Recent studies have shown that methanogens can reductively dissolve FeS₂ (Payne et al., 2021a,b). Given that (1) methanogens can produce H₂ during methanogenesis (Lupa et al., 2008; Kulkarni et al., 2018) combined with (2) prior studies that have demonstrated abiotic FeS₂ reduction by H₂, albeit at high temperature (>90°C) and high H₂ partial pressure (>8 bar, equivalent to >7 mM aqueous H₂; Truche et al., 2010), it was necessary to first evaluate the potential for H₂ to abiotically reduce FeS₂ at lower temperature. High temperature (>90°C) abiotic reduction of FeS₂ occurs according to Eq. 1:

where the value of x can range from 0 to 0.17 (Hall, 1986; Rickard and Luther, 2007; Truche et al., 2010). The thermodynamics of Eq. 1 were evaluated and abiotic reduction experiments with H₂ were conducted using laboratory-synthesized nanoparticulate FeS₂ to determine if abiotic FeS₂ reduction can occur at temperatures lower than 90°C, to identify the potential products of FeS₂ dissolution, to quantify the threshold concentration of H₂ required to generate detectable products of this reaction, and to determine the sensitivity of the reaction to product (i.e., HS^–) accumulation.

Thermodynamic calculations were performed at pH 7.0 and an ionic strength of 0.05 M, a temperature of 38°C, and a free energy of formation for FeS₂ of –160.2 kJ mol^–1 (Chareev et al., 2014). These conditions were chosen because they are similar to those used or measured in methanogen cultures demonstrated to reduce FeS₂ (Payne et al., 2021b). At 38°C, reduction of FeS₂ by H₂ was shown to be favorable across a range of aqueous H₂ concentrations as low as 10^–10 M and HS^– concentrations (10^–7 to 10^–2 M), if one assumes that Fe_1–xS [Fe_0.86S; free energy of formation of –136 kJ mol^–1 (Chareev et al., 2014)] is the end product of the reaction (Truche et al., 2010) and using the Davies equation (Stumm and Morgan, 1996) to account for the influence of ionic strength on soluble ion activities (Figure 1). Interestingly, the reaction was not thermodynamically favorable if the end product, Fe_1–xS, was replaced with the iron monosulfide phase mackinawite [FeS; free energy of formation of –89.2 kJ mol^–1 (Berner, 1967)] as a product (data not shown). This observation is consistent with the detection of Fe_1–xS as the primary product of abiotic FeS₂ reduction, albeit at high temperature [>90°C; (Truche et al., 2010)]. These calculations confirm that abiotic reduction of FeS₂ to HS^– and Fe_1–xS, but not FeS, by H₂ is thermodynamically feasible under the conditions of prior experiments (Payne et al., 2021a,b).

Abiotic experiments were conducted using synthetic FeS₂ (2 mM in FeS₂ formula unit) nanoparticles in reactors incubated at 38°C in low salinity, carbonate-buffered medium (2 g L^–1 NaHCO₃) at pH 7.0, similar to the conditions under which biological FeS₂ reduction has been demonstrated (Payne et al., 2021a,b) and for which the thermodynamic parameters above were calculated. In reactors with a 100% H₂ headspace (1.98 × 10^–3 M aqueous H₂), rapid abiotic reduction of FeS₂ was observed, as determined by the production of total sulfide (sum of aqueous HS^– and gaseous H₂S) according to Eq. 1 (Figure 2). Significantly less, albeit still quantifiable, total sulfide was generated in abiotic reactors with a headspace comprising 10% H₂ (1.98 × 10^–4 M aqueous). Reactors with headspace H₂ of 1% (1.98 × 10^–5 M aqueous) or lower did not yield detectable total sulfide (minimum detection limit = 1.5 μM aqueous concentration) during the 4-day incubation. Nonetheless, this indicates that synthetic FeS₂ can be reduced at temperatures as low as 38°C, but the products of this reaction are only detectable by colorimetric methods above a threshold H₂ concentration that exists between 1.98 × 10^–5 and 1.98 × 10^–4 M. To confirm that abiotic reduction by H₂ is not limited to synthetic nanoparticulate FeS₂, 1.5 g of ground (63–125 μm size fraction) specimen-grade, high-purity FeS₂ was incubated in 75 mL of base salts medium under 100% H₂ (1.98 × 10^–3 M aqueous concentration) or 100% N₂. After 5 days of incubation at 38°C, significant total sulfide (7.19 ± 0.01 μmol) was generated through abiotic reduction by H₂ and no sulfide was detected in N₂ control reactors (data not shown). The concentration of dissolved H₂ required to reduce FeS₂ in abiotic experiments was far higher than the minimum (10^–10 M) estimated by thermodynamic calculations. This suggests that factors other than the free energy for the reduction alone [e.g., particle interfacial energy (Stumm, 1992)] likely play a role in controlling the favorability of the reduction reaction.

In line with previous high-temperature studies of FeS₂ reduction (Truche et al., 2010), thermodynamic calculations predicted that the favored product of FeS₂ reduction is Fe_1–xS. To examine the potential solid phase(s) of Fe(II) and S formed during low temperature FeS₂ reduction in the presence of H₂, X-ray diffraction (XRD) analyses of FeS₂ reacted abiotically under 100% H₂ headspace for 24 h at 38°C were conducted. The XRD spectra for the H₂-reacted FeS₂ was nearly identical to that for the unreacted FeS₂ (Supplementary Figures 1A,B). Reference peaks for Fe_1–xS (PDF #97-015-1765) and FeS_mack (PDF #97-063-3302) were manually searched against the XRD spectra, for both reacted and unreacted FeS₂, without matches. It is likely that the abundance of Fe_1–xS, if produced during abiotic H₂-driven FeS₂ reduction, remains below the detection limit of XRD (∼10% relative abundance by weight). Additional characterization of FeS₂ surfaces following reduction by H₂ using higher resolution spectroscopic methods, such as X-ray absorption spectroscopy, are needed to identify the low abundance secondary mineral(s) that may form on the FeS₂ surface.

While XRD analyses did not identify Fe_1–xS produced during abiotic reduction of FeS₂ reduction by H₂, it was still the favored secondary mineral to form on the surface of FeS₂ based on thermodynamic calculations and observations from studies conducted at higher temperature (Truche et al., 2010). Thus, the solubility of Fe_1–xS was examined to identify its plausibility as a source of soluble Fe and S. Specimen Fe_1–xS obtained from Aymar quarry, Gualba mines, Gualba, Montseny, Barcelona Spain, which has previously been shown to be of high purity (de Aldecoa et al., 2013), was used in experiments. XRD of the mineral indicated that the only FeS phase present was Fe_1–xS (∼80%) with the balance as quartz (Supplementary Figure 1D). Dissolved (<0.2 μm filtered) Fe(II) was detected in abiotic incubations of 0.1 g of ground (63–125 μm) Fe_1–xS following 7 days incubation (Figure 3). While this amount of Fe_1–xS is greater than what one would expect to form in cultures of M. barkeri actively reducing FeS₂, it does demonstrate that Fe(II) can be solubilized from Fe_1–xS and is a possible source of Fe used by cells reducing FeS₂. Interestingly, HS^– was not detected (detection limit of 1.5 μM) in solution following this incubation period. This observation is consistent with a previous study that indicated that Fe [as Fe(II)], but not S (as HS^–), in Fe_1–xS is mobile and can solubilized from the mineral surface leaving behind a metal-deplete surface layer (Mikhlin, 2000).

Importantly, during FeS₂ reduction, the total number of moles of HS^– formed should be equivalent to the total moles of Fe_1–xS formed per Eq. 1. Since the dissolution of Fe from Fe_1–xS is incomplete, the yield of Fe(II) in solution from Fe_1–xS dissolution is far lower than the yield of total sulfide (>1 μM) during FeS₂ reduction. Based on thermodynamic data (Luther and Rickard, 2005; Rickard and Luther, 2007), aqueous solutions with HS^– in excess of Fe(II) favors hydrated FeS_aq clusters as the predominant form of Fe(II) in solution. Experiments have shown that the methanogen M. voltae, when grown with FeS₂ as the sole source of Fe and S, simultaneously exhibited evidence indicative of Fe limitation [i.e., up-expression of the Fe(II) transporter FeoB and the metal regulator DtxR] but at the same time hyperaccumulated Fe as a thioferrate-like mineral (Payne et al., 2021a). To explain this paradox, it was suggested that cells incorrectly sensed Fe(II) limitation during growth due to assimilation of Fe(II) complexed with sulfide (i.e., FeS_aq). In this model, excess Fe(II) that was assimilated as hydrated FeS_aq clusters (cells require more S than Fe) was then sequestered as thioferrate-like nanoparticles to limit toxicity (Payne et al., 2021a). FeS_aq is predicted to be uncharged at the circumneutral pH of the culture medium used to cultivate M. voltae and M. barkeri (Jordan et al., 2019a,b). Thus, it is plausible that slightly hydrophobic and uncharged FeS_aq can either passively diffuse or be actively transported across cellular membranes to provide Fe and S to cells. Additional growth experiments with Fe_1–xS are described below.

The demonstration of abiotic reduction of synthetic and specimen FeS₂ by H₂, even at low temperature and low H₂ concentrations, led to an investigation into the potential role of biogenic H₂ in FeS₂ reduction by M. barkeri strains Fusaro and MS. M. barkeri Fusaro was capable of reducing and growing on FeS₂ as the sole Fe and S source, as indicated by the production of CH₄, DNA, and total sulfide (Figures 4A,C,D). Cultures of FeS₂-grown M. barkeri Fusaro generated H₂ at concentrations that were similar to cultures grown with Fe(II) and HS^– (Figure 4B). M. barkeri strain MS produced slightly more H₂ during growth on FeS₂ compared to growth on Fe(II) and cysteine, consistent with increased growth as indicated by higher DNA and CH₄ production (Supplementary Figure 3B). Regardless, the maximum amount of H₂ generated by M. barkeri strains Fusaro and MS grown with FeS₂ (2.65 and 2.79 μM aqueous, respectively) was nearly two orders of magnitude below the lowest concentration of H₂ (198 μM) that was experimentally found to drive abiotic FeS₂ reduction (Figure 2). While this points to biogenic H₂ not acting as the mediator of FeS₂ reduction, it cannot be ruled out that the interface between FeS₂ minerals and actively growing M. barkeri cells could sustain a locally elevated concentration of H₂ that facilitates indirect, abiotic reduction of the mineral. This is particularly true considering that thermodynamic calculations indicate that far lower concentrations of H₂ (10^–10 M aqueous) may be able to drive abiotic FeS₂ reduction with Fe_1–xS as the end-product of the reaction (Eq. 1).

To further test whether FeS₂ reduction is mediated by biogenic H₂, a strain of M. barkeri Fusaro with mutations in all [NiFe]-hydrogenase operons encoded in its genome, rendering the strain unable to consume or produce H₂ (Mand et al., 2018), was evaluated for its ability to reduce FeS₂. The strain has mutations in one membrane-associated energy converting [NiFe]-hydrogenase (Ech), two F₄₂₀-reducing [NiFe]-hydrogenases (Frh and Fre), and two membrane-associated methanophenazine-reducing [NiFe]-hydrogenases (Vht and Vhx; Kulkarni et al., 2018). The M. barkeri Fusaro [NiFe]-hydrogenase mutant maintained the ability to reduce FeS₂ as indicated by significant accumulation of HS^– in the medium during growth with methanol as the methanogenesis substrate and with FeS₂ as the sole Fe and S source (Figure 5B). Growth was not detected when a source of Fe or S was not provided (Figures 5A,C). Further, growth of the M. barkeri Fusaro [NiFe]-hydrogenase mutant was slightly, albeit significantly (p < 0.05), enhanced on FeS₂ relative to non-mineral sources of Fe and S [Fe(II) and HS^–] (Figure 5C).

Throughout the growth experiment with the M. barkeri Fusaro [NiFe]-hydrogenase mutant, H₂ remained below the limit of detection (0.1 μM aqueous). This indicates the [NiFe]-hydrogenase mutant strain of M. barkeri Fusaro did not produce H₂, yet it maintained the ability to reduce FeS₂ and assimilate dissolution products to meet Fe and S biosynthetic demands. It follows that indirect reduction of FeS₂ through biogenic H₂ is not the mechanism of biological FeS₂ reduction. Further, these results indicate that [NiFe]-hydrogenases (Ech, Frh, Fre, Vht, and/or Vhx) themselves are not involved in FeS₂ reduction, ruling out H₂ or [NiFe]-hydrogenase mediated EET as a mechanism to drive FeS₂ reduction. This is potentially consistent with previous results suggesting that H₂ and [NiFe]-hydrogenases do not play a role in EET from the cells to external electron acceptors in M. barkeri strain Fusaro (Rowe et al., 2019). However, the directionality of EET from M. barkeri cells to FeS₂ differs from that of the previous study which focused on electron acquisition from cathodes to M. barkeri Fusaro cells via EET (Rowe et al., 2019). Further, while electron acquisition from cathodes via EET was apparently respiratory in nature (Rowe et al., 2019), the reduction of FeS₂ as described herein fulfills the purpose of generating bioavailable forms of Fe and S to meet nutritional demands. It is thus possible that the mechanisms of EET for M. barkeri in the growth conditions described herein, versus those described previously, differ.

Prior work showed that M. voltae A3 could not grow or reduce synthetic FeS₂ when physical access to the mineral was restricted using dialysis tubing (Payne et al., 2021b). However, it remained unclear whether cells required direct access to the mineral to (1) carryout reduction and/or to (2) acquire Fe and S from the Fe_1–xS that is predicted to precipitate on the surface of FeS₂ following reduction. Field emission microscopy (FEM) was used to visualize physical associations between M. barkeri Fusaro cells and minerals during growth with FeS₂ as the sole source of Fe and S. FEM shows that M. barkeri cells directly associated with FeS₂ surfaces during growth (Supplementary Figure 5). M. barkeri is known to associate with other solid phases, such as cathodes, via extracellular polymeric substances (EPS; Rowe et al., 2019), thus a similar mechanism may be employed to attach to FeS₂.

To examine the requirement for M. barkeri Fusaro cells to directly contact FeS₂ in order to grow, synthetic FeS₂ was sequestered in dialysis tubing with 100 kDa diameter pore size to prohibit physical contact of the cells with the mineral surface. This also excluded extracellular organic or inorganic molecular complexes with diameters larger than 100 kDa from interacting with the mineral surface. Under a subset of conditions, the synthetic quinone, AQDS, previously shown to act as an electron shuttle between M. barkeri and Fe-oxide minerals (Bond and Lovley, 2002), was provided to cells grown with FeS₂ in solution and sequestered in dialysis tubing. This allowed for the determination of whether (1) direct electron transfer from the methanogen cell surface was required to reduce the mineral, (2) AQDS could enable FeS₂ reduction by acting as an electron shuttle between the cell and FeS₂, and/or (3) the products of FeS₂ reduction could pass through the dialysis membrane to support growth. Cells that were provided with direct access to FeS₂ rapidly reduced the mineral as indicated by total sulfide accumulation up to 3.53 μmol (30.5 μM aqueous sulfide) before the concentration slowly started to decrease, presumably due to its ultimate utilization by the cells (Figure 6B). Concurrently, cells generated CH₄ and biomass, as determined by DNA production (Figures 6A,C) and protein production (Supplementary Figure 3). When FeS₂ was sequestered in dialysis membranes and no AQDS was provided, M. barkeri Fusaro could not reduce FeS₂ as indicated by the lack of sulfide accumulation in the growth medium (Figure 6B). Furthermore, production of CH₄ and biomass (DNA and protein) were 1–2 orders of magnitude less when access to FeS₂ was restricted by sequestering FeS₂ in dialysis tubing when compared to when FeS₂ was not sequestered and were not significantly different from controls where no Fe or S were provided (Figure 6A).

Methanogens can transfer electrons to AQDS, generating the reduced form anthrahydroquinone-2,6-disulfonate (AH₂QDS). The conversion of AQDS to AH₂QDS was monitored spectrophotometrically by observing a shift in absorbance from 325 to 450 nm (Bond and Lovley, 2002) in cultures provided with direct access to FeS₂ or when physical contact between cells and FeS₂ was prevented using dialysis tubing. M. barkeri rapidly reduced AQDS to AH₂QDS in cultures with FeS₂ free in solution (Figure 6D), and this corresponded to accelerated HS^– production from FeS₂ reduction and enhanced cell growth and CH₄ production activity (Figures 6A–C). Furthermore, the concentration of HS^– remained steady during growth and even increased slightly toward the end of the methanogen growth phase, suggesting that cells continued to transfer electrons to AQDS, which in turn continued to reduce FeS₂ even after cell growth had ceased. When FeS₂ was sequestered in dialysis tubing, the conversion of AQDS to AH₂QDS by M. barkeri lagged by approximately 2–4 days. However, following this lag phase, AQDS was reduced, and production of sulfide and cells occurred (Figures 6B,C). Despite the lag in activity and growth, the addition of AQDS as a soluble electron shuttle enabled comparable CH₄ and biomass (DNA and protein) production in cultures where physical contact between cells and FeS₂ was limited to those provided with FeS₂ free in solution. This indicated that M. barkeri requires physical contact with FeS₂ to directly transfer electrons to the mineral surface invoking EET as the primary mechanism for FeS₂ reduction. However, soluble electron shuttles such as AQDS (a humic acid analog) can allow for mineral reduction when direct contact between cells and minerals is restricted. Further, this indicates that M. barkeri does not produce endogenous soluble electron shuttles capable of FeS₂ reduction but may take advantage of exogenous shuttles (e.g., humic acids, quinones) in natural systems if they are available.

Thermodynamics experiments described above predicted that the byproducts of FeS₂ reduction are HS^– and a secondary mineral, Fe_1–xS, which is consistent with previous observations of abiotic FeS₂ reduction products at high temperature (Truche et al., 2010). To evaluate if M. barkeri Fusaro can use Fe_1–xS as a source of Fe and/or S for growth and to determine if direct contact with this mineral is required for its acquisition and assimilation, dialysis experiments were conducted using ground (63–125 μm) specimen grade Fe_1–xS particles as the sole Fe source. Ground Fe_1–xS particles were sequestered in 100 kDa dialysis tubing and reactors were then amended with either 0 or 500 μM HS^– as a S source, since abiotic dissolution of Fe_1–xS was shown to release Fe(II) but not HS^–, as discussed above. When 0 μM HS^– was provided, no CH₄ production or growth was detected in any condition (Figure 7). However, when cultures were provided with 500 μM HS^–, activity and growth with sequestered Fe_1–xS was comparable to growth with 20 μM Fe(II) and 500 μM HS^– (Figure 7). Cultures provided with no Fe and 500 μM HS^– demonstrated activity and growth, albeit CH₄ production was 53–59% lower and biomass 40–41% lower than in cultures grown with 20 μM Fe(II) or provided with Fe_1–xS sequestered in dialysis tubing, respectively. This apparent growth under Fe-limiting conditions is comparable to what has been reported previously for methanogens and is attributed to trace Fe contamination in reagents, in particular Na₂S, despite its American Chemical Society (ACS) grade (Payne et al., 2021b). Collectively, these data indicate that cells require direct access to the surface of FeS₂ to reduce the mineral but not to acquire dissolution products, as the solubility of Fe_1–xS with respect to Fe(II) can support growth of M. barkeri. Further, these observations indicate that the source of S for growth is HS^– from the initial FeS₂ reduction step, not Fe_1–xS. As stated above, the predominant form of Fe(II) solubilized from Fe_1–xS in sulfidic solutions would be as FeS_aq, which is readily produced in aqueous solutions containing excess HS^– relative to Fe(II), making this the likely form of Fe and S that is assimilated.

Because FeS₂ is insoluble in water (equilibrium solubility at 25°C, 0.05 M ionic strength, and pH 7 = ca. 0.1 μM), as summarized by Rickard and Luther (2007), the reduction of persulfides in FeS₂ by methanogens must occur extracellularly through EET mediated by large (>100 kDa) redox-active molecules or direct electron transfer (e.g., involving electronic conduits within pili, membrane-bound oxidoreductases, secreted molecules, etc.). Genes encoding proteins involved in reducing the persulfide in FeS₂ through EET might be expected to be upregulated during growth on FeS₂ as the sole source of Fe and S relative to growth soluble Fe and S sources. Shotgun transcriptomics data collected from M. barkeri strain MS were examined to identify protein-encoding genes whose expression was differentially regulated during growth on FeS₂ (Supplementary Figures 4A–D). Strain MS was selected as it can use cysteine as its sole sulfur source in addition to HS^–. This provides a suitable control for examining possible deleterious effects of minimal HS^– generated from FeS₂ reduction on growth relative to growth in the presence of 2 mM Na₂S [Fe(II)/HS^–], an important consideration given reports of HS^– toxicity to cells (Edgcomb et al., 2004). M. barkeri strain MS encodes the same complement of [NiFe]-hydrogenases as the wild-type M. barkeri strain Fusaro (Mand et al., 2018; Mand and Metcalf, 2019). Consistent with the results described above indicating that H₂ and [NiFe]-hydrogenases are not involved in FeS₂ reduction, none of the five [NiFe]-hydrogenases were differentially expressed during growth on FeS₂ compared to growth on Fe(II)/Cys or Fe(II)/HS^– (Supplementary Figure 4E).

Of the 3,413 genes covered in the differential transcriptomics experiment, transcripts of only 30 were significantly upregulated [p value < 0.05, log₂ fold change (LFC) > 0.5] and met abundance thresholds (>0.005% of total normalized read count) when cells were grown on FeS₂ relative to those grown with either Fe(II)/cysteine or Fe(II)/HS^– (Figure 8A and Supplementary Table 1). This collection of genes included those that code for 12 hypothetical proteins and seven proteins annotated as being involved in translation, ribosomal processing, and other growth-related functions. Nine of the identified FeS₂-upregulated genes coded for proteins that are predicted to be membrane-associated. This included genes encoding ABC transporters of organic compounds (MSBRM_0980, MSBRM_0837, and MSBRM_2090) and ferrous iron in its hexaquo Fe(II) form (FeoB; MSBRM_0201; Lau et al., 2016). A gene coding for the transcriptional regulator for FeoB, feoA (MSBRM_0200; Lau et al., 2016), was also identified in this subset of genes upregulated during growth on FeS₂. A model to rationalize the upregulation of transcripts of FeoAB when FeS₂ is provided as the sole Fe source versus Fe(II) was outlined in a recently published paper that focused on M. voltae A3 (Payne et al., 2021a). Briefly, and as stated above, FeS₂-grown M. voltae cells (and possibly M. barkeri cells) were proposed to transport Fe(II) that is complexed with sulfide (FeS_aq; Truche et al., 2010). It was hypothesized that assimilation of Fe(II) as FeS_aq led cells to sense limited availability of free/dissociable Fe(II) inside the cell via Fe(II)-binding transcription factors (i.e., DtxR, FeoA). As such, cells upregulated the hexaquo Fe(II) transporter (FeoB), and to a lesser extent, its transcriptional regulator (FeoA) and global metal regulator (DtxR), to overcome perceived Fe(II) limitation. Evidence indicated that excess transported Fe (cells require more S than Fe) was sequestered intracellularly as thioferrate-like molecules (Payne et al., 2021a). To this end, FeoAB are not hypothesized to be directly involved in FeS₂ reduction, FeS_aq assimilation, or in supplying reducing equivalents for EET reactions involving FeS₂.

A cluster of differentially expressed genes that are co-localized on the M. barkeri MS genome that included several oxidoreductases was identified (Figure 8B). Among the 15 genes that comprise this cluster, six were significantly upregulated (p value < 0.05, LFC > 0.5) during growth on FeS₂. An additional eight genes were moderately upregulated with a LFC below 0.5, and one additional gene was upregulated though its expression was not significantly different from Fe(II)/HS^–- or Fe(II)/cysteine-grown cells. This cluster of genes codes for several enzymes predicted to be cytoplasmic, including one ferredoxin-domain containing oxidoreductase (Fdx; MSBRM_0264), three paralogous flavodoxin domain-containing oxidoreductases (Fld; MSBRM_0267, MSBRM_0270, MSBRM_0273), and three paralogous aldo-keto reductases (AKR; MSBRM_0269, MSBRM_0274 and MSBRM_0280). Two proteins were predicted to be membrane-associated including a putative alpha-beta hydrolase (mABH; MSBRM_0279) and an additional membrane-associated protein of unknown function (MSBRM_0268). Furthermore, a recently published shotgun proteomics study of M. voltae A3 cells grown with 2 mM FeS₂ as the sole Fe and S source compared to cells grown with 26 μM Fe(II) and 2 mM HS^– revealed significant up-expression (p < 0.05, LFC > 1) of two homologs of Fld (Mvol_0029, Mvol_0089) and three homologs of AKR (Mvol_0066, Mvol_0126, and Mvol_0416) in FeS₂-grown cells (Payne et al., 2021a), all of which were predicted to be cytoplasmic. Homologs of mABH and the protein of unknown function (MSBRM_0268) were not upregulated in FeS₂-grown M. voltae cells. The Fld and AKR homologs are not co-localized in the genome of M. voltae A3 although the majority are co-localized in the genome of M. barkeri Fusaro. Yet, the similarities in their regulation and expression in M. barkeri MS and M. voltae A3 and their cytoplasmic cellular location point to the potential role of these putatively redox-active enzymes in FeS₂-dependent growth, as discussed in detail below.

To date, five strains representing four methanogen species, M. voltae A3 (Payne et al., 2021b), M. barkeri MS (Payne et al., 2021b), M. barkeri Fusaro (this work), an unnamed Methanothermobacter strain isolated from a New Zealand hot spring (provided by Matthew Stott, unpublished data), and M. maripaludis S2 (unpublished data) have been shown to reduce FeS₂ to meet biosynthetic demands for Fe and S. These four species represent evolutionarily and ecologically distinct lineages of methanogens, including more deeply diverging members (M. voltae A3, M. maripaludis S2, uncharacterized Methanothermobacter) and more recently diverging members (M. barkeri MS, M. barkeri Fusaro; Bapteste et al., 2005; Brochier-Armanet et al., 2011; Petitjean et al., 2015). This suggests that the ability of methanogens to reduce FeS₂ may be widespread among this group of microorganisms. As such, homologs of the 30 genes upregulated on FeS₂ relative to Fe(II)/cysteine and Fe(II)/HS^– were surveyed in the genomes of four of the five known FeS₂ reducers (a genome is not available for the uncharacterized Methanothermobacter strain).

Homologs for 19 of the 30 upregulated FeS₂ genes were detected in all four FeS₂-reducing methanogens with available genomes (Supplementary Table 1). Five of these 19 genes encoded proteins that were related to ribosomal function, five were related to membrane transport, and two were related to energy or metabolism (Gene IDs provided in Supplementary Table 1). Five of the remaining 19 conserved genes were predicted to encode oxidoreductases (Fdx, Fld, and two AKR) and one of the genes is predicted to encode the hydrolase, mABH (Figure 8). Collectively, these data suggest that one or more proteins encoded in this gene cluster (i.e., Fdx, Fld, AKR, and/or mABH) may be involved in a function associated with FeS₂ reduction as they are conserved across methanogens capable of this functionality.

Several dedicated mechanisms for EET have been proposed to explain the ability of microorganisms to transfer electrons to minerals, cathodes, or other microorganisms (Shi et al., 2016; Jorgensen, 2021; Rotaru et al., 2021). Notably, all the examples of EET to date have been described for their role in cellular respiration. However, EET to FeS₂ by methanogens for mineral reduction is likely related to nutrient acquisition as FeS₂ can serve as the sole source of Fe and S.

Methanogens, including those from the Methanosarcina genus, are capable of Fe(III) oxide mineral reduction as a means for energy conservation (Bond and Lovley, 2002; Sivan et al., 2016; Shang et al., 2020). It has been suggested that the mechanism of Fe(III)-oxide reduction mimics that for bacterial species, whereby Methanosarcina acetivorans cells use transmembrane, multiheme c-type cytochromes (MHC) encoded by mmcA to transfer electrons extracellularly to Fe(III) oxides (Holmes et al., 2019), thereby diminishing the rate of methanogenesis, possibly due to diversion of electrons away from methanogenic pathways. However, neither of the genomes of the M. barkeri strains used in this study (Fusaro or MS) or the genome of M. voltae A3 (Payne et al., 2021b) encode homologs of MHC (data not shown). Further, it has recently been shown that while MHC have been implicated in direct interspecies electron transfer between syntrophic Geobacter and Methanosarcinales, these molecular complexes are not actually required (Yee and Rotaru, 2020). The possibility exists for indirect EET via electron shuttling molecules such as flavins or enzymes such as [NiFe]-hydrogenase and formate dehydrogenase that are excreted from methanogen cells that then drive FeS₂ reduction. Indeed, M. maripaludis has been shown to excrete extracellular enzymes, including both [NiFe]-hydrogenases and formate dehydrogenases, that are capable of driving corrosion of metallic iron (Deutzmann et al., 2015). However, in the present study it was shown that when M. barkeri was grown in medium with FeS₂ sequestered in dialysis tubing with a pore size of 100 kDa, cells were unable to reduce FeS₂, as evinced by the lack of HS^– production and growth (Figure 5). This was also observed for M. voltae grown with formate as methanogenesis substrate (Payne et al., 2021b). Thus, a secreted electron shuttling molecule or enzyme would need to have a hydrodynamic diameter large enough to be blocked by the 100 kDa pore, which excludes involvement of low-molecular-weight endogenously produced electron shuttling molecules like flavins. Growth studies with the [NiFe]-hydrogenase M. barkeri Fusaro mutant rule out involvement of hydrogenase enzymes in FeS₂ reduction. Nevertheless, at this time, enzymes and enzyme complexes with large molecular weights (>100 kDa) cannot necessarily be ruled out as being involved in FeS₂ reduction.

A recent study showed that M. barkeri is capable of assimilating and/or precipitating nanoparticulate magnetite intracellularly, and these conductive nanoparticles enhanced methanogenesis activity presumably by acting as solid electron shuttles across the cell membrane (Fu et al., 2019). Given the production of FeS phases (e.g., Fe_1–xS and/or FeS) during FeS₂ reduction (Eq. 1), it seems plausible that they may participate in EET in methanogens in a manner similar to that of Shewanella (Kondo et al., 2015). Follow-up studies using high-resolution imaging and spectroscopic approaches combined with electrochemical approaches will advance understanding of the likely role of FeS nanoparticles in EET and FeS₂ reduction.

Without a clear mechanism to describe biological FeS₂ reduction emerging from experimental data, a model for spontaneous electron transfer from the cell surface to FeS₂ is considered. In this model, reduction of FeS₂ may occur by diversion of low potential electrons from the membrane, extracellular proteins, or other components of the extracellular milieu. While it is not yet clear what membrane and extracellular components are involved in FeS₂ reduction, experiments that showed both AH₂QDS and H₂ can abiotically reduce FeS₂ allow for an estimate of the reduction potential of electrons involved in mineral reduction.

The calculated non-standard state electrochemical potential (E) of H₂ [E°= –414 mV (Thauer et al., 1977)] at a headspace partial pressure of 0.25 bar (198 μM dissolved H₂) in pH 7.0 medium at 38°C is –318 mV, indicating that electrons under such conditions would need to have a midpoint potential equal to or lower than this to reduce FeS₂. However, thermodynamic calculations conducted herein indicate that FeS₂ reduction is favorable at H₂ concentrations as low as 1 μbar (10^–10 M), which equates to electrons with midpoint potentials of –123 mV. Furthermore, the standard state reduction potential of AH₂QDS has been estimated to be –184 mV (Clark, 1960), which is lower than that for H₂ at a concentration of 10^–10 M, consistent with presented evidence indicating it can facilitate EET to FeS₂ (Figure 6). Numerous membrane-associated oxidoreductases in methanogens are involved in redox reactions involving electrons that are far more reduced than –184 mV (Thauer et al., 1977), suggesting that their diversion via an undefined or even undedicated mechanism toward FeS₂ is feasible. Alternatively, small redox-active molecules produced by methanogens, such as methanophenazine [E = –165 mV (Beifuss and Tietze, 2005)] may be involved in shuttling electrons across the cell membrane to the FeS₂ surface. This would imply, however, that the mechanism involved in biological FeS₂ reduction is not universal across FeS₂-reducing methanogens since M. voltae does not produce methanophenazine (Deppenmeier et al., 1999).

Synthetic nanoparticulate pyrite (FeS₂) was prepared as previously described (Payne et al., 2021b) and XRD analysis is provided in Supplementary Figure 1. ACS grade chemicals were used in mineral syntheses (and preparation of cultivation medium, described below) and glassware was washed with 10% trace-metal grade nitric acid (HNO₃) to remove potential trace metal impurities. Briefly, the initial FeS₂ synthesis reaction was conducted in an anaerobic chamber (2.5% H₂/balance N₂) and was then bubbled with sterile N₂ gas passed over a heated (200°C) and H₂-reduced column containing native copper shavings for 1 h L^–1. Following bubbling, FeS₂ was incubated anoxically (N₂ headspace) in a sealed, serum bottle for 4 days at 65°C followed by an additional 4 days at 85°C to complete the synthesis. Finally, to remove unreacted HS^–, Fe(II), FeS, and S⁰, the synthesized nanoparticulate FeS₂ was washed (via centrifugation and decanting) four times with 1N HCl, once with boiling 6N HCl, twice with MilliQ (MQ) H₂O, three times with >99.5% acetone, and finally three times with sterile, anoxic MQ H₂O inside of an anaerobic chamber. Following the final wash step, the FeS₂ was resuspended in sterile, anoxic MQ H₂O and transferred to a sterile serum bottle that was then capped and sealed prior to removal from the anaerobic chamber. Upon removal from the anaerobic chamber, the headspace of the serum bottle was flushed with 0.22 μm filtered N₂ gas. To determine the concentration of the synthesized FeS₂ slurry, triplicate 1 mL aliquots were transferred to pre-weighed serum bottles, dried anoxically under N₂ gas, and weighed to calculate the weight percent and molar concentration. The concentration of the FeS₂ stock was adjusted to 0.2 M by diluting into sterile, anoxic MQ H₂O.

Specimen grade FeS₂ was obtained from Zacatecas, Mexico (Ward’s Science, Rochester, NY, United States) and specimen grade Fe_1–xS was obtained from Aymar quarry, Gualba mines, Gualba, Montseny, Barcelona Spain. Within a laminar flow hood, specimen minerals were crushed with a MQ-cleaned and ethanol-sterilized jaw crusher (Gilson, Lewis Center, OH, United States) then applied to a cleaned and sterile sieve stack (United States Standard #10/2,000 μm, United States Standard #35/500 μm, United States Standard #60/250 μm, United States Standard #120/125 μm, United States Standard #230/63 μm, and catch pan; all 8” diameter). The 63–125 μm fractions were collected and washed. Specimen FeS₂ was washed as described above for synthetic FeS₂ while specimen Fe_1–xS was washed three times with >99.5% acetone and then three times with sterile, anoxic MQ within an anaerobic chamber. The washed minerals were transferred to sterilized, N₂ purged, sealed serum bottles then dried under a stream of 0.22 μm filtered N₂ gas. The dried minerals were characterized using a SCINTAG X-1 system X-ray powder diffraction (XRD) spectrometer (XRD Eigenmann GmbH, Mannheim, Germany).

Synthetic, nanoparticulate FeS₂ (2 mM) was added to triplicate reactors that contained 35 mL of low salinity basal medium (in g L^–1: NaCl, 1.00; MgCl₂ ⋅ 6H₂O, 0.40; NH₄Cl, 0.50; KCl, 0.50; CaCl₂ ⋅ 2H₂O, 0.10; KH₂PO₄, 0.15_; NaHCO₃, 2.0) in 70 mL glass serum bottles capped with blue butyl rubber stoppers (Bellco, Vineland, NJ, United States). This medium is similar to growth medium for M. barkeri MS described below, but without methanol, acetate, vitamins, or trace metal solutions. The reactors were then purged for 20 min with 0.2 μm filtered N₂ that had been passed over a heated (200°C) and H₂-reduced column containing native copper shavings before pressurizing the reactors by adding H₂ to the specified concentration, replacing the appropriate gas volume, and then pressurizing to 2.5 bar with N₂. In the case of the 100% H₂ condition, the headspace was purged with H₂ for 20 min and was then pressurized to 2.5 bar with H₂. The reactors were incubated at 38°C statically in the dark.

Reactors were sampled for HS^– via the methylene blue assay (Fogo and Popowsky, 1949) and Fe(II) via the ferrozine assay (Stookey, 1970). Samples for HS^– determination were analyzed immediately whereas samples for Fe(II)/(III) determination were first incubated for 16 h in 1N HCl at room temperature (∼20°C) before centrifugation (5,000 × g, 5 min, 4°C) to pellet the mineral. The supernatant was split in half, with one half used to measure the concentration of Fe(II). The other half of the sample was used to quantify Fe(III) by first reducing the sample with 0.2 M hydroxylamine hydrochloride for 30 min and then determining total Fe, from which Fe(II) was subtracted to arrive at Fe(III) concentration. Fe(III) was not detected in any incubation experiment.

To further examine the mineral product that forms during reductive dissolution of FeS₂, 65 μmol of laboratory synthesized nanoparticulate FeS₂ was incubated in 5 mL (6.5 mM final concentration) in MQ H₂O undisturbed at 38°C for 24 h under 2.5 bar H₂ (1.98 × 10^–3 M aqueous). The pelleted mineral was dried anoxically under a stream of N₂ gas (∼10 psi delivered through a 22-gauge needle) for ∼2 h while sitting in a warm water bath on low heat to promote drying. The dried mineral was characterized using XRD, as described above, within 2 h of preparation.

Methanosarcina barkeri strain MS and M. barkeri strain Fusaro were obtained from the American Type Culture Collection (ATCC-43569 and ATCC-BAA-2329, respectively). A mutant strain of M. barkeri strain Fusaro with deletions in four operons encoding for its five [NiFe]-hydrogenases was constructed as part of a prior study and was used herein (Mand et al., 2018). For all strains, growth medium was prepared without added Fe and S in MQ water and using acid (10% HNO₃) washed glassware. Specifically, M. barkeri was grown in low-salinity medium, as described above for abiotic experiments. The base salts solution was boiled for 10 min then purged with N₂ gas passed over the heated copper column containing reduced copper shavings for 1 h L^–1. After sparging, base salts medium was moved to an anaerobic chamber and allowed to cool to room temperature. Once cooled, NaHCO₃ (2.00 g L^–1) was added, and the pH was adjusted to 7.0 with anoxic 2N HCl. Next, 75 mL of the base salts medium were dispensed into 165 mL serum bottles, sealed with blue butyl stoppers, and capped with aluminum crimp caps. Sealed serum bottles were removed from the anaerobic chamber, the headspace was exchanged for 15 min with N₂:CO₂ (80%:20%) gas passed over a heated copper column, and then they were autoclaved. After autoclaving, a sterile and anoxic 100X phosphate solution containing 0.35 g L^–1 K₂HPO₄ and 0.23 g L^–1 KH₂PO₄ was added to each bottle of base salts medium to a final dilution of 1X. Prior to inoculation, the base salts medium used to cultivate both methanogen strains was amended with 1% (v/v) Wolfe’s vitamins and 1% SL-10 trace metals (no added Fe containing components). All M. barkeri cultures were provided with 0.5% (v/v) methanol and 40 mM acetate as a methanogenesis substrates and grown anaerobically with a N₂:CO₂ (80%:20%) headspace pressurized to 2.5 bar.

Maintenance cultures were transferred every 4–7 days when they reached late log-phase. Cultures were maintained by providing aqueous sources of Fe and S, which consisted of 20 μM FeCl₂ and 2 mM HS^–. The M. barkeri Fusaro mutant strain was also provided with 2 mM cysteine to maintain cultures. To inoculate experimental cultures, mid-log phase grown cells were washed by centrifugation (4,696 × g for 20 min at 4°C) under anoxic conditions. Spent medium was decanted and washed cells were resuspended in sterile, anoxic base salts medium in a sealed serum bottle. Cultures were handled within an anaerobic chamber throughout the washing procedure. A 10% (v/v) transfer of washed cells was used to inoculate freshly prepared medium. All cultures were grown statically on their sides at 38°C in the dark.

Growth of M. barkeri was monitored by quantification of DNA and protein. M. barkeri grows in aggregates and strongly associates with FeS₂ (Supplementary Figure 5), which prohibits accurate cell enumeration. Further, the presence of FeS₂ minerals in the growth medium interferes with OD measurements, commonly used to quantify growth of M. barkeri (Sowers et al., 1993). Therefore, DNA and protein were used as proxies for cell production. In the case of DNA quantification, it is important to note that other Methanosarcina species contain fewer genome copies per cell during slow growth compared to fast growth (Hildenbrand et al., 2011) and thus DNA could underestimate biomass production during early and late log phase. It is also not known if cells produce excess EPS or protein during attachment to mineral phases, which could influence protein-based estimates of biomass production. Nonetheless, to quantify DNA, 2 mL of M. barkeri MS or Fusaro cultures were centrifuged at 20,000 × g for 20 min at 4°C, the supernatant was removed, and the cell pellet was resuspended in a lysis buffer solution containing 489 μL sodium phosphate buffer (MP Biomedicals, Irvine, CA, United States) and 61 μL MT buffer (MP Biomedicals). Once the lysis buffer was added to the cell pellet, the solution was mixed by gentle agitation and subjected to three rounds of freezing at –80°C and heating/thawing at 70°C in a heat block. The mixture was then transferred to a Lysis E tube (MP Biomedicals) and homogenized on a bead beater (Biospec Products, Bartlesville, OK, United States) for 40 s. Finally, the tube and its contents were centrifuged for 15 min at 14,000 × g to separate cell debris from DNA. The concentration of DNA in the supernatant was quantified fluorometrically with a Qubit HS Double Stranded DNA kit and Qubit fluorimeter (Invitrogen, Carlsbad, CA, United States). For protein quantification, a 1 mL aliquot of M. barkeri cultures was centrifuged at 20,000 × g for 20 min at 4°C, the supernatant was removed, and the cell pellet was resuspended in 0.5 M NaOH. The solution was incubated at 99°C in a heat block for 10 min, allowed to cool to room temperature, and then centrifuged for 15 min at 14,000 × g to separate debris from protein. The concentration of protein was quantified fluorometrically with a Qubit Protein Assay kit and Qubit fluorometer (Invitrogen, Carlsbad, CA, United States).

Headspace gas from cultures was sampled with a N₂-flushed syringe to monitor CH₄ and H₂ and was diluted with ultra-high purity N₂ into gas-tight CaliBond bags (Calibrated Instruments Inc., Manhasset, NY, United States). CH₄ and H₂ were determined by gas chromatography via injection of a 5 mL of sample into an injector valve set at 55°C on an SRI 8610C gas chromatograph (SRI instruments, Torrance, CA, United States). The gas chromatroph was equipped with a 4.5 m × 0.125″ OD Hayesep DB 100/120 packed column with the oven set to 44°C (Valco Instrument Company Inc., Houston, TX, United States). CH₄ was detected by a flame-ionization detection at 156°C with ultra-high purity He as carrier gas and H₂ was measured by a pulse-discharge He-ionizer detector at 100°C. Methane and H₂ peak area values were converted to ppm using pre-mixed gas standards (EGAS Depot, Nampa, ID, United States). Dissolved HS^– and headspace CH₄ measurements were converted to total sulfide and total CH₄ (dissolved and gas phase) using Henry’s Law.

For dialysis experiments, 100 kDa dialysis tubing was prepared for FeS₂ and Fe_1–xS experiments, as previously described (Payne et al., 2021b) using a series of ethanol and MQ H₂O rinses to wash and sterilize the membranes. Once sterilized, tied dialysis membranes were transferred to an anaerobic chamber where synthetic FeS₂ (final concentration of 2 mM Fe) or specimen Fe_1–xS (0.1 g) was added to each dialysis bag then the ends were secured using monofilament lines. The same concentrations of FeS₂ and Fe_1–xS were added to reactors where minerals were not sequestered in dialysis membranes. The dialysis bags were rinsed with sterile, anoxic MQ H₂O before transferring to a prepared bottle of medium amended with methanogenesis substrates, trace metals, and vitamins, which was then capped and sealed before removing from the anaerobic chamber. Unamended controls contained dialysis tubing prepared the same as above but without any mineral or Fe added. Upon removing the sealed medium bottles from the anaerobic chamber, the headspace was purged with an 80:20 mixture of 0.2 μm filtered N₂:CO₂ for 45 min to remove gas originating from the anaerobic chamber and were then pressurized with N₂:CO₂ (final pressure of 1.72 atm) before inoculating with mid-log M. barkeri Fusaro wild-type cells grown with sulfide and FeCl₂ (10% v/v inoculum) that were washed by pelleting cells via centrifugation and gently resuspending in anoxic base salts medium within an anaerobic chamber. Anoxic AQDS was added to a final concentration of 2 mM and monitored spectrophotometrically at 325 nm for AQDS and 450 nm for AH₂QDS.

Cultures of M. barkeri MS were grown in quadruplicate on three different sources of Fe and S for RNA-Seq analysis: 2 mM FeS₂, 20 μM FeCl₂ and 2 mM L-cysteine, or 20 μM FeCl₂ and 2 mM Na₂S. Quadruplicate cultures for each condition were kept separate throughout sample collection, RNA sequencing, and analysis. Cells were harvested at mid-log growth phase, as determined by CH₄ production, and final biomass was estimated by DNA quantification. To harvest biomass, M. barkeri MS cells were subjected to vacuum filtration at low pressure (5 psi) onto 47 mm 0.2-μm Supor 200 PES filters (Pall, Port Washington, NY, United States) within an anaerobic chamber. Using flame-sterilized scissors, each filter was cut in half and each half transferred to a 1.5 mL cryotube that was then sealed before being removed from the anaerobic chamber. The tube and its contents were immediately flash frozen in liquid nitrogen and frozen cells were stored at –80°C until processing.

Total RNA from M. barkeri strain MS was extracted using TRIzol reagent (Invitrogen) following the manufacturer’s protocol with slight modification. Individual frozen half filters containing M. barkeri cells were transferred to a Lysis E tube (MP Biomedicals) on ice and 1 mL of TRIzol was immediately added to the frozen filter. TRIzol-treated M. barkeri cells were subjected to three cycles of 40 s of bead beating followed by resting for 5 min at room temperature (∼20°C). Two hundred microliter of molecular grade chloroform was added to each tube and each tube was inverted to mix, incubated at room temperature for 3 min, and centrifuged for 15 min at 12,000 × g at 4°C. The upper aqueous phase containing RNA was carefully transferred to a clean 2 mL tube. To precipitate RNA, 0.5 mL of pre-chilled (4°C) 100% molecular grade isopropanol was added to each tube and tubes and their contents were incubated on ice for 10 min followed by centrifugation for 10 min at 12,000 × g at 4°C. The supernatant was discarded, and 1 mL of 75% molecular grade ethanol was added to wash the RNA. After centrifugation for 5 min at 7,500 × g at 4°C, the supernatant was removed, and the RNA pellet was air dried for 5–10 min. Total RNA was dissolved in 50 μL RNA-grade water (Thermo Fisher Scientific, Waltham, MA, United States) by incubating in a 55°C heat block for 10 min. RNA was treated with Turbo DNase (Invitrogen) to remove residual DNA per manufacturer’s instructions. After DNase treatment, the remaining total RNA was subjected to a second round of isopropanol precipitation, ethanol wash, and resuspension as described above. The concentration of total RNA was quantified fluorometrically using the Qubit BR RNA kit and a Qubit fluorometer (Invitrogen) and the quality checked using a NanoDrop ND-1000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, United States). Absence of DNA was verified by subjecting the RNA extract to 40 cycles of PCR using archaeal-specific 16S rRNA primers (344F/915R) as previously described (Boyd et al., 2013) and checking for amplification products via gel electrophoresis. Total RNA was sent to the University of Wisconsin’s Genome Expression Center for quality control, rRNA depletion using custom M. barkeri strain MS-specific oligos designed using the sequences for M. barkeri MS’s large and small ribosomal subunits, and sequencing on an Illumina NovaSeq 2 × 150 bp.

Paired-end reads were processed using default settings in TrimGalore!,¹ a wrapper that implements CutAdapt (Martin, 2011) and FastQC² to remove adapter sequences and filter reads, respectively. Reads were aligned to the reference M. barkeri MS genome (ASM97002v1) using Bowtie2 (Langmead and Salzberg, 2012). Reads were counted for each locus using HTSeq (Anders et al., 2015) followed by normalization and analysis in DESeq2 (Love et al., 2014) implemented in R v3.6.0.