Five short-term experiments (10 to 14 days) were conducted in this study. A pure culture of M. barkeri (strain DSM-800 isolated from an anaerobic sewage digester) was obtained from DSMZ (Braunschweig, Germany). Cultures were grown at 37°C and 125 rpm in 110 mL crimp top bottles with rubber stoppers (either Chemglass CLS-4209-14 or GMT stoppers). First, the bottles were prepared with 45 mL of modified phosphate buffer media for methanogens, limited organics (modified after Boone et al., 1989, Supplementary methods S1) and 1% (v/v) vitamin solution (DSMZ medium 141). All media were prepared anaerobically and sterilized under N₂ purge (all gasses used for purging were of 99.999% grade). To eliminate O₂ traces, 0.5 mL of titanium-citrate (Ti-Ct) was injected to reach a final concentration of 2.5 mM. Bottles were then purged by H₂ for 20 min and injected with CO₂ resulting in a pressure increase of 20% and an H₂:CO₂ ratio of 80:20 in the headspace. Media bottles were then inoculated with 2 mL of M. barkeri strain DSM-800 at the exponential growth stage. Bottles were kept under the same conditions (37°C and 125 rpm) for approximately two weeks until exponential growth was obtained, as indicated by increasing CH₄ concentration in the headspace. Then the experiments started by switching to N₂:CO₂ (ideally 80:20) atmosphere. Treatments consisted of iron oxides additions in the form of hematite (powder ≥99% <5 μm, Sigma 310,050), magnetite (nano-powder 97%, Sigma 637,106) or poorly crystalline ferrihydrite (amorphous iron; see Supplementary Figure S1 for treatment bottles and SEM images). Amorphous iron was prepared by titrating FeCl₃ solution with NaOH 10 M until pH 7 was obtained. All additives were prepared with 0.22 μm filtered distilled water and purged with N₂ for 20 min.

An additional experiment (V) was conducted to ascertain whether M. barkeri requires direct contact with the iron oxide in order to reduce it. Hematite was enclosed within predrilled cryogenic vials (2 mL), where the cap was replaced by 0.22 μm PVDF membrane (obtained from Merck Millex® filter unit). The experiment was set up with a 100% N₂ atmosphere. During the experiment, bottles were kept under 37°C in an incubator and shaken gently each morning (to avoid tearing the cryogenic vials membrane). The full experimental design appears in Supplementary data, Suppelmentary Table S1. Samples were analyzed to determine the concentration of dissolved Fe²⁺, CH₄, dissolved inorganic carbon (DIC) and δ¹³C of DIC and CH₄.

Additional control experiments were conducted, comparing treatments without culture (medium was autoclaved and cooled) followed by the addition of vitamins, PCA, and iron oxides (amorphous iron, hematite, or magnetite), as previously described. It should be noted that in all the autoclaved control experiments, iron oxides were added several hours after autoclave, allowing cultures to cool down to ambient temperature to avoid abiotic reactions involving the added iron.

M. barkeri cultures containing 10 mM hematite, 100 μM PCA, both of them or none of them (natural cultures) were centrifuged (Allegra™ X-22R Benchtop Centrifuge, Beckman Coulter) for 10 min in maximal RPM in 4°C and filtered with 0.22 μm PVDF filters (Merck Millex®). The filtered supernatant was extracted in solid phase extraction (SPE) method by the manufacture protocol using Biotage® VacMaster™ processing manifold. Briefly, SPE columns (Strata™-X 33 μm Polymeric Reversed Phase 30 mg, phenomenex®) were conditioned and equilibrated with 1 mL MeOH and 1 mL water, respectively. Then, the supernatant was loaded onto the columns; the columns were washed with 1 mL water and dried. The metabolites were eluted from the column with 500 μL 2% formic acid in MeOH twice. The samples were concentrated in vacuum (CentriVap Concentrators, LABCONCO) and kept at-20°C until further analysis.

The samples were re-suspended in 200 μL of pre-cooled 1:1 H₂O:MeOH solution, vortexed, and placed in ice for 15 min. The samples were sonicated for 8 min and centrifuged (Centrifuge 5,424, Eppendorf) for 5 min at 14,680 rpm at room temperature. 150 μL of the supernatant was transferred to mass spectrometry (MS) analysis, and 10 μL were transferred to quality control (QC) vail. The samples were examined by liquid chromatography–tandem mass spectrometry (LC–MS/MS) using an UltiMate 3,000 UHPLC+ focused LC–MS system (Thermo Scientific™) with 2.6 μm Accucore™ C18 (100 × 2.1 mm) HPLC column (Thermo Scientific™), followed by Q Exactive™ Focus Hybrid Quadrupole-Orbitrap™ Mass Spectrometer (Thermo Scientific™) analysis. 5 μL was injected from each sample while the column oven was heated to 40°C. The flow rate was 0.4 mL/min when solvent A was 100% water with 0.1% formic acid, and solvent B was 100% with 0.1% formic acid. The gradients started with 90% of solvent A for 1 min. After 8 min, the gradient was increased to 90% solvent B, and stayed in this composition until 11 min. The gradient changed to 90% A until 12 min, and was maintained that way until 15 min. The scan range was set to 120–1,500 m/z, with a normalized stepped collision energy of 10, 20, 40 eV. The capillary temperature was 275°C, and the spray voltage was 3.5 kV. The MS analysis was conducted in positive ionization modes. To evaluate the instrument operations, the three first and last samples were a test mix sample containing 5-aminofluorescein, caffeine, and theophylline. After the test mix, the QC sample was measured five times, again every six measurements, and five more times in the end. All MS spectra were analyzed by Xcalibur and freestyle software (Thermo Scientific™).

Acquired MS .raw files were converted to .mzXML files with GNPS (Global Natural Products Social Molecular Networking) Vendor Conversion. The converted files were uploaded to mzMine3 (Pluskal et al., 2010) for feature detection and feature alignment. MzMine3 output was uploaded to GNPS and Sirius for further analysis. To learn more about the relationships and structures of the metabolites in the cultures, a molecular network was created in GNPS (Wang et al., 2016; Schmid et al., 2021) with precursor ion mass tolerance and fragment ion mass tolerance of 0.02 Da, and cosine score of 0.7 to form a network edge. An analysis with MolNetEnhancer followed the molecular networking (Ernst et al., 2019) to assess the chemical family classification of the samples’ metabolites. We analyzed the results based on the direct parent MolNetEnhancer classification. The Sirius tool (Dührkop et al., 2015, 2019; Djoumbou Feunang et al., 2016; Kim et al., 2021), with CSI:FingerID and CANOPUS, was used to annotate a putative identification of the compounds’ structure. MetaboAnalyst (Pang et al., 2022) was used to create principal component analysis (PCA) based on the MS data.

Samples for Fe²⁺ and DIC were prefiltered (0.22 μm). One milliliter was immediately fixed using the Ferrozine method (Stookey, 1970) for Fe²⁺ measurement. Ferrous iron absorbance after fixation was measured on a spectrophotometer (562 nm) with a detection limit of 1 μM. Additional 1 mL samples were kept to determine DIC concentrations and δ¹³C_DIC analyses at 4°C for up to 1 week prior to measurement. The δ¹³C_DIC was measured on a continuous flow Isotopic Ratio Mass Spectrometer (IRMS, DeltaV, Thermo), interfaced to Gas Bench II (GBII), and is reported in per mille units relative to Vienna Pee Dee Belemnite (VPDB) with a precision of 0.2 ‰. DIC was measured on the same IRMS using the peak area with a precision of 0.1 mM. Methane and 13 C CH 4 were measured from the headspace. Concentrations were measured on a Gas Chromatograph (Focus GC, Thermo) equipped with a flame ionization detector (FID) and packed column (Shincarbon ST) with a detection limit of 1 nmol. The 13 C CH 4 was measured on the same IRMS with PreCon device addition to GBII.

Gene copies of 16S rRNA were measured by sampling 10 mL culture from experiment IV bottles on day 11. Samples were centrifugated at 9,000 g for 10 min and the pellet was used for DNA extraction (DNeasy PowerLyzer PowerSoil Kit, QIAGEN). DNA concentrations in samples were evaluated using a nanodrop spectrophotometer (ND1000). Real-time PCR amplifications were conducted using an Azure Cielo 3 system (Azure biosystems, United States) in 20 μL volumes, with each standard reaction mix containing 10 μL of SYBR™ Green PCR Master Mix (Applied Biosystems™), 0.8 μL of each primer (F341 and R518) (Muyzer et al., 1993), 6.4 μL distilled water and 2 μL of either DNA or standard (plasmid) as a template. Thermocycling was as follows: 95°C for 5 min, 30 cycles of 10 s at 95°C, 20 s at 62°C, 30 s at 72°C, 2 s at 78°C. Runs were completed with a melting analysis (65°C to 95°C, ramp, 0.5°C/min) to check for product specificity and primer dimer formation. The assay included a standard curve of six standards (from 2.5·10⁸ to 2.5·10³ copies per μL) that were freshly prepared from a stock solution, with an efficiency of 101.5 and an R² = 0.996. All standards were run in duplicate and samples in triplicates.

Methanosarcina barkeri is known as a hydrogenotrophic methanogen, which can also use acetate and methanol as substrates (Bond and Lovley, 2002). Our results show that M. barkeri is also able to use other metabolism and transfer electrons to iron minerals. It can reduce several iron-oxide minerals at close to natural conditions, with the organic substrate in the medium (Figures 1, 2). The concentrations of the most abundant iron oxides can vary widely depending on the settings. The concentration used in the study were similar to those found in natural sediments (Canfield, 1989; Bar-Or et al., 2017; Boyko et al., 2021). As was anticipated by other studies (Munch and Ottow, 1983), amorphous iron was the most reactive iron-oxide in our experiments (Figure 2). Magnetite and hematite showed much lower reactivity, where more iron was reduced in the presence of magnetite, leaving hematite as the least reactive metal oxide in this experiment (Figure 2). In the killed cultures and media control experiments the amount of reduced iron was negligible. The question is how does M. barkeri overcome the kinetic barrier that is usually involved in the reduction of low-solubility iron minerals?

Microbial reduction of iron oxides is a challenging process. The environmentally abundant iron-oxides are insoluble (Kappler and Straub, 2005), and the cell membrane forms a highly impermeable barrier (Shi et al., 2007). Microorganisms, however, have developed several mechanisms to overcome this challenge: (i) Electron carriers from the c-type cytochrome family are used to transfer electrons from the inner side of the membrane to the cell surface, and eventually to the mineral (Mehta et al., 2005); (ii) extension of the surface that can take part in this process by conductive pili or nanowires enable the electron transfer from the microbe to the mineral (Gralnick and Newman, 2007); (iii) secretion of iron chelators (siderophores) that can chelate Fe(III) tightly, convert this relatively insoluble iron to a bioavailable substance, which can then be utilized by the cell (Lippard and Berg, 1994), and (iv) reduction of electron shuttles that are subsequently oxidized by mineral iron oxides, effectively recycling the electron shuttles.

Several known compounds can participate in the last mechanism, e.g., cysteine (Doong and Schink, 2002), humic substances, and various redox-active compounds that can be produced by microbes, such as phenazines (Hernandez et al., 2004). A major class of redox molecules used as electron shuttles is the class of quinones. Quinones can be found in all living domains (Nowicka and Kruk, 2010). However, in methanogens they are usually found only in very low amounts (Hughes and Tove, 1982; Abken et al., 1998). Microbes, however, also often use redox compounds that are not produced by themselves, explaining M. barkeri’s ability to utilize both humic acids and the related quinone analog AQDS (Bond and Lovley, 2002).

M. barkeri has been shown to use these electron shuttles to increase the availability of Fe(III) in amorphous iron oxides (Roden and Zachara, 1996; Scott et al., 1998). Here, we confirm that M. barkeri can exploit AQDS to reduce not only a high-reactivity iron species such as amorphous iron, but also less reactive iron minerals (Figures 1, 2) and under close to natural conditions (N₂:CO₂ atmosphere). The reaction between the AQDS and available iron species is an abiotic process. Some iron reduction was observed in control cultures containing dead archaea. Nevertheless, in the absence of reducing organisms the AQDS cannot be recycled, limiting the amount of reduced iron. The archaea reduce more iron than the killed control in the presence of only hematite (without AQDS), validating the crucial role that the living microbes play in the iron-reduction process. When AQDS was added to live cultures containing hematite, the iron reduction increased dramatically, supporting the assumption that this process can indeed occur in nature.

Although methanogens do not produce quinones, they possess other membrane-bound redox electron carriers, such as methanophenazines (Abken et al., 1998; Beifuss et al., 2000). It is possible that electron shuttling by methanophenazines can be utilized for additional energy generation processes, such as iron reduction. Methanophenazines and cytochromes were recently found to be also involved in ferrihydrite-dependent AOM (Yan et al., 2018; Yu et al., 2022), but AOM did not seem to occur in our cultures. Phenazines are known to promote mineral iron reduction (Hernandez et al., 2004) thanks to their favorable redox potential. Due to their lipid character, methanophenazines are highly insoluble in aqueous environments (such as the culture media). Therefore, in order to examine their potential more soluble analogs such as PCA are added (Abken et al., 1998). PCA was also added to mimic free phenazines that are abundant in sediments (Thomashow, 2013). The addition of PCA enhanced iron reduction in all samples (Figures 1, 3). It did not affect the reactivity order of the different minerals, i.e., more iron-reduction occurred in the presence of amorphous iron and PCA, while the lowest amount was observed when hematite was added to PCA containing cultures. Some iron reduction was observed in the killed control cultures due to abiotic reactions, emphasizing the crucial role of microorganisms in cycling the electron shuttle, thus enabling the process to continue.

Moreover, the addition of PCA to the media with the iron mineral released more metabolites than without it, but no unique metabolites were detected in cultures containing both hematite and PCA. This change in the metabolite levels, but not in their identity, suggests that the combination of the iron mineral and the electron shuttle enhances existing pathways. This observation supports the possibility that M. barkeri’s own phenazines participate in the iron-reduction. As methanophenazines are strongly associated with the microbial membranes, their potential role in iron reduction by M. barkeri versus free phenazines was supported by the experiments where hematite was added to the media directly or inside a vail, which enabled only the transfer of dissolved species. The results indicate clearly that direct contact between the mineral and the cells is needed for electron transfer.

One of the major metabolite groups that was found is the indole family. Indoles are known for their role in biological processes such as signaling (Lee and Lee, 2010; Tomberlin et al., 2017) and regulation (Hu et al., 2010; Duca and Glick, 2020), and can strongly affect the growth and behavior of organisms from all kingdoms (Bennett et al., 2004; Lee et al., 2015). Therefore, it was not surprising to identify indoles in our assays. One of the metabolites found in higher concentrations in cultures that contained PCA and hematite together is structurally related to 5-hydroxyindole. Although a comparison to a commercial standard showed that this unidentified compound is not identical to 5-hydroxyindole, the similar MS2 fragmentation implies that it is related to this indole, and further investigation of its role in biological pathways in M. barkeri is needed.

Amorphous iron reduction by M. barkeri was shown to inhibit methanogenesis (Bond and Lovley, 2002; Van Bodegom et al., 2004; Sivan et al., 2016). The switch between these different processes was also observed in goethite (Liu et al., 2011b). Our study evaluated the archaea’s ability to switch between those two environmentally significant processes, with less reactive minerals than amorphous iron, in a non-H₂ atmosphere. In the presence of PCA, M. barkeri switched between methanogenesis and iron reduction with all three iron-oxides tested in this study (Figure 4; Supplementary Figure S2). As the electron shuttle dramatically increased the iron-reduction and decreased the methanogenesis (Figure 3), it seems that PCA enables the archaea to favor iron-reduction over methanogenesis. Iron reduction is favored only when there is direct contact between the cells and the mineral, as shown nicely in Figure 7, hinting to the roles of methanophenazines, as mentioned above. When the conditions are less optimal for iron reduction, M. barkeri will carry out both methanogenesis and iron reduction.

The two processes (methanogenesis and iron reduction) thus display an inverse relationship (Supplementary Figure S2). Accordingly, while the magnitude of iron reduction was largest with amorphous iron, then magnetite, and eventually, with hematite, methane production showed the inverse trend: highest with hematite, then magnetite, and lowest with amorphous iron. The methane production in the hematite and magnetite-containing cultures without PCA, however, is not significantly different from the mineral free cultures. This indicates that iron does affect methanogenesis mainly in the presence of the electron shuttle. These observations also support the hypothesis that methanogens may respire iron oxides in nature, due to their methanophenazines.

Previous studies have shown that methanogens are able to utilize iron reduction for energy cell growth (Prakash et al., 2019; Yang and Lu, 2022). Our results also show that although M. barkeri did not thrive on iron reduction (Supplementary Figure S3), Fe²⁺ accumulation during the experiments suggested that the cultures are still active. The iron reduction rates were close to those of Shewanella putrefaciens (Supplementary Table S2). It should be noted that we cannot enterally rule out the possibility of cell lysis as a source for non-viable iron reduction, and the iron reduced in the control treatment may be attributed to this process. Nevertheless, the amount of iron reduced in our control experiments is less than 10% of the amount of iron reduced in live treatments.

Methanogenesis is a dominant reaction in anoxic environments, resulting in significant isotopic fractionation (Whiticar et al., 1986). The carbon isotopic composition of methane ( δ 13 C CH 4 ) excepted at the end of the second experiment also supports this observation of inhibition of methanogenesis by iron oxides. As a result of methanogenesis, the substrate pool in the methane-producing cultures becomes heavier along the experiment because the methanogens keep using more ¹²C than ¹³C. At some point, the archaea will exploit a significant portion of the light substrate present in the culture, leaving them to utilize the heavier substrate, resulting in heavier methane. The final δ 13 C CH 4 of methane-producing cultures was indeed ~10‰ higher than in the cultures where methanogenesis was inhibited almost completely.

In order to investigate the kinetics of the iron reduction reaction we introduced different concentrations of hematite, in the range of 0 to 10 mM (Figure 6A). Many parameters affect iron reduction rates, e.g., crystallinity, surface area (Roden, 2003), and the presence of electron shuttles (Lovley et al., 1996; Doong and Schink, 2002). The amount of reduced iron increased in almost all samples during the experiments. Interestingly, regardless of the concentration of iron in the samples, about 1% of hematite was reduced, indicating that the reaction rate is dependent on the iron concentration, resulting in a pseudo-first order reaction.

Another indication of the crucial role of the iron concentration in the reaction kinetics can be observed by the linear relationship of the added hematite to the produced Fe²⁺ (Figure 6B). This linear relationship indicates again that the iron concentration constitutes the barrier for this reaction and no other substrates, fitting first order reaction. The reduction rate coefficient was found to be 3.60 × 10⁻³ day⁻¹. Although this reaction rate coefficient is not high, the reaction still occurs with hematite, the least reactive mineral examined in this experiment, and in a wide range of hematite concentrations, providing additional proof that iron reduction by methanogens potentially takes place in natural sediments. Interestingly, our observed iron reduction rate by the methanogens is only slightly lower than the known iron reducers Shewanella putrefaciens (and one order of magnitude less than the Geobacter genus; Supplementary Table S2). The fact that M. barkeri can reduce hematite with similar reaction rates as iron reducer establishes the potential role of this methanogen in iron reduction in deep sediment, and its ability to outcompete with bacterial iron reducers.

The substrate for the iron reduction is probably not methane, as ¹³C-labeled methane was not converted to enriched ¹³C-DIC. In addition, the slight methane concentration decrease in the living cultures without PCA cannot be related to AOM, because this decrease was also observed in cultures without iron oxides (Figure 4). The electron donor is hence probably from the inoculum. We flushed the vials very carefully, so a residual amount of hydrogen from the pre-culture is less likely, but we cannot rule it out. The reducing agent (Ti-Ct) is less likely based on our recent work (Shang et al., 2020), but also cannot be ruled out. The most probable electron donor is the yeast extract itself (Supplementary methods S1). However, the exact organics in the yeast are less relevant to this study.

Deep sediments have complex microbial redox couplings that affect global biogeochemical cycles, such as carbon, and as a result, it affects all living species on Earth. Methane, a significant greenhouse gas, is consumed and produced in those sediments and can be released into the atmosphere. Iron reduction by lowly reactive iron minerals in the methanogenic zone was previously demonstrated (Gault et al., 2011; Egger et al., 2017; Vigderovich et al., 2019; Amiel et al., 2020) but the involved dominant microbes have not been identified yet. It has been suggested that archaeal methanogens may play a role in this reduction, based on geochemical and microbial studies in natural sediments (Elul et al., 2021). Here, we indeed established that the naturally abundant archaeon methanogen M. barkeri might actually switch its metabolic pathway from methanogenesis to iron reduction, with abundant and low-reactivity minerals such as magnetite and hematite, and in the absence of hydrogen. Further proof that the metabolic switch indeed occurs in nature is still required, but the fact that this process readily takes place in pure cultures without H₂ addition can shed light on the advantage of iron reduction for methanogens such as M. barkeri in nature. The thermodynamics of iron reduction of magnetite and hematite are similar to those related to methane production, so favoring this specific reaction is probably due to kinetic limitations and not thermodynamics. We also showed that hematite reduction by M. barkeri is a pseudo-first order reaction, indicating that in deep sediments, where iron minerals are abundant, M. barkeri could serve as an iron reducer, depending only on the available Fe(III) concentration. M. barkeri can thus switch from methanogenesis to reduce abundant iron oxides (Bar-Or et al., 2017; Vigderovich et al., 2019; Amiel et al., 2020) to survive in this environment. Further studies are needed to establish and characterize the switch between iron reduction and methane production by methanogens in nature. A better understanding of the processes in deep sediments will provide important insights into the biogeochemical cycles of iron and the greenhouse gas methane.