Ecological and metabolic restructuring of anaerobic microbiomes under sulfate stress via magnetite-enhanced cooperative networks

Abstract

Background

Anaerobic digestion systems with elevated sulfate often suffer reduced methane yields, challenged by the competition between sulfate-reducing bacteria and methanogens, and inhibited by hydrogen sulfide introduction. The present work explores the role of magnetite in improving anaerobic digestion performance under elevated sulfate conditions by chemically influencing the anaerobic system and reshaping microbial interaction patterns.

Results

Magnetite addition mitigated hydrogen sulfide toxicity via precipitation and increased methane production by 19%. Genome-centric metagenomics revealed a notable proliferation of the methanogenic population in the magnetite-amended reactors, consistent with the elevated methane output in the presence of both magnetite and sulfate, without suppressing sulfate-reducing, homoacetogenic, or syntrophic acetate-oxidizing activity. Magnetite was associated with enhanced methanogenesis and a strengthened cooperative syntrophic network among the four microbial guilds, in line with more efficient carbon and electron flow despite sulfate stress. Community genome-scale metabolic modeling supported these trends, validating the feasibility of the proposed interaction network and indicating that interspecies metabolite transfer between partners is stoichiometrically feasible, supporting the observed community behavior.

Conclusions

This study demonstrates the role of magnetite not only as a hydrogen sulfide scavenger but also as a community modulator, promoting resilient direct electron transfer-based networks, ultimately unlocking higher-efficiency biogas production in sulfate-impacted digesters. Our findings support the concept that interactions between sulfate-reducers and hydrogenotrophic methanogens are not purely competitive, and that conductive materials such as magnetite can enhance their metabolic coupling even under sulfate stress.

Download video file ^{(40.1MB, mp4)}

Video Abstract

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s40168-026-02443-4.

Introduction

Anaerobic digestion (AD) is a microbiologically driven process in which a complex microbial community coordinates the sequential steps of hydrolysis, acidogenesis, acetogenesis, and methanogenesis, converting organic wastes into renewable energy in the form of biogas, composed primarily of methane (CH₄) and carbon dioxide (CO₂) [1]. Although a minor component of the produced biogas, hydrogen sulfide (H₂S) poses a significant obstacle to the efficiency of AD. Formed as a by-product of sulfate (SO₄^2 −) reduction by sulfate-reducing bacteria (SRB) [2, 3], H₂S is highly toxic and corrosive, threatening both methanogenic activity and infrastructure integrity [2, 3].

H₂S toxicity is not the sole reason SRB disrupt the delicate balance of AD. A major issue arises from their competition with methanogens for key energy sources such as acetate and hydrogen (H₂), which SRB use during sulfate reduction, producing H₂S in the process [4, 5]. With sulfate reduction being thermodynamically more favorable than methanogenesis [4, 5], SRB not only introduce toxicity but also create a complex relation with methanogenic archaea [6], involving competition or cooperation depending on system conditions, posing a significant challenge to the stability of AD systems. The outcome of this interaction depends on the availability of organic matter, contributing to the chemical oxygen demand (COD), and the sulfate level in the system [2, 4, 7].

Recent studies suggest that conductive materials (CMs) may promote methanogenesis over sulfate reduction by reinforcing direct interspecies electron transfer (DIET) [8–10]. Unlike indirect interspecies electron transfer (IIET), DIET allows electron transfer directly from exoelectrogenic to electrotrophic microorganisms via microbial bioelectrical structures, like electrically conductive pili (e-pili) and multiheme c-type cytochromes (MHCs) [11–13]. In the absence of such biological means, microorganisms can perform DIET by exploiting direct electron flow through CMs [8, 13–15]. CMs are commonly classified as carbon-based and non-carbon-based, both playing a pivotal role in supporting cooperative metabolism and helping to overcome thermodynamic limitations [8, 9, 15]. Carbon-based CMs, such as activated carbon, graphite, and carbon nanotubes, are widely used for their high conductivity [8, 9, 15]. Due to their large surface area, carbon-based CMs provide a stable surface for microbial attachment, thereby enhancing the conductivity within the system [9, 15]. Non-carbon-based CMs, including iron oxides, stainless steel, and conductive polymers, facilitate electron transfer by providing high electrical conductivity, though with material-specific trade-offs in surface area and biocompatibility [8, 15]. The selection of CMs depends on factors like conductivity, biocompatibility, and stability within the specific microbial system, which together shape AD performance. Among them, magnetite, an iron oxide containing both Fe²⁺ and Fe³⁺, can chemically react with H₂S under mesophilic anoxic conditions, precipitating sulfide and thereby lowering H₂S toxicity, potentially benefiting the overall microbial community [4, 6]. Moreover, magnetite has been reported to compensate for c-type cytochromes embedded in e-pili structure, facilitating electron transfer between bioelectrically syntrophic partners [8, 14, 16].

While the competition between acetoclastic methanogens and SRB is well documented [17–19], the nature of the interactions between hydrogenotrophic methanogens and SRB remains controversial. Traditionally, these microbial guilds have been considered competitors; however, acetate-oxidizing SRB produce CO₂ during sulfate reduction, while in some cases, may also contribute to H₂ availability via the WL pathway-linked metabolism or hydrogenase activity [6, 20]. These products can be used by hydrogenotrophic methanogens, providing a plausible route for metabolic coupling rather than strict competition. This cross-feeding may be further strengthened in the presence of CMs, which can facilitate electron exchange and help maintain reducing equivalents that support methanogenesis. This perspective moves beyond a purely competitive framework and motivates investigation into how CMs could modulate these microbial interactions.

In this study, hydrogenotrophic methanogenesis by a mixed microbial culture fed with H₂:CO₂ was tested under two sulfate concentrations, with and without the addition of magnetite, a naturally occurring, widely available, and low-cost material with potential for recovery due to its magnetic properties [21–23], previously shown to facilitate electron transfer and enhance AD in systems involving SRB and methanogens [6]. Magnetite’s interplay with dissolved H₂S was also examined in a dedicated abiotic experiment, enabling quantitative evaluation and providing mechanistic insights into their interaction within the system. Genome-centric metagenomic analysis was performed to resolve community structure and dynamics across conditions, with a focus on identifying key interactions and putative DIET-performing species. Community genome-scale metabolic modeling and flux balance analysis were applied to test the feasibility of the hypothesized interactions and identify candidate intermediate metabolite exchanges between partners, providing a mechanistic framework to interpret potential cooperation. Compared to previous studies focusing mainly on methane yield enhancement using conductive materials, this integrative strategy links experimentally observed sulfur dynamics and methane production to specific genome-resolved guilds and interaction patterns to provide mechanistic insight into how magnetite promotes cooperative, electron-mediated networks under sulfate stress. Overall, the study establishes magnetite as a dual-action amendment in sulfate-rich AD systems, simultaneously mitigating H₂S toxicity and strengthening electron-mediated cooperative networks that underpin enhanced and sustained CH₄ production.

Materials and methods

Experimental setup

Mesophilic inoculum from Hashøj biogas plant (Zealand, Denmark) was prepared as previously described [24]. Inoculum characteristics are provided in Supplementary material 1. Magnetite (Fe²⁺/Fe³⁺ oxide powder, < 5 μm, 95%, Sigma-Aldrich) was added as conductive material at 10 g/L [6, 25]. Anoxic serum bottles (250 mL; working volume 100 mL) were purged with pure N₂ (Air Liquide, Danmark A/S), each containing 80 mL basal anaerobic (BA) medium and 20 mL inoculum [26]. A blank test without a carbon source was included. Five conditions were tested using H₂/CO₂ gas (4:1, Air Liquide, Danmark A/S) as feed in a final volume of 200 mL: a control without sulfate (CTRL), and four treatments with 0.6 or 1.2 g/L sulfate (sodium sulfate ≥ 99.0%, Sigma-Aldrich), as in our previous work [6], in the presence or absence of magnetite (SULF_0.6, SULF_1.2, MAGN_0.6, MAGN_1.2). All conditions were prepared in triplicate, incubated under mesophilic conditions at 100 rpm, and monitored until the stationary phase.

DNA extraction, quantification, and sequencing

At the end of the incubation period, DNA extraction was performed on samples selected to capture the most relevant differences between treatments with and without magnetite addition. Specifically, a 10 mL sample was collected from one biological replicate for each of CTRL, SULF_0.6, and MAGN_0.6. Genomic DNA was then isolated and quantified, and Illumina library preparation and sequencing were carried out according to a previously described protocol [6].

Genome-centric metagenomic analysis and comparative genomics

Raw sequencing data were analyzed as previously described [25]. Briefly, reads were processed with Trimmomatic (v0.39) [27], assembled with MEGAHIT (v1.2.9) [28], aligned with Bowtie2 (v2.2.5) [29], and sorted with SAMtools (v1.19) [30]. Binned metagenome-assembled genomes were screened with CheckM (v1.1.6) [31], dereplicated with dRep (v3.4.5) [32], and filtered according to the MIMAG standards [33]. Taxonomy was assigned with GTDB-Tk (v2.1.0) [34], and gene and functional annotation were performed with Prodigal (v2.6.3) [35] and eggNOG (v2.1.12) [36], respectively. Multi-alignment for putative e-pilins was performed using CLUSTALW, while e-pilins assessment followed previously established criteria as described in Giangeri and colleagues [25, 37]. Sequence data reported in this study have been submitted to the National Centre for Biotechnology Information under the BioProject PRJNA1242289.

Genome-scale metabolic modeling

MAGs corresponding to the highest-quality representatives of the four key functional groups (i.e., methanogens, SRB, homoacetogens, and SAOB) were selected for in-silico metabolic modeling. Individual genome-scale metabolic models (GEMs) were reconstructed and gap-filled with gapseq (v1.4.0) [38] to enable growth, using default parameters, and manually curated to ensure that each strain could perform its expected metabolic function within the reactor/specific medium (Supplementary material 3). Specifically, for each GEM, an expected metabolic role was assigned based on its functional group and inferred physiology (e.g., methanogenesis, sulfate reduction). Model validation was then performed by testing whether the reconstructed GEM could achieve the corresponding phenotype, including growth and production or consumption of key metabolites, under reactor-defined media constraints using Flux Balance Analysis (FBA) in COBRApy [39]. Models that failed to satisfy these requirements were further curated by inspecting pathway completeness and restoring connectivity from substrate uptake to product formation through the addition of missing reactions with gapseq, including transport steps or key enzymatic conversions, followed by manual curation in COBRApy. Missing reactions were added only when required to complete well-established canonical pathways supported by genome annotations and biochemical databases. This procedure ensured that each GEM was functionally consistent with its expected metabolic role while remaining constrained by genome-derived annotations and experimentally defined medium conditions, thereby limiting overestimation of metabolic potential.

Curated GEMs were then integrated into a multi-species community model using MICOM (v0.37.1) [40], with each species exchanging metabolites via a shared extracellular pool. Experimental measurements (CO₂ uptake, CH₄ production) were used to get the substrate uptake and metabolites secretion rates that were then used to constrain the community model, ensuring that simulated fluxes were consistent with reactor performance. Phenotype-based constraints were also applied to individual models and community models. The trade-off solution (value = 0.7) was applied to balance community-level and individual growth. To characterize the feasible flux space, flux variability analysis (FVA) was performed to constrain the solution space, which was subsequently sampled using the artificial centering hit-and-run (ACHR) algorithm (100,000 samples) [41]. The mean of the sampled flux distributions was used as a representative community flux state, thereby reducing the impact of solution degeneracy and uncertainty. Data visualization was performed using the Python package matplotlib (v3.10.0).

Analytical methods

The analytical methods used in this study, including total solids (TS), volatile solids (VS), pH, and volatile fatty acid (VFA) concentrations, are described by Vayena and colleagues [24]. For gas composition monitoring, a Thermo Fisher Scientific GC-TRACE 1310 gas chromatograph was used, with an HP-Molesieve column (Agilent Technologies, USA; 30 m length, 0.53 mm inner diameter, 50 μm film thickness) for H₂ determination and a Thermo (P/N 26004–6030) column (Thermo Fisher Scientific, USA; 30 m length, 0.320 mm inner diameter, and 10 µm film thickness) for CH₄ and CO₂. Sulfate concentration was measured after microfiltration (0.2 μm) of the sample, using LCK153 and LCK353 kits (Hach, USA) and a DR3900 spectrophotometer (Hach, USA).

Physical-chemical magnetite-H₂S interaction and sulfur detection

An independent abiotic experiment was performed at mesophilic conditions to assess the interaction between magnetite and dissolved H₂S. Serum vials (60 mL) containing 12 mL BA medium were purged with N₂/CO₂ gas (4:1, Air Liquide, Danmark A/S). Magnetite was added at 10 g/L (MAGN), and H₂S was injected to achieve an overpressure of 1 bar. A control without magnetite (CTRL) was also set up to estimate the dissolution of H₂S in the liquid phase. Both assays were performed in duplicate. After gas addition, the overpressure and headspace H2 were monitored at specific time intervals over 24 h using the CPH6400 pressure indicator (WIKA, Germany) and GC-TRACE 1310 gas chromatograph, respectively. At the end of the experiment, magnetite was recovered, rinsed with distilled water, and analyzed using an elemental analyzer (EA3000, Eurovector, Italy) to quantify potential sulfur incorporation and precipitation. Elemental analysis (CHNS) was performed using 2.5-bis(5-tert-butyl-2-benzoxazolyl) thiophene (BBOT, 99%, Thermo Fisher Scientific, USA) as a calibration standard and vanadium pentoxide as a combustion aid, ensuring complete sulfur conversion (Thermo Fisher Scientific, USA). Dissolved sulfide species (H₂S, HS ⁻, S^2 −) in the liquid phase were quantified colorimetrically using the methylene blue method [42].

Statistical analysis

Methane production data were analyzed in R (v4.3.2, R Core Team, 2023) using Welch’s t-tests for pairwise comparisons, assuming unequal variances. Comparisons were performed between each condition and the control, as well as between selected condition pairs. Statistical significance was determined at p < 0.05. All analyses were performed using base R functions.

Results

Magnetite-induced H₂S sequestration alleviated sulfate stress in methanogenic systems

After 5 days, all tested conditions reached stationary phase (Fig. 1a). As expected, methanogenesis prevailed in the control (CTRL), with final CH₄ production reaching approximately 40 mL. In contrast, the addition of sulfate led to a significant reduction in CH₄ production, which was approximately 40% lower in both SULF_0.6 and SULF_1.2 (p < 0.05). In the sulfate-free CTRL vials, SRB remained inactive due to the lack of an electron acceptor. Meanwhile, sulfate addition activated SRB, which utilized available electron donors for their metabolism, ultimately leading to significantly lower CH₄ production compared to sulfate-free conditions (p < 0.05) [4, 5, 43]. Interestingly, the presence of magnetite reversed the sulfate-induced inhibition at 0.6 g/L sulfate, resulting in a 19% increase in CH₄ production in MAGN_0.6 compared to SULF_0.6 (p < 0.05, Fig. 1a). Additionally, magnetite enhanced daily CH₄ production, with statistically significant differences observed after day 2 (p < 0.05). However, these trends were not observed at 1.2 g/L sulfate (MAGN_1.2). This concentration-dependent effect suggests that while this level of magnetite can alleviate moderate sulfate stress, it is insufficient to counteract the dominant SRB-driven competition at higher sulfate loads.

Fig. 1.

Temporal changes in a CH₄ production, b acetate, and c sulfate during the batch experiment. Sulfate was monitored to assess reduction activity and potential competition between SRB and methanogens. Acetate, the only detected VFA, represents fermentation intermediates and acetogenic activity. CH₄ reflects methanogenic performance. Data points represent mean values from biological replicates, and error bars represent standard deviations

Acetate, the only VFA detected in the inoculum, though present in minor amounts (< 0.03 g/L), appeared relatively stable across most conditions, except for a pronounced accumulation in CTRL (Fig. 1b), consistent with enhanced acetogenic activity alongside methanogenesis. In sulfate-amended systems, acetate accumulation was less evident, likely due to SRB competition for shared substrates and/or subsequent acetate consumption. Consistently, a comparison of CH₄ and CO₂ profiles (Fig. 1a, Supplementary material 1) showed that CO₂ uptake exceeded what could be explained by CH₄ production, suggesting diversion of CO₂ to non-methanogenic sinks, including acetate formation via homoacetogenesis. Together, these trends support the hypothesis that the acetate produced was re-oxidized via the Wood-Ljungdahl (WL) pathway and channelled back to H₂ and CO₂, thereby supporting methanogenic growth.

The time course of sulfate showed that magnetite did not inhibit sulfate reduction at any tested level, with similar sulfate consumption observed in its presence or absence, indicating no adverse effect on SRB metabolism. Previous studies have demonstrated that magnetite can mitigate the toxicity of dissolved H₂S by promoting its oxidation and precipitation as zerovalent sulfur (S0) or sulfur compounds [4, 44], suggesting an indirect interaction of magnetite with the sulfate reduction pathway. Results of the abiotic test further support this hypothesis, as a clear reduction in overpressure was observed in the abiotic MAGN test, which was substantially higher than the reduction caused by H₂S dissolution in the liquid phase in the abiotic CTRL (Fig. 2). Additional evidence includes the measured concentrations of dissolved sulfide species (H₂S, HS⁻, S^2 −) at the end of the abiotic test: 880 ± 51 mg/L in CTRL and 1.45 ± 0.05 mg/L in MAGN, as well as the elemental sulfur that was detected in the extracted magnetite via elemental analysis, with an average content of 31.3 ± 2.9% (Fig. 2). These findings confirm that H₂S was effectively removed in the presence of magnetite. Elemental analysis, however, cannot distinguish between sulfur compounds and, therefore, cannot clarify the exact form in which sulfur was precipitated. The conversion of H₂S in anoxic systems follows a pathway dependent on the availability of dissolved H₂S and the pH. Under acidic conditions, excess dissolved H₂S favors the formation of iron disulfide (FeS₂) and H₂, whereas neutral to slightly alkaline pH promotes the formation of iron monosulfide (FeS) and zerovalent sulfur (S⁰) [44]. If H₂ were produced, the interaction between magnetite and H₂S could offer a dual benefit to the metabolic system: removing toxicity while increasing the pool of available electron donors, thus promoting further substrate degradation. However, the abiotic test was conducted in a N₂/CO₂-purged (4:1) system, maintaining a neutral pH [26], and so no H₂ production was expected. This was confirmed by gas analysis, which showed no H₂ accumulation in the MAGN test (Fig. 2). Similarly, pH measurements from the different biotic experiments (Supplementary material 1) consistently showed slightly alkaline values (7.45-8.36) throughout the experimental period, ruling out the possibility of H₂ production due to magnetite-H₂S chemical interaction. Thus, under the tested conditions, H₂S precipitation along magnetite primarily contributed to toxicity mitigation rather than supplying additional electron donors for further biomethanation.

Fig. 2.

Workflow for the abiotic test on magnetite-H₂S interaction, showing the experimental setup, analytical methods employed, and key results, including headspace overpressure, H₂ content over time, dissolved sulfide species (shown as H₂S), and elemental sulfur accumulation on magnetite. Full datasets are available in Supplementary material 1. Created with BioRender.com, 2025

Microbial community shifts driven by sulfate and magnetite interactions

Genome-centric metagenomic analysis was performed for CTRL, SULF_0.6, and MAGN_0.6. The 1.2 g/L sulfate level was excluded due to the absence of notable differences between treatments with and without magnetite. A total of 61 metagenome-assembled genomes (MAGs) were reconstructed, of which 57 met the MIMAG criteria for medium-quality (completeness > 70%, contamination < 10%) or high-quality genomes (completeness > 90%, contamination < 5%) [33], with an average reads-alignment of 97%, hence qualifying for further analysis (Supplementary material 2). Taxonomic classification assigned 9 MAGs at the genus and 4 at the species level, reflecting the limited number of fully characterized species [45]. The microbial community resembled that of conventional AD systems [45, 46], with Firmicutes dominating, followed by Bacteroidetes and Proteobacteria (26, 9, and 6 MAGs, respectively). Archaea, represented by three Euryarchaeota species, accounted for a small fraction of the community.

The detected methanogens were assigned to the Methanoculleus genus, with Candidatus Methanoculleus thermohydrogenotrophicum sp. DTU8, Methanoculleus bourgensis sp. DTU41, and Methanoculleus sp. DTU48. All three MAGs harbored marker genes of the Wood-Ljungdahl (WL) pathway (i.e., cdhC, cdhD, and cdhE) and acetate synthase (ACS), consistent with CO₂ assimilation and subsequent biosynthetic processes (Supplementary material 2) [47, 48]. Sulfate, in SULF_0.6, increased the relative abundance of methanogens from 1.6 to 3.5% (Fig. 3), while magnetite led to a more extensive increase up to 5.4%.

Fig. 3.

Visual representation of the relative abundances of microorganisms present in all tested conditions. The heatmap displays the relative abundances of the 57 MAGs retrieved across the three tested conditions (CTRL, SULF_0.6, MAGN_0.6). Abundance values range from 0% (white) to 10% (purple), with values exceeding 10% shown as saturated

The substantial sulfate reduction in all sulfate-amended assays indicated strong SRB activity (Fig. 1c). Yet functional annotation identified only partial coverage of the dissimilatory sulfate reduction pathway, with six MAGs carrying at least one related gene but none encoding more than half of the pathway, preventing definitive SRB classification (Supplementary material 2). However, taxonomic profiling showed 3.2- and 2.9-fold increases in the relative abundance of Halopseudomonas formosensis sp. DTU25 and Sulfurospirillaceae sp. DTU53 after sulfate addition (Fig. 3), taxa previously reported to perform sulfate reduction [5, 6]. Both MAGs harbored genes linked to the pathway, while their proliferation aligned with the observed sulfate reduction (Fig. 1c). Similarly, Myxococcota sp. DTU51, which exhibited a tenfold increase (Fig. 3), has also been reported to perform dissimilatory sulfate reduction [49], and although it lacked genes directly associated with this pathway, it may still be considered a putative SRB. Together, these discrepancies highlight the difficulty in identifying SRB solely from genomic data, likely due to reference database gaps [50], high gene diversity, incomplete MAGs, and low SRB abundance in mixed communities. The presence of magnetite primarily impaired H. formosensis sp. DTU25, which showed a 1.5-fold decrease in relative abundance, and slightly affected Myxococcota sp. DTU51 (onefold decrease), while Sulfurospirillaceae sp. DTU53 remained unaffected.

Out of the 57 MAGs, 14 possessed more than 50% of the genes encoding the WL pathway, supporting CO₂ assimilation for acetate production and vice versa [47, 51]. Several of these species have previously been reported to engage in syntrophic associations with methanogens [46]. Among them, Limnochordia sp. DTU31, DTU18, and DTU38, species previously linked to WL pathway activity [51, 52], showed notable changes. Sulfate addition reduced the relative abundances of Limnochordia sp. DTU31 and DTU38 from 5.1% and 10.3% to 4.8% and 6.5%, respectively, while Limnochordia sp. DTU18 increased from 0.7 to 1.4%. The combination of sulfate and magnetite favored the Limnochordia population, particularly Limnochordia sp. DTU31 and Limnochordia sp. DTU38, which increased to relative abundances of 7.8% and 18.9%. Notably, two additional MAGs, Dethiobacteria sp. DTU36, abundant in all conditions, and Dethiobacteria sp. DTU56, also harbored WL pathway genes. These species have been identified as potential syntrophic acetate-oxidizing bacteria (SAOB) [53], a relevant trait given that SAOB can establish syntrophic interactions, potentially via DIET, with both methanogens and SRB [54].

Magnetite-stimulated methanogenesis and redox cycling

In the CTRL system, the absence of sulfate prevented SRB growth, avoiding interference with methanogenesis. As a result, CH₄ production remained high despite the low relative abundance of the methanogenic population. The presence of Limnochordia species with acetogenic potential likely accounted for acetate accumulation (Fig. 1b). However, the co-occurrence and activity of homoacetogenesis and methanogenesis, competing for common substrates, raised questions about their functional compatibility. Both groups contributed to CO₂ removal, with part of the carbon converted to CH₄ and part to acetate. This partitioning explains the faster decline in CO₂ compared to CH₄ accumulation (Fig. 1, Supplementary material 1). With SAOB also present, a fraction of the produced acetate could then be reconverted into CO₂ and H₂, making both substrates available again for methane production and thereby enabling a gradual increase in CH₄ levels.

Given that sulfate reduction is kinetically and thermodynamically more favorable than methanogenesis, SRB should outcompete methanogens in SULF_0.6, as indicated by sulfate-stimulated SRB activity (Figs. 1c and 3) [7, 10]. Accordingly, CH₄ production declined (Fig. 1a), and a decrease in methanogenic abundance was expected. Instead, methanogenic MAGs increased in relative abundance, an increase that should be interpreted with caution, as it could reflect a stronger decline of other microbial groups under sulfate stress, rather than an increase in absolute methanogenic abundance. This pattern suggests that sulfate suppresses methanogenesis without fully eliminating it, allowing hydrogenotrophic methanogens to persist under sulfate-reducing conditions. SRB are expected to capture a substantial fraction of electron flow by efficiently competing for H₂, thereby limiting both hydrogenotrophic methanogenesis and homoacetogenesis [7, 10]. Nevertheless, some homoacetogenic activity may persist, contributing to acetate accumulation despite competition for H₂. SRB display considerable metabolic flexibility, shifting between H₂ and acetate oxidation depending on substrate availability and energetic constraints [50], whereas hydrogenotrophic methanogens are restricted to H₂ and CO₂. Thus, as H₂ concentrations decrease, acetate becomes an important intermediate. Under such conditions, SRB and, where present, SAOB increasingly oxidize acetate via sulfate reduction and the WL pathway, to CO₂, and in the case of SAOB, limited H₂. Some SRB may also oxidize acetate via the WL pathway [6, 20], further contributing to the CO₂ and H₂ turnover. Meanwhile, hydrogenotrophic methanogens, although requiring a minimum H₂ partial pressure for growth, remain active at comparatively low H₂ levels [55, 56] and may therefore scavenge residual H₂, particularly when sulfate is being depleted by SRB activity and not replenished [57]. These interactions could sustain low but detectable CH₄ production even when methanogenesis is strongly constrained. Additionally, sulfate may support overall microbial growth by supplying essential sulfur for cellular functions [58], indirectly contributing to the increased abundance of methanogens. Nevertheless, the kinetic and thermodynamic advantages of SRB, together with H₂S toxicity, divert electrons and carbon away from methanogenesis, thereby limiting overall CH₄ production (Fig. 1b).

In MAGN_0.6, magnetite likely acted as an electrically conductive medium, accommodating electron transfer between microbial partners, thereby strengthening metabolic cooperation among the four microbial groups. Although SRB showed modestly reduced relative abundances in the presence of magnetite (Fig. 3), sulfate consumption remained unchanged, indicating sustained SRB activity. The lower SRB relative abundance, therefore, likely reflects the relative expansion of other microbial guilds, rather than impaired SRB performance. Magnetite ultimately favored methanogens, which, with access to ample electrons and CO₂, reached their highest relative abundance under this condition. The combined presence of sulfate and magnetite created an environment conducive to methanogenic proliferation, as evidenced by the increased CH₄ yield and daily CH₄ production rate compared to SULF_0.6 (“Magnetite-induced H₂S sequestration alleviated sulfate stress in methanogenic systems” section). Magnetite may have acted as an electron conduit, enabling direct electron transfer and reducing the reliance of hydrogenotrophic methanogenesis on H₂ diffusion. This could accelerate CH₄ production from CO₂, easing competition for shared electron donors and reinforcing metabolic coupling with SRB. At the same time, the removal of H₂S (“Magnetite-induced H₂S sequestration alleviated sulfate stress in methanogenic systems” section) further improved microbial growth, with both methanogenesis and Limnochordia species reaching their most proliferative phases (Fig. 3). Together, magnetite’s presence and acetate recycling by SAOB and SRB maintained efficient substrate turnover and high electron availability within the microbial network, explaining the increased CH₄ production in MAGN_0.6.

Overall, magnetite further promoted a redox cycle between acetate and CO₂ within the system. Acetate was oxidized to CO₂, with the released electrons being transferred via magnetite rather than being used to generate H₂. These electrons could then be reused to reduce CO₂ to either acetate or CH₄. Redox cycles are not uncommon as an energy conversion mechanism in microbial systems involving methanogens and SRB [43, 59]. Likewise, magnetite has previously been reported to facilitate DIET-based cooperative interactions by establishing an electrical network that mediates long-range extracellular electron transfer [60]. Thus, magnetite-mediated electron transfer likely enabled redox cycling between the two carbon forms, supporting the concurrent proliferation of homoacetogens and methanogens, as reflected by their increased relative abundances and higher CH₄ yield. The persistent presence and activity of both SRB and SAOB indicated that multiple guilds benefited from this dynamic. Ultimately, magnetite fostered a cooperative network among these populations, contributing to improved methanogenesis even under elevated sulfate conditions.

Potential interplay between magnetite and biological conductive structures

Metagenomics analysis revealed two MAGs, Halopseudomonas formosensis sp. DTU25 and Alcaligenaceae sp. DTU13, with genes for cytochrome c biosynthesis, suggesting their potential involvement in DIET (Supplementary material 2). In parallel, comparative genomics identified potential species carrying genes for e-pilin. Of the two retrieved protein sequences, only Acholeplasmataceae sp. DTU23 fulfilled all the e-pilin classifications except having a longer sequence length (262 amino acids) and three aromatic-free gaps that exceeded 22 amino acids (Supplementary material 2). In the presence of magnetite, the relative abundance of both Halopseudomonas formosensis sp. DTU25 and Alcaligenaceae sp. DTU13 was reduced (Fig. 3). Beyond possible metabolic limitations due to substrate availability, this reduction could be related to the electron transfer capabilities of magnetite, allowing it to substitute for the multiheme cytochrome in the presence of e-pilins [8, 14], potentially making these species less essential to the microbial network. In contrast, Acholeplasmataceae sp. DTU23 remained unaffected by magnetite, possibly due to magnetite’s capability to associate with e-pilin, enhancing its effectiveness for interspecies electrical connections [8]. In principle, the presence of e-pilins and cytochromes is often associated with DIET [13, 61]. Even if most of the above species were not directly involved in the main carbon cycle and methanogenesis, they may still have contributed to the overall electron flow within the system by acting as electrical conduits for DIET through their conductive structures (Figs. 4 and 5) [8].

Fig. 4.

Schematic representation of the metabolic pathways occurring in a SRB, b methanogens, and c homoacetogens, including relative abundances of the corresponding MAGs, along with heatmaps portraying the copy number of the functional genes associated with each metabolic pathway. Acronyms: glutathione disulfide (GSSG), glutathione (GSH), Phosphoadenosine phosphate (PAP), methanofuran (MFR), Tetrahydromethanopterin (THMPT), reduced cofactor F420 (Red Co F420), oxidized cofactor F420 (Ox Co F420), coenzyme A (CoA), coenzyme B (CoB), coenzyme M (CoM), tetrahydrofolate (THF). Created with BioRender.com, 2025

Fig. 5.

Potential interactions among key microbial groups influencing methanogenesis under the three experimental conditions: CTRL, SULF_0.6, and MAGN_0.6. Panels depict the interactions between methanogens (Mthg), SRB, and their potential syntrophic partners, SAOB and HomoAcetogens (Actg). Cell sizes reflect relative abundances, complemented by bar plots for quantitative comparison. Substrates and products are denoted by names and chemical formulas; dashed arrows indicate uptake, solid arrows indicate production. Created with BioRender.com, 2025

Model-based insights into community interaction networks

To evaluate the feasibility of the hypothesized cooperative interactions, genome-scale metabolic models (GEMs) were reconstructed for four representative community members selected based on high relative abundance and genome quality. Selected members included the methanogen Candidatus Methanoculleus thermohydrogenotrophicum sp. DTU8 (completeness 97.71%, contamination 2.61%), and Halopseudomonas formosensis sp. DTU25 (completeness 93.86%, contamination 2.35%), Limnochordia sp. DTU38 (completeness 89.80%, contamination 4.84%), and Dethiobacteria sp. DTU36 (completeness 90.47%, contamination 2.97%), putative SRB, homoacetogen, and SAOB, respectively. The curated GEMs of the individual species were then integrated into multispecies community networks, one per condition, i.e., for CTRL, SULF_0.6, and MAGN_0.6.

Flux balance analysis (FBA) confirmed the theoretical feasibility of the proposed networks under all conditions, with each species able to grow and perform its target metabolic functions under the imposed constraints while reproducing the emergent behaviors hypothesized from the biochemical and metagenomic data (Fig. 6). Solution-space sampling further identified acetate as a central intermediate, produced and consumed by multiple guilds, with its turnover redistributing carbon and reducing equivalents across the community and thereby influencing methane formation. Specifically, the homoacetogen uptakes H₂ and CO₂, secreting acetate. This acetate is then either oxidized by the SAOB back to H₂ and CO₂, or used by the SRB as an electron donor for sulfate reduction, producing CO₂ and H₂S. In both cases, acetate turnover controls the supply of H₂ and CO₂ to the shared metabolite pool, which feeds hydrogenotrophic methanogens. At elevated sulfate levels, in SULF_0.6 and MAGN_0.6, the solutions further indicated increased sulfate uptake by multiple community members, in line with its proposed role as a sulfur source for biomass synthesis, including amino acids and other sulfur-containing cell components [58]. Together, these model-based predictions support the experimental interpretation that the observed H₂, acetate, and sulfate flux patterns arise from a feasible division of labor within the community.

Fig. 6.

Predicted metabolic exchanges of key extracellular compounds in the three experimental conditions (CTRL, SULF_0.6, and MAGN_0.6). Heatmaps show exchange fluxes for each functional group (methanogen (Mthg), SAOB, homoacetogen (Actg), and SRB) and for the overall system. Compounds correspond to substrates supplied in the medium or major products measured in the reactors. The color scale indicates the direction and magnitude of the predicted fluxes (mmol·gDW⁻¹·h⁻¹), with green denoting uptake, purple secretion, and white indicating negligible net flux

Flux-space sampling of the community model further provided a comprehensive view of feasible interspecies exchange patterns, not easily measurable in vivo (Supplementary material 3), suggesting extensive exchange of amino acids and short-chain organic acids among the four functional groups (Fig. 7). The main amino acids (mean exchange flux > 0.1 mmol·gDW⁻¹·h⁻¹) were L-Alanine, L-Aspartate, L-Cysteine, L-Glutamine, L-Glutamate, Glycine, L-Serine, and L-Threonine (Fig. 7). Among the species, the homoacetogen, the SAOB, and the methanogen act as major hubs that both secrete and consume several of these compounds. The SRB, in turn, was predicted to take up multiple amino acids, suggesting a reliance on external nitrogen and sulfur-containing building blocks to support growth. Notably, all these amino acids are glucogenic, feeding into pyruvate or other central carbon intermediates, and so their frequent exchange suggests that their interspecies transfer could funnel carbon into central metabolism and support the synthesis of more energetically expensive biomass precursors in partner species [62]. At SULF_0.6 and MAGN_0.6, where additional sulfate is present in the system, the model predicts increased production and uptake of L-Cysteine, indicating a higher demand for reduced sulfur for biomass formation, consistent with the observed increased sulfate uptake by several community members (Fig. 6). Other short-chain carboxylates and organic acids, including formate, succinate, pyruvate, lactate, citrate, fumarate, malate, and propionate, were also predicted to be actively exchanged (Fig. 7). These compounds could act as complementary electron and carbon carriers that buffer flux dynamics and help maintain methanogenic activity even under sulfate stress, as shown by the shifts in the metabolic interactions network after the addition of sulfate. Overall, the model indicates that amino acids and carboxylic acids may serve as important mediators of interspecies metabolic interactions, offering multiple alternative routes to sustain co-existence.

Fig. 7.

Predicted cross-feeding of metabolic intermediates among community members in the three experimental conditions (CTRL, SULF_0.6, MAGN_0.6). Heatmaps depict exchange fluxes of amino acids and organic acids with mean fluxes > 0.1 mmol·gDW⁻¹·h⁻¹ between the methanogen (Mthg), SAOB, homoacetogen (Actg), and SRB, internally produced and consumed within the community. The color scale indicates the direction and magnitude of the predicted fluxes (mmol·gDW⁻¹·h⁻¹), with green denoting uptake, purple secretion, and white indicating negligible net flux

Discussion

Sulfate-rich environments are a persistent challenge in anaerobic digestion, as sulfate reduction diverts electron flow away from methanogenesis while generating H₂S, further inhibiting key community members and destabilizing syntrophic interactions. Finding strategies to alleviate sulfate stress is therefore critical for improving the efficiency and robustness of biogas production systems. In this study, we observed that magnetite supplementation can partially overcome sulfate-induced inhibition. Methane production under moderate sulfate conditions (0.6 g/L) was increased in the presence of magnetite, accompanied by sustained activity of methanogens, sulfate reducers, syntrophic acetate-oxidizers, and homoacetogens. Magnetite addition was associated with enhanced daily CH₄ production, reflecting accelerated methanogenesis driven by the reduced reliance on slower H₂ diffusion and partially relieved competition for diffusible electron donors that followed the increased electron availability within the community. Magnetite did not impede sulfate reduction but was associated with microbial cooperation compatible with DIET-like interactions and redox cycling between acetate and CO₂, while alleviating H₂S-associated constraints. Direct validation of DIET requires additional targeted experimental approaches beyond the scope of this study. Still, its outcomes suggest that even under competitive and potentially inhibitory conditions, magnetite can stabilize and enhance methanogenic function.

The insights from this work extend beyond the immediate experimental system. Low-abundance methanogens can persist and remain functionally relevant when electron transfer is facilitated, highlighting a mechanism for resilience in complex microbial networks. Material-mediated electron transfer provides an additional tool for managing microbial interactions, supporting syntrophic cooperation, and promoting efficient substrate turnover. Furthermore, the genome-scale metabolic models introduced in this study provide a platform for hypothesis testing and prediction of community-level fluxes. With further experimental data, these models can support the mechanistic evaluation of interspecies exchange and guide strategies to optimize microbial networks in both laboratory and applied settings. The emergence of acetate and amino acids as central intermediates further emphasizes their role as key currencies for maintaining energy and carbon flow under sulfate stress. These findings have practical implications for AD and related bioenergy applications. CMs such as magnetite could be incorporated into sulfate-impacted digesters to mitigate H₂S-related toxicity and increase methane yield, while the concept of using magnetite for the modulation of microbial electron exchange may be relevant to other engineered anaerobic ecosystems and bioelectrochemical applications where metabolic cooperation is critical for performance.

At the same time, there are important limitations to consider. Experiments were performed under controlled batch conditions, primarily to provide mechanistic insight into magnetite-microbe interactions, which may not fully reflect continuous or industrial-scale reactors where long-term dynamics, operational stability, material-related effects, and spatial gradients can influence outcomes. Moreover, high sulfate concentrations (1.2 g/L) still constrained magnetite’s effectiveness at the tested dose, indicating that additional measures may be needed under extreme sulfate loads. Furthermore, direct measurements of electron transfer processes, along with assessment of long-term reactor performance, including continuous-flow stability and magnetite retention or recovery, will be necessary to translate these findings into operational practice.

When it comes to genome-scale metabolic modeling, its application provides an alternative framework to interpret genome annotation, where instead of analyzing reactions individually, metabolism is represented as a connected, mass-balanced network. The model enables quantitative predictions in the form of optimal yields, rather than only qualitative functional assignments derived from annotation. However, such models represent the theoretical maximum metabolic potential under the assumption that only metabolic reactions operate without additional biological constraints. In reality, microbial activity is influenced by multiple factors, including signalling, regulation, protein limitation, and kinetic constraints. Moreover, such models inherently involve uncertainty, as there is not a single unique solution but rather a large feasible solution space that can satisfy the same constraints and objective, meaning that multiple flux distributions can be consistent with the model predictions. To address this, flux sampling was applied in the present work allowing to explore the solution space. Finally, model predictions can be further refined by incorporating additional data types, such as transcriptomic, proteomic, or metabolomic information, to better constrain flux distributions to physiologically relevant states.

Overall, this study highlights the potential of magnetite to both alleviate sulfate-associated stress and support microbial cooperation that sustain methanogenic function, providing mechanistic insights and practical strategies that can improve AD. By linking experimental observations with genome-centric analysis and genome-scale metabolic models, our work offers tools that can guide future research and applied interventions. Further experiments, ideally integrated with multi-omics measurements such as metatranscriptomics and metabolomics, could help link predicted exchanges to expressed functions and in situ fluxes, enabling optimization of microbial networks, improved prediction of system performance, and the design of more resilient, efficient, and controllable bioenergy systems. Simple material amendments, together with predictive modeling, can thus have profound effects on microbial network function and ecosystem performance.

Conclusions

This study demonstrates that magnetite addition enhances methanogenic performance in sulfate-rich anoxic systems through combined chemical and community-level modulations. Magnetite reduced dissolved sulfide species accumulation via precipitation and increased methane production by 19% at 0.6 g/L sulfate addition without limiting sulfate utilization. Genome-centric analysis associated magnetite with increased growth and metabolic activity of methanogens and homoacetogens while retaining SRB and SAOB functionality. This shift coincided with a more strongly connected cooperative syntrophic network, consistent with more efficient carbon and electron flow even under elevated sulfate conditions. Serving as an electron conduit facilitating direct electron transfer-based interactions, magnetite likely reshaped community structure and function in favor of methanogenesis. Combined sulfate and magnetite addition led to greater methanogenic enrichment (from 3.5 to 5.4%), indicating that SRB and methanogens can engage in cooperative rather than strictly competitive interactions when magnetite is present. Community GEMs supported the feasibility of the proposed network and provided insight into stoichiometrically plausible interspecies intermediate exchange of amino acids and organic acids, further strengthening the evidence for cooperative interactions in the community. These findings highlighted magnetite’s dual function in steering microbial metabolism towards higher biogas yields and reducing corrosive gas formation in AD systems. Future research should explore the long-term stability and scalability of these effects, as well as magnetite recovery strategies, and evaluate the enrichment or engineering of magnetite-responsive microbial consortia.

Authors’ contributions

G.V., M.G., P.G.I., P.G.K., and I.A. contributed to the conceptualization of the study. G.V., G.G., M.G., P.G.I., P.T., and P.G.K. developed the methodology. G.V. developed the software. G.V. and M.G. performed validation. G.V., G.G., M.G., and P.T. conducted the formal analysis. G.V., M.G., and P.G.I. carried out the investigation. G.V. and G.G. curated the data and prepared the visualizations. G.V. and G.G. wrote the original draft. G.V., G.G., P.G.I., P.T., P.G.K., and I.A. reviewed and edited the manuscript. P.G.I., P.T., P.G.K., and I.A. provided supervision. I.A. provided resources, administered the project, and acquired funding. All authors reviewed and approved the final manuscript.

Funding

Funding for this work was provided by the European Research Council (ERC Advanced Grant 101098064-ANAEROB), Villum Investigator programme (Grant 00073410-BioDesign), and the Novo Nordisk Foundation (Grant NNF21OC007139-DIET-Syntrophy).

Data availability

The datasets supporting the conclusions of this article are included within the article and its additional files. Sequence data have been submitted to the National Centre for Biotechnology Information (NCBI) under BioProject PRJNA1242289.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.