The possible occurrence of iron-dependent anaerobic methane oxidation in an Archean Ocean analogue

In the ferruginous and anoxic early Earth oceans, photoferrotrophy drove most of the biological production before the advent of oxygenic photosynthesis, but its association with ferric iron (Fe3+) dependent anaerobic methane (CH4) oxidation (AOM) has been poorly investigated. We studied AOM in Kabuno Bay, a modern analogue to the Archean Ocean (anoxic bottom waters and dissolved Fe concentrations > 600 µmol L−1). Aerobic and anaerobic CH4 oxidation rates up to 0.12 ± 0.03 and 51 ± 1 µmol L−1 d−1, respectively, were put in evidence. In the Fe oxidation–reduction zone, we observed high concentration of Bacteriochlorophyll e (biomarker of the anoxygenic photoautotrophs), which co-occurred with the maximum CH4 oxidation peaks, and a high abundance of Candidatus Methanoperedens, which can couple AOM to Fe3+ reduction. In addition, comparison of measured CH4 oxidation rates with electron acceptor fluxes suggest that AOM could mainly rely on Fe3+ produced by photoferrotrophs. Further experiments specifically targeted to investigate the interactions between photoferrotrophs and AOM would be of considerable interest. Indeed, ferric Fe3+-driven AOM has been poorly envisaged as a possible metabolic process in the Archean ocean, but this can potentially change the conceptualization and modelling of metabolic and geochemical processes controlling climate conditions in the Early Earth.


Material and methods
Description of the study site and the sampling device. Kabuno Bay (− 1.6216°N, 29.0497°E; Figure S1) was sampled in May 2013 (late rainy season), September 2013 (dry season) and August 2014 (dry season) 15 . Vertical profiles of temperature, conductivity, pH and oxygen were obtained with a Yellow Springs Instrument (YSI) 6600 V2 multiparameter probe, with a detection limit for dissolved oxygen of 0.01 mg L −1 . High amounts of dissolved gases (in particular CO 2 ) were present in superficial and deep waters of Kabuno Bay, causing losses of CH 4 when samples were brought to the surface. To avoid that, a home-made sampler (Figure S2) was used; sealed N 2 -flushed 60 mL glass serum bottles (SUPELCO, Sigma Aldrich, 33109-U) were fixed on a two-meter high plate, every 0.25 m. Thin needles (0.6 × 25 mm) equipped with non-return valves (valves allowing the water to fill in the bottles but preventing gases to escape from the bottles) penetrated the grey butyl stoppers (WHEATON, USA). The non-return valves were sealed by a butyl stopper. A string was connected to stoppers in series (all stoppers were connected to the same string). The device was immersed at the sampling depth, and the stoppers were removed from the non-return valves by pulling the string, allowing water to enter the bottles through the needle. The system was left under water 10 min to fill the serum bottles. Once the sampling device was brought back to the surface, the needles were removed from the butyl stoppers and further processed as described below. Serum bottles were half-filled with water, and the other half was a N 2 headspace.
The five main rivers of Kabuno Bay ( Figure S1) were sampled every month from November 2013 to June 2014, by the mean of a Niskin bottle. Samples for total Fe and Mn concentrations determination were taken in plastic vials, stored at 4 °C and analyzed as described hereafter.
Chemical analyses. Samples for CH 4 concentrations were collected in sealed (with butyl stoppers previously boiled in milli-Q water in the laboratory, and aluminium caps) N 2 -flushed 60 mL glass serum bottles, as described above. Two bottles were directly poisoned with 100 µL of HgCl 2 . CH 4 concentrations were determined via the headspace equilibration technique and measured by gas chromatography (GC) 16 , as described by Borges et al. 1 . The precision of measurements was ± 3.9% and the detection limit of the method is 0.5 nmol L −1 .
Samples for nutrient analyses were collected into 250 mL borosilicate bottles, using the same sampling device as for the gas sampling. Water was then collected from the bottles with a 50 mL-syringe, filtered through a 0.22 µm syringe filter (polyethylsulfone), preserved with 200 µL of H 2 SO 4 5 N, and stored frozen. Nitrite (NO 2 − ) and NO 3 − concentrations were measured by spectrophotometry, by the sulfanilamide method 17 and the vanadium reduction to NO 2 − method 18 , respectively, while ammonia (NH 4 + ) was determined with the dichloroisocyanurate-salicylate-nitroprussiate colorimetric method 19 . NO 2 − and NH 4 + were quantified on a Thermo Spectronic Genesys 10vis spectrophotometer using a 5-cm light path, and NO 3 was determined with a Multiskan Ascent Thermo Scientific multi-well plate reader. The detection limits for these methods were 0.03, 0.15 and 0.3 µmol L −1 for NO 2 − , NO 3 − and NH 4 + , respectively. The concentrations for NO 3 − and NO 2 − are reported here as NO x concentrations (NO 3 − + NO 2 − ). Nutrients concentrations are not available in May 2013, due to a problem during samples preservation.
Samples for SO 4 2− and sulfide (HS − ) concentrations were collected in N 2 -flushed 60 mL serum bottles, by the same sampling method as described above. Water was rapidly filtered after collection through a 0.22 µm syringe filter, and collected in 5 mL Cryotube vials and 50 mL plastic vials for SO 4 2− and HS − , respectively. Samples were preserved with 20 µL of 20% zinc acetate (ZnAc), for SO 4 2− and 200 µL of ZnAc for HS − ; both samples were then stored frozen. SO 4 2− concentrations were quantified by ion chromatography (Dionex ICS-1500, with an autosampler Dionex AS50, a guard column Dionex AG22 and an analytical column Dionex IonPac AS22) and HS − concentrations were determined with a Thermo Spectronic Genesys 10vis spectrophotometer, using a 5-cm light path, according to the method described by Cline 20 . The detection limits were 0.5 and 0.25 µmol L −1 for SO 4 2− and HS − , respectively. Samples for Fe and Mn measurements were collected into sealed N 2 -flushed 60 mL glass serum bottles, with the sampler described above. Water was rapidly transferred from the bottles to the filtration set with a syringe equipped with a tube, and was passed through 25 mm glass fiber filters 15 . Filters were collected in 2 mL Eppendorf vials and preserved with 1 mL of a HNO 3 − 2% solution, while filtrates were collected into four 2 mL Eppendorf vials and preserved with 20 µL of a HNO 3 65% solution. Particulate Fe and Mn concentrations were determined from the filters, which were digested with nitric acid in Teflon bombs in a microwave digestion apparatus (Ethos D, Milestone Inc.) and diluted with milli-Q water to a final volume of 50 mL. Dissolved Mn and Fe concentrations were determined from the filtrates, which were diluted with milli-Q water to a final volume of 50 mL. Fe and Mn concentrations were determined by inductively coupled plasma mass spectrometry (ICP-MS) using dynamic reaction cell (DRC) technology (ICP-MS SCIEX ELAN DRC II, PerkinElmer inc.). Analytical accuracy was verified by a certified reference material (BCR 715, Industrial Effluent Wastewater). CH 4 oxidation rate measurements. Samples for CH 4 oxidation incubations were collected in N 2 -flushed 60 mL glass serum bottles sealed with butyl stoppers previously boiled in milli-Q water in the laboratory, and aluminum caps, using the sampling device described above. CH 4 oxidation was determined following the methodology described by Roland et al. 21 . Briefly, two bottles were immediately poisoned with 100 µL of HgCl 2 after collection (T 0 ), five bottles received an inhibitor of sulfate-reducing bacteria (sodium molybdate, + Mo) and five other did not receive any amendment (− Mo). In May 2013, the molybdate (Mo) solution was prepared directly on the field with milli-Q water stored at ambient temperature and was not flushed, while it was prepared with fresh milli-Q water and was flushed before the transport in September 2013 and August 2014. The bottles of both  16 , as described by Borges et al. 1 . The precision of measurements was ± 3.9% and the detection limit of the method is 0.5 nmol L −1 . The precision was calculated based on the analysis of the two T 0 bottles sampled in duplicate for each depth and then accounted for the variability induced by the handling of samples (samples collection, storage) and our analytical method. CH 4 oxidation rates were calculated as a linear regression of CH 4 concentrations over time during the course of the incubation. Table S1 shows standard deviations, initial CH 4 concentrations, percentage of CH 4 consumed and the time lapse during which the CH 4 oxidation rates were calculated for each depth 15 .
A correction of the CH 4 oxidation rates has been applied taking into account the potential oxygen supply through the injection of the Mo solution. We considered that a maximum of 2.5 µmol L −1 of O 2 were added to each bottle (250 µL of the solution were added to 30 mL of water). The calculations were made according to Roland et al. 21 . Only the time change in dissolved CH 4 concentration and not the concomitant decrease in the concentration of the electron acceptors potentially involved in CH 4 oxidation processes were monitored during the incubation. This is due to the interference caused by the HgCl 2 poison addition with the analytical methods used to determine the electron acceptors concentrations. where D turbulent is the turbulent diffusion coefficient, Grad is the vertical concentration gradient of each element, C is the concentration of the element at a given depth, and Adv is the upwelling velocity. Vertical fluxes were computed by using a range of turbulent diffusion coefficient and of upwelling velocity of 1.4 × 10 -7 -1.0 × 10 -6 m 2 s −1 and 6.3 × 10 -9 -6.3 × 10 -8 m s −1 , respectively 14 . Bacteriochlorophyll pigments analyses. Samples for pigments analyses were collected every 0.25 m, from 9 to 13 m depth in September 2013, and from 8 to 12 m depth in August 2014. Water was collected with the sampler described above, and filtered through Whatman GF/F 47 mm diameter filters. The filtration volume depended on the depth sampled, but was on average 0.3 L. Filters were preserved in 5 mL Cryotube vials and stored frozen. The pigment extraction was made in 4 mL of 90% HPLC grade acetone. Two 15-min sonication steps separated by an overnight period at 4 °C in dark were applied, and extracts were stored in 2 mL-amber borosilicate vials. HPLC analyses were carried out as described by Sarmento et al. 24 .

Contribution to CH 4 oxidation.
Archaeal diversity. Samples for DNA analyses were collected as detailed by Inceoğlu et al. 25 . Genomic DNA was extracted as previously described 26 and further subjected to FLX-titanium amplicon pyrosequencing 27 from collected filters using ARCH 349F (5′-GYGCASCAGKCGMGAAW-3′) and ARCH 806R (5′-GGA CTA CVSGGG TAT CTAAT-3′) as sequencing primers targeting the 16S rRNA V3-V4 region 27 . Reads used in the present study can be accessed through sequencing read archive (SRX349388). Archaeal 454-pyrosequencing sequences from Kabuno Bay water samples tentatively belonging to the Methanosarcinales order were subsequently analysed for further taxonomic refinement. All sequences were aligned using the SINA aligner 28 and then imported into the latest SILVA 16S rRNA-ARB-compatible database (SSURef-132_NR_99_13_12_17_opt. arb; http://www.arb-silva .de) in ARB 29 . Two base frequency filters ("termini" and "ssuref:archaea"; positional variability by parsimony) were applied to exclude highly variable positions before adding sequences to the original database using the "parsimony quick add marked" tool from ARB. Sulfate concentrations were relatively high in the chemocline and did not change substantially within the CH 4 gradient, while HS − concentrations were always two orders of magnitude lower than SO 4 2− (i.e., lower than 1.0 µmol L −1 ), with the exception of a maximum peak (10 µmol L −1 ) detected at 9.5 m depth in September 2013 (Fig. 2). The vertical fluxes of the potential electron acceptors for AOM and their reduced forms are shown in Table S2.
Five main rivers enter Kabuno Bay and are sources of Fe and Mn to the lake. Based on data gathered during a monthly monitoring of these rivers, we estimated that they supplied the lake with 3.3 mmol m −2 d −1 of Fe 3+ oxide and 0.1 mmol m −2 d −1 of Mn 4+ oxide.  (Fig. 1). Methane removal in anoxic waters by aerobic organisms has been found to be supported by oxygenic photosynthesis in the well-illuminated (10% of incident PAR) chemocline of Lake Cadagno 30 . Kabuno Bay chemocline is located below the photic zone so that light conditions (0.1-1% PAR) do not appear suitable to support significant phytoplankton activity. Pigments analysis   31,32 , was at least one order of magnitude higher than chlorophyll a in the chemocline (Figs. 1 and S4). Also, Morana et al. 14 showed that 74 ± 13% of particulate biomass in the chemocline derive from anoxygenic CO 2 fixation by GSB. These multiple lines of evidence indicate that biological primary production in the chemocline is largely dominated by anoxygenic photoautotrophs so that a cryptic oxygen cycle sustained by oxygenic photosynthesis in Kabuno Bay anaerobic waters can be ruled out. Nevertheless, we cannot rule out the possibility of punctual and/or episodic oxygen incursions in the anoxic waters, which could explain the presence of NO x below the oxycline in September 2013. This oxygen could be rapidly used by aerobic CH 4 oxidation, and the part of aerobic CH 4 oxidation in the water column of Kabuno Bay might be more significant than reflected by the results of our in vitro incubations. However, in the particular conditions of our in vitro incubations, we could estimate that 89-98% of the CH 4 was oxidized under anoxic conditions, raising the question which electron acceptors supported AOM. The water column was particularly rich in SO 4 2− , but SO 4 2− concentrations did not decrease substantially with depth in the chemocline (Fig. 2), neither in anoxic waters. Indeed, SO 4 2− reduction rates reported by Llirós et al. 11 were relatively low compared to SO 4 2− concentrations and were ca. 10 times lower than the AOM rates. Furthermore, except in May 2013 where AOM rates were lower with Mo added (what can be linked to the slightly different methodology used), our in situ incubation experiments revealed that AOM rates were up to 6 times higher (September 2013 and August 2014) in presence of Mo, an inhibitor of SO 4 2− reduction ( Fig. 1), contrary to what would have been expected if CH 4 oxidation depended on SO 4 2− reduction. All these multiple lines of evidence suggest that SO 4 2− reduction did not sustain a significant part of the CH 4 oxidation in the chemocline of Kabuno Bay. Similarly, it seems unlikely that Mn oxides and NO 3 − would fuel a substantial part of AOM. Their concentrations were always low, and the upward fluxes of Mn 2+ and NH 4 + could only have sustained at most 4% and 25% of the AOM rates, considering an extreme scenario where the totality of Mn 2+ and NH 4 + fluxes, once oxidized, would be exclusively involved in Mn oxides or NO 3 dependent AOM, which is unlikely. Instead, it has been showed that denitrification and NO 3 − reduction to NH 4 + (DNRA) linked to Fe oxidation occur in the water column of Kabuno Bay 33 . As both these processes are thermodynamically more favorable than NO 3 driven AOM, we hypothesized that the NO 3 − formed in the water column of Kabuno Bay is rapidly consumed by DNRA and denitrification coupled to Fe oxidation rather than by NO 3 driven AOM, which can also explain the low NO 3 − concentrations. Therefore, it is likely that AOM coupled to NO 3 − reduction did not occur at a significant extent in Kabuno Bay. On the other hand, Llirós et al. 11 showed that Fe 3+ reduction in Kabuno Bay was ca. 24 times higher than SO 4 2− reduction. This study also reported the existence of an important Fe-related bacterial community in the water column of Kabuno Bay, among which Chlorobium phaeoferrooxidans, a GSB capable of Fe oxidation 34 was the dominant member. Here, we showed that Bchl e, a specific biomarker of the GSB, was mainly located in the Fe oxidation-reduction zone, which co-occurred with the maximum CH 4 oxidation peaks (Fig. 1). 16S-RNA gene based pyrosequencing data from February 2012 showed the presence of an archaeal community related to Candidatus Methanoperedens nitroreducens (the most abundant retrieved OTUs showed 94.6 and 93.7% sequence similarity, respectively, against Candidatus Methanoperedens nitroreducens (JMIY01000002); Fig. 3 and Table S3). The peaks of abundance of these archaea co-occurred within the Fe oxidation/reduction zone and the maximum occurrence of photoferrotrophy previously reported by Llirós et al. 11 (Fig. 4). Overall, putative AOM-related archaea in Kabuno bay represented ca. 16% of the whole community in February 2012. Recent studies have  Fe-linked AOM alone, without a bacterial partner, via the "reverse methanogenesis" pathway and possibly using a extracellular electron transport pathway 37 . Despite that these prokaryotes diversity data were acquired during a different sampling campaign 25 than the AOM measurement reported here, the good agreement between (1) high Candidatus methanoperedens abundance and high Bchl e concentrations in the Fe oxidation-reduction zone in 2012, and (2) high AOM rates and high Bchl e concentrations in the Fe oxidation-reduction zone during this study, allow to hypothesize that Candidatus methanoperedens could represent the dominant microbes thriving AOM during the present study. Accordingly, these AOM-associated archaea (AAA 5 ) could therefore be responsible for the Fe-related AOM measured in Kabuno bay, taking into account that AOM coupled to Fe reduction is thermodynamically more favorable than coupling with other electron acceptors 38 . Iron-driven AOM has already been suggested as a dominant CH 4 removal pathway in other ferruginous environments (i.e. [38][39][40]. Bacterial data reported from samples collected in 2012 evidenced that photoferrotrophs were responsible for the oxidation of 37 mmol Fe m −2 d −1 11 . Comparison of this Fe oxidation rate with AOM showed that Fe 3+ released via photoferrotrophy could potentially fuel an important fraction of the AOM (up to 31% in September 2013), in contrast to NO 3 − and SO 4 2− reduction rates 11,33 , which could only fuel up to 5% of the AOM observed (Fig. 5). Furthermore, the observation of higher AOM rates when Mo was added may result from the higher availability of Fe oxides when sulfate-reducing bacteria (SRB) activity was inhibited. Indeed, SO 4 2− reduction produces HS − that can rapidly reduce Fe oxides to form iron sulfides, which could precipitate and then be unavailable for Fe-dependent CH 4 oxidizers. Fe is known to be efficiently recycled in the chemocline of ferruginous lakes 38 or bioturbated marine sediments 41 .
Overall, process rate measurements and metagenomics data suggest that an intense biotic regeneration of Fe 3+ mediated by photoferrotrophs could provide an important fraction of the electron acceptors required to oxidize CH 4 anaerobically in Kabuno Bay. Following the same approach as Jones et al. 42 and assuming steady state conditions, the rate of Fe or Mn leaving the water column via sedimentation must equals the rate of Fe or Mn input in Kabuno bay's water column via the rivers. Assuming that Mn and Fe leave the water column as a Mn 2+ or Fe 2+ mineral, all the oxidized Mn or Fe must ultimately be reduced, and the downward flux of particulate Fe 3+ should be equivalent to the input of Fe via the rivers (3 mmol m −2 d −1 ) and the upward flux of Fe 2+ (4-33 mmol m −2 d −1 ). If we consider that the totality of this downward flux of particulate Fe 3+ is reduced within Kabuno bay's chemocline, but that only 3 mmol Fe m −2 d −1 finally leave the mixolimnion, we estimated that Fe would be recycled up to 11 times before removal by sedimentation.

Consequences on representation of the Archean ocean metabolism. A strong interaction
between photoferrotrophs and CH 4 oxidizers has been hypothesized in several other modern Archean ocean analogues, such as Lake Matano 38 and Lake La Cruz 43 . Under the ferruginous conditions of Archean oceans, photoferrotrophs would have been responsible for most of the primary production of the primitive Earth 12 . It is also generally assumed that a much larger fraction of the organic matter generated by primary producers would have been processed by methanogens than nowadays, given the absence of oxic and SO 4 2− driven mineralization of organic matter in the O 2 and SO 4 2− depleted waters of the early Earth Ocean 44,45 . Pavlov et al. 46 49 who showed that the combined effects of competition between methanogens and SO 4 2− reducers and occurrence of SO 4 2− driven CH 4 oxidation would have effectively reduced the CH 4 fluxes from the ocean even at modest SO 4 2− concentrations, and regardless of O 2 concentrations. Similarly, we propose that strong interactions between photoferrotrophs and Fe-dependent CH 4 oxidizers might have exerted an important control on the Archean-Proterozoic CH 4 cycling. For instance, modelled photoferrotrophic primary production in early Earth ferruginous ocean (3.8 mmol m −2 d −1 50 ) would have produced Fe oxides at a rate of 15.3 mmol m −2 d −1 , assuming a 4:1 ratio between Fe oxidation and C fixation by photoferrotrophs. Although simplistic given the complexity of the C and Fe biogeochemical cycle, comparison of this rate of Fe oxides production with the seato-air CH 4 flux (1.9 mmol m −2 d −1 ) proposed by Pavlov et al. 46 suggests that photoferrotrophy would have been high enough to potentially support the oxidation of 100% of the CH 4 flux to the early atmosphere. In Kabuno Bay, a modern analog to early Earth ocean, measured photoferrotrophy rates would have been sufficient to sustain a smaller (31 ± 18%, n = 3), but still important, part of the oxidation of the CH 4 flux measured during this study (19 ± 11 mmol m −2 d −1 , n = 3).
It has been hypothesized that the rise of atmospheric oxygen about 2.4 Gyr ago (Great Oxidation Event) was triggered by a decrease of atmospheric CH 4 levels 51 . This has been attributed to an increased importance of SO 4 2 reduction in the oceans that outcompeted the methanogenic organisms, although this explanation is not consistent with geological records 52 . It has also been hypothesized that the production of methane decreased in response to the decrease of the nickel inputs to oceans, as this is a key metal for methanogen enzymes 53 . Alternatively, it was showed that supplying a culture media with nickel does not stimulate methanogenesis 54 . A widespread occurrence of AOM coupled to Fe oxidation offers an alternative explanation that would have led to the decrease of dissolved CH 4 in the ocean and consequently the emission of CH 4 to the atmosphere, thus going in the direction of the study of Riedinger et al. 39 . This is consistent with the revision of the putative composition of the Archean atmosphere 55 that suggests that the amount of greenhouse warming by CH 4 was more limited than previously thought 49,56 .
While we can hypothesize that Fe(OH) 3 sustaining AOM was initially provided by photoferrotrophy, the onset of oxygenic photosynthesis would also have been an ample supply of Fe(OH) 3 , available to sustain AOM, from the oxidation of Fe 2+ with dissolved O 2 released by oxygenic photoautotrophs (Fig. 6). This might also provide an explanation to the modest but significant increase in Fe-using genes from the Archean to the Proterozoic given by the phylogenomic analyses 57 . A decrease of dissolved CH 4 concentrations due to Fe-dependent AOM could have had several other consequences on the timing and sequence of Archean and Proterozoic events. The decrease of atmospheric CH 4 would have led to a decrease of H 2 and CO generation by photolysis, which in turn would have led to a decrease of supply of H 2 and CO to surface oceans, contributing to the demise of pelagic H 2 -methanogens and CO-acetogens (Fig. 6), that is usually attributed exclusively to the inhibition of these anaerobic organisms by increasing O 2 level. www.nature.com/scientificreports/ Finally, the great oxygenation event was shortly afterwards followed by a low latitude glaciation that would be associated to lower atmospheric CH 4 concentrations. However, the trigger of the collapse of CH 4 has not been clearly identified, some authors arguing that CH 4 decreased as a consequence of rising O 2 58 , while others proposed that low CH 4 level preceded the transition to an O 2 -richer atmosphere 49,51 . Co-occurrence of photoferrotrophy and Fe-dependent CH 4 oxidation in the Archean would support the latter hypothesis and the view of Olson et al. 48 who recently proposed an alternative mechanism for the initiation of low-latitude glaciation with low baseline atmospheric CH 4 levels.  Figure 6. Potential role of iron-oxide AOM in Archean Ocean metabolism. In the Archean Ocean, anaerobic metabolism (photoferrotrophy, methanogenesis and CO-acetogenesis in dark blue) dominated before the advent of oxygenic photosynthesis (in light blue). In red, we propose anaerobic methane oxidation (AOM) using Fe oxides as electron acceptor as additional metabolic process. Before the advent of oxygenic photosynthesis, Fe 3+ dependent AOM linked photoferrotrophy and H 2 -methanogenesis that so far have been seen as two parallel and unconnected processes. After the advent of oxygenic photosynthesis, Fe oxides-dependent AOM might have used the abundantly produced Fe(OH) 3 from the oxidation of Fe 2+ by O 2 to further remove CH 4 from the water column, facilitating the increase of O 2 in the atmosphere and the great oxidation event.