the structural hemes may have contributed to their loss of function. Clearly, detailed comparative enzymatic assays will be required to demonstrate the structural, as opposed to functional, role of the “bottom” siroheme”.
Collectively, the combination of phylogenetic and structural bioinformatics studies of Dsr conducted herein point to the need for comparative experimental studies to test the present hypotheses in order to fully understand the functional differences encompassed within the diversity of these enzymes. This is particularly true for the two most-basal branching lineages of Dsr from SRO that are most reminiscent of the ancestral Dsr enzymes that likely shaped sulfur and carbon biogeochemical cycles on early Earth and that continue to shape these cycles in thermal environments. The vast majority of these taxa (inclusive of MV2-Eury) are either known only from cultivation-independent environmental genomics studies or have very-limited cultivation information. Thus, future efforts should first be made to domesticate these SROs and optimize cultivation conditions to enable more thorough investigations of their physiology, ecology, and enzymology. In particular, these efforts should focus on SROs that are physiologically unlike canonical SROs that conduct SO42- reduction but that rather are limited to SO32- reduction. Such investigations would enable a better understanding of the evolution of Dsr as it transitioned from early SO32- respiring organisms to the SO42- reducers that are widespread in anoxic environments on Earth today.