The surface soil samples were collected on 25 July 2012 from an open fen close to the Wetland National Nature Reserve of Zoige located in Qinghai-Tibetan Plateau (33°47^′ N, 102°57^′ E) (Fu et al., 2015). Enrichment cultivation was conducted in the same way as previously described (Fu et al., 2018). Enrichment incubation was initiated by inoculating 4% (v/v) pre-incubated soil slurry (21 days) into 60-mL vessels containing 25 mL of Hepes-buffered (30 mM, pH 7.0) fresh medium under a headspace of N₂/CO₂ (80/20). The basal medium contained MgCl₂⋅6H₂O (0.4 g L^-1), CaCl₂⋅H₂O (0.1 g L^-1), NH₄Cl (0.1 g L^-1), KH₂PO₄ (0.2 g L^-1), KCl (0.5 g L^-1), and resazurin (0.0005 g L^-1), and was supplemented with Na₂S⋅9H₂O (1.0 mM), vitamin and trace element solutions as described previously (Lü and Lu, 2012). Sodium acetate was added to a final concentration of 5 mM in the initial four transfers and then increased to 10 mM thereafter. Cysteine was not added to avoid the possible effect of electron shuttle molecules. NanoFe₃O₄ were synthesized as described previously (Kang et al., 1996). The first transfer was inoculated from the pre-incubated soil slurry, the effect of nanoFe₃O₄ concentration (2.32, 4.64, and 6.96 mM of Fe in the medium) was determined. Continuous transfers were conducted in the presence of nanoFe₃O₄ (4.64 mM of Fe in the medium). The inocula for every transfer were taken from the later nanoFe₃O₄-amended cultivation. For a comparison, the same inocula were used to make parallel preparations without nanoFe₃O₄ in the medium (i.e., the control).

The final cultivation (after thirteen transfers) was used to extract microbial DNA following the previous protocol (Ma et al., 2012). Briefly, 10 mL of enrichment was extracted sequentially with TPMS buffer [50 mM Tris–HCl (pH 7.0), 1.7% (wt/vol) polyvinylpyrrolidone K25, 20 mM MgCl₂, 1% (wt/vol) sodium dodecyl sulfate] and phenol-based lysis (PBL) buffer [5 mM Tris–HCl (pH 7.0), 5 mM Na₂EDTA, 1% (wt/vol) sodium dodecyl sulfate, 6% (vol/vol) water-saturated phenol]. Beads-beating was performed in FastPrep-24 (MP Biomedicals, United States). The supernatants were further extracted with water-saturated phenol, phenol-chloroform-isoamyl alcohol [25:24:1 (vol/vol/vol)], and chloroform-isoamyl alcohol [24:1 (vol/vol)]. The extracts were purified by cold ethanol and sodium acetate.

DNA samples from both the control and nanoFe₃O₄ treatment were used to construct bacterial and archaeal clone libraries. The PCR amplification, cloning and sequencing followed the previous procedure (Kumar et al., 2016). Phylogenetic trees were constructed using the neighbor-joining algorithm of the MEGA7 program (Kumar et al., 2016), and bootstrap analysis was implemented with 1000 replicates. DNA-SIP was performed using the same cultivation. For this purpose, the sodium acetate-2-¹³C (99 atom%; Sigma-Aldrich) was added as the substrate. At the end of incubation, the carbon isotopic ratios (δ¹³C values) of CH₄ and CO₂ were analyzed by a gas chromatography-isotope ratio mass spectrometry system (Fu et al., 2018). DNA was extracted from the ¹³C-labeled and non-labeled cultivations and subjected to DNA-SIP procedure through the isopycnic centrifugation and density gradient fractionation of DNA as described previously (Rui et al., 2011). Centrifugation medium was prepared by mixing cesium trifluoroacetate (CsTFA) (Amersham Pharmacia Biotech) with gradient buffer (0.1 M Tris–HCl, pH 8; 0.1 M KCl; 1 mM EDTA). The mixtures were centrifuged in a Ti90 vertical rotor (Beckman) at 177000 g, 20°C for >36 h using Beckman Optima 2-80XP Ultracentrifuge (Beckman Coulter, United States). The density-resolved DNA was fractionated, and the buoyant density of each fraction was determined by refractometer.

The density-resolved DNA gradients were quantified for total bacteria and archaea using real-time quantitative PCR (Gan et al., 2012). Quantitative PCR of archaeal and bacterial 16S rRNA genes were carried out in a 7500 real-time PCR system (Applied Biosystems) using the primer pair Ar364/Ar934, and Ba519f/Ba907r, respectively. The fingerprinting of the DNA gradients was conducted using the terminal restriction fragment length polymorphism analysis (T-RFLP) following the protocol described previously (Liu et al., 2011). PCR amplification was performed using the primer pairs of Ba27f/Ba907r for bacteria and Ar109f/Ar934r for archaea. The 5^′ end of the Ba27f and Ar934r primers were labeled with 6-carboxyfluorescein (FAM). PCR products were purified using an agarose gel DNA extraction kit (TaKaRa) and digested with Msp I (Takara) for bacteria and Taq I (Takara) for archaea, respectively.

Methanosarcina barkeri (DSM800) were purchased from German culture collection DSMZ (Braunschweig, Germany). The basal medium was consistent with that described in the previous enrichment culture experiment. Sodium acetate was added to a final concentration of 10 mM. The effect of nanoFe₃O₄ was tested for pure culture strains.

The redox activity was analyzed by cyclic voltammetry (CV) and electrochemical impedance spectroscopy (EIS) using the CHI660 as described previously (Zhou et al., 2018). Graphite plates (1.0 cm × 1.5 cm) and saturated calomel reference electrodes (SCE) were used as the working electrode, counter electrode and reference electrode, respectively. The CV scan parameters were as follows: initial potential was 0.6 V vs. SCE, maximum potential was 0.6 V vs. SCE, minimum potential was -0.8 V vs. SCE, termination potential was 0.6 V vs. SCE, and the scan rate was 0.015 V/s. The EIS scan had a high frequency value of 100000 Hz, a low frequency value of 0.1 Hz, and an amplitude of 0.005 V. In the electrochemical impedance test, the protoplast part of the biofilm was considered to be insulated under low frequency electrical perturbations, and the biofilm was considered as a multi-element circuit.

For XRD analysis, the culture was centrifuged (4000 g, 2 min). Then the supernatant was removed, and the solids were lyophilized in vacuum for 48 h. The dried sample was analyzed by a D/MAX-2400 X-ray diffractometer. The monochromatic Cu target Kα was irradiated with a current of 80 mA and a voltage of 40 kV. The acquisition range is 3–70° with an accuracy of 1.2°/min and a time of 6 s (Li et al., 2015). For Raman analysis, the centrifuged sample was air-dried under nitrogen protection until no significant aqueous layer was visible. Mineral structure and composition were analyzed using a LabRam HR800 laser confocal micro-Raman spectrometer. The excitation wavelength is 532 nm, the grating is 600 gr/mm, the objective magnification is 50 times, and the focused spot size is about 1.5 μm. The entire process is carried out under nitrogen protection.

For scanning electron microscope (SEM) observation, cells were carefully harvested without stir. Following centrifugation (4000 g, 2 min), the supernatant was removed and the pellet was fixed with 2.5% (wt/vol) glutaraldehyde, dehydrated using a graded series of ethanol solutions, and dried with t-butanol. The samples were mounted on copper stubs, coated with platinum, and then imagined using JEOL S-4800 (JEOL, Japan). For transmission electron microscope (TEM) analysis, the fixed cells were sent to Beijing Zhongkebaice Technology Service Co., Ltd. to make ultrathin sections, and the morphology and structure of the samples were observed by G2F20 (200 KV) for the regions or particles of interest.

Gas samples (0.1 mL) were regularly taken from headspace of incubations with a pressure-lock precision analytical syringe (Baton Rouge, LA, United States). The concentrations of CH₄ and CO₂ were analyzed using gas chromatographs GC-7890 (Agilent Technologies, United States) equipped with a thermal conductivity detector (Fu et al., 2018). Liquid samples (0.5 mL) were taken with sterile syringes and centrifuged for 15 min at 17,949 ×g at 4°C. The supernatant was collected, passed through 0.22-μm-pore-size filters, and analyzed for the concentrations of acetate and butyrate with an HPLC-1200 using a Zorbax SB-AQ C₁₈ column (Agilent Technologies, United States) (Fu et al., 2018). Redox potential (ORP) were measured with an Unisense redox microelectrode RD 100 (Unisense, Denmark).

The production of CH₄ occurred without lag in the first transfer indicating the ready activity of acetate oxidation in this wetland sediment (Figure 1A). Addition of 5 mM acetate yielded about 3.78 ± 0.13 mM CH₄ (normalized to liquid volume) in the first transfer (Figure 1A). When the concentration of acetate was increased to 10 mM, about 8.13 ± 0.12 mM of CH₄ was obtained in the later transfers (Figure 1B–D). These results indicated that CH₄ accumulated in the headspace corresponded to 80% of theoretical stoichiometric prediction of complete conversion of acetate to CH₄ and CO₂. Addition of nanoFe₃O₄ did not influence this conversion efficiency. However, nanoFe₃O₄ significantly accelerated the CH₄ production rate, with shorter lags and greater maximal rates compared with the control. For the 4th transfer, CH₄ production displayed a long lag in the control while it took less than a week before the onset of rapid production in the presence of nanoFe₃O₄ (Figure 1B–D).

Isotopic and chemical analysis was conducted during the labeling experiment at the 14th transfer. Incubations with or without labeling showed identical patterns of acetate consumption and CH₄ production (Figure 2A without and Figure 2C with isotopic labeling). In consistence with the early transfer incubation experiments, the addition of 10 mM acetate produced about 8.1–8.9 mM CH₄ (Figure 2A,C). When the sodium acetate-2-¹³C (99 atom%; Sigma-Aldrich) was added as the substrate, the carbon isotopic ratios (δ¹³C values) of CH₄ increased rapidly in the presence of nanoFe₃O₄, peaked at 40 ± 2.5% on the 9th day of culture (Figure 2D). However, the δ¹³C values of CO₂ for the control (1.3 ± 0.04%) only slightly increased (Figure 2B). These results indicated that CH₄ was mainly produced from direct acetate cleavage. In the direct pathway, acetotrophic methanogens like Methanosarcina and Methanosaeta convert acetate into CH₄ and CO₂ (¹³CH₃COOH →¹³CH₄ + CO₂). However, in the syntrophic pathway, acetate is oxidized to H₂ and CO₂ (¹³CH₃COOH + 2H₂O → 4H₂ + ¹³CO₂ + ¹²CO₂). H₂ is consumed by hydrogenotrophic methanogens to generate CH₄ (4H₂ + ^12/13CO₂ →^12/13CH₄ + 2H₂O). If the system is dominated by the syntrophic pathway, considering that the headspace of the serum bottle was filled with N₂:CO₂ (80:20), the carbon isotopic ratios (δ¹³C values) of CO₂ will increase, and the carbon isotopic ratios (δ¹³C values) of CH₄ should be even lower than the carbon isotopic ratios (δ¹³C values) of CO₂.

Scanning electron micrograph showed the aggregated cells in the control (Figure 3A) at the 14th transfer. With the addition of nanoFe₃O₄, the shell-like coat formed on microbial aggregates with nanoFe₃O_4-coated surfaces (Figure 3B).

DNA-SIP, T-RFLP and clone sequence analyses were used to determine microbial composition of the enrichment cultures. DNA-SIP was performed by applying ¹³C-labeled acetate (¹³CH₃COONa). Almost identical pattern was observed in the distribution of the density-resolved DNA fragments along the buoyant density gradient for the control (Figure 4A,C) and the nanoFe₃O₄ treatment (Figure 4B,D). The distribution of the archaeal (Figure 4A,B) and bacterial (Figure 4C,D) DNA shifted to the heavier fractions in the labeled samples compared with the non-labeled control (Figure 4). These results indicated that both the archaeal and bacterial populations assimilated ¹³C-labeled acetate.

Three major T-RFs were detected in the archaeal T-RFLP fingerprints (Figure 5A,B) of the density-resolved DNA. The 184 bp T-RF was predominant in all T-RFLP profiles, and its relative abundance was reduced only in heavier layers. Analysis of the clone sequences indicated that these T-RFs belonged to Methanosarcina (Figure 6A,B). The T-RFs detected in the bacterial T-RFLP fingerprints include: 167 bp, representing Anaerovorax; 182 bp, representing Syntrophomonas; 284 bp, representing Veillonellaceae; 305 and 470 bp, both representing Clostridium; 363 bp, representing Cytophaga; 430 bp, representing Azonexus. The 430 bp was predominant in all experimental treatments (Figure 5C,D). For the control without nanoFe₃O₄, 430 bp showed a trend of increasing first and then decreasing with the DNA buoyant density (Figure 5C); the 305 and 167 bp was secondary dominated in the non-labeling and labeling treatments, respectively. For the nanoFe₃O₄ treatments, the 430 bp T-RF was predominant across density gradients; other T-RFs either declined or did not change with the DNA buoyant density (Figure 5D).

Two bacterial and two archaeal clone libraries were constructed, with one each for the control and nanoFe₃O₄ treatment, respectively. All the archaeal clone sequences from both the nanoFe₃O₄ treatment and the control were affiliated to the Methanosarcinales order, with Methanosarcina barkri as the closest pure culture relative (Figure 6A,B). Clone sequences indicated that the bacterial communities in the enrichments consisted mainly of Syntrophomonas, Veillonellaceae, Anaerovorax, Clostridium, Cytophaga, Azonexus, Desulfovibrionaceae, and Saccharofermentans acetigenes (Figure 6C,D). The sequences affiliated to Azonexus accounted for approximately 59% of total sequences in the nanoFe₃O₄ library (Figure 6D). By comparison, Clostridium accounted for 49%, followed by Veillonellaceae (15%), Anaerovorax (10%) and Azonexus (10%) in the control library (Figure 6C).

Pure culture systems of M. barkeri were constructed with acetate as the sole substrate. Various tests were carried out to investigate the effect of nanoFe₃O₄ on the methangenesis of pure culture strains. NanoFe₃O₄ significantly promoted the production of methane and the consumption of acetate by M. barkeri (Figure 7A). The maximum methane production rate (V_max) was 6.15 ± 0.34 mmol/L⋅d in the presence of nanoFe₃O₄, which was about two times that of the control (2.91 ± 0.21 mmol/L⋅d). This rapid methanogenesis process occurred 10 days earlier than the control. However, the addition of nano-graphite (particle diameter: 35 nm) did not show any promoting effect. On the other hand, the addition of CNTs (outside diameter: 10–20 nm, length: 10–30 μm) retained a stimulatory effect that was less significant compared with nanoFe₃O₄ (Figure 7F).

In order to clarify the action sites of nanoFe₃O₄ on M. barkeri, we performed ultrathin sections of M. barkeri and TEM analysis. The results showed that the nanoFe₃O₄ existed in extracellular and intercellular spaces, cell surfaces, cell membranes, and the cytoplasms (Figure 7B). XRD was used to determine the possible structural changes of nanoFe₃O₄ during the culture process. The XRD pattern data showed no change in the crystal structure of nanoFe₃O₄ after culture (Figure 7C). The structure and composition of nanoFe₃O₄ were further analyzed with Raman spectroscopy on a microscopic scale. The Raman spectrum detected the production of hematite (compared to a Raman spectrum of hematite reported in the RRUFF data base¹) after the incubation with M. barkeri (Figure 7D).

After the addition of nanoFe₃O₄, the oxidation-reduction potential (ORP) of the culture system was increased from -350 to -300 mV, but did not continue to increase as the concentration of nanoFe₃O₄ increased (Figure 7E). We performed a cyclic voltammogram (CV) scan of all pure culture methanogenic systems to analyze the redox activity changes of the systems. CV can reflect the redox activity and capacitance in the electrochemical reactor. The larger the capacitance, the more charge is stored in the system. As shown in Figure 8A–D, the capacitance of the bioreactors with nanoFe₃O₄ was higher than that of the control reactors, suggesting an enhanced redox activity. The capacitance of the bioreactors with nanoFe₃O₄ increased to the maximum on day 10 (Figure 8B) and then gradually decreased to the minimum on day 30 (Figure 8D), which was consistent with the methanogenic activity. By contrast, the capacitance of the control bioreactors without nanoFe₃O₄ kept relatively stable during the 30-days cultivation. Moreover, a pair of oxidation and reduction peaks with a mid-point potential of approximately -200 mV vs. SHE appeared in the CV scans for both treatments. However, the peaks disappeared for the bioreactors with nanoFe₃O₄ on day 30. This indicated the redox transformation of certain redox-active species in the bioreactors with nanoFe₃O₄. EIS was used to evaluate the electron transfer resistance of the reactors, which was crucial for the charge transfer efficiency. As shown in Figure 8E–H, the addition of nanoFe₃O₄ significantly decreased the electron transfer resistance (R_ct = 37.7 Ω) compared with the control reactors (R_ct = 353.5 Ω) at day 0. Although the electron transfer resistances of the control reactors decreased to 68.9 Ω after 30-days cultivation it was still higher than that for the nanoFe₃O₄ treatment (27.6 Ω). The decreasing R_ct over time might be due to the adsorption of M. barkeri on the electrode surfaces.

Previous literatures have reported that conductive materials such as nanoFe₃O₄, graphite, CNTs, activated carbon, etc., can promote the methane production via mediating DIET. However, few studies have focused on the effects of these materials on methanogens themselves. This work evaluated the effects of nanoFe₃O₄ in enrichment cultures initiated with wetland soils with acetate as a substrate. Compared with the control, the addition of nanoFe₃O₄ consistently shortened the lag period and enhanced the maximum rate of CH₄ production. Highly enriched methanogens were obtained through continuous transfers in the presence of nanoFe₃O₄. Mass balance and isotopic labeling indicated that the conversion of acetate conformed to the formula: CH₃COOH → CH₄ + CO₂. Molecular analyses revealed that Methanosarcina closely related to a M. bakeri strain were left as the only methanogen in the enrichment. These results suggested that nanoFe₃O₄ played an important role in the acetotrophic methanogenesis (Figure 9). Such a promoting effect of nanoFe₃O₄ was further confirmed in pure culture of M. bakeri.

Different mechanisms may be involved in the promoting effect of nanoFe₃O₄/M. bakeri system. Firstly, nanoFe₃O₄ has a relatively low redox potential (-314 mV) (Straub et al., 2001). It has been argued that the stimulatory effect by nanomaterials like CNTs on syntrophic coculture and pure culture of methanogens is caused by the decrease in redox potential (Salvador et al., 2017). Similar effect may be postulated for nanoFe₃O₄. However, the experimental results showed that the ORP of the system increased about 50 mV after the addition of nanoFe₃O₄ (Figure 7E). Therefore, this inference was untenable here.

Our previous research suggested that the electrical conductivity of nanomaterials played the key role in promoting the syntrophic oxidation of butyrate (Li et al., 2015; Fu et al., 2018). However, the results of this study showed that graphite and CNTs with higher conductivity did not have the similar promoting effect like nanoFe₃O₄ (Figure 7F). Apart from the common property in electric conductivity, nanoFe₃O₄, CNTs and graphite are chemically and physically different. It was probable that a certain special property of nanoFe₃O₄ rather than the conductivity played a decisive role.

The system’s CV scan results showed that the redox activity was highly consistent with the methanogenic activity. The addition of nanoFe₃O₄ significantly increased the redox activity of the system (Figure 8A–D). Since soluble electron shuttle molecule was not present in our system, the CV measurement could reflect the activities of the enzymes that catalyzed the electron transport in the membrane. In other words, the higher catalytic currents could be an indicator of a better redox enzyme activity. As shown in Figure 8A–D, the catalytic currents for the nanoFe₃O₄ treatment increased in the first 10 days and then gradually decreased to the background at day 30. This means that the redox enzyme activities in the membrane were the highest in the middle period of the experiments. The detection of hematite by Raman further indicated that nanoFe₃O₄ did undergo a redox transformation on the microcosmic scale (Figure 7D). Based on this, we can infer that nanoFe₃O₄ participate in the electron transfer on the membrane of M. bakeri.

Acetate is the worst substrate for methanogenesis with a standard free energy change of only -36 kJ mol^-1. It can be only used by cytochrome-containing methanogens such as Methanosarcina acetivorans, Methanosarcina mazei, and Methanosarcina barkeri (Thauer et al., 2008; Welte and Deppenmeier, 2014). To conserve the little energy available as much as possible, acetotrophic methanogens have evolved a sophisticated metabolic pathway (Figure 9). In the pathway, CH₃COOH reacts with sulfhydryl-containing coenzyme A (SH-CoA) to form CH₃-CO-S-CoA. This is an endergonic reaction. Then, the carboxyl group combines with the oxidized ferredoxin (Fd_ox) to form a reduced ferredoxin (Fd_red), and releases CO₂ and regenerates SH-CoA. The methyl moiety combines with tetrahydromethanopterin (H₄MPT) to form CH₃-H4MPT, which reacts with sulfhydryl-containing coenzyme M (HS-CoM) to form CH₃-S-MPT and H₄MPT. Then CH₃-S-MPT reacts with sulfhydryl-containing coenzyme B (HS-CoB) to form CH₄ and heterodisulfide of CoM and CoB (CoM-S-S-CoB) (Costa and Leigh, 2014; Welte and Deppenmeier, 2014). In Methanosarcina species, Fd_red is used to produce H₂ by ferredoxin-dependent hydrogenase (Ech). The H₂ diffuses across the cell membrane and is oxidized by methanophenazine-dependent hydrogenase (Vht) to reduce a membrane-bound methanophenazine (Mph). The reduced Mph delivers electrons to the membrane-bound heterodisulfide reductase (HdrDE) to regenerate free CoM and CoB from the CoM-S-S-CoB (Kulkarni et al., 2018; Lovley, 2018; Mand et al., 2018).

Mph is a unique membrane electron carrier found in Methanosarcina species (Beifuss et al., 2000). The Mph-catalyzed step is rate limiting in the central metabolism of acetate. Mph has a lower redox potential (-150 mV). The content of Mph in M. acetivorans were threefold higher than that in M. barkeri (Duszenko and Buan, 2017). This suggests the cell membrane of M. acetivorans is electrically quantized as if it were a single conductive metal sheet and near optimal for electron transport (Duszenko and Buan, 2017). Therefore, in theory, by adding an electron carrier with a function similar to Mph to the medium, the acetotrophic methanogenic rate of M. barkeri can be increased. Beckmann et al. have demonstrated that soluble neutral red (-375 mV) delivers reducing equivalents directly to the membrane bound HdrED of Methanosarcina species, and thus increases the rates of proton translocation and regeneration of the methanogenic cofactors CoM-SH and CoB-SH (Beckmann et al., 2016). This was the first report that an artificial electron shuttle can mimic the membrane integrated electron shuttle Mph.

Although nanoFe₃O₄ (-314 mV) existed in solid state, the results of ultrathin sections proved that nanoFe₃O₄ passed through the cell membrane of M. barkeri, so nanoFe₃O₄ could function as electron shuttles like soluble neutral red. It was worth mentioning that nanoFe₃O₄ did not show any effect on CH₄ production by two hydrogenotrophic methanogens Methanococcus maripaludis and Methanocella conradii (Fu et al., 2018). The most essential difference between them and M. barkeri was the lack of electron transport chain on the membrane (Thauer et al., 2008; Costa and Leigh, 2014; Welte and Deppenmeier, 2014). Based on the characteristics and performance of nanoFe₃O₄, we propose the hypothesis that nanoFe₃O₄ may act as solid-state electron shuttles that performs a similar function like Mph. Here we also emphasize that this may not be the only mechanism by which nanoFe₃O₄ stimulates methane production. Other unknown biochemical activities may influence methanogenesis. Our experimental evidence has clarified that a large amount of nanoFe₃O₄ passes through the cell membrane and enters the cytoplasm. What changes may occur in this part of the nanoFe₃O₄, what reactions they may participate in, and what kind of effects on the organism, we don’t know anything about these issues and deserve further study.