The most common methane-producing microorganisms have a high demand for sulfur due to their specific enzymes and metabolism. While most of these methanogens use sulfides (HS), some have been shown to metabolize higher oxidation states of sulfur or even metal sulfides (for example, FeS2) for sulfur acquisition1,2,3,4,5. However, Methanothermococcus thermolithotrophicus is the only known methanogen capable of growing on sulfate (\({{\rm{SO}}}_{4}^{2-}\)) as its sole sulfur source4,6. The metabolism of this marine hydrogenotroph, isolated from geothermally heated sea sediments near Naples (Italy), is paradoxical, as \({{\rm{SO}}}_{4}^{2-}\) reduction should lead to several physiological obstacles for a methane-producing microbe. First, methanogens commonly thrive in reduced sulfidic environments where all electron acceptors other than CO2 are depleted, including \({{\rm{SO}}}_{4}^{2-}\) (refs. 7,8). Second, at the interface where methanogens and \({{\rm{SO}}}_{4}^{2-}\) ions coexist, hydrogenotrophic methanogens must compete with dissimilatory \({{\rm{SO}}}_{4}^{2-}\)-reducing microorganisms for the common substrate dihydrogen (H2)9. Third, methanogens live at the thermodynamic limits of life and the adenosine triphosphate (ATP) hydrolysis coupled to \({{\rm{SO}}}_{4}^{2-}\) reduction would be a substantial investment for such energy-limited microorganisms8,10. Finally, the \({{\rm{SO}}}_{4}^{2-}\) reduction pathway generates toxic intermediates that would interfere with cellular processes.

To assimilate \({{\rm{SO}}}_{4}^{2-}\), the organism would have to capture the anion and transport it into the cell using a transporter. Inside the cell, \({{\rm{SO}}}_{4}^{2-}\) is activated by an ATP sulfurylase (ATPS) to generate adenosine 5′-phosphosulfate (APS)11,12,13. From there, organisms can use different strategies (Extended Data Fig. 1, routes a–c): (1a) APS is directly reduced by an APS reductase (APSR) to generate AMP and \({{\rm{SO}}}_{3}^{2-}\). (1b) Alternatively, APS can be further phosphorylated to 3′-phosphoadenosine 5′-phosphosulfate (PAPS) by the APS kinase (APSK). A PAPS reductase (PAPSR) will reduce PAPS to \({{\rm{SO}}}_{3}^{2-}\) and the toxic nucleotide 3’-phosphoadenosine 5′-phosphate (PAP). PAP must be quickly hydrolysed to AMP and inorganic phosphate by a PAP phosphatase (PAPP). In both scenarios, the final step is carried out by a siroheme-containing sulfite reductase, which reduces the \({{\rm{SO}}}_{3}^{2-}\) into HS. The latter can then be incorporated into biomass. (1c) In a different pathway, the sulfite group of PAPS is transferred to another acceptor to build up sulfated metabolites. Route 1a is very similar to the dissimilatory pathway (Extended Data Fig. 1, route 2). However, dissimilatory APSRs and dissimilatory sulfite reductases are structurally and phylogenetically distinct from their assimilatory counterparts and indirectly couple their reactions to membrane pumps, allowing for energy conservation14,15,16.

Genes encoding putative enzymes associated with \({{\rm{SO}}}_{4}^{2-}\) reduction have been found in the genomes of multiple methanogens13, including M. thermolithotrophicus. For this methanogen, a theoretical, albeit incomplete, \({{\rm{SO}}}_{4}^{2-}\) assimilation pathway can be hypothesized. Here we elucidated the complete \({{\rm{SO}}}_{4}^{2-}\) reduction machinery of this archaeon and describe how this one methanogen can convert \({{\rm{SO}}}_{4}^{2-}\) into an elementary block of life.


A marine methanogen consuming \({\mathbf{SO}}_{\mathbf{4}}^{\mathbf{2-}}\)

Cultures grown on Na2S were successively transferred to a sulfur-free medium until no growth was observed. M. thermolithotrophicus showed robust growth when at least 100 µM of Na2SO4 was supplemented in the medium and reached similar cell yields as the Na2S-grown culture. Under these cultivation conditions, \({{\rm{SO}}}_{4}^{2-}\) is consumed over time as cell density increases (Fig. 1a). When cells are grown only on Na2S, no \({{\rm{SO}}}_{4}^{2-}\) could be detected (Fig. 1a), indicating that M. thermolithotrophicus is not performing sulfide oxidation.

Fig. 1: \({\mathbf{SO}}_{\mathbf{4}}^{{\mathbf{2-}}}\) growth dependency of M. thermolithotrophicus.
figure 1

a, M. thermolithotrophicus cultures grown on Na2S (black squares, 0.5 mM) and Na2SO4 (grey squares, 0.5 mM). The consumption or release of \({{\rm{SO}}}_{4}^{2-}\) in Na2S or Na2SO4 cultures are shown by black and grey triangles, respectively. Data are presented as mean ± s.d. and individual values are shown as spheres (n = 3 replicates). Differences between expected (0.5 mM) and measured (0.13 mM) \({{\rm{SO}}}_{4}^{2-}\) concentration for the initial point are considered to be due to an artefact from the medium (see Methods). b, M. thermolithotrophicus grown on 10 mM Na2SO4 in three independent fermenters. The sampling points are represented by grey squares. c, Molybdate (Na2MoO4) inhibition of Na2SO4 assimilatory and dissimilatory archaea. Growth experiments for M. thermolithotrophicus were performed in duplicates and for A. fulgidus in quadruplicates (left) or triplicates (right). Data are represented as mean and for the triplicates and quadruplicates ± s.d. d, Predicted operon for \({{\rm{SO}}}_{4}^{2-}\) reduction from the whole-genome shotgun sequence of M. thermolithotrophicus.

Source data

We then challenged the \({{\rm{SO}}}_{4}^{2-}\)-grown culture by switching from batch to fermenter conditions, where H2S can escape to the gas phase and does not accumulate compared to flask conditions. In this open system with temperature and pH controlled, M. thermolithotrophicus grew to a maximum optical density (OD)600nm of 6.45 within 19 h (Fig. 1b).

One way to determine whether M. thermolithotrophicus relies on canonical enzymes of the \({{\rm{SO}}}_{4}^{2-}\) reduction pathway is to use molybdate (\({{\rm{MoO}}}_{4}^{2-}\)). The structural analogue of \({{\rm{SO}}}_{4}^{2-}\) binds to the ATPS and triggers molybdolysis, which hydrolyses ATP to AMP and pyrophosphate (PPi), resulting in cellular energy depletion17,18. A \({{\rm{MoO}}}_{4}^{2-}\):\({{\rm{SO}}}_{4}^{2-}\) molar ratio of 0.004:1 is sufficient to inhibit the activity of dissimilatory \({{\rm{SO}}}_{4}^{2-}\)-reducing bacteria for 168 h, an effect mainly due to molybdolysis by ATPS19,20,21. \({{\rm{SO}}}_{4}^{2-}\) assimilation is also affected by \({{\rm{MoO}}}_{4}^{2-}\), as demonstrated by studies on plants22. In the latter, growth inhibition occurred when \({{\rm{MoO}}}_{4}^{2-}\) was in excess compared to \({{\rm{SO}}}_{4}^{2-}\) and the ATPS activity was notably affected at a 1:1 ratio18. When applied on M. thermolithotrophicus, a high \({{\rm{MoO}}}_{4}^{2-}\):Na2S ratio of 12.5:1 did not disturb growth of the Na2S culture, indicating that \({{\rm{MoO}}}_{4}^{2-}\) is not interfering with their basal metabolism. In contrast, a \({{\rm{MoO}}}_{4}^{2-}\):\({{\rm{SO}}}_{4}^{2-}\) ratio of 6.25:1 was inhibitory to the \({{\rm{SO}}}_{4}^{2-}\)-grown culture, while a 1:1 ratio was not (Fig. 1c and Extended Data Fig. 2a). \({{\rm{SO}}}_{4}^{2-}\) addition to the \({{\rm{MoO}}}_{4}^{2-}\)-inhibited culture restored growth (Extended Data Fig. 2b), indicating the reversibility of inhibition and its strict control by the \({{\rm{MoO}}}_{4}^{2-}\):\({{\rm{SO}}}_{4}^{2-}\) ratio rather than the \({{\rm{MoO}}}_{4}^{2-}\) concentration. In comparison, in Archaeoglobus fulgidus, an archaeon that performs dissimilatory \({{\rm{SO}}}_{4}^{2-}\) reduction to conserve energy, we observed growth inhibition at a \({{\rm{MoO}}}_{4}^{2-}\):\({{\rm{SO}}}_{4}^{2-}\) ratio of 0.001:1 (Fig. 1c and Extended Data Fig. 2c). These results suggest that M. thermolithotrophicus reduces \({{\rm{SO}}}_{4}^{2-}\) via an assimilatory pathway containing a functional ATPS. Genes coding for putative standalone ATPS and APSK were indeed on the same locus in the genome of the strain DSM2095 that we had re-sequenced (Fig. 1d) (refs. 13,23). To confirm their functions, the ATPS and APSK from M. thermolithotrophicus (MtATPS and MtAPSK, respectively) were further characterized.

A classic ATPS/APSK to activate \({\mathbf{SO}}_{\mathbf{4}}^{\mathbf{2-}}\)

The activity of the recombinantly expressed MtATPS and MtAPSK was tested via a coupled assay (Fig. 2a and Supplementary Fig. 1) and a specific activity of 0.070 ± 0.004 µmol of oxidized NADH min−1 mg−1 of MtATPS was measured. Under these conditions, the rate-limiting step was the pyrophosphatase activity. This highlights the need for rapid pyrophosphate degradation (Fig. 2a) to avoid a retro-inhibition as previously shown for other ATPS24. A \({{\rm{MoO}}}_{4}^{2-}\):\({{\rm{SO}}}_{4}^{2-}\) ratio of 1:1.25 decreased the activity by half (see Methods), corroborating that ATPS is also reacting with \({{\rm{MoO}}}_{4}^{2-}\) as shown in other homologues18,21.

Fig. 2: MtATPS and MtAPSK catalyse the first steps of the \({\mathbf{SO}}_{\mathbf{4}}^{{\mathbf{2-}}}\) reduction pathway.
figure 2

a, Top panel, reactions catalyzed by MtATPS and MtAPSK; Bottom panel, the specific activity of MtATPS and MtAPSK, determined via a coupled enzyme assay. ‘-’ indicates the absence of the indicated reactant. Data are presented as mean ± s.d. and individual values are shown as grey spheres (n = 3 replicates). b, MtATPS homodimeric structure in which one monomer is shown in light yellow surface and the other one in cartoon. c, Active site of MtATPS (yellow) superposed on the ATPS from Thermus thermophilus HB8 (PDB: 1V47, grey) containing the APS shown as balls and sticks with carbons coloured cyan. Residues involved in substrate binding are highlighted in sticks and only the ones from MtATPS are labelled. Hydrogen bonds between the ATPS from T. thermophilus and APS are represented as dashed lines. Nitrogen, oxygen, phosphorus and sulfur are coloured in blue, red, orange and yellow, respectively. d, Sequence conservation across ATPS homologues. e, MtAPSK homodimeric structure in which one monomer is shown in light orange surface and the other one in cartoon. The flexible loop illustrated by the dashed line could not be modelled. In all structures, the N and C termini are shown by a blue and red sphere, respectively. f, Sequence conservation across APSK homologues. For d and f, red bold residues are involved in substrate binding, while red and black stars are perfectly and well-conserved residues, respectively.

Source data

The structure of MtATPS was refined to 1.97 Å resolution and obtained in an apo state despite co-crystallization with APS and \({{\rm{SO}}}_{4}^{2-}\) (Extended Data Table 1). While the crystal packing suggests a homotetrameric assembly in two crystalline forms, size exclusion chromatography and surface analysis using PISA ( confirmed a homodimeric state similar to bacterial homologues (Extended Data Fig. 3a and Supplementary Fig. 2). The structure exhibits the typical ATPS fold comprising three domains (domain I, 1–156; domain II, 164–314 and domain III, 320–382; Fig. 2b). The dimeric interface is mainly organized by domain III as observed in T. thermophilus, a notable difference compared with other structural homologues (Extended Data Fig. 3a,b) (refs. 25,26,27). Similar to many thermophilic bacteria and archaea, the domain III contains a zinc-binding domain (320–343; Extended Data Fig. 3c,d) that might contribute to thermal stability27. MtATPS superposition with structural homologues shows a slight domain rearrangement probably due to the absence of substrate (Extended Data Fig. 3b). All residues critical for the reaction are conserved in MtATPS, arguing for a conserved reaction mechanism (Fig. 2c,d, Extended Data Figs. 3e,f and 4a, and Supplementary Fig. 3).

The APS-kinase model from M. thermolithotrophicus, MtAPSK, was refined to 1.77 Å. MtAPSK forms a homodimer with an organization very similar to bacterial enzymes, which was expected due to its high sequence conservation (Extended Data Figs. 4b and 5a). Despite co-crystallization and soaking the crystals with APS and MgCl2, the MtAPSK structure was obtained in its apo state with a bound phosphate at the expected position of the ATP β-phosphate (Fig. 2e and Extended Data Fig. 5b,c). The N terminus and region 125–152 (the latter being involved in substrate binding28,29) could not be modelled due to the lack of electron density. However, the residues binding the substrates and Mg2+ are conserved (Fig. 2f, Extended Data Fig. 5b,c and Supplementary Fig. 4), suggesting that MtAPSK should be functional, as confirmed by the coupled enzyme assay.

An exonuclease-derived PAP phosphatase

If the ATPS and APSK are active, they will produce PAPS, an intermediate that could follow the metabolic routes 1b or 1c (Extended Data Fig. 1). Both routes will lead to the production of the toxic product PAP, which inhibits sulfotransferases and exoribonucleases, and disrupts RNA catabolism30,31. Therefore, it needs to be efficiently hydrolysed by a PAP phosphatase. While the genome did not contain any related PAP phosphatase, a gene coding for a putative phosphoesterase (Fig. 1d) was found in the genomic environment harbouring the ATPS and APSK genes. This PAP-phosphatase candidate, belonging to the DHH family, was recombinantly expressed and produced inorganic phosphate (Pi) from PAP at fast rates (50.2 ± 5.9 µmol of Pi released min−1 mg−1 of purified enzyme with manganese). The activity was stimulated by manganese addition and showed a high specificity towards PAP (Fig. 3a).

Fig. 3: A unique type of PAP phosphatase.
figure 3

a, Top panel, reaction catalysed by MtPAPP; Bottom panel, specific activity of the MtPAPP determined via the production of Pi (left) and relative enzymatic activity towards different nucleotides (right). Data are presented as mean ± s.d. and individual values are shown as grey spheres (n = 3 replicates). b, Organization of MtPAPP shown in cartoon representation. The N and C termini are highlighted as blue and red balls, respectively. Carbon, nitrogen, oxygen and phosphorus of AMP is coloured as cyan, blue, red and orange, respectively. c, Close-up view of the active site of MtPAPP. The residues coordinating the AMP and Mn2+ ion are highlighted by sticks and coloured as in b, with the residues from the DHH motif coloured in pink. d, Cut-through view of MtPAPP structure shown in surface representation and coloured by its sequence conservation across 168 archaeal homologues. The colour gradient ranges from variable (teal) to conserved (magenta). e, Secondary structure representation was done with ESPript 3.0 (ref. 61). The coloured frame corresponds to the different domains: DHH domain in light green, linker in grey and DHHA1 domain in darker green. Perfectly and well-conserved residues across 168 archaeal homologues are highlighted in red and yellow, respectively. The secondary structures composing MtPAPP are labelled, in which β-sheets, α-helices, and β-turns are highlighted as arrows, springs and bold TT, respectively.

Source data

To decipher the mechanism of this uncanonical PAP phosphatase (named MtPAPP), the enzyme was co-crystallized with manganese and PAP. The structure, solved by molecular replacement with a template generated by AlphaFold2 (refs. 32,33), was refined to 3.1 Å resolution and contained the product AMP and an ion in its active site, modelled as a partially occupied Mn2+ (Extended Data Table 1). While the MtPAPP sequence does not align with homologues belonging to the DHH family (except for the DHH motif), it shares an overall fold similar to the exonuclease RecJ or the oligoribonuclease NrnA from Bacillus subtilis (BsNrnA, which also exhibits PAP-phosphatase activity; Extended Data Fig. 6a) (refs. 34,35,36). The monomer is composed of an N-terminal (DHH, residues 1–180) and a C-terminal domain (DHHA1, residues 211–315) interconnected by a linker region (residues 181–210), forming a central groove (Fig. 3b). The DHH domain contains the catalytic site and the DHHA1 domain serves as a scaffold to bind the substrate with high specificity (Fig. 3b,c and Extended Data Fig. 6b). The motif coordinating the Mn2+ ion in RecJ and BsNrnA is perfectly conserved in MtPAPP34,36, therefore we expect that in its active state, MtPAPP would be loaded with two Mn2+. The first one, partially observed in the structure, is coordinated by four aspartates (Asp8, Asp10, Asp57, Asp127) and a long-range interaction with His6. The absent second Mn2+ would be coordinated by the Asp10, Asp57, Asp127, the DHH motif (His76, His77) as well as by water molecules (Extended Data Fig. 6c). While the AMP shares a similar localization with structural homologues (β9β10β11), it is bound by a different interaction with the protein (Extended Data Fig. 6b and Supplementary Fig. 5). The nucleotide binding site would ideally place the 3′-phosphate of the PAP in front of the manganese when the enzyme is in its closed state (Extended Data Fig. 6c). The inter-domain movement, allowed by the linker, would facilitate a rapid exchange of the substrate/product, increasing the turnover of MtPAPP. The complete sequence of this PAP phosphatase was found in the genome of 168 archaea in which the nucleotide binding site is conserved (Fig. 3d,e and Supplementary Fig. 6). This suggests a common enzyme in archaea to detoxify PAP (Extended Data Fig. 7a).

A dissimilatory APSR-like enzyme reduces PAPS

No genes encoding for a canonical PAPS reductase (route 1b) or sulfo-transferase (route 1c) were found in the M. thermolithotrophicus genome. However, genes annotated as dissimilatory APS reductase (α and β subunit, APSR; Extended Data Fig. 7b and Supplementary Fig. 7) are present and co-occur with the previously described genes (Fig. 1d).

To experimentally confirm the activity and substrate specificity of this APS-reductase-like enzyme, both subunits were co-expressed in Escherichia coli, purified and tested for enzyme activity assays (Fig. 4a). In contrast to dissimilatory APSRs which catalyse the reversible reduction of APS to AMP and \({{\rm{SO}}}_{3}^{2-}\) (refs. 37,38), we could not measure the reverse reaction (that is, AMP and \({{\rm{SO}}}_{3}^{2-}\), or PAP and \({{\rm{SO}}}_{3}^{2-}\) as substrates) for M. thermolithotrophicus enzyme by using K3Fe(CN)6 as an electron acceptor. Instead, we used a coupled enzyme assay to reconstitute the pathway in vitro (Extended Data Fig. 8). MtATPS, a pyrophosphatase and the MtAPSK were used to generate PAPS and MtPAPP was added to remove PAP, a potential retro-inhibitor of the reaction39. The activity was monitored via the oxidation of reduced methyl viologen (MVred). When all components were present, a specific enzymatic activity of 0.114 ± 0.007 µmol of oxidized MV min−1 mg−1 of the APS-reductase-like enzyme was measured. A fivefold excess of the APS-reductase-like enzyme resulted in a 220% increase of the specific enzyme activity, indicating that the enzyme was the rate-limiting step of the reaction (Extended Data Fig. 8c). However, the accumulation of PAP (induced by the removal of MtPAPP or Mn2+) strongly inhibited the activity. The specific enzymatic activity with APS as a substrate (that is, removal of MtAPSK) was 0.007 ± 0.001 µmol of oxidized MV min−1 mg−1 of the APS-reductase-like enzyme (Fig. 4a). Considering the complexity of this coupled enzyme assay, kinetic parameters could not be determined. However, the assay did provide insights about the substrate specificity and confirmed that the APS-reductase-like enzyme from M. thermolithotrophicus exhibits traits of a PAPS reductase.

Fig. 4: MtPAPSR has a dissimilatory APS-reductase architecture but is specific to PAPS.
figure 4

a, Top panel, reaction catalyzed by MtPAPSR; Bottom panel, relative enzyme activity of MtPAPSR, determined via a coupled enzyme assay (Extended Data Fig. 8). Data are presented as mean ± s.d. and individual values are shown as grey spheres (n = 3 replicates). b, MtPAPSR organization with one heterodimer in surface representation and the other in cartoon. N and C termini of both subunits are shown as balls and coloured in blue and red, respectively. Heterodimeric partners are labelled with a prime. Carbon, nitrogen, oxygen, phosphorus, iron and sulfur are coloured in lemon, blue, red, orange, brown and yellow, respectively. c, Close-up of cofactors and the electron flow. [4Fe-4S] clusters and cysteines coordinating them, FAD and the Trp42 proposed to participate in the electron transfer are shown in sticks and balls and coloured as in b. d,e, Active sites of APSR (d) from A. fulgidus containing APS (AfAPSR, PDB: 2FJA) and MtPAPSR with an artificially modelled PAPS (e) shown with a transparent surface. Residues involved in substrate recognition (based on modelled PAPS) are in balls and sticks and coloured as in b. A red arrow points to where PAPS would clash. f, Sequence conservation across the alpha subunit of MtPAPSR, AfAPSR, Megalodesulfovibrio gigas (DgAPSR, PDB: 3GYX) and the putative APSR from Caldanaerobius fijiensis (CfAPSR, WP_073344903), which shares 68% sequence identity with MtPAPSR. Residues involved in APS binding for APSR are in bold and red; perfectly and well-conserved residues are highlighted with red and black stars, respectively. Trp206 and Tyr207 are involved in FAD binding. The sequence alignment was done with MUSCLE62.

Source data

To gain further molecular insights into the unconventional APS-reductase-like enzyme from M. thermolithotrophicus, the enzyme was crystallized under anaerobic conditions. The structure was solved by a single-wavelength anomalous dispersion experiment measured at the Fe K-edge and refined to 1.45 Å resolution (Extended Data Table 1). The complex organizes as an α2β2 heterotetramer, with the same assembly as dissimilatory APS reductases (Fig. 4b and Extended Data Fig. 9). It is, however, drastically different from characterized single-domain assimilatory APS/PAPS reductases, which are thioredoxin/glutathione-dependent. Assimilatory APS/PAPS reductases share no sequence or structural homology with the M. thermolithotrophicus enzyme, and several motifs that are proposed to mediate substrate binding and catalytic activity in assimilatory APS/PAPS reductases are absent (Extended Data Fig. 9) (ref. 40).

In M. thermolithotrophicus, the α subunit, containing the flavin adenine dinucleotide (FAD), is a member of the fumarate reductase family37,41,42 and the β subunit is mainly composed of a ferredoxin-like domain in which two [4Fe-4S] clusters are coordinated by eight cysteine residues (Fig. 4c). While there are no assimilatory P/APS-reductase homologues to M. thermolithotrophicus enzyme, it shares 38% sequence identity with the α subunit of the dissimilatory APS reductase from A. fulgidus (AfAPSR PDB: 2FJA, rmsd of 1.02 Å for 437 Cα aligned on the α subunit). The residues coordinating APS, invariable in the dissimilatory family, differ in M. thermolithotrophicus and might provoke a switch of specificity from APS to PAPS. Despite a short soak with PAP, the putative substrate pocket contains only solvent, and we used the AfAPSR to artificially model PAPS in the active site of the enzyme (Fig. 4d,e). The different substitutions mainly carried by loop 104–123 would accommodate the additional 3′-phosphate group by salt-bridge interactions and hydrogen bonds (Fig. 4d–f). In APS reductases, however, a conserved glutamine (α145 in A. fulgidus) would clash with this phosphate group. The catalytic residues proposed in dissimilatory APS reductases are retained in the enzyme of M. thermolithotrophicus (Extended Data Fig. 9 and Supplementary Fig. 7). We therefore propose an identical reaction mechanism on the basis of a nucleophilic attack of the atom N5 of FAD on the sulfur PAPS, which creates a FAD-PAPS intermediate that decays to PAP and FAD-\({{\rm{SO}}}_{3}^{2-}\) (refs. 37,42). Taking together the enzyme rates and the structural analysis, we propose that M. thermolithotrophicus harbours a unique class of PAPS reductase (MtPAPSR) used to convert PAPS into \({{\rm{SO}}}_{3}^{2-}\) and PAP.

F420-dependent sulfite reductase catalyses the last step of the pathway

The \({{\rm{SO}}}_{3}^{2-}\) generated by MtPAPSR must be further reduced to HS. In hydrogenotrophic methanogens, \({{\rm{SO}}}_{3}^{2-}\) damages the methane-generating machinery and must be detoxified by the F420-dependent sulfite reductase (Fsr)23,43. We previously identified and characterized Group I Fsr in M. thermolithotrophicus (MtFsr) and determined a robust enzymatic activity towards \({{\rm{SO}}}_{3}^{2-}\) (ref. 23). Besides a second Fsr isoform, M. thermolithotrophicus does not contain other potential sulfite reductases. While mass spectrometry confirmed that the Fsr isolated from \({{\rm{SO}}}_{4}^{2-}\)-grown cells is the characterized Group I MtFsr, the physiological role of the second Fsr isoform remains unknown23. Therefore, MtFsr is the best candidate to catalyse the final reduction of \({{\rm{SO}}}_{3}^{2-}\) to HS. Native polyacrylamide gel electrophoresis (native PAGE) with cell extracts of cultures grown on different sulfur substrates confirmed the absence of MtFsr from cells grown on Na2S and its high abundance in cells grown on \({{\rm{SO}}}_{3}^{2-}\) (refs. 23,43).

We determined a specific sulfite reductase activity of 18.42 ± 0.13 µmol of oxidized MV min−1 mg−1 of cell extract from Na2SO3-grown cells, in comparison to 7.31 ± 0.63 µmol of oxidized MV min−1 mg−1 of cell extract from Na2SO4-grown cells, whereas cell extract from an Na2S-grown culture had a specific sulfite reductase activity of 3.04 ± 0.25 µmol of oxidized MV min−1 mg−1 (Fig. 5a). In agreement, we observed a band compatible with Fsr on the native PAGE for the \({{\rm{SO}}}_{4}^{2-}\)-grown culture but in lower amounts compared with \({{\rm{SO}}}_{3}^{2-}\) conditions (Fig. 5b). The MtFsr structure recently published by our group was obtained from \({{\rm{SO}}}_{4}^{2-}\)-grown cells, which confirmed that it is the same enzyme expressed as under \({{\rm{SO}}}_{3}^{2-}\) conditions23. Taken together, these results argue that MtFsr is used as the last enzyme in the \({{\rm{SO}}}_{4}^{2-}\) reduction pathway (Fig. 5c).

Fig. 5: Proposed \({\mathbf{SO}}_{\mathbf{4}}^{{\mathbf{2-}}}\) assimilation pathway in a methanogen.
figure 5

a, Sulfite-reductase activity in cell extract from M. thermolithotrophicus grown on different sulfur sources. Data are presented as mean ± s.d. and individual values are shown as grey spheres (n = 3 biologically independent replicates). b, hrCN gel with M. thermolithotrophicus cell extract grown on different sulfur sources (15 µg protein loaded per sample, n = 2 biologically independent duplicates). c, Proposed \({{\rm{SO}}}_{4}^{2-}\) assimilation pathway in M. thermolithotrophicus. Yellow and grey backgrounds highlight the \({{\rm{SO}}}_{4}^{2-}\) reduction and methanogenesis pathways, respectively. Thick arrows indicate high metabolic fluxes. The structures of the enzymes operating the \({{\rm{SO}}}_{4}^{2-}\)-assimilation pathway are shown in surface representation, with ligands as balls and sticks. Enzymes are abbreviated as follows: Fwd/Fmd, formylmethanofuran dehydrogenases; Ftr, tetrahydromethanopterin (H4MPT) formyltransferase; Mch, methenyl-H4MPT cyclohydrolase; Mtd, methylene-H4MPT dehydrogenase; Mer, 5,10-methylene-H4MPT reductase; Mtr, N5-CH3-H4MPT: coenzyme M methyl-transferase; Mcr, methyl-coenzyme M reductase; Adk, adenylate kinase; Frh, F420-reducing [NiFe]-hydrogenase; Eha/Ehb, energy-converting hydrogenase. The putative \({{\rm{SO}}}_{4}^{2-}\) transporters belonging to the class DASS/SUIP are proposed to be WP_018154444/WP_018154062 and the pyrophosphatase WP_018154121.

Source data

Genetic potential is not enough to sustain \({\mathbf{SO}}_{\mathbf{4}}^{\mathbf{2-}}\) growth

Methanogens commonly use HS as a sulfur source, and the ones that express Fsr type I can also grow on \({{\rm{SO}}}_{3}^{2-}\) (refs. 13,23,43). Interestingly, some methanogens have genes that encode for proteins of the complete or partial \({{\rm{SO}}}_{4}^{2-}\) reduction pathway (Supplementary Fig. 8) (ref. 13). So why is M. thermolithotrophicus the only methanogen so far that has been proven to grow on \({{\rm{SO}}}_{4}^{2-}\)? We used Methanocaldococcus infernus as a model organism to investigate this further. M. infernus is a marine hyperthermophile that shares a very similar physiology with M. thermolithotrophicus and can grow in the same medium. It contains all genes coding for the enzymes characterized in this study except for the described PAPSR. However, the M. infernus genome encodes for a putative thioredoxin-dependent PAPSR and APSR, which share high sequence identities with the biochemically characterized assimilatory APSR and PAPSR from M. jannaschii (Fig. 6a and Extended Data Fig. 7b) (refs. 44,45). Therefore, based on genomic information, M. infernus should be able to assimilate \({{\rm{SO}}}_{4}^{2-}\).

Fig. 6: \({\mathbf{SO}}_{\mathbf{4}}^{{\mathbf{2-}}}\) reduction potential in Methanococcales.
figure 6

a, Methanocaldococcus infernus has the genomic potential to perform the whole \({{\rm{SO}}}_{4}^{2-}\) assimilation pathway. WP_013099421 has 59.68% amino acid sequence identity with the MtPAPP, WP_013099422 has 70.60% sequence identity with MtATPS and WP_157198836 has 75.44% sequence identity with MtAPSK. The APSR (WP_013100115) and PAPSR (WP_013099852) are similar to the biochemically characterized APSR and PAPSR from M. jannaschii (68.64% and 58.35% amino acid sequence identity, respectively), which have been shown to reduce APS/PAPS44,45. WP_013099852 is not homologous to MtPAPSR but homologous to WP_018154242, a putative PAPS reductase in M. thermolithotrophicus. WP_013100746 has 65.30% sequence identity to Group I MtFsr. b, Growth of M. thermolithotrophicus and M. infernus on 2 mM Na2S, without an additional sulfur source, 2 mM Na2SO3, 2 mM Na2SO4 or 2 mM Na2SO4 with 2 mM Na2S. The \({{\rm{SO}}}_{3}^{2-}\) in brackets indicates that it was used as the sulfur substrate for the inoculum. Represented are the maximum OD600nm of the cultures, in triplicates, shown as mean ± s.d. M. infernus cultures grown without sulfur,Na2SO4 and on Na2S with Na2SO4 are in duplicates. The individual data point of each replicate is shown as a sphere.

Source data

M. thermolithotrophicus and M. infernus were grown in the same medium and under the same cultivation conditions except that M. infernus was kept at 75 °C and M. thermolithotrophicus at 65 °C. M. infernus grew on 2 mM Na2S and Na2SO3 but was unable to use \({{\rm{SO}}}_{4}^{2-}\) as a sole source of sulfur in contrast to M. thermolithotrophicus (Fig. 6b).

This raises the question about the physiological function of the genes related to \({{\rm{SO}}}_{4}^{2-}\) assimilation in methanogenic archaea. Based on our data, it could be that other methanogens still require these enzymes to acquire sulfur via the route 1c (Extended Data Fig. 1). The sulfur group would be transferred to an acceptor by a non-canonical sulfo-transferase, which might be important for uncharted biosynthetic pathway(s). This could explain why the gene coding for the PAPP is still present in methanogens also harbouring the genes encoding an ATPS as well as an APSK. A counterargument to this hypothesis is the presence of the thioredoxin-dependent PAPSR or APSR, characterized in M. jannaschii, which rather argues for route 1b (refs. 13,44,45). It is worth noting that the gene coding for this putative assimilatory APSR also exists in M. thermolithotrophicus (WP_018154242.1). Therefore, further biochemical investigations will be needed to elucidate the physiological roles of these enzymes in methanogens.


This work unveiled the unique \({{\rm{SO}}}_{4}^{2-}\)-assimilation metabolism of a methanogenic archaeon, offering a molecular snapshot of the complete set of enzymes involved in the pathway. M. thermolithotrophicus activates \({{\rm{SO}}}_{4}^{2-}\) by conventional ATPS and APSK, but transforms it further by uncanonical enzymes (Fig. 5c).

PAPS produced by the APSK is usually metabolized by thioredoxin- or glutathione-dependent assimilatory PAPS reductases, which are organized as homo-oligomers. In contrast, MtPAPSR inherited the heterotetrameric organization and FAD-based catalytic mechanism from dissimilatory APS reductases (Fig. 4, and Extended Data Figs. 1 and 9). We propose that the substitution of only a few amino acids switched the specificity towards PAPS (Fig. 4d–f), which might have been the result of a fine-tuned evolutionary adaptation to promote assimilatory \({{\rm{SO}}}_{4}^{2-}\) reduction. It would be worthwhile to exchange the residues that confer PAPSR traits at the active site (Ser122, Lys120, Arg121) with those of APSR and observe the effects on substrate affinity.

The generated PAP is efficiently hydrolysed by MtPAPP. This PAP phosphatase belongs to the DHH family of phosphoesterases and shares structural homology with exonucleases but has no sequence homology with them. In comparison, conventional PAP phosphatases (part of the FIG superfamily) have a different fold (that is, CysQ) and use three magnesium ions to hydrolyse the 3′-phosphate of PAP30. MtPAPP appears to be a remarkable example of convergent evolution, illustrating how archaea developed their own apparatus to detoxify PAP efficiently.

Group I Fsr catalyses the final step of the \({{\rm{SO}}}_{4}^{2-}\) reduction pathway. This enzyme shows distinct traits of dissimilatory sulfite reductases, with the active site composition of an assimilatory one23. By encoding the fsr gene on a different locus and most probably under a different regulator for its expression (for example, sulfite sensor), the methanogen is able to uncouple rapid \({{\rm{SO}}}_{3}^{2-}\) detoxification from expressing the whole \({{\rm{SO}}}_{4}^{2-}\) assimilation machinery. While the MtATPS, MtAPSK and MtPAPSR show rather slow catalytic rates (see Supplementary Discussion), MtFsr and the MtPAPP have high specific activities compared with the first steps, triggering the equilibrium towards HS production and efficiently eliminating toxic intermediates. Although our proposed pathway (Fig. 5c) would allow favourable thermodynamics, the first reactions should be regulated to avoid unnecessary ATP hydrolysis. We suspect that MtATPS, MtAPSK and MtPAPSR are cross-regulated by the accumulation of their own products, as already shown for homologues39,46,47, which would allow direct retro-control to harmonize the intracellular sulfur flux.

M. thermolithotrophicus lives at the thermodynamic limit of life but the described \({{\rm{SO}}}_{4}^{2-}\) assimilation requires the hydrolysis of three ATP to ADP for one processed \({{\rm{SO}}}_{4}^{2-}\). Nevertheless, it is expected that under natural conditions, the benefit of fixing \({{\rm{SO}}}_{4}^{2-}\) counterbalances the energy expenditure. The \({{\rm{SO}}}_{4}^{2-}\)-grown cultures are not hampered by the additional energy requirement, which can be explained by our cultivation conditions that provide a high and constant H2 partial pressure. Under environmental conditions, with a lower and fluctuating H2 partial pressure, growth on \({{\rm{SO}}}_{4}^{2-}\) is likely to be more challenging for M. thermolithotrophicus. While the methanogen cannot avoid the ATP investment, it may have found an energy-saving strategy for the 8-electron reduction reaction from PAPS to HS. Fsr oxidizes F420H2, which is reduced back by the F420-reducing hydrogenase23,48. F420H2 or NAD(P)H would be advantageous electron donors for MtPAPSR, but it would require the assistance of an oxidase partner that has not yet been identified. Alternatively, the standalone MtPAPSR may depend on reduced ferredoxin, which could be obtained from the H2-dependent ferredoxin reduction via the Eha/Ehb complex, another advantageous strategy of hydrogenotrophs to provide reducing power to fuel anabolic reactions (proposed in Fig. 5c) (ref. 49).

So far, it appears that the concomitant process of methanogenesis and complete \({{\rm{SO}}}_{4}^{2-}\) reduction to HS is restricted to M. thermolithotrophicus. Strikingly, the only apparent difference between M. thermolithotrophicus and other methanogens with the genomic potential to perform \({{\rm{SO}}}_{4}^{2-}\) reduction is the acquisition of a PAPS reductase, which appears to belong to the dissimilatory family (Supplementary Fig. 8, Extended Data Fig. 7b and Supplementary Discussion). The physiological function of these \({{\rm{SO}}}_{4}^{2-}\)-reduction-associated genes in other methanogens remains to be uncovered, as well as the advantages of assimilating \({{\rm{SO}}}_{4}^{2-}\) for M. thermolithotrophicus. From an ecological point of view, it might be beneficial, if not essential, for M. thermolithotrophicus survival to be able to switch from H2S uptake to \({{\rm{SO}}}_{4}^{2-}\) reduction under environmental conditions (see Supplementary Discussion).

The transplantation of the M. thermolithotrophicus \({{\rm{SO}}}_{4}^{2-}\) reduction system into methanogenic hosts, which are already used as gas converters (for example, Methanothermobacter), would circumvent the need for highly toxic and explosive H2S by using inexpensive and abundant \({{\rm{SO}}}_{4}^{2-}\). Beyond opening fantastic possibilities for safer biotechnological applications, a \({{\rm{SO}}}_{4}^{2-}\)-reducing hydrogenotrophic methanogen also reinforces the question about the extent of an intertwined methanogenesis and sulfate reduction pathway during the evolution of early archaea. M. thermolithotrophicus has most probably assembled the entire \({{\rm{SO}}}_{4}^{2-}\) reduction pathway progressively via a ‘mix-and-match’ scenario, providing a competitive advantage under fluctuating sulfur-source conditions and expanding its ecological niches.


Archaea strains and cultivation media

M. thermolithotrophicus (DSM 2095), M. infernus (DSM 11812) and A. fulgidus (DSM 4304) cells were obtained from the Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures (Braunschweig, Germany). M. thermolithotrophicus and M. infernus were cultivated in the same previously described minimal medium with some modifications23 (see Extended Data for the complete composition of the media).

Anaerobic growth of Archaea

Cell growth was followed spectrophotometrically by measuring the OD600. The purity of the culture was checked by light microscopy. The methanogens were cultivated with 1 × 105 Pa of H2:CO2 at an 80:20 ratio in the gas phase. M. infernus was cultivated at 75 °C in 250 ml glass serum flasks and M. thermolithotrophicus was grown at 65 °C in flasks or fermenters. The serum flasks were not shaken but standing. A. fulgidus was cultivated in anaerobic and sealed 22 ml Hungate tubes, with 0.8 × 105 Pa N2:CO2. DSM 4304 culture (0.5 ml) was grown in 10 ml of classic media (see Supplementary Materials for the complete composition of the media composition) containing a final concentration of 20 mM d/l-lactate. The culture was incubated at 80 °C, standing. All cultures were stored at room temperature in the dark under anaerobic conditions. For the A. fulgidus medium, we found that high molybdate concentrations made it unstable. One of the bottles with a high MoO42− concentration turned yellow (unrelated to O2 contamination) and was omitted, resulting in triplicate instead of quadruplicate cultures (Fig. 1c, right panel).

Adaptation of M. thermolithotrophicus to \({\mathbf{SO}}_{\mathbf{4}}^{\mathbf{2-}}\) and minimal \({\mathbf{SO}}_{\mathbf{4}}^{\mathbf{2-}}\) requirement

M. thermolithotrophicus cells grown on 2 mM Na2S were successively transferred to 10 ml sulfur-free cultivation medium. After two transfers, the carry-over sulfur concentration of the inoculum did not support growth of M. thermolithotrophicus. By supplementing 2 mM Na2SO4, M. thermolithotrophicus growth was resumed. No reducing agent was added to cope with the absence of HS, which normally establishes a suitable reducing environment. Incubation without shaking is particularly important for reproducibility. Therefore, after inoculation, the cultures were incubated at 65 °C, standing for one night followed by shaking at 180 revolutions per minute (r.p.m.) until they reached their maximum OD600. The gas phase was refreshed after the overnight incubation to maintain the pressure at 1 × 105 Pa of H2:CO2. To measure the minimal \({{\rm{SO}}}_{4}^{2-}\) concentration required to sustain growth, sulfur-limited M. thermolithotrophicus cells (using an inoculum to medium ratio of 1:20) were provided with 2 mM, 1 mM, 0.5 mM, 0.25 mM, 0.1 mM and 0.04 mM Na2SO4. Growth was still observable for cells grown on 0.1 mM but not on 0.04 mM Na2SO4.

\({\mathbf{SO}}_{\mathbf{4}}^{\mathbf{2-}}\) measurements via ion chromatography

Ion chromatography (Methrom ion chromatograph) was used to measure the \({{\rm{SO}}}_{4}^{2-}\) concentrations, analysed via the software IC MagIC Net 3.2. A volume of 8 ml per sample was required, with a maximum concentration of 0.5 mM \({{\rm{SO}}}_{4}^{2-}\). \({{\rm{SO}}}_{4}^{2-}\)-reducing M. thermolithotrophicus cells were therefore grown in 1 l Duran bottles with 100 ml sulfur-free media, which was supplemented with 0.5 mM \({{\rm{SO}}}_{4}^{2-}\) before inoculation. As a negative control, 0.5 mM Na2S-grown M. thermolithotrophicus cells were used, inoculated and collected similarly as the \({{\rm{SO}}}_{4}^{2-}\)-reducing cultures. All samples were taken aerobically and were passed through a 0.45 µM filter (Sartorius). If the cell densities were too high to be filtered, the samples were centrifuged at 13,000 × g for 7 min at 4 °C and the supernatant was taken for ion chromatography measurements. The samples were stored at 4 °C if the measurements were not immediately performed.

Growth of M. thermolithotrophicus in a fermenter

M. thermolithotrophicus was grown in three independent fermenters at 60 °C, with 10 mM Na2SO4 as sole sulfur source. For each fermenter, 7 l of anaerobic cultivation medium (see Sulfur-free cultivation medium for Methanococcales) supplemented with 10 mM Na2SO4 was continuously bubbled with H2:CO2 (80:20, 3 l min−1). Under stirring (220 r.p.m.), the medium was inoculated with 360 ml preculture (with an OD600 higher than 3). One hour after inoculation, the culture was stirred at 800 r.p.m. NaOH (1 M) was used as a base to readjust the pH upon acidification, which was controlled using a pH probe. The cells were grown until late exponential phase (OD600 of 6.25–6.8) and then immediately transferred in an anaerobic tent (N2:CO2 atmosphere at a ratio of 90:10). Cells were collected by anaerobic centrifugation for 30 min at 6,000 × g at 4 °C. The highest OD600nm recorded for M. thermolithotrophicus in a SO42−-grown fermenter was 6.8 after 20 h. \({{\rm{SO}}}_{4}^{2-}\) Culture (7 l) with an OD600 of 6.8 yielded 54 g of cells (wet weight). The cell pellet was transferred in a sealed bottle, gassed with 0.3 × 105 Pa N2, flash frozen in liquid N2 and stored at −80 °C.

Synthetic gene constructs

The DNA sequences of the ATP sulfurylase, the APS kinase, the PAP phosphatase and the PAPS reductase α and β subunits from M. thermolithotrophicus were codon optimized for E. coli, synthesized and cloned into pET-28a(+) vectors. For MtATPS, MtAPSK and MtPAPP, the restriction sites NdeI and BamHI were used, with a stop codon (TGA) incorporated before BamHI. For MtPAPSR, a His-tag was placed at the C terminus of the α subunit and a ribosome binding site was inserted between the coding sequences of the α and β subunits. The MtPAPSR construct had the restriction sites NcoI and BamHI, with one stop codon incorporated after the His-tag for the α subunit and one stop codon before BamHI for the β subunit. These steps were performed by GenScript (GenScript). All sequences used are detailed in Supplementary Information under Constructs and gene codon optimization.

Enzyme overexpression and purification

All constructs were overexpressed and purified under aerobic conditions following a similar protocol, except for MtPAPSR which was overexpressed and purified under an anaerobic atmosphere. All enzymes were passed on a HisTrap high-performance column (GE Healthcare), followed, if necessary, by tag cleavage and gel filtration (see Supplementary Materials for the complete protocol).

Protein crystallization

Purified MtATPS, MtAPSK and MtPAPP were kept in 25 mM Tris/HCl pH 7.6, 10% v/v glycerol, 2 mM dithiothreitol and 150 mM NaCl. MtPAPSR was kept in the same buffer without NaCl. Freshly prepared unfrozen samples were immediately used for crystallization. MtATPS, MtAPSK and MtPAPP crystals were obtained under aerobic conditions at 18 °C. MtPAPSR crystals were obtained anaerobically (N2:H2, gas ratio of 97:3) by initial screening at 20 °C. The sitting drop method was performed on 96-well MRC 2-drop crystallization plates in polystyrene (SWISSCI) containing 90 µl of crystallization solution in the reservoir.

Crystallization of MtATPS

MtATPS (0.7 µl) at a concentration of 14 mg ml−1 (MtATPS form 1, Extended Data Table 1) or at a concentration of 27 mg ml−1 (MtATPS form 2) was mixed with 0.7 µl reservoir solution. MtATPS at 27 mg ml−1 was co-crystallized with 2 mM AMPcPP as well as 2 mM Na2SO4. For MtATPS form 1, transparent star-shaped crystals appeared after a few weeks in the following crystallization condition: 35% w/v pentaerythritol ethoxylate (15/4 EO/OH) and 100 mM 2-(N-morpholino)ethanesulfonic acid (MES) pH 6.5. For MtATPS form 2, transparent, long but thin plate-shaped crystals appeared after a few weeks in the following crystallization condition: 20% w/v polyethylene glycol 8000, 100 mM MES pH 6.0 and 200 mM calcium acetate.

Crystallization of MtAPSK

MtAPSK (0.7 µl) at a concentration of 17.6 mg ml−1 was mixed with 0.7 µl reservoir solution and co-crystallized with 2 mM MgCl2. Transparent, plate-shaped crystals appeared after a few weeks in the following crystallization condition: 20% w/v polyethylene glycol 3350 and 100 mM tri-sodium citrate pH 5.5. MtAPSK was also crystallized with 2 mM MgCl2 and 2 mM APS but the obtained structures of those crystals were of lower resolution and without any substrate or product present in the active site.

Crystallization of MtPAPSR

MtPAPSR (0.7 µl) at a concentration of 20 mg ml−1 was mixed with 0.7 µl reservoir solution and co-crystallized with FAD (0.5 mM final concentration). The crystal used for phasing was a brown flat square and appeared after a few days in the following crystallization condition: 40% v/v 2-methyl-2,4-pentanediol and 100 mM Tris/HCl pH 8.0.

The crystal used to refine at high resolution was brown with an elongated plate shape. It appeared after a few days in the following crystallization condition: 35% v/v 2-methyl-2,4-pentanediol, 100 mM Tris pH 7.0 and 200 mM NaCl. Before transfer to liquid N2, the crystal was soaked in 10 mM disodium 3’-phosphoadenosine 5’-phosphate for 7 min.

Crystallization of MtPAPP

MtPAPP (0.7 µl) at a concentration of 20 mg ml−1 was mixed with 0.7 µl reservoir solution and co-crystallized with Tb-Xo4 (10 mM final concentration), MnCl2 (2 mM final concentration) and 2 mM PAP. The Tb-Xo4 is a nucleating/phasing agent50, which should increase the crystallization performance; however, in this case, the same crystalline form was obtained in the absence of the compound and diffracted to similar resolution. Transparent, bipyramid crystals appeared after a few weeks in the following crystallization condition: 1.6 M tri-sodium citrate.

X-ray crystallography and structural analysis

MtPAPSR crystal handling was done inside the Coy tent under anaerobic atmosphere (N2:H2, 97:3); the other crystals were handled under aerobic conditions. The crystals were directly plunged in liquid nitrogen or were soaked for 5–30 s in their crystallization solution supplemented with a cryoprotectant before being frozen in liquid nitrogen. For MtATPS form 2, 30% glycerol was used as cryoprotectant. For MtAPSK, 25% ethylene glycol was used as cryoprotectant.

Crystals were tested and collected at 100 K at different synchrotrons (Extended Data Table 1). Data were processed with autoPROC51 except for MtPAPP, which gave better statistics with indexation by the X-ray Detector Software (XDS) and the scaling step performed with SCALA52. All data collection statistics are provided in Extended Data Table 1. MtATPS forms 1 and 2, MtAPSK and MtPAPP were solved by using PHENIX with the following templates: 1V47 (ATPS from T. thermophilus) for MtATPS form 1, MtATPS form 1 for MtATPS form 2 and 5CB6 (APS kinase from Synechocystis sp.) for MtAPSK. For MtPAPP, the template was created de novo using AlphaFold 2 (ref. 32).

For MtPAPSR, an X-ray fluorescence spectrum on the Fe K-edge was measured to optimize the data collection at the appropriate wavelength. Datasets were collected at 1.73646 Å for the single-wavelength anomalous dispersion experiment. Native datasets were collected at a wavelength of 0.97625 Å on another crystal. Data were processed and scaled with autoPROC51. Phasing, density modification and automatic building were performed with CRANK-2 (ref. 53).

All models were manually rebuilt with COOT and further refined with PHENIX54,55. During the refinement, non-crystallographic symmetry and translational-libration screw were applied. For all structures except for ATPS form 1, hydrogens were added in riding position in the last refinement cycle. Hydrogens were removed in the final deposited models.

All models were validated using MolProbity56. Data collection and refinement statistics, as well as PDB identification codes for the deposited models and structure factors are listed in Extended Data Table 1. Figures were generated with PyMOL (Schrödinger). The metal in MtATPS was modelled as zinc using CheckMyMetal57.

High-resolution clear native PAGE (hrCN PAGE)

To visualize the expression levels of MtFsr when cells were grown on different sulfur sources, hrCN PAGE was performed. M. thermolithotrophicus cultures (2 × 10 ml) were supplemented with either 2 mM Na2S, 2 mM Na2SO3, 2 mM Na2S and 2 mM Na2SO4, or 2 mM Na2SO4 as sulfur substrates and grown for one night at 65 °C, standing. Cells were collected by anaerobic centrifugation at 6,000 × g for 20 min at room temperature and the cell pellets were resuspended in 2 ml lysis buffer (50 mM tricine pH 8.0 and 2 mM sodium dithionite). The cells were sonicated 4× at 70% intensity for 10 s, followed by a 30 s break (MS 73 probe, SONOPULS Bandelin). The hrCN PAGE was run anaerobically and the protocol is detailed in Extended Data under hrCN PAGE preparation. One gel with an 8–15% acrylamide gradient was run (shown in Fig. 5b) and another one with a 5–15% acrylamide gradient (see Source Data Fig. 5).

Coupled enzyme activity of MtATPS/MtAPSK

The activity of both enzymes was determined by the production of ADP which was coupled to NADH oxidation via pyruvate kinase and lactate dehydrogenase58. The assays were performed in a final volume of 100 µl 96-well deep-well plates and spectrophotometrically monitored (Omega multimode microplate reader) at 360 nm at 35 °C. KH2PO4 (100 mM) at pH 7.0, supplemented with 1.5 mM MgCl2 and 100 mM KCl, was used as a buffer. For NADH, a molar extinction coefficient of 4,546.7 cm−1 M−1 was experimentally determined for the above-named conditions. To the buffer, 1 mM NADH, 2.5 mM Na2SO4, 1 mM phosphoenolpyruvate (PEP), 2 mM ATP, 2 U inorganic pyrophosphatase (Saccharomyces cerevisiae, 10108987001, Sigma-Aldrich), 1.1 U ml−1 lactate dehydrogenase, 0.8 U ml−1 pyruvate kinase (rabbit muscle, P0294, Sigma-Aldrich) and 0.5 mg ml−1 MtAPSK (all final concentrations) were added. The reaction was started by the addition of 0.5 mg ml−1 MtATPS. Addition of 0.02 mM Na2MoO4 did not affect activity (0.116 ± 0.027 µmol of oxidized NADH min−1 mg−1), but the addition of 2 mM Na2MoO4 resulted in a decrease (0.068 ± 0.019 µmol of oxidized NADH min−1 mg−1). All assays were performed in triplicates.

MtPAPP enzyme assay

The activity of the MtPAPP was determined by the production of orthophosphate, which was quantified using the malachite green phosphate assay kit (Sigma-Aldrich) by the formation of a green complex. The assays were performed in 96-well deep-well plates and the absorbance at 620 nm was spectrophotometrically followed (Omega multimode microplate reader). Tris/HCl (25 mM) at pH 7.64 was used as a buffer. Buffer, 40 µM PAP or 90 µM of AMP/ADP/ATP/APS or PPi, 1 mg ml−1 bovine serum albumin, 50 µM MnCl2 and/or 50 µM MgCl2 (final concentration) were mixed in a 1.5 ml Eppendorf tube on ice. Previously frozen MtPAPP (0.5 µg ml−1 final concentration) was added and the mixture (final volume of 40 µl) was immediately incubated for 5 min at 40 °C. Next, 14 µl of the reaction mix was diluted in 66 µl of filtered Milli-Q H2O and immediately flash frozen in liquid N2 to quench the reaction. Then, 20 µl of malachite green reagent was added to the samples, the mixture was incubated at room temperature for 30 min and the formation of the green complex was measured at 620 nm. All assays were performed in triplicates. The measurements presented in Fig. 3a come from two different experiments (left and right subpanels). Both experiments were performed at two different days with the same enzyme preparation.

Coupled MtPAPSR assay

Since PAPS is unstable at high temperatures, we first tried to determine the activity of MtPAPSR in the direction of PAPS production, as previously described for dissimilatory APS reductases for APS production38. PAPS oxidation was determined in 50 mM Tris/HCl buffer (pH 7.5) containing 5 mM Na2SO3, 2 mM PAP or 2 mM AMP (final concentrations) and 3.27 µg ml−1 MtPAPSR. The reaction was started with a final concentration of 0.5 mM K3Fe(CN)6. The decrease in absorbance at 420 nm was measured and corrected for the background reaction without enzyme. No activity was detected. Therefore, we used the physiological reaction to monitor MtPAPSR activity. To perform the coupled MtPAPSR assay, the enzymes needed to be purified at the same time and immediately used for the assay (see Supplementary Materials for the detailed purification protocol for the enzymes used in this assay).

MtPAPSR activity assays were carried out in an anaerobic atmosphere (100% N2) at 45 °C. The assays were performed in 200 µl final volume in 96-well deep-well plates and spectrophotometrically monitored on a SPECTROstar Nano microplate reader. HEPES (50 mM, pH 7.0) supplemented with 50 mM KCl, 1.5 mM MnCl2 and 1.5 mM MgCl2 was used as a buffer. Reduced methyl viologen (MVred, 0.5 mM) served as an electron donor for MtPAPSR. The molar extinction coefficient (ε600nm = 8,133.3 cm−1 M−1) was experimentally determined using the above-named conditions and by reducing methyl viologen with 2 mM sodium dithionite. For the assay, methyl viologen was reduced with carbon monoxide by the CO-dehydrogenase from Clostridium autoethanogenum according to a previously published protocol59. CO was exchanged for N2 and the MVred was immediately used for the assay. To the buffer and MVred, 5 mM ATP, 1 mM sodium dithionite, 0.2 U pyrophosphatase (E. coli, MFCD00131379, Sigma-Aldrich), 0.127 mg ml−1 MtATPS, 0.12 mg ml−1 MtAPSK, 0.1 mg ml−1 MtPAPP and 0.0645 mg ml−1 MtPAPSR were added. The reaction was started by the addition of 5 mM Na2SO4 and followed by oxidation of MVred at 600 nm. All assays were performed in triplicates.

Sulfite reductase activity in cell extracts

To determine the sulfite reductase activity from M. thermolithotrophicus, cultures were grown on either 2 mM Na2S, 2 mM Na2SO3 or Na2SO4 in 10 ml of the above-mentioned medium in serum flasks. Cells (9 ml) were collected in late exponential phase (OD600: 3.45 for 2 mM Na2S, 3.91 for 2 mM Na2SO3, 3.37 for Na2SO4) by centrifugation at 6,000 × g for 10 min at 4 °C. The supernatant was discarded and the cell pellets were frozen in liquid N2. The pellets were then resuspended in 1 ml 0.5 M KH2PO4 pH 7.0. The cells were lysed by sonication (2× 10 s at 50% intensity, probe MS73, SONOPULS Bandelin), followed by centrifugation at 4 °C at 15,600 × g. The supernatant was passed through a 0.2 µm filter and the protein concentration was determined by the Bradford method (6.63 mg ml−1 for 2 mM Na2S, 6.14 mg ml−1 for 2 mM Na2SO3 and 6.31 mg ml−1 for Na2SO4). The activity assays were performed under an anaerobic atmosphere (100% N2) at 50 °C in 96-well deep-well plates and spectrophotometrically monitored (SPECTROstar Nano microplate reader). The assay mixture contained 0.5 M KH2PO4 pH 7.0, 118 µM MVred (final concentration, previously reduced with the equimolar amount of sodium dithionite) and 30 µM Na2SO3 (final concentration). Under these conditions, a molar extinction coefficient of ε600nm = 9,840 cm−1 M−1 was experimentally determined. The reaction was started by the addition of 0.05 µg of cell extract, followed by oxidation of MVred at 600 nm. All assays were performed in triplicates.

Phylogenetic trees

For a detailed description of the phylogenetic analysis, see Supplementary Materials60.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.