Global diversity and inferred ecophysiology of microorganisms with the potential for dissimilatory sulfate/sulfite reduction

Abstract Sulfate/sulfite-reducing microorganisms (SRM) are ubiquitous in nature, driving the global sulfur cycle. A hallmark of SRM is the dissimilatory sulfite reductase encoded by the genes dsrAB. Based on analysis of 950 mainly metagenome-derived dsrAB-carrying genomes, we redefine the global diversity of microorganisms with the potential for dissimilatory sulfate/sulfite reduction and uncover genetic repertoires that challenge earlier generalizations regarding their mode of energy metabolism. We show: (i) 19 out of 23 bacterial and 2 out of 4 archaeal phyla harbor uncharacterized SRM, (ii) four phyla including the Desulfobacterota harbor microorganisms with the genetic potential to switch between sulfate/sulfite reduction and sulfur oxidation, and (iii) the combination as well as presence/absence of different dsrAB-types, dsrL-types and dsrD provides guidance on the inferred direction of dissimilatory sulfur metabolism. We further provide an updated dsrAB database including > 60% taxonomically resolved, uncultured family-level lineages and recommendations on existing dsrAB-targeted primers for environmental surveys. Our work summarizes insights into the inferred ecophysiology of newly discovered SRM, puts SRM diversity into context of the major recent changes in bacterial and archaeal taxonomy, and provides an up-to-date framework to study SRM in a global context.


Introduction
The sulfur cycle is one of the most important biogeochemical cycles on Earth (Canfield and Farquhar 2012 ) tightly interacting with carbon, nitrogen, and metal cycling (Jørgensen 2021 ).It is mainly regulated by activities of sulfate/sulfite-reducing microorganisms (SRM) and sulfur-oxidizing micr oor ganisms (SOM) as their counterparts (Dahl et al. 2008, Rabus et al. 2013, Rabus et al. 2015, Wasm und et al. 2017, Jør gensen 2021 ), whic h c ycle sulfur betw een its most oxidized (sulfate, + VI) and its most reduced state (sulfide, -II).On a global scale, sulfate reduction is one of the dominant processes driving the mineralization of organic matter in anoxic en vironments .Of the estimated 260 Tmol C org r eac hing the global seabed each year, one third is mineralized through sulfate reduction in marine sediments (Jørgensen 2021 ).About 90% of the end product, sulfide, is re-oxidized to sulfate either directly or indir ectl y at the expense of oxygen (Jørgensen 2021 ).This r epr esents 25% of global oxygen consumption in sediments and has a direct impact on the redox state of Earth's surface .T he relevance of sulfur cycling increases further in coastal sediments, where sulfate reduction accounts for 50% of C org mineralization and reoxidation of sulfide consumes 50% of the oxygen entering the sediment (Jørgensen 2021 ).In addition to marine sediments, marine oxygen minimum zones represent environments of active sulfur cycling.In these o xygen-de pleted waters, sulfide produced by sulfate reduction is rapidly re-oxidized to sulfate by sulfide oxidation coupled to nitrate reduction (Canfield et al. 2010, Johnston et al 2014, Callbeck et al. 2018, van Vliet et al. 2021 ).Here, the term "cryptic sulfur c ycle" w as coined for the first time-"cryptic" because it was not evident from the spatial concentration profiles of inorganic sulfur compounds, in particular sulfide (Canfield et al. 2010 ).
While the importance of sulfate reduction in marine environments is well explained by the high availability of sulfate (ca.28 mM), its role in biogeochemical cycling of anoxic freshwater envir onments suc h as lake sediments , groundwater, peatlands , or rice pad d y fields is less obvious because of the low pr e v ailing sulfate concentr ations (typicall y 10-300 μM) (Pester et al. 2012 ).Nevertheless, the rates at which sulfate reduction proceeds can be equally high in marine and freshwater settings, resulting in rapid cycling of sulfur species in anoxic freshwater en vironments .Because of its less ob vious r ele v ance and high variability in space and time, the sulfur cycle in freshwater systems is often r eferr ed to as a cryptic or hidden sulfur cycle as well (Pester et al. 2012 ).The contribution of sulfate reduction to C org mineralization in anoxic fr eshwater envir onments has not been e v aluated so systematically as in marine en vironments , but single studies report values of 17-35% in lake sediments (Urban et al. 1994, Thomsen et al. 2004 ) and 36-50% in peatlands (r e vie w ed in P ester et al. 2012 ).Yet another low-sulfate environment with cryptic sulfur cycling are deep marine subsurface sediments below the sulfate-methane tr ansition zone.Her e, sulfur cycling operates at very slow sulfate r eduction r ates .T hese slow r ates ar e maintained by the r e-suppl y of sulfate mediated by Fe(III)-driven sulfide oxidation (Holmkvist et al. 2011a,b , Pellerin et al. 2018, Jørgensen et al. 2019, Findlay et al. 2020 ).
Besides their r ele v ance for biogeoc hemical cycling, SRM r epr esent an important symbiotic guild in the mammalian intestinal tract (Barton et al. 2017 ) and are also beneficial in bioremediation, such as degrading hydrocarbons and removing heavy metals from sulfate-containing groundwater and wastewater (Muyzer andStams 2008 , Qian et al. 2019 ).Ho w e v er, they can also be an economic burden by causing steel corrosion or oil souring (Muyzer and Stams 2008, Rey et al. 2013, Rabus et al. 2015, Singh and Lin 2015, Wolf et al. 2022 ).In the context of climate change and human-induced eutrophication, it is noteworthy that oxygen concentr ations in pela gic zones of the global ocean, coastal waters, and lakes have been declining for decades (Jenny et al. 2016, Breitburg et al. 2018 ).The resulting oxygen-deficient zones can turn euxinic (anoxic conditions with > 0.1 μM sulfide) upon release of toxic sulfide by SRM, which further aggravates the negative effects of oxygen shortage causing death to fauna including economically r ele v ant fish, shrimp and crabs (Diaz and Rosenberg 2008, Jenny et al. 2016, Bush et al. 2017, Diao et al. 2018, van Vliet et al. 2021 ).On the other hand, SRM can also exert positive climate change effects.Especially in low-sulfate habitats with active cryptic sulfur cycling, such as rice pad d y fields, peatlands and lake sediments, SRM compete for substrates with microorganisms involved in the methanogenic degradation network (Pester et al. 2012, Wörner et al. 2016, Wörner and Pester 2019 ).This leads to a partial diversion of the carbon flux from CH 4 to CO 2 , which is the less potent greenhouse gas on a per molecule basis (Pester et al. 2012 ).Stimulation of cryptic sulfur c ycling, e.g.b y the addition or intrinsic activity of sulfide-oxidizing cable bacteria can thus exert positive effects on mitigation of methane emissions (Sandfeld et al. 2020, Scholz et al. 2020 ) or delay the de v elopment of euxinia (Seitaj et al. 2015 ).
Most SRM share a canonical core enzyme repertoire for carrying out dissimilatory sulfate reduction (Fig. 1 ).This intracellular pathway includes the enzymes sulfate aden yl yltr ansfer ase (Sat), aden yl yl phosphosulfate r eductase (AprAB), dissimilatory (bi)sulfite reductase (DsrAB), and the sulfide releasing DsrC.The complexes QmoAB(C) and DsrMK(JOP) complement the pathway by transferring reducing equivalents to w ar ds AprAB and DsrC, respectiv el y (Per eir a et al. 2011, Ramos et al. 2012, Santos et al. 2015 ).Her eafter, we r efer to this pathw ay as the Dsr-pathw ay.Most SRM (with the exception of early diverging archaea) and microorganisms r el ying on a partial sulfate r eduction pathway suc h as sulfite-, thiosulfate-, and or ganosulfonate r educers as well as sulfur dispr oportionating micr oor ganisms utilize in addition DsrD, which is an allosteric activator of DsrAB (Ferreira et al. 2022 ).Among these enzymes, DsrAB can be used not only as a functional but, with some limitations, also as a phylogenetic marker for SRM.Phylogenetically, this enzyme comprises three major lineages that largely differentiate between (i) reductively-operating DsrAB of archaeal origin, (ii) r eductiv el y-oper ating DsrAB of bacterial origin, and (iii) o xidati v el y or r e v erse-oper ating DsrAB (rDsrAB), whic h occur in a variety of phototrophic and chemotrophic SOM (Loy et al. 2009, Müller et al. 2015 ).SOM that r el y on rDsrAB for sulfur oxidation also share a number of additional enzymes with SRM, including Sat, AprAB, QmoABC , DsrC , and DsrMKJOP (Dahl 2017 , Tanabe andDahl 2022 ).
The phylogenetic distinction of r eductiv el y and o xidati v el y operating DsrAB was initially also supported by the presence of additional, pr esumabl y SOM-specific enzymes .T hese include DsrEFH as a sulfur donor protein for DsrC in SOM (Stoc kdr eher et al. 2012 ) and DsrL as an essential oxidoreductase in sulfur oxidation (Lübbe et al. 2006 ) that tr ansfers r educing equiv alents from rDsrAB to NAD + (Löffler et al. 2020 ).Howe v er, meta genomeassembled genomes (MAGs) from a variety of habitats questioned this clear distinction, with DsrEFH, DsrL, or both being co-encoded together with r eductiv e DsrAB (Ananthar aman et al. 2018, Hausmann et al. 2018, Tan et al. 2019, Thiel et al. 2019, Ye et al. 2022 ).The recent identification of two discrete DsrL types, with DsrL-1 occurring only in SOM, while DsrL-2 occurring in organisms with either a r eductiv e/dispr oportionating or o xidati ve sulfur metabolism (Löffler et al. 2020 ), highlights the difficulties in delineating the energy metabolism solely from genomic data.Functional gene prediction is further complicated by the examples of Desulfurivibrio alkaliphilus (Thorup et al. 2017 ) and the socalled cable bacteria affiliated to the Desulfobulbaceae (Pfeffer et al. 2012, Risgaar d-P etersen et al. 2015 ).Both can oxidize sulfide by operating the canonical pathway of sulfate reduction in r e v erse, including a r eductiv e-type DsrAB, and couple this either with intr acellular nitr ate r eduction in the case of D. alkaliphilus (Thorup et al. 2017 ) or to electrogenic oxygen or nitrate reduction in spatiall y separ ated cells along filaments in the case of cable bacteria (Kjeldsen et al. 2019 ).
Despite these constraints, dsrAB gene-based molecular appr oac hes hav e become an important tool for stud ying the di versity and ecology of SRM in the environment.First introduced by Wagner et al. 1998 , cumulative evidence from a large variety of marine, terrestrial, and man-made environments revealed that the diversity of SRM extends massiv el y beyond cultur ed r epr esentatives in the four bacterial phyla Desulfobacterota (formerly known as Deltaproteobacteria and Thermodesulfobacteria , Waite et al. 2020 ), Bacillota (formerly known as Firmicutes, Oren and Garrity 2021 ), Thermodesulfobiota (Frolov et al. 2023 ), and Nitrospirota (Oren and Garrity 2021 ) as well as the tw o ar chaeal phyla Thermoproteota (formerly known as Cr enarc haeota, Or en and Garrity 2021 ) and Halobacterota (formerly part of the Eury ar chaeota, Rinke et al. 2021 ).A systematic r e vie w of envir onmental dsrAB genes encoding the r eductiv e bacterial-type DsrAB r e v ealed at least 13 lineages at the a ppr oximate famil y le v el that could not be r elated to an y cultured SRM or higher-rank taxa (Pester et al. 2012, Müller et al. 2015 ).At the species le v el, a br oad census based on dsrB gene amplicon sequencing identified 167 397 species-le v el oper ational taxonomic units (OTUs) across 14 differ ent envir onments (Vigner on et al. 2018 ).If compared to the a ppr oximatel y 460 described SRM listed in the LPSN database (lpsn.dsmz.de),this means that > 99% of SRM diversity is represented by uncultured microorganisms without taxonomic assignment.
Members of well c har acterized Desulfobacterota ( Desulfobacteraceae , Syntrophobacteraceae , Desulfovibrionaceae , Desulfobulbaceae ) often dominate the SRM community in marine and freshwater surface sediments (Vigneron et al. 2018, Wörner and Pester 2019, Jørgensen 2021 ) and the uncharted dsrAB gene sequence space lar gel y r epr esents lo w-abundance taxa.Ho w e v er, in certain envir onments r epr esentativ es of uncultur ed dsrAB linea ges can constitute numericall y r ele v ant members of the SRM comm unity (Vigneron et al. 2018 ), including coastal sediments in the Arctic (Flieder et al. 2021 ), wetlands (Pester et al. 2012 ) subsurface marine sediments with active but cryptic sulfur cycling (Leloup et al. 2009 ), to name a fe w.Ther efor e, ther e is a need to identify these yet unknown SRM and to understand their ecophysiology and evolution.In recent years, an increasing number of new DsrAB-encoding taxa have been discovered by metagenomic surveys of environmental samples and the delineation of MAGs.Her e, we pr ovide a systematic r e vie w of these nov el findings, giv e insights into the increased diversity of (putative) SRM, and place this in the context of the r ecentl y pr oposed ov er arc hing c hanges to bacterial and archaeal taxonomy (Parks et al. 2018, Parks et al. 2020, Oren and Garrity 2021, Rinke et al. 2021 ).Detailed overviews of well-studied phyla harboring SRM, including cultured and envir onmental r epr esentativ es , ha v e been pr ovided in excellent r evie ws else wher e (Rabus et al. 2013, Rabus et al. 2015, Langwig et al. 2022 ).

Unprecedented diversity of Bacteria and Archaea with the potential for dissimilatory sulfate/sulfite metabolism
The number of genomes of uncultivated microorganisms assembled from metagenomes is rapidly growing.In recent years, thousands of MAGs from poorly characterized bacterial and archaeal phyla, including those that still lac k cultur ed r epr esentativ es (candidate phyla), wer e r ecov er ed fr om a lar ge v ariety of envir onments (Ananthar aman et al. 2016, P arks et al. 2017, Parks et al. 2018, Rinke et al. 2021 ).The vast number of novel MAGs allo w ed resear chers to screen for the genomic potential of a dissimilatory sulfur metabolism in microbial lineages that were pr e viousl y not linked to such processes.In addition, bioinformatics tools were developed lately to identify genes related to sulfur compound dissimilation, transport and intracellular transfer with confidence and in a high throughput manner (Mendler et al. 2019, Neukirchen and Sousa 2021, Tanabe and Dahl 2022, Zhou et al. 2022 ).This resulted in a burst of discoveries since 2018.For example, a study by Anantharaman et al. ( 2018 ) sub-stantially expanded the known diversity of bacterial and archaeal phyla with the capacity for sulfite/sulfate reduction from 7 to 20 phylum-le v el linea ges.Specificall y, members of the Acidobacteriota , Armatimonadota , Bacteroidota , Verrucomicrobiota , UBA9089 ( Ca .Desantisbacteria), SAR324 ( Ca .Lambda pr oteobacteria), Ca .Zixibacteriota and Ca .Hydr othermarc haeota contained Dsr-pathway genes to perform sulfate/sulfite reduction (Anantharaman et al. 2018 ).Chloroflexota associated with marine sediments (Wasmund et al. 2016 ) and freshwater subsurface sediments (Hug et al. 2016 ), ne wl y described members of the Nitrospirota r ecov er ed fr om rice pad d y soil (Zecchin et al. 2018 ), enigmatic bacteria of novel candidate phyla such as SZUA-79 ( Ca .Acidulodesulfobacterales) from a mine dr aina ge with pH ∼2 (Tan et al. 2019 ), and cryptic candidate phyla like UBA9089 and CG2-30-53-67 (Probst et al. 2017 ) contribute further to the diversity of bacteria with the potential to reduce sulfate/sulfite.Anaerobic oxidation of methane (AOM) or other alkanes coupled to sulfate reduction was suggested to be performed by microbial consortia of methanotrophic archaea and sulfate-reducing bacteria (Boetius et al. 2000 , Knittel andBoetius 2009 ).The recent finding that some Halobacterota ( Archaeoglobaceae ) and Thermoproteota ( Ca .Methanodesulfokores washburnensis) encode both a methanogenesis pathway and the capability to perform sulfate or sulfite r eduction, r espectiv el y, suggests that sulfur-dependent AOM can be carried out in a single organism, independent from syntrophic interactions (McKay et al. 2019, Wang et al. 2019 ).
The enormous increase in the phylogenetic breadth of bacteria and archaea with the potential for DsrAB-based dissimilatory sulfate/sulfite reduction is currently missing a systematic ov ervie w.Her e, we scr eened all publicly available and functionally pre-annotated genomes and MAGs summarized on the Annotree platform v.1.2.0 (Mendler et al. 2019 ) for the presence of dsrAB genes ( http://annotree.uwaterloo.ca, accessed on April 3 rd , 2023, for bacteria and February 6 th , 2023, for archaea).This resulted in a total of 902 bacterial and 48 archaeal genomes distributed across 27 and 4 phyla according to the GTDB-Tk taxonomy (Parks et al. 2018, Rinke et al. 2021 ), r espectiv el y (Fig. 2 , Fig. 5 ).Phyla provisionally split by the GTDB release 214 into different sublineages such as Bacillota and Bacillota_A to Bacillota_H were counted as one phylum.The r etrie v ed 950 genomes r epr esented 370 isolated species (353 bacterial and 17 archaeal species) and 936 species in total according to the GTDB-Tk taxonomy ( Table S1 ).Since Annotree (Mendler et al. 2019 ) is based on the Genome Taxonomy Database, whic h consider ed onl y MAGs of > 50% completeness and < 10% contamination (Parks et al. 2018, Rinke et al. 2021 ), MAGs that did not fulfill these quality criteria were omitted from our analysis .T hese included, for example , a Verrucomicrobiota MAG with earl y div er ging DsrAB (Verrucomicr obia bacterium SbV1) as well as r epr esentativ es of the Schekmanbacteria (Schekmanbacteria bacterium RBG_13_48_7) and Chloroflexota (Chloroflexi bacteria RBG_13_60_13 and RBG_13_52_14) described pr e viousl y in the liter atur e (Ananthar aman et al. 2018 ).
The remaining 11% of DsrAB-encoding genomes were spread ov er 20 differ ent bacterial phyla, with the Acidobacteriota (19 MAGs) and Actinomycetota (14 MAGs) r epr esenting the most prominent gr oups (Fig. 2 ).Repr esentativ es of the Acidobacteriota, Zixibacteria, Bdellovibrionota, Armatimonadota, and the candidate phyla UBA9089 (Desantisbacteria), SZUA-79, OLB16, and AABM5-125-24 were all characterized by the full set of Dsr -pathway genes, including dsrD as indicator for a r eductiv el y oper ating metabolism and dsrL of type 2 (Fig. 2 ), which is present in organisms with either a r eductiv e/dispr oportionating or oxidativ e sulfur metabolism.The same was true for MAGs within the Actinomycetota and Myxococcota , with three exceptions which are described in more detail below.Re presentati ves of the Chloroflexota, Deferribacterota , and candidate phylum RBG-13-61-14 also encoded DsrD but lacked dsrL genes.Since most of the other Dsr -pathway encoding genes could be r ecov er ed for the latter three phyla, a reductiv el y oper ating pathway is indicated her e as well.Members of the Methylomirabilota (pr e viousl y assigned to Candidatus Rokubac-teria, Anantharaman et al. 2018 ) lacked both dsrD and dsrL genes but belong to the group of microorganisms with the earliest div er ging DsrAB (Fig. 2 ).The gr oup of bacteria and arc haea with earl y div er ging DsrAB consistentl y lac ks dsrD and dsrL genes but contains cultured re presentati ves with a reductively operating Dsr -pathway (Anantharaman et al. 2018, Ferr eir a et al. 2022 ).In summary, members of fourteen phyla without cultured SRM (representing 7% of all r ecov er ed bacterial genomes) encode the full enzyme complement to perform dissimilatory sulfate/sulfite reduction.In contr ast, r epr esentativ es of the Verrucomicrobiota and candidate phylum CG2-30-70-394 lacked the dsrD gene but encoded DsrL-1, which resembles the situation in canonical SOM and is indicative of an oxidativ el y oper ating sulfur metabolism.
The situation was more complex in members of the Nitrospirota, Nitrospinota , Spirochaetota , Bacteroidota, and SAR324.In these five phyla, different MAGs of the same phylum carried different gene combinations of the Dsr -pathway.Within the Nitrospirota , the majority of MAGs encoded DsrD but lacked genes encoding DsrL, including cultured SRM of the genus Thermodesulfovibrio (Zecchin et al. 2018 ).Ho w e v er, ther e wer e two MAGs that lacked the dsrD gene but encoded either  or DsrL-2 (f_9FT-COMBO-42-15) indicating an o xidati v el y oper ating Dsr-pathway (Fig. 3 , Supplementary Table S1 ).The opposite was true for the Bacteroidota .Here, most MAGs and genomes of cultured representatives belonged to the class Chlorobia (family Chlorobiaceae ), which r epr esent canonical SOM and carried gene combinations of the Dsr-pathway typical for SOM (no dsrD , dsrL-1 or dsrL-2).Howe v er, fiv e r epr esentativ es of the Bacteroidota family UBA2268 (class Kapabacteria ), encoded both DsrD and DsrL-2 (which was clearly distinct from DsrL of the Chlorobiaceae ) indicating a r eductiv el y operating Dsr-pathway.In situ transcriptional profiles of UBA2268 MAGs r ecov er ed fr om hot spring micr obial mats clearl y supported a r eductiv el y oper ating Dsr-pathway activ ated under anoxic conditions (for details see below, Thiel et al. 2019 ).For members of the Nitrospinota and Spirochaetota , MAGs were more evenly distributed and either carried gene combinations indicative of a r eductiv e ( dsrD along with dsrL 2 or no dsrL ) or o xidati ve sulfur metabolism (no dsrD but dsrL -1 or dsrL -2).Furthermor e, our anal ysis r ecov er ed six SAR324 members with different gene combinations of the Dsrpathw ay.Tw o SAR324 MAGs, which affiliated to the provisional famil y XYD2_FULL-50-16, wer e r ecov er ed fr om the terr estrial subsurface with an indicated r eductiv e sulfur metabolism ( dsrD along with dsrL 2).The remaining four MAGs (f_NAC60-12) were recover ed fr om marine envir onments with an indicated o xidati ve sulfur metabolism (no dsrD , dsrL -1).The latter coincides well with reports of Dsr-pathway encoding SAR324 from oxygenated deep ocean w aters (Sw an et al. 2011 ), hydr othermal v ent plumes (Sheik et al. 2014 ), and marine oxygen minimum zones (van Vliet et al. 2021 ).
An unusual gene combination was observed for eight MAGs spr ead ov er the phyla Actinom ycetota , Myxococcota , CG2-30-53-67 (one MAG each) and the Desulfobacterota (five MAGs within the famil y Desulfocapsaceae ).Her e, at least two differ ent bacterial-type DsrAB , including always one reductive and one o xidati ve one, were encoded on the same genome (Fig. 4 , Table S1 ).The most striking example was Actinomycetota MAG GCA_003 599 855, which was r ecov er ed fr om a 1.5 km deep terr estrial aquifer (Momper et al. 2017 ) (Parks et al. 2018 ).The phylogenomic tree was inferred from 902 bacterial (metagenome-assembled) genomes .T he scale bar indicates 20% sequence div er gence .T he maximum likelihood tree was constructed with IQ-TREE (Nguyen et al. 2015 ) using automatic substitution model selection (LG + F + R10) and ultr afast bootstr a p anal ysis (n = 1000).Bootstr a p support is indicated by blac k dots ( ≥90%) or blac k circles (70-90%).Within each lineage, the presence of Dsr-pathway encoding genes was indicated if > 30% of dsrAB -containing genomes carried the r espectiv e genes as inferred by an automated hmm search (Zhou et al. 2022 , custom- (Rinke et al. 2021 ).The phylogenetic tree was inferred from 48 archaeal (metagenome assembled) genomes .T he scale bar indicates 20% sequence div er gence .T he maximum likelihood tree was constructed with IQ-TREE (Nguyen et al. 2015 ) using automatic substitution model selection (LG + F + R6) and ultrafast bootstrap analysis (n = 1000).Bootstrap support is indiacted by black dots ( ≥90%) or black circles (70%-90%).Within each lineage, the presence of Dsr-pathway encoding genes was indicated if > 30% of dsrAB -containing genomes carried the r espectiv e genes as inferred by an automated pHMM search (Zhou et al. 2022 ).
dsrCTMKJOP and on a separate contig oxidative DsrAB just upstr eam of dsrEFHCMKLLJOP .Inter estingl y, DsrL was of type 1 and encoded by two gene copies in direct vicinity.An additional Myxococcota MAG (GCA_003 153 055) encoded o xidati ve DsrAB as well but was missing further genes indicative of a r eductiv e or oxidative metabolism.Yet another unusual r epr esentativ e of the terrestrial deep subsurface was the single MAG r epr esenting candidate phylum CG2-30-53-67 ( Table S1 ), which was r ecov er ed fr om deep gr oundwater (Pr obst et al. 2017 ).Also here, the reductive DsrAB was encoded upstream of DsrD and DsrL -2, with the latter falling into the major DsrL-2 cluster.The o xidati ve DsrAB was encoded upstream of a second copy of DsrL -2, which clustered together with DsrL-2 of Magnetococcia (Figs 3 and 4 ) .As suggested before (Löffler et al. 2020 ), this bacterium is likely capable of switching the direction of dissimilatory sulfur metabolism by regulating the different types of DsrABL.The same is likely true for the Actinomycetota MAG described abo ve .
The five unusual MAGs from the family Desulfocapsaceae within the Desulfobacterota were spread over the three genera Desulforhopalus , Desulfomarina , and Desulfopila (Fig. 4 , Table S1 ).A uniting feature of these MAGs was their r ecov ery fr om o xic-ano xic transition zones in marine surface sediments, either at methane seeps or tidal sediments.A second uniting feature was the genomic organization of the different dsrAB gene sets.Reductive DsrAB was always encoded upstream of DsrD and o xidati ve DsrAB always upstream of DsrL-2 (Fig. 4 ), which clustered together with DsrL-2 of canonical SOM of the genus Chlorobium (Fig. 3 ).Both dsrAB gene sets were always recovered on separate contigs and did not form operons with other Dsr-pathway genes such as dsrMKJOP , aprAB , qmoABC , and sat .Here, a switch of the direction of dis-similatory sulfur metabolism might be regulated by differ entiall y forming DsrABD or DsrABL-2.Inter estingl y, thr ee additional Desulfobacterota MAGs affiliated to the provisional genus UBA2156 encoded o xidati v e DsrAB onl y a gain along with DsrL-2, whic h clustered with DsrL-2 of canonical SOM of the genus Chlorobium (Fig. 3 ).Since one of these MAGs also encoded DsrD, it remains unclear whether the genes encoding r eductiv e DsrAB wer e missed by the assembly and binning process.Vice versa, a misassembly or false binning cannot be excluded for any of the above-mentioned MAGs carrying unusual dsrAB gene combinations.Ho w e v er, all these MAGs were of high quality with estimated contaminations ranging from 0.7% to 6.4% (3.0 ± 2.0%, av er a ge ± standard deviation) and estimated completeness ranging from 77% to 99% (92 ± 8%, av er a ge ± standard deviation).The combination of this high binning quality with the r ecov ery of such unusual dsrAB gene combinations in multiple studies from various environments and sever al phylogenetic linea ges mak es the lik elihood quite high that these MAGs r epr esent r eal micr oor ganisms a waiting disco very using cultivation approaches.
Based on the findings described abo ve , we propose to further subdivide the DsrL-2 cluster into three phylogenetically distinct subclusters to guide genome annotations.Subclusters DsrL-2A and DsrL-2B encompass (i) canonical SOM of the Bacteroidota and Pseudomonadota , (ii) Nitrospinota and Nitrospirota MAGs with an indicated o xidati ve sulfur metabolism, (iii) and MAGs encoding oxidativ e and r eductiv e DsrAB on the same genome.For the latter gr oup, DsrL-2A and DsrL-2B wer e al w ays encoded just do wnstream of o xidati ve DsrAB, with the respective genes being part of the same operon (Fig. 4 ).T herefore , we propose DsrL-2A and DsrL-2B to function as an indicator for an o xidati ve sulfur metabolism if encoded in close proximity to oxidative DsrAB.In contrast, subcluster DsrL-2C encompasses all MAGs with an indicated reductive sulfur metabolism as evidenced by encoding DsrD and reductive DsrAB but not oxidative DsrAB.Here, we propose DsrL-2C to function as an indicator for a r eductiv e sulfur metabolism if encoded in close proximity to reductive DsrAB.DsrL-1 subclusters A and B were already defined by Löffler et al. 2020 , and encompass canonical SOM of the Alpha -and Gammaproteobacteria for subcluster DsrL-1A and canonical SOM of the Chlorobia and Magnetococcia as well as MAGs with an indicated o xidati ve sulfur metabolism for subcluster DsrL-1B.Along these lines, we examined the occurrence of DsrEFH among all recovered MAGs.In the overwhelming majority, DsrEFH was encoded in SOM and MAGs with an indicated o xidati ve sulfur metabolism and absent in SRM and MAGs with an indicated r eductiv e sulfur metabolism (Fig. 2 ).Also, in MAGs that encoded both r eductiv e and oxidative DsrAB, DsrEFH w as alw ays encoded along with o xidati ve DsrAB (Fig. 4 ).Ho w e v er, ther e wer e two linea ges that r epr esented exceptions to this general pattern.Two Actinomycetota within the genus Aquicultor and two SAR324 affiliated to the provisional family XYD2_FULL-50-16 encoded both r eductiv e DsrAB and DsrEFH on the same genome ( Table S1 ), making DsrEFH an imperfect predictor for an indicated r eductiv e or oxidative sulfur metabolism.
Archaea encoding the Dsr -pathway are currently characterized by a solely reductively operating Dsr-pathway.Among the 48 archaeal genomes studied, 67% belonged to cultured r epr esentativ es with a known sulfate-, sulfite-, or thiosulfatereducing metabolism within the phyla Thermoproteota (members of the genera Caldivirga , Thermoproteus , Thermocladium , Vulcanisaeta , and Pyrobaculum ) and Halobacteriota ( Archaeoglobus spp.).The remaining MAGs expanded the diversity of archaea encoding a Dsr -pathway to include two additional phyla ( Hydrothermarchaeota and Thermoplasmatota ) and four additional families in the Thermoproteota (three families) and Halobacterota (one family).Most of the DsrAB-encoding archaea either represented cultured thermophiles or their MAGs were retrieved from hightemper atur e envir onments (e .g. hot springs , hydr othermal v ent fluids) including their de posits (e.g.dee p sea hydrothermal vent field site).The only exceptions w ere tw o Halobacteriota MAGs (GC A_002507545, GC A_002494625) and one Thermoplasmatota MAG (GCA_002503985) assembled from marine sediment or soil meta genomes, r espectiv el y (P arks et al. 2017 ).Notably, at least two of the families encoding the Dsr-pathway ( Kor archaeaceae , ph ylum Thermoproteota; Archaeoglobaceae , phylum Halobacteriota ) were r epr esented by MAGs (McKay et al. 2019, Wang et al. 2019 ) that additionally encode the complete pathway for (r e v erse) methanogenesis.Based on these findings, it was postulated that anaerobic methane oxidation coupled to sulfate/sulfite reduction might also operate in single microorganisms (McKay et al. 2019, Wang et al. 2019 ) as opposed to the standard model of syntrophic associations (Knittel and Boetius 2009 ).

Genome-centric metagenomics anchors and expands dsrAB -based functional and taxonomic assignment
Appr oac hes based on dsrAB gene sequence analysis have been extensiv el y used to study the ecology of SRM and in part also SOM.These surveys were based on the assumption that there is a clear phylogenetic separation between DsrAB present in archaea with a r eductiv e sulfur metabolism, bacteria with a r eductiv e sulfur metabolism, and bacteria with an o xidati ve sulfur metabolism.We used the expanded diversity of Dsr-pathway encoding micr oor ganisms described above along with the indicated reductiv el y or oxidativ el y oper ating dir ection of their sulfur metabolism to explore their DsrAB phylogeny.Our analysis sho w ed that the distinction of archaeal reductively-operating DsrAB, bacterial reductiv el y oper ating DsrAB, and bacterial o xidati v el y-oper ating DsrAB still lar gel y holds true.While arc haeal r eductiv el y operating DsrAB and bacterial o xidati v el y-oper ating DsrAB formed monophyletic clusters in our analysis (with the exception of later all y acquir ed dsrAB genes, see below), bacterial r eductiv el yoper ating DsrAB wer e spr ead betw een these tw o clusters in a bush-like manner (Fig. 6 ).This is consistent with a recent phylogenetic analysis of DsrAB including already parts of the novel dsrAB gene sequence space discov er ed in MAGs fr om v arious habitats (Anantharaman et al. 2018 ).Ho w ever, it does not reproduce an ymor e the monophyletic separation of reductively-operating DsrAB observed in phylogenetic analyses based mainly on canonical SRM and PCR-derived dsrAB gene sequences of environmental studies (e.g.Müller et al. 2015 ).Ne v ertheless, the distinction of r eductiv e bacterial-type DsrAB in our (metagenome assembled) genome census was not onl y anc hor ed by canonical sulfate/sulfite-reducing bacteria but contained in addition exclusiv el y phyla whose r epr esentativ e MAGs encoded DsrD (in combination with or without DsrL-2C) str ongl y indicating a r eductiv etype sulfur metabolism as well.The only exception were the few MAGs encoding both reductive and o xidati ve bacterial-type DsrAB.Ho w e v er, also in these MAGs DsrD w as alw ays encoded downstr eam of r eductiv e bacterial-type DsrAB (Fig. 4 ).Furthermore, the demarcation to o xidati ve bacterial-type DsrAB was well supported by the latest div er ging r eductiv e DsrAB cluster that contained DsrAB of Desulfurella amilsii ( Campylobacterota ) as an organism capable of growth by thiosulfate reduction (Fig. 6 ).
Pr e vious PCR-based dsrAB gene studies described thirteen lineages of reductive bacterial-type DsrAB that r epr esented a ppr oximate famil y le v el gr oups of uncultur ed micr oor ganisms that could not be assigned to known taxa (Pester et al. 2012, Müller et al. 2015 ).Members of some of these groups were identified as abundant and active in different marine and freshwater habitats (e.g.Pester et al. 2012, Müller et al. 2015, Pelikan et al. 2016, Wörner and Pester 2019 ).Based on pr e vious findings and the phylogenomic survey of this study, we summarize the curr entl y known taxonomic classification of these uncultured family-level DsrAB lineages (Fig. 6 , Table 1 ).DsrAB sequences of uncultured lineages 3 and 5 were found in members of the Chloroflexota .Lineage 3 members were represented by a single-cell amplified genome (SAG) recov er ed fr om deep marine subsurface sediments (Wasmund et al. 2016 ).Because of the low cov er a ge of this SAG (46%, Wasmund et al. 2016 ), it was not part of our genome collection but was considered in our DsrAB analysis (Fig. 6 , class Dehalococcoidia ).Chloroflexota r epr esenting uncultur ed famil y linea ge 5 wer e r ecov er ed fr om hydr othermal sediments and a bior eactor (MAG collection of Parks et al. 2017 andZhou et al. 2020 ).Uncultured family lineage 6 belongs to the Desulfobacterota (order Syntrophales ) and DsrAB sequences of uncultured lineages 8 and 9 were found in MAGs of terrestrial and marine Acidobacterota , respectively (Hausmann et al. 2018, Flieder et al. 2021 ) .Sequences of DsrAB lineages 10 and 13 ha ve been unco vered in Nitrospirota genomes from an aquifer system (Anantharaman et al. 2016 ).Furthermore, uncultured family lineages 1 and 11 cluster within the Desulfobacterota.While uncultur ed famil y linea ges 1 still has an unr esolv ed famil y affiliation, uncultur ed famil y linea ge 11 gr ouped next to later all y acquir ed dsrAB genes of Nitrospirota affiliated to provisional family SM23-35.In summary, the affiliation of 8 of the 13 uncultured family-level Figure 6.Maximum likelihood phylogeny of DsrAB sequences derived from (metagenome assembled) genomes and environmental surveys.Clades r epr esented by a majority of DsrAB-encoding (metagenome assembled) genomes not affiliated to canonical SRM or SOM are shown in magenta.The cov er a ge of inferred phylogenetic clades by published broad-range PCR primers ( ≥75% of sequences in a clade; 1 mismatch allo w ed) is indicated by colored dots .T he binding positions of the e v aluated primers is indicated at the bottom using dsrAB of Desulfovibrio vulgaris or Allochromatiom vinosum as model organism of dsrAB primers designed to target bacterial-type dsrAB encoding the reductive or o xidati ve enzyme version, respectively.The maximum likelihood tree was constructed using deduced DsrAB amino acid sequences with IQ-TREE (Nguyen et al. 2015 ) using automatic substitution model selection (LG + R10) and ultrafast bootstrap analysis (n = 1000).Bootstr a p support is indicated by black dots ( ≥90%) or black circles (70-90%).The tree was inferred from 613 re presentati ve DsrAB sequences with an indel filter covering 571 amino-acid positions: 346 re presentati ve DsrAB sequences were taken from a curated DsrAB database including 7921 pure culture and environmental sequences (as based on Müller et al. 2015 ) and amended with 267 DsrAB sequences derived from (metagenome assembled) genomes representing novel phylogenetic clades.Scale bar indicates 50% sequence div er gence.Clades containing taxonomically resolved uncultured family-level DsrAB lineages are indicated by a superscript number based on the following denomination: 1, uncultured family-lineage 1; 3, uncultured family-lineage 3; 5, uncultured family-lineage 5; 6, uncultur ed famil y-linea ge 6; 8, uncultur ed famil y-linea ge 8; 9, uncultur ed famil y-linea ge 9; 10, uncultur ed famil y-linea ge 10; 11, uncultur ed famil y-linea ge 11; 13 uncultured family-lineage 13.Please note that the numbers in brackets behind candidate phylum CG2-30-53-67 represent the two div er ging dsrAB copies carried by the single MAG r epr esenting this phylum.LA-dsrAB , later all y acquir ed dsrAB .

Uncultured famil y-le vel lineages GTDB-Tk taxonomy Reference
Uncultur DsrAB lineages could be resolved using (meta-)genome targeted a ppr oac hes.Ho w e v er, the affiliation of uncultur ed famil y linea ges 2, 4, 7, and 12 still awaits its discovery.Se v er al studies hav e pr ovided e vidence that the distribution of dsrAB genes among extant micr oor ganisms is r epr esented by a combination of div er gence thr ough speciation, functional div ersification and later al gene tr ansfer (LGT) (Klein et al. 2001, Zverlov et al. 2005, Loy et al. 2008, Müller et al. 2015, Anantharaman et al. 2018 ).Well documented examples are the later all y acquired dsrAB genes of a group of Desulfotomaculum spp.( Bacillota ) from Desulfobacterota (Klein et al. 2001, Zverlov et al. 2005 ) or the bacterial origin of r eductiv e DsrAB in members of the archaeal gen us Ar chaeoglobus ( Halobacteriota ) (Müller et al. 2015 ).Based on our extended analysis, we conclude that 14 major taxa of Dsrpathway encoding micr oor ganisms likel y acquir ed dsrAB genes in m ultiple later al gene tr ansfer e v ents .T hese encompass besides the Bacillota and Halobacterota also the Methylomirabilota, Chloroflexota , Alphaproteobacteria , Magnetococcia , Hydrothermarchaeota , Desulfobacterota , Actinomycetota , Nitrospirota , Nitrospinota , Bacteroidota , Spirochaetota , Myxococcota , and candidate phyla CG2-30-53-67 and SAR324 (Fig. 7 ).The latter se v en ar e especiall y inter esting because they harbor members with either r eductiv e DsrAB, oxidative DsrAB, or both (Fig. 6 , Fig. 7 ).Analysis of our extended dsrAB gene dataset could not r epr oduce the postulated LGT of bacterial dsrAB to archaea of the families Wolfr amiir aptor aceae (pr e viousl y r eferr ed to as Aigarc haeota pSL4) and Korarc haeaceae ( Ca. Methanodesulfok or es washburnensis) (Müller et al. 2015, McKay et al. 2019 ) despite them showing higher similarity to r eductiv e bacterial-type DsrAB of Bacillota than to r eductiv e DsrAB of all other arc haea (Fig. 6 ).Mor e in-depth phylogenetic studies will have to show whether lateral gene transfer of dsrAB occurred in Wolfr amiir aptor aceae and the Korarchaeaceae as well.
With the updated dsrAB gene database provided in this review, dsrAB -based marker gene surveys will greatly benefit as sequences can be better taxonomically anchored.To this end, we provide an updated dsrA and dsrB gene r efer ence database including the sequences from 902 bacterial and 48 archaeal genomes and MAGs analyzed in this study (available under https://www.arb-silv a.de/pr ojects/dsr absilv a/), whic h will be useful for dsrAB gene amplicon sequencing analyses (e.g.Müller et al. 2015, Pelikan et al. 2016, Vigneron et al. 2018, Wörner and Pester 2019 ).
We further e v aluated the cov er a ge of those genomes and MAGs with primer sets designed to target the dsrA and dsrB gene encoding either r eductiv e bacterial-type (Pelikan et al. 2016 ) or o xidati ve bacterial-type (Loy et al. 2009, Müller et al. 2015 ) DsrAB ( Tables S2 ).
A good cov er a ge of near full-length bacterial-type dsrAB encoding the r eductiv e v ersion of the enzyme can be ac hie v ed using the primers DSR190f mix (88%) and DSR2107r mix (88%).This covers the great majority of (putative) SRM (Fig. 6 ; see Table S3 for details).For amplicon sequencing, we can confirm the recommendation of Pelikan et al. ( 2016 ) to use the primer pair DSR1762f mix and DSR2107r mix.T hey co ver most (97% and 88%, respectively) of the bacterial-type dsrB genes that encode the r eductiv e enzyme version (Fig. 6 ; Table S3 ).Ho w ever, for extended coverage of the few MAGs within the Bdellovibrionota , Campylobacterota , Deferribacterota , SAR324, SZUA-79, OLB16, or AABM5-125-24 new primer variants will need to be designed for both near-full length and short dsr(A)B amplicons ( Table S3 ).Primer pairs rDSR1f mix and rDSR4r mix were designed to amplify near full-length bacterial-type dsrAB encoding the o xidati v e enzyme v ersion (Loy et al. 2009 ).T hey ha ve an acceptable cov er a ge (68% and 95%, r espectiv el y), but do not cov er a consider able fr action of the o xidati ve DsrAB-encoding Alphaproteobacteria, Gammaproteobacteria , and Magnetococcia as well as o xidati ve DsrAB-encoding members of the Nitrospinota , Nitrospirota , and candidate phylum CG2-30-53-67.Ho w e v er , the majority of bacterial-type dsrB genes encoding the o xidati v e enzyme v ersion is cov er ed by the primer pairs DSR1762f mix (94%) and rDSR4r mix (95%), which can be used for short-read amplicon sequencing ( Table S3 ).No primers were published so far to specifically amplify dsrAB of archaeal origin.

Insights into the ecophysiology of newly discovered SRM
For most of the ne wl y discov er ed micr oor ganisms possessing a Dsr-pathway only a (partial) genome sequence is available so far.Even though their genomic context provides clues about a reductiv el y or oxidativ el y oper ating dissimilatory sulfur metabolism, we still miss a large part of their actual physiology.Enriching and isolating these micr oor ganisms into cultur e will r emain the best a ppr oac h to understand their biology but will also take time.An alternative to cultivation is to understand the ecophysiology of these ne wl y discov er ed SRM (and SOM) in their natural setting using controlled experimental setups along with studying their activity responses at the transcriptome and/or proteome level or using isotope labeling techniques at the population or single-cell le v el.In the follo wing, w e describe se v er al examples, wher e this has been ac hie v ed. or black circles (70%-90%).For both trees, the scale bars indicates 20% sequence div er gence.

Diao et al. | 13
Acidobacteriota encoding a Dsr-pathway were first discovered in pristine low-sulfate environments including peatlands (Hausmann et al. 2018 ) and aquifers (Anantharaman et al. 2018 ) but also in acidic sulfide mine waste r oc k sites (Ananthar aman et al. 2018 ) and later in marine surface sediments (Coskun et al. 2019, Flieder et al. 2021 ).Clone library-based studies targeting dsrAB genes extend the habitat range to deep marine sediments below the sulfate-methane transition zone and high-temperature envir onments (r e vie w ed in P ester et al. 2012in P ester et al. , Müller et al. 2015 ) ).A few studies succeeded to investigate the encoded metabolic potential, the transcriptional activity, as well as the abundance and distribution of Dsr-pathway encoding Acidobacteriota in more detail.
Peatland Acidobacteriota encoding a Dsr-pathway were studied in detail in a small acidic fen in the Fichtel mountains located in Centr al Eur ope.Her e, the y mak e up r oughl y tw o thir ds of microorganisms encoding r eductiv e bacterial-type DsrAB (Pelikan et al. 2016 ) and contribute a consider able fr action to the ov er all micr obial peat soil transcriptome ( > 2% of all mRNA reads; Hausmann et al. 2018 ), impl ying a pr edominant r ole in the cryptic sulfur cycle of this habitat.They are affiliated to four different families ( Acidobacteriaceae , SBA1, Bryobacteraceae , and UBA7540) within the class T erriglobia (comprising former uncultured dsrAB family-level lineage 8) with some r ecov er ed MAGs encoding the full canonical pathway of sulfate reduction while others harboring only genes for sulfite reduction.The latter encoded in addition enzymes that can liberate sulfite from organosulfonates, implying organic sulfur compounds as complementary energy sources.Interestingly, these Acidobacteriota encoded also the full r espir atory c hain for aer obic r espir ation including low and high affinity terminal oxidases as well as a large enzymatic repertoire for polysaccharide degradation and sugar utilization.In addition, capabilities for a fermentative lifestyle and hydrogen oxidation were encoded as well.This "Swiss army knife"-array of potential energy metabolism variants opens up a lot of possibilities how these Dsr-pathway encoding Acidobacteriota may cope with the fluctuating redox conditions in peat soils .T he r edox state of these typicall y water-satur ated soils can c hange dr amaticall y and is mainl y driv en by c hanges in the water table through rainfall and droughts.In addition, lateral flow of water can heavily influence the topography of redox gradients in peat soils through space and time (Frei et al. 2012, Pester et al. 2012 ).In response , Dsr-pathwa y encoding Acidobacteriota were postulated to be capable of switching from a sulfate-reducing or, in case of sulfate shorta ge, fermentativ e lifestyle under anoxic conditions to aerobic respiration under oxic conditions using polysaccharides or low-molecular weight organic compounds as substrates (Hausmann et al. 2018 ).Especially the potential use of pol ysacc harides under sulfate reducing conditions would differentiate them from canonical SRM, which are not able to degrade or ganic pol ymers (Rabus et al. 2013 ).Also, a r e v ersal of the Dsrpathway for sulfur oxidation in combination with aerobic respiration was proposed (Hausmann et al. 2018 ), although the genomic context suggests rather a reductively operating sulfur metabolism (see above).
Peat soil incubations under controlled substrate and sulfate supply may provide insights into the postulated metabolism of DsrAB-encoding Acidobacteriota .When peat soil was incubated ano xically with indi vidual fermentation intermediates (formate, acetate , propionate , lactate or butyrate) with and without externally supplied sulfate , Dsr-pathwa y encoding Acidobacteriota sho w ed a steady transcriptional activity including all genes of the Dsr -pathw ay.Ho w e v er, ther e was no significant incr ease of transcriptional activity triggered by either one of the individually supplied lo w-molecular w eight compounds indicating that the ac-tivity of the r espectiv e Acidobacteriota r ather r elied on or ganic substances already present in the peat itself (Hausmann et al. 2018 ).Extending upon these initial results, Dyksma and Pester ( 2023 ) incubated peat soil in a bioreactor setting under alternating oxic (50% air saturation) and anoxic conditions and a steady supply of pectin as an abundant terrestrial plant polysaccharide.Indeed, a Dsr-pathway encoding Acidobacterium differ entiall y expr essed the full canonical pathway of sulfate reduction under anoxic conditions and the full r espir atory c hain under oxic conditions providing experimental evidence that facultatively anaerobic SRM within the Acidobacteriota exist (Dyksma and Pester, 2023 ).Similar results were already indicated in studies on model SRM within the Desulfobacterota , i.e.Desulfovibrio species, albeit at m uc h lo w er oxygen concentrations.Desulfovibrio spp.typically encode highaffinity bd -type terminal oxidases onl y, whic h ar e implied to function in oxygen detoxification rather than aerobic growth (Santana 2008, Ramel et al. 2015 ).When grown in semi-solid media within an oxygen gradient, Desulfovibrio magneticus formed a visible band at the o xic-ano xic interface in the absence of sulfate and the authors inter pr eted this as micr o-oxic gr owth coupled to oxygen respir ation (Lefèvr e et al. 2016 ).Aerotactic band formation was also observed for Desulfovibrio desulfuricans in oxygen gradients within a diffusion c hamber (Fisc her and Cypionka, 2006 ) and active oxygen reduction as a defense strategy to re-establish anoxic conditions have been reported for Desulfovibrio , Desulfomicrobium and Desulfobulbus spp.(Brune et al. 2000, Cypionka 2000, Sass et al. 2002, Mogensen et al. 2005 ).In a more detailed study, a strain of Desulfovibrio vulgaris Hildenborough was exposed to O 2 -driven labor atory ada ptiv e e volution and acquir ed via point m utations as well as deletions/insertions the ability to gain energy from oxygen r espir ation under micr ooxic conditions (0.65% O 2 , Sc hoeffler et al. 2019 ).Since the enzymatic systems r equir ed for both sulfate and oxygen r espir ation wer e alr eady pr esent in the genome of D. vulgaris , only a limited number of mutations were apparently requir ed to r edir ect the flow of r educing equiv alents to w ar ds aerobic r espir ation coupled to growth (Schoeffler et al. 2019 ).
Marine Acidobacteriota were first indicated in dsrAB gene-based surveys by Müller et al. 2015 and recently their respective genomes could be r ecov er ed fr om marine surface sediments in the Arctic off the coast of Svalbard (Flieder et al. 2021 ).Here, they comprised the second most abundant DsrAB-encoding phylum after the Desulfobacterota (on av er a ge 13%) and r epr esented 4% of dsrB transcripts, emphasizing their in situ activity.When expanded to a global marine dsrAB gene dataset, acidobacterial dsrAB genes av er a ged 15% in marine sediments worldwide (Flieder et al. 2021 ).They are affiliated to a different class ( Thermoanaerobaculia; family FEB-10) than peatland Dsr-pathway encoding Acidobacteriota and comprise former uncultured dsrAB family-level lineage 9. Detailed annotation of their genomes r e v ealed the metabolic potential for various respiratory pathways based on oxygen, nitrous o xide, metal-o xide, tetrathionate, sulfur and sulfate/sulfite as terminal electron acceptor.Potential electron donors comprised cellulose , proteins , c y anoph ycin, h ydrogen, and acetate (Flieder et al. 2021 ).In summary, both terrestrial and marine Dsr-pathway encoding Acidobacteriota likely represent an ecologically important but so far overlooked group of SRM with a large metabolic versatility in respect to potential substrates including organic polymers and alternative electron acceptors including oxygen.
The metabolic flexibility to switch between sulfate reduction and aerobic respiration was also indicated in metagenomic and metatranscriptomic surveys of microbial mat-inhabiting members of the Bacteroidota family UBA2268 (Kapabacteria).In contrast to their well-studied phototrophic and sulfur-oxidizing r elativ es within the Chlorobiaceae , UBA2268-related MAGs retrie v ed fr om micr obial mats of hot springs or groundwater encode r eductiv e DsrAB as well as DsrD and DsrL-2C.For one of these MAGs ( Candidatus Thermonerobacter thiotrophicus), the genome was annotated in greater detail and its transcriptional profile c har acterized during the diel cycle in the microbial mat of the thermal outflow of Mushroom Spring in Yellowstone National Park, USA (Thiel et al. 2019 ).Despite being a low-sulfate environment ( < 200 μM), the phototrophic microbial mat was characterized by high sulfate reduction rates ( > 5 μmol cm −3 d −1 ) during the night, which cease during daytime because of oxygen production by cyanobacteria-driven photosynthesis (Dillon et al. 2007 ).Accordingly, Ca .T. thiotrophicus sho w ed str ong expr ession of all Dsr-pathway genes during the night and a sharp decrease in its transcript levels during da ytime .Ca .T. thiotrophicus also encoded a full r espir atory c hain including alternativ e complex III, an aa 3 -type low-affinity terminal oxidase as well as a bd -type high-affinity terminal oxidase.Inter estingl y, genes encoding the aa 3 -type low-affinity terminal oxidase were differentiall y expr essed as compar ed to Dsr-pathwa y genes .T heir highest expr ession le v els wer e observ ed at light-dark tr ansitions in the morning and e v ening (Thiel et al. 2019 ) corresponding to incr easing and decr easing oxygen le v els in the mat, r espectiv el y, but avoiding times of oxygen (ov er)satur ation during daytime (Dillon et al. 2007 ), when genes encoding o xidati v e str ess r esponse dominated the transcriptional profile (Thiel et al. 2019 ).In contrast, genes encoding the bd -type terminal oxidase sho w ed highest expression during the night implying a role in oxygen detoxification at low oxygen le v els during active sulfate reduction.The absence of encoded CO 2 -fixation pathways and the increased expression of genes involved in gl ycol ysis/gluconeogenesis , the TC A cycle , and acetate-related metabolism during the night indicated a heter otr ophic lifestyle based on small organic molecules, which primaril y corr elated with sulfate r eduction (Thiel et al. 2019 ).
Another intriguing group are the many DsrAB-encoding Nitrospirota members with an indicated r eductiv e sulfur metabolism, whic h wer e encounter ed in envir onments of mainl y moderate temper atur es and that are distinct from their thermophilic, sulfate-r educing r elativ es within the genus Thermodesulfovibrio .When excluding Thermodesulfovibrio spp., 37 additional MAGs encoding an indicated r eductiv el y oper ating Dsr-pathway were r ecov er ed r epr esenting 10 candidate families within the phylum Nitrospirota .Typically, these MAGs were recovered from lowsulfate environments encompassing rice pad d y soil (Zecchin et al. 2018 ), permafrost soils (Woodcroft et al. 2018 ), freshwater sediments , aquifer sediments , gr oundwater, the terr estrial and marine deep subsurface (Jungbluth et al. 2017, Anantharaman et al. 2018, Probst et al. 2018 ).Ho w ever, a few w ere also recovered from br ac kish (Arshad et al. 2017 ) and saline marine environments (Kato et al. 2018 ).Among the mesophilic, low-sulfate adapted Nitrospirota, r epr esentativ es fr om rice paddies wer e studied in mor e detail.From pad d y soil that w as used to gro w rice in the presence and absence of gypsum (CaSO 4 •2 H 2 O), the partial genome of Dsr-pathway encoding Candidatus Sulfobium mesophilum ( Nitrospirota family UBA6898) could be recovered.Parallel metaproteomics r e v ealed activ e expr ession of its Dsr-pathway under gypsum amendment in support of a sulfate-reducing lifestyle.Interestingly, Ca .S. mesophilum also encoded the full pathway of dissimilatory nitr ate r eduction to ammonia, whic h was expr essed in the treatment without gypsum amendment.The r elativ e abundance of Ca .S. mesophilum was similar under both treatments, indicating that it maintains a stable population in rice pad d y soils while shifting its primary ener gy metabolism.In contr ast to the Acidobacteriota described abo ve , Ca .S. mesophilum was rather adapted to the breakdown of classical substrates of SRM covering the metabolic potential to utilize butyrate , formate , H 2 , and acetate as an electron donor (Zecchin et al. 2018 ).
The Actinomycetota represented yet another unusual phylum harboring Dsr-pathw ay encoding members (Müller et al. 2015 ).Besides the unusual Actinom ycetota MAG GCA_003 599 855, whic h encoded r eductiv e and o xidati v e DsrAB, all other r etrie v ed Actinomycetota could be split into two major groups based on the completeness of their Dsr-pathway and habitat pr efer ence.All members of the class Coriobacteriia (five genomes/MAGs within the genera Gordonibacter , Rubneribacter , Berryella , and UBA8131) encoded only the genetic potential to reduce sulfite to sulfide including DsrD, lacked the dsrL gene, and were so far isolated or encountered in intestines of humans and animals including pig, c hic ken, and termites (Würdemann et al. 2009, Selma et al. 2014, Medv ec ky et al. 2018, Parks et al. 2018, Wylensek et al. 2020 ).Cultur ed r epr esentativ es fr om the genera Gordonibacter and Berryella ar e strict anaer obes, supporting the notion that the encoded reductive bacterial-type DsrAB and DsrD point to w ar ds a reductive sulfur metabolism.Ho w ever, dissimilatory sulfite reduction by these micr oor ganisms still awaits experimental v alidation.Sulfite in the gut environment is lik ely deri ved from sulfonates , i.e .organic sulfur compounds with a SO 3 2 − moiety, such as the amino acid taurine (Wei and Zhang 2021 ) or the sugar sulfoquinovose (Hanson et al. 2021 ).The second major group within the Dsrpathway encoding Actinomycetota encodes the full Dsr-pathway including DsrD and DsrL-2C.The y re present uncultured members of the classes Thermoleophilia (Kato et al. 2018 ) and Aquicultoria (Jiao et al. 2021 ), with the latter encoding the o xygen-sensiti ve Wood-Ljungdahl pathway pointing towards a strictly anaerobic lifestyle and a r eductiv el y oper ating sulfur metabolism.In contr ast to the intestinal and incomplete Dsr-pathway encoding members of the Coriobacteriia , they were found in terrestrial and marine environments including groundwater, the terrestrial subsurface (Jiao et al. 2021 ), and deep-sea massive sulfide deposits (Kato et al. 2018 ).

Conclusion
Meta genome-driv en discov eries hav e opened a new window into the hid den di versity of SRM.We can now start to a ppr eciate that besides the four bacterial and two archaeal phyla harboring cultured SRM, the potential to perform dissimilatory sulfate/sulfite reduction extends to a total of 23 bacterial and 4 archaeal phyla.Many of the phyla now recognized to play a role in sulfur cycling wer e r epr esented by DsrAB-encoding MAGs r ecov er ed fr om low-sulfate en vironments , supporting the notion that hidden or cryptic sulfur cycling in low-sulfate environments is an understudied area.For a few of these potential SRM, such as members of the Acidobacteriota , mesophilic Nitrospirota, and Bacteriodata family UBA2268 (Kapabacteria), meta-omics based studies under constr ained envir onmental conditions could pr ovide str ong e vidence of a sulfate-reducing lifestyle.Ho w e v er, the lar ge majority of nov el, putativ e SRM still await experimental confirmation of their physiology.Furthermore, we could show that the primers used in dsrAB gene-based a ppr oac hes cov er a lar ge fr action of the novel diversity of SRM, with many of the previously taxonomicall y unr esolv ed DsrAB linea ges now anc hor ed by DsrABencoding MAGs.As such, dsrAB gene-based surveys can be used with confidence in the future to explore the enigmatic world of a functional microbial guild that has sha ped biogeoc hemical cycling on Earth since the Archaean (Shen et al. 2001, Wacey et al. 2011 ).

Figure 3 .
Figure3.Phylogenetic analysis of bacterial DsrL proteins .T he maximum likelihood tree was inferred from 438 DsrL proteins and constructed with IQ-TREE(Nguyen et al. 2015 ) using automatic substitution model selection (Q.pfam + I + R9) and ultrafast bootstrap analysis (n = 1000).Bootstrap support is shown by black dots ( ≥90%).DsrL sequences marked with an asterisk were not detected by the custom-made pHMM for DsrL ( https:// github.com/AnantharamanLab/ Diao _ et _ al _ 2023 ), but have been previously described in the literature(Hausmann et al. 2018, Löffler et al. 2020 ).Additional DsrL sequences ( * 1) were collected from MAGs with low completeness, which were not included in our MAG analysis.

Figure 4 .Figure 5 .
Figure 4. Organization of dsr gene clusters in MAGs encoding both reductive and o xidati ve bacterial-type DsrAB.

Figure 7 .
Figure 7.Comparison of phylogenomic and DsrAB trees for microorganisms representing all inferred DsrAB-encoding archaeal and bacterial lineages .T he phylogenomic tree was inferred from a set of 43 conserved single-copy marker genes obtained with CheckM (Parks et al. 2015 ) using 38 re presentati ve archaeal and 207 re presentati ve bacterial (metagenome-assembled) genomes .T he phylogenomic maximum likelihood tree was constructed with IQ-TREE (Nguyen et al. 2015 ) using automatic substitution model selection (LG + R10) and ultrafast bootstrap analysis (n = 1000).The DsrAB tree was constructed using 269 deduced DsrAB amino acid sequences, which were extracted from 245 re presentati ve (metagenome-assembled) genomes .T he DsrAB maximum likelihood tree was constructed with IQ-TREE (Nguyen et al. 2015 ) using automatic substitution model selection (LG + R8) and ultrafast bootstrap analysis (n = 1000).Bootstrap support is indicated by black dots ( ≥90%)