The methane cycle in ferruginous Lake Matano

In Lake Matano, Indonesia, the world’s largest known ferruginous basin, more than 50% of authigenic organic matter is degraded through methanogenesis, despite high abundances of Fe (hydr)oxides in the lake sediments. Biogenic CH4 accumulates to high concentrations (up to 1.4 mmol L−1) in the anoxic bottom waters, which contain a total of 7.4 × 105 tons of CH4. Profiles of dissolved inorganic carbon (ΣCO2) and carbon isotopes (δ13C) show that CH4 is oxidized in the vicinity of the persistent pycnocline and that some of this CH4 is likely oxidized anaerobically. The dearth of NO3− and SO42− in Lake Matano waters suggests that anaerobic methane oxidation may be coupled to the reduction of Fe (and/or Mn) (hydr)oxides. Thermodynamic considerations reveal that CH4 oxidation coupled to Fe(III) or Mn(III/IV) reduction would yield sufficient free energy to support microbial growth at the substrate levels present in Lake Matano. Flux calculations imply that Fe and Mn must be recycled several times directly within the water column to balance the upward flux of CH4. 16S gene cloning identified methanogens in the anoxic water column, and these methanogens belong to groups capable of both acetoclastic and hydrogenotrophic methanogenesis. We find that methane is important in C cycling, even in this very Fe‐rich environment. Such Fe‐rich environments are rare on Earth today, but they are analogous to conditions in the ferruginous oceans thought to prevail during much of the Archean Eon. By analogy, methanogens and methanotrophs could have formed an important part of the Archean Ocean ecosystem.


INTRODUCTION
Nearly 600 m deep, Lake Matano in Indonesia is one of the deepest lakes on Earth, and its anoxic waters comprise the world's largest known Fe(II)-rich (ferruginous) basin (Crowe et al., 2008a,b). Lateritic soils consisting of up to 60 wt.% Fe (hydr)oxides dominate the surficial geology of Lake Matano's catchment basin (Golightly, 1981;Crowe et al., 2008b) and erosion of these soils supplies Fe-rich particulate matter to Lake Matano. The water column of Lake Matano exhibits weak thermal gradients, yet a persistent pycnocline is located at around 100 m depth, forming a boundary between the well-mixed, oxic surface water and the anoxic bottom water (Crowe et al., 2008b;Katsev et al., 2010). Exchange of water across this pycnocline is slow and occurs on a millennial time scale, similar to temperate meromictic lakes (Crowe et al., 2008b;Katsev et al., 2010). Owing to this slow exchange and the very low concentrations of dissolved S (i.e., SO 4 2) and RS(-II)), ferrous Fe has accumulated to high concentrations in the anoxic bottom waters (Crowe et al., 2008b). Slow exchange of water also favors the accumulation of dissolved gases. Indeed, dissolved gases (CH 4 and CO 2 ) are known to accumulate to high concentrations in a number of the world's permanently stratified lakes, including Lake Kivu in East Africa (Schmid et al., 2005) and Lake Pavin in France (Lehours et al., 2009). Our study of the biogeochemical cycling of CH 4 (and CO 2 ) in Lake Matano was initially prompted by concern that these gases might reach levels that could cause a limnic eruption, similar to the catastrophic events in Lakes Nyos and Lagos in Africa (Zhang & Kling, 2006). Additional motivation for this study arose from Lake Matano's status as an end member sulfur-poor, iron-rich environment. These chemical conditions and the stratified nature of Lake Matano support a novel and unique microbial ecology, including recently discovered low-light adapted Chlorobiaceae (Crowe et al., 2008a). The ubiquity of SO 4   2) in most modern environments intrinsically links the CH 4 and S cycles because the rates and pathways of both methanogenesis and methanotrophy are directly influenced by SO 4 2) availability (Capone & Kiene, 1988;Reeburgh, 2007). The scarcity of dissolved SO 4 2) in Lake Matano may foster an atypical CH 4 cycle, likely precluding SO 4   2) -based anaerobic CH 4 oxidation and favoring aerobic CH 4 oxidation or anaerobic CH 4 oxidation via alternative electron acceptors (Zehnder & Brock, 1980;Konhauser et al., 2005;Caldwell et al., 2008;Beal et al., 2009). In this study, we used geochemical and microbiological techniques to elucidate the pathways of methane production and consumption, characterize Archaeal ecology, quantify the importance of CH 4 to carbon cycling in a SO 4 2) -poor environment, and evaluate the potential risk of methane-driven limnic eruption in Lake Matano. We also discuss the potential for CH 4 cycling in other ferruginous environments, including the Archean Ocean.

Sampling and sample storage
Sampling was conducted at a central, deep-water location (2°28¢00¢¢S and 121°17¢00¢¢E) (Fig. 1) in July 2006 (end of the wet season) and February 2007 (beginning of the wet season) using local fishing boats. All water samples were collected with 5 L Go-Flow (Niskin; General Oceanics, Miami, FL, USA) bottles attached in series to a stainless steel cable and a hand-operated winch. The bottles were placed at depth to an accuracy of ±1 m with the help of a commercial fish finder (Furuno,FCV 585;Furuno Electric Co.,Nishinomiya,Japan). Water samples for methane determination were drawn directly from the spigot of the Niskin bottle using a 60 mL syringe and immediately transferred from the syringe to pre-evacuated 14-mL crimp-sealed serum vials that had been flushed with N 2 gas and contained 100 lL of a saturated HgCl 2 solution to prevent further microbial methane production or oxidation. The partially filled serum vials were stored upside down to prevent potential leakage of CH 4 around the stopper. Total alkalinity was determined on water samples transferred to glass bottles and preserved with HgCl 2 . Samples for the determination of total dissolved inorganic carbon (RCO 2 ) and NH 4 + were drawn from the spigots with a 1 mL syringe and injected directly into a flow-injection analysis system (see below).

Analyses
Water temperature and conductivity profiles were collected in situ using a submersible Conductivity-Temperature-Depth probe (CTD) (Sea-Bird, SBE-19; Sea-Bird Electronics, Bellevue, WA, USA). Methane concentrations in the headspace of the crimp-sealed vials were determined by gas chromatography (Agilent 6890N; Agilent Technologies, Santa Clara, CA, USA, with a HayeSep Q 80 ⁄ 100 column) with flame ionization (detection limit of 10 )7 mmol L )1 and a precision of <1%, 2006) or thermal conductivity (detection limit of 2 · 10 )5 mmol L )1 and a precision of <1%, 2007) detection. Total alkalinity was determined by potentiometric titration with a dilute HCl solution (Gieskes & Rogers, 1973) using an automated Radiometer titrator (Titrilab 865; Radiometer Analytical SAS, Lyon, France) with a reproducibility of better than 0.5%. The RCO 2 and dissolved NH 4 + concentrations were measured by flow injection analysis (Hall & Aller, 1992). Detection limits were <50 lmol L )1 and 0.5 lmol L )1 for RCO 2 and NH 4 + , respectively, with a relative precision of better than 5%. The isotopic composition (d 13 C) of CH 4 and CO 2 in headspace samples was determined by gas chromatography-combustion-isotope ratio mass spectrometry (GC-C-IRMS) using a ThermoFinnigan MAT 253 (Thermo Fisher Scientific, Waltham, MA, USA) mass spectrometer interfaced to a ThermoElemental Trace GC Ultra with a Rt-QPLOT column (Restek; Restek Corporation, Bellefonte, PA, USA). d 13 C RCO2 was determined by converting RCO 2 to gaseous CO 2 following the addition of concentrated orthophosphoric acid to the samples under a 15-mTorr vacuum. Water vapor was eliminated by double trapping, first with a liquid N 2 -trap and followed by a dry-ice isopropyl alcohol trap. d 13 C measurements were made using a VG-PrismÔ triple-collector mass spectrometer (Isoprime, Cheadle Hulme, UK). All stable C isotopic compositions are reported in per mil (&) relative to the V-PDB standard with a precision of better than 0.2&. Duplicate measurements of d 13 C RCO2 in water recovered from 110 m depth yielded values of )8.70& and )8.85& and were within the reported uncertainty. Samples of particulate organic matter (POM) were collected on pre-combusted 0.45 lm glass fiber filters. Portions of these filters were analyzed for nitrogen and carbon using an elemental analyzer.
Dissolved oxygen (DO) measurements were made by linear sweep voltammetry using an Analytical Instrument Systems (AIS) (Flemington, NJ, USA) DLK-60 with cell stand. Water samples were transferred directly from the Niskin bottles to a glass voltammetric cell via Tygon tubing. The cell was sealed from the atmosphere using rubber stoppers and o-rings and was flushed with three times the cell volume of water before measurement. Measurements were conducted with a threeelectrode configuration using a Au-Hg amalgam microelectrode (50 lm radius), a saturated Ag-AgCl electrode, and a platinum wire as the working, reference, and counter electrodes, respectively (Luther et al., 2003). The DO detection limit was 2.5 lmol L )1 based on three times the standard deviation of five measurements conducted in lake water purged for 15 min with N 2 gas that had been scrubbed through a catechol solution to remove traces of O 2 from the commercial, compressed N 2 . Fe, Mn, and S species were determined as previously described (Crowe et al., 2008a,b). NO 3 ) and NO 2 ) concentrations were determined spectrophotometrically using a 5-m long cell path (Yao et al., 1998). Unfortunately, high reagent blanks precluded us from reaching the single-digit nmol L )1 detection limits previously attained (Yao et al., 1998). Samples used to measure the natural radiocarbon content of CH 4 were collected in 15-mL serum vials with an average nitrogen headspace of 5.5 mL from depths of 105,112,118,130,140,160,200,250,350,450, and 550 m. To collect enough CH 4 for radiocarbon measurements, the entire headspace was extracted and combined in two distinct, pooled samples. One of these pooled samples comprised headspace gases from depths of 105, 112, 118, 130, 140, 160, and 200 m, and the other from 250, 350, 450, and 550 m. The CH 4 was then prepared for radiocarbon analyses as previously described by Kessler & Reeburgh (2005). The samples were analyzed using 14 C-Accelerator Mass Spectrometry at the University of California, Irvine.
The pH of the water column was determined on individual samples transferred to plastic bottles without headspace gas using a hand held pH-meter (VWR, West Chester, PA, USA) and an Accumet (Cole-Parmer, Vernon Hills, IL, USA) combination electrode calibrated with three NIST-traceable buffers at the ambient surface water temperature ($29.5°C). The electrode diameter was adjusted with a piece of Tygon tubing and Parafilm to form an air-tight seal upon its insertion in the sampling bottles. Speciation and saturation state calculations were carried out using the JCHESS geochemical modeling software and the Chess thermodynamic database (van der Lee, 1993).

Microbiology
Samples of cellular material were collected from 2 L of water on 0.45 lm glass fiber filters and preserved at )80°C. DNA was extracted from these filters using a Mo-Bio PowerSoil extraction kit (MO BIO Laboratories, Carlsbad, CA, USA). 16S rDNA was amplified using polymerase chain reaction (PCR) with Platinum Ò Taq DNA polymerase and primers designed to target specific groups (Table 1). All PCR cycles were performed according to the specifications given by Invitrogen (Carlsbad, CA, USA) for Platinum Ò Taq DNA polymerase. To test the reliability of the PCR reactions, positive and negative controls were conducted for each primer set with the exception of the ANME. The PCR products were imaged by gel electrophoresis to check for the positive amplification of the template and to estimate the length of the amplified fragments. The amplified DNA was cloned into the TOPO cloning vector and transformed into chemically competent One Shot TOP10 E. coli. Individual clone colonies were randomly selected for sequencing after blue-white b-galactosidase screening to confirm positive ligation. Plasmid Methane cycle in ferruginous Lake Matano 63 DNA was purified using a PureLink Quick plasmid Miniprep Kit (Invitrogen, Carlsbad, CA, USA). Forward and reverse sequencing of the purified DNA was performed by Macrogen (Rockville, MD, USA). The number of clones sequenced is presented in Table 3.
Contiguous sequences were assembled using the assembly function in Geneious Ò (Auckland, New Zealand). BLAST (Basic Local Alignment Search Tool) queries identified all clones generated with amplicons from primers designed to target Type-1 aerobic methanotrophs as similar to Thiobacillus sp. Species of the genus Thiobacillus are not known aerobic methanotrophs but are known to yield false-positive amplicons with the primer set used (Wise et al., 1999). Phylogenetic analyses of these sequences confirm their classification within the beta-proteobacteria (see phylogenetic tree in Fig. SI11, Appendix 4). These clones were not considered further. Amplicons generated using the Archaeal and ANME-1,2,3 targeting primer sets were also cloned and sequenced. Chimeric sequences in this library were identified with Bellerophon (v.3) and eliminated from the dataset (DeSantis et al., 2006). Nearest non-chimeric neighbors and isolates were identified using Simrank and included in subsequent phylogenetic analyses. Additional sequences were culled from the Greengenes database (Hugenholtz taxonomy) (DeSantis et al., 2006). Sequence alignments were performed using the MUSCLE plugin in Geneious Ò . Phylogenetic trees were constructed by neighbor joining, maximum likelihood, and Bayesian inference methods using the Geneious tree builder, and the PHYML (Guindon & Gascuel, 2003) and MrBAYES (Huelsenbeck & Ronquist, 2001) plugins, respectively. Branching topologies in the neighbor joining and maximum likelihood trees were verified by conducting 1000 and 100 bootstrap analyses, respectively. Lake Matano clone sequences were submitted to GenBank (accession numbers: HM583588-HM583606).

Physical properties of Lake Matano
From 2004 to 2007, Lake Matano remained stratified with a persistent pycnocline between 100 and 120 m depth and non-persistent (seasonal) pycnoclines in the upper 100-m surface layer ( Fig. 2) (Crowe et al., 2008b;Katsev et al., 2010). The persistent pycnocline largely results from a decrease in water temperature with increasing depth and is marginally enhanced by higher solute concentrations in the bottom waters. It persists over at least decadal and likely much greater time scales, seemingly unaffected by seasonality (Fig. 2, inset) (Crowe et al., 2008b;Katsev et al., 2010). The seasonal pycnocline was located between 60 and 90 m depth in July 2006, and between 30 and 60 m in February 2007 (Fig. 2). The density gradient in this seasonal pycnocline is largest during the rainy season (February) and is completely eroded by the middle of the dry season (September), likely as a result of evaporative cooling of the surface waters (Crowe et al., 2008b;Katsev et al., 2010). A third, diurnal pycnocline often forms within the upper 30 m as a result of surface water heating by solar radiation (not seen in  the profiles shown in Fig. 2). Thermal profiles, recorded between 2004 and 2007 at numerous locations throughout the lake (not shown), failed to detect a deep heat source, and provide no evidence for hydrothermal activity.

Distributions of major redox active species
The persistent stratification is reflected in the distribution of several redox active species (Figs 3 and 7A). DO in the surface waters was nearly at saturation with respect to the atmosphere and decreased with depth. Hypoxic conditions (<62.5 lmol L )1 O 2 ) (Diaz & Rosenberg, 1995) prevailed between 95 and 110 m, and sub-to anoxic conditions were reached below 110 m. Based on the bathymetry of Lake Matano (Crowe et al., 2008b), the anoxic basin covers 107 km 2 and contains 35 km 3 of anoxic water. The vertical distributions of other redox active species (Mn, Fe, SO 4   2) , and HS ) ) were consistent with their respective chemical properties. There was some overlap between dissolved Mn and O 2 within the hypoxic zone: Mn was detected (0.45 lmol L )1 ) at 105 m depth where O 2 concentrations of up to 37 lmol L )1 were measured. The coexistence of these two species in the water column is consistent with the metastability and relatively slow oxidation kinetics of Mn(II) in the presence of O 2 (Luther, 2005;Morgan, 2005). In contrast, Fe was undetectable (<0.1 lmol L )1 ) within the hypoxic zone but measurable (0.29 lmol L )1 ) at 110 m, the depth at which O 2 became undetectable (<2.5 lmol L )1 ). Sulfate concentrations in the surface mixed layer were low (<20 lmol L )1 ), and trace sulfide (peak of <0.6 lmol L )1 at $140 m; Fig. 3) was detected in the anoxic bottom water. Nitrate and nitrite were undetectable (<100 nmol L )1 ) within the water column with the exception of a small peak ($200 nmol L )1 ) of NO 3 ) between 90 and 100 m depth (Fig. 4A). Particulate Fe and Mn concentrations also peaked near the persistent pycnocline ( Fig. 4B). Nearly identical vertical distributions of these redox sensitive species between 2004 and 2007 (Crowe et al., 2008b) illustrate the near steady-state nature of redox cycling within the pycnocline of Lake Matano. Importantly, the profiles of Fe(II) and NH 4 + reveal that they originate from the sediment and ⁄ or deep bottom waters (i.e., below 200 m where the concentration profiles are vertical), likely due to the anaerobic degradation of organic matter (Katsev et al., 2010). Without sediment porewater profiles, we cannot distinguish between the two potential sources. Given the highly active regional tectonics and the close proximity of volcanic activity, we had previously postulated a hydrothermal source for Fe(II) but, as pointed out above, we find no evidence for local hydrothermal activity in our vertical temperature profiles.

Distributions of dissolved CH 4 and RCO 2
Methane concentrations in Lake Matano ranged from 0.003 mmol L )1 in the surface mixed layer to 1.4 mmol L )1 in the anoxic bottom water (Fig. 5A): the 35 km 3 anoxic basin of Lake Matano contains 4.6 · 10 10 mol or 7.4 · 10 5 metric tons of CH 4 . In comparison, the world's largest anoxic basin, the Black Sea (545 000 km 3 ), contains 9.6 · 10 7 metric tons of CH 4 (Reeburgh et al., 1991), but at concentrations less than 0.012 mmol L )1 (Schubert et al., 2006). Vertical profiles of methane concentrations constructed from samples taken in both 2006 and 2007 ( Fig. 5A) are nearly identical, suggesting that CH 4 concentrations are close to steady state, as are other redox active species (Crowe et al., 2008B). At 0.003 mmol L )1 , CH 4 in the surface water is supersaturated with respect to the atmosphere, thus the lake acts as a net source of CH 4 (methane in the surface water diffuses out) to the overlying atmosphere. Below 200 m depth, the concentrations are equivalent to the solubility limit of CH 4 in water at atmospheric pressure (Henry's law constant = 1.3 · 10 )3 mol L )1 atm )1 ; Canfield et al., 2005). Hence, samples collected from below 200 m may have degassed during recovery, and actual concentrations may be higher than reported. RCO 2 profiles (computed from total alkalinity and pH measurements using CO 2 SYS; Lewis & Wallace, 1998) are given in Fig. 5B. The RCO 2 concentrations calculated from the alkalinity measurements are in agreement with measured RCO 2 concentrations (data not shown), but the former carry smaller analytical uncertainties. The concentrations of CO 2(aq) in the bottom water of Lake Matano are less than 1 mmol L )1 (partial pressures up to 0.023 atm) and well below saturation at in situ temperatures and pressures ($19 atm at 200 m depth; Table 2). The total partial pressure of gases in the bottom waters did not exceed 5.2% of saturation at the in situ temperatures and pressures (Table 2), rendering limnic eruption unlikely.
depth) (Stuiver & Polach, 1977). Measurements of 14 C CH4 yielded conventional (not corrected for reservoir effects) ages of 2043 ± 4 and 2255 ± 5 year before present (BP) for carbon in CH 4 from the transition zone (105-200 m) and deep waters (>200 m), respectively. Thus, the age of the CH 4 accumulating in the deep water is relatively young. If some of the CH 4 were produced from the reduction of hydrothermal RCO 2 , this methane would not contain 14 C and could yield a deceivingly old age for the CH 4 itself. The young age suggests that ancient or distant sources are unlikely. It should be noted that methane might also originate from percolating groundwater, as it may at one of the rare and narrow lake beaches (i.e., Butterfly beach, which is situated on a large block of variably metamorphosed carbonate rocks on the south shore of Lake Matano, Fig. 1).

Microbial ecology
To obtain a preliminary characterization of the microbial community related to methane cycling in Lake Matano, we conducted culture-independent 16S rRNA gene cloning using primers designed to target specific groups (Table 1). Table 3 shows the results of PCR amplification using these primer sets. PCR failed to amplify DNA from known aerobic methanotrophs, indicating that they were absent from the water column or too small to be captured by our 0.45 lm pore-size filters. Phylogenetic analyses of Archaeal amplicons were conducted, and a maximum likelihood tree is shown in Fig. 8. This tree illustrates the phylogenetic relationships between the Lake Matano sequences, other sequences closely related to the Lake Matano clones, and sequences from CH 4rich environments. Clone sequences, obtained with primers purportedly targeting ANME-1, 2, and 3 (MWC118AM), cluster within the order Methanomicrobiales and are not closely affiliated with any of the ANME. Other studies have also found that the ANME-F primer can lack specificity for known ANME (Schubert et al., 2006). Clone sequences MWC118AM 2, 3, 4, 7, 8, 9, and 10 cluster tightly with each other and with additional clones obtained with primers targeting Archaea in general (MWC118AR 3, 8, 10, 14, and 11) (Fig. 8). Isolated organisms most closely related to these clone sequences are known methanogens of the family Methanomicrobiaceae, which use either hydrogenotrophic CO 2 reduction or formate disproportionation to generate methane. One other clone, MWC118AR clone 7, clusters within the order Methanosarcinales and is closely related to known methanogens of the family Methanosaetaceae. The remaining Archaeal clone sequences are widely distributed within the kingdom Crenarchaeota (data not shown). Many of the previously isolated Crenarchaeota that are closely related to these Matano clones are deeply rooted and thermo-  ANMEf-915r ANME-1, 2, 3 118 + 10 21fa-EelMS932r ANME-2 118 ) Archaeal anaerobic lake sediment clone SL-Pro (AB236081)

MWC118AR Clone 7
Layered structure granule clone UASB26 (AB329663)  3,8,10,11,14,16,22 Fig. 8 Phylogenetic tree displaying the affiliations of 16S rRNA sequences obtained from 118 m depth in Lake Matano using both PCR primers to target Archaea in general (MWC118AR) and the ANME (MWC118AM). The tree includes reference species in addition to clones obtained from methane-rich environments and the nearest non-chimeric neighbors identified using the Greengenes Simrank tool. The tree was constructed using the PHYML Maximum Likelihood algorithm. The numbers at the branching nodes represent bootstrap support for neighbor joining (1000 bootstrap replications) and maximum likelihood (100 bootstrap replications) in addition to probability from Bayesian inference (NJ ⁄ ML ⁄ BI). Bootstrap support of less than 60% and probabilities of less than 0.6 are not shown. The clone sequences from Lake Matano are emphasized with a bold font. Aquifex pyrophilus (M83548) was used as an outgroup to root the tree. The inset shows the distribution of the main Archaeal phylotypes as a percentage of the number of clones sequenced. phyllic. Of the clones sequenced from DNA amplified with the general Archaeal primers, 79% belonged to the order Methanomicrobiales, 4% to the order Methanosarcinales, and the remaining 17% were Crenarchaeota. It is noteworthy that our PCR with Archaeal primers failed to generate products at depths of 105, 115, and 140 m. It is possible that the PCR reaction was possibly inhibited for these samples, but our ability to amplify bacterial DNA argues against this. Alternatively, Archaeal cell abundances may have been lower in water from these depths.

DISCUSSION
According to the vertical distribution of CH 4 in the water column of Lake Matano (Fig. 5A), net CH 4 production occurs in the sediments or bottom waters, whereas net CH 4 consumption occurs within the 100-200 m depth interval. The sharp methane concentration gradient between 200 and 100 m depth implies the presence of a methane sink in the vicinity of the persistent pycnocline. In the following discussion, we constrain the rates and pathways of CH 4 production and consumption to develop a mechanistic understanding of CH 4 cycling in Lake Matano.

Methane production pathway
The carbon isotopic composition of CH 4 (d 13 C CH4 ) (less than )70&) clearly demonstrates its biological origin (Whiticar, 1999). This biogenic CH 4 has accumulated to relatively high concentrations despite the abundance of Fe and Mn (hydr)oxides, which, according to thermodynamic arguments, should preclude methanogenesis (Reeburgh, 2007). Standard-state thermodynamic calculations, however, ignore bioavailability and the kinetics of substrate uptake (Roden & Wetzel, 2003). For example, Roden (2003) suggested that the capacity of microbial Fe(III) oxide reduction to suppress other terminal electron acceptor pathways in anoxic soils and sediments is controlled by the reactive (i.e., microbially accessible) surface site density rather than the thermodynamic properties of the bulk (hydr)oxide minerals. Accumulation of sorbed and ⁄ or surface-precipitated Fe(II) on Fe(III) (hydr)oxide surfaces renders these surfaces progressively inaccessible to Fe-reducing bacteria and, thereby, alters the long-term reactivity of Fe(III) (hydr)oxides (Roden & Urrutia, 1999, 2002. Given the high concentrations of dissolved Fe(II) in the anoxic bottom waters (Fig. 3) and sediment porewaters (Fowle et al., 2006) of Lake Matano, surface deactivation of Fe(III) (hydr)oxides that sink through the bottom waters and accumulate in the lake sediments in high abundances (up to 20 wt.%; Crowe et al., 2004) is plausible. Furthermore, the absence of bioturbation in the anoxic bottom sediments limits recycling and regeneration of reactive Fe (hydr)oxide surfaces, promoting methanogenesis relative to Fe reduction (Roden & Wetzel, 1996).
The difference between the carbon isotopic compositions of methane and RCO 2 in the deep anoxic bottom waters yields an isotope separation factor (e C ), which is commonly used to fingerprint the different methanogenic pathways (Whiticar, 1999). The e C in Lake Matano, $65&, is within the range of fractionation factors observed for pure cultures of methylotrophic and hydrogenotrophic methanogens and is much higher than the 24-27& observed in pure cultures of acetate disproportionating organisms (Whiticar, 1999). Hence, this large isotope separation factor may indicate that CH 4 production in Lake Matano is dominated by hydrogenotrophic methanogens. This is broadly consistent with the microbial ecology of the water column, which identifies most of the methanogen clone sequences as members of the order Methanomicrobiales. The Methanomicrobiales isolated to date utilize H 2 ⁄ CO 2 , formate, and alcohols as substrates, and they are not known to conduct acetate disproportionation (Canfield et al., 2005). Nevertheless, there is some disparity between the true fractionation factors observed in culturebased experiments and the isotope separation factors observed in natural systems dominated by one form of methanogenesis or another. Given that the Lake Matano e C value falls at the very upper end of those observed in systems dominated by acetate disproportionation and at the lower end of values reported for systems dominated by CO 2 reduction (Whiticar, 1999), methanogenesis in Lake Matano may occur via both pathways. Indeed, the water column clone MWC118AR 7 groups within the Methanosaetaceae and all known Methanosaetaceae are obligate acetate disproportionators (Canfield et al., 2005). A more detailed microbiological study will be needed to resolve the specific methanogenic pathways in Lake Matano.
Regardless of the specific pathway, the presence of methanogens within the chemocline of Lake Matano indicates that methanogenesis may occur directly within the water column, despite the fact that the concentration profiles implicate the sediment as the principal CH 4 source. DNA sequences and biomarkers affiliated with methanogens have been detected in the water column of the Black Sea (Vetriani et al., 2003;Schubert et al., 2006). Due to the high abundance of sulfate in the Black Sea water column, it was concluded that these methanogen signatures originated from anaerobic methane oxidizing Archaea and that methanogenesis was not operating directly in the water column. A similar argument could be made with respect to Fe in the water column of Lake Matano, and this is discussed below (see next section). Though measured in the water column of a meromictic lake (Iversen et al., 1987), water column methane production appears to be rare in modern anoxic water columns (Reeburgh, 2007). Lake Matano provides a new environment in which to study pelagic methanogenesis under persistently anoxic conditions.
We also note that most of Lake Matano's Euryarchaeota water column clones cluster tightly with each other (Fig. 8), and their closest isolated relatives belong to newly described Methane cycle in ferruginous Lake Matano 69 genera of the order Methanomicrobiales, including Methanolinea (Imachi et al., 2008), Methanoregula (Brauer et al., 2006a), and Methanosphaerula (Cadillo-Quiroz et al., 2009). Our clones, the aforementioned isolates and additional clone sequences, form a coherent clade previously referred to as the 'R-10 or Fen cluster' (Brauer et al., 2006b). This clade is clearly separated (bootstrap support of 100 for neighbor joining and maximum likelihood trees, and a probability of 1 for the Bayesian tree) from the other families of the Methanomicrobiales, and our work extends the environmental distribution of the R-10 ⁄ Fen cluster from wetlands and wastewater sludges to stratified, freshwater lakes.

Methane oxidation pathway(s)
Methane oxidation in aquatic environments is catalyzed by bacteria and Achaea (Reeburgh, 2007;Bastviken et al., 2008). Microbial CH 4 oxidation is known to occur both aerobically (i.e., with O 2 ) and anaerobically with either SO 4 2) (Hinrichs et al., 1999;Boetius et al., 2000;Orphan et al., 2001) or NO 3 ) (Raghoebarsing et al., 2006;Ettwig et al., 2008) as the electron acceptor. Sulfate-dependent anaerobic oxidation of methane (AOM) is responsible for most CH 4 oxidation in marine sediments (Strous & Jetten, 2004) where it consumes more than 90% of the CH 4 produced (Reeburgh, 1996;Knittel & Boetius, 2009). As lake water sulfate concentrations are typically low, lakes emit globally more methane to the atmosphere than the ocean, and they are responsible for between 6 and 16% of total non-anthropogenic methane emissions (Bastviken et al., 2004). AOM coupled to denitrification has been observed in laboratory cultures (e.g. Ettwig et al., 2008), but its environmental significance is unknown. Under standard state, AOM with the (hydr)oxides of Fe and Mn is thermodynamically favorable, but conclusive evidence for Fe-or Mn-dependent AOM in the environment remains elusive (Zehnder & Brock, 1980;Gal'chenko, 2004;Konhauser et al., 2005;Nauhaus et al., 2005;Caldwell et al., 2008;Beal et al., 2009). Recently, Mn-and Fe-dependent AOM was postulated in laboratory incubations of S-poor marine sediments (Beal et al., 2009), but the results of this study do not constrain the role of Fe and Mn (hydr)oxides in methane oxidation. As noted by the authors, SO 4 2) may have been regenerated through the oxidation of sulfide coupled to reduction of the added Fe and Mn (hydr)oxides. In Lake Matano, the dearth of SO 4 2) (Crowe et al., 2008a,b) and NO 3 ) might preclude AOM, but rapid cycling of Fe and Mn within the chemocline may allow CH 4 oxidation to take place through Fe(III) or Mn(III ⁄ IV) reduction. We consider this possibility in detail below. A sharp gradient in the methane concentration (Fig. 5) between 100 and 200 m depth implies that methane oxidation occurs at or immediately below the persistent pycnocline. The oxidation of isotopically light, biogenic CH 4 should be reflected in the C isotopic composition of its oxidation prod-uct, RCO 2 . This is supported by our data, which show two distinct minima in the d 13 C RCO2 (points I and II in Fig. 6B) profile at depths of 104 and 122 m that, together, delineate the depth interval of methane oxidation. Similarly, heavier CH 4 (d 13 C CH4 shifted to less negative values) is observed between 110 and 115 m (Fig. 6A). The concentrations of methane and RCO 2 in the deep water are such that a complete oxidation of methane to CO 2 would produce a d 13 C RCO2 as low as )23&. The more positive values observed likely reflect a combination of dilution by heavier epilimnetic and deep water RCO 2 and the escape of a small fraction of CH 4 to the atmosphere.
The d 13 C RCO2 values observed at 104 and 122 m and the relatively heavier RCO 2 observed at 110 m generates a 'sawtooth' pattern that can only be explained by sources of isotopically light RCO 2 at 104 and 122 m and a sink of isotopically light RCO 2 at 110 m (Figs 6B and 7B for more detail). The obvious source for isotopically light RCO 2 in the vicinity of the persistent pycnocline is the oxidation of CH 4 . The sink for isotopically light RCO 2 at 110 m is likely generated by the preferential uptake of 12 C RCO2 during microbial C fixation or methanogenesis. Indeed, this positive excursion is coincident with the presence of a community of putative photoferrotrophic Chlorobiaceae capable of photosynthetic C-fixation (Crowe et al., 2008a). Regardless of its origin, the sink for isotopically light RCO 2 clearly delineates two spatially distinct zones of methane oxidation and calls for a discussion of the potential oxidants available within these zones (see Appendix 1 for a quantitative examination of the C fluxes and the resulting isotope ratios).
The peak in d 13 C RCO2 at point I (104 m, Fig. 6B) likely results from aerobic methane oxidation. Interpolation between O 2 concentrations measured at 102 m (54 lmol L )1 ) and 105 m (37 lmol L )1 ) yields an estimated O 2 concentration of 43 lmol L )1 at 104 m, which is well above typical half-saturation constants for aerobic methanotrophs (van Bodegom et al., 2001). Oxygen in Lake Matano becomes undetectable (£2.5 lmol L )1 ) below 110 m (Figs 3 and 7A), and the peak in d 13 C RCO2 at point II (122 m, Fig. 6B) indicates that CH 4 oxidation occurs in the presence of little or no oxygen. Aerobic methane oxidation below our oxygen detection level may be possible, but it should be kinetically inhibited by the high Fe(II) concentrations (Appendix 2). Our calculations (Appendix 2) reveal that the kinetics of abiotic Fe(II) oxidation at circum-neutral pH is sufficiently rapid that Fe(II) competes with methanotrophic bacteria for molecular oxygen, potentially favoring anaerobic methane oxidation. Based on the known kinetics of aerobic methanotrophy (van Bodegom et al., 2001(van Bodegom et al., , 2004 and abiotic Fe(II) oxidation (Millero et al., 1987), the maximum potential rate of aerobic methane oxidation at $120 m depth is 4.5 · 10 )8 mol L )1 day )1 . This rate is one order of magnitude lower than the methane oxidation rate (5 · 10 )7 mol L )1 day )1 ) necessary to sustain the d 13 C RCO2 peak at point II (122 m) against the diffusion of heavier RCO 2 (see Appendix 1 for details on the calculation of CH 4 oxidation rates from d 13 C RCO2 profiles). Furthermore, the maximum rate of aerobic methane oxidation calculated in Appendix 2 assumes that aerobic methanotrophic bacteria occur at moderately high cell densities of 10 5 cells mL )1 . PCR failed to amplify DNA of known aerobic methanotrophs from cellular material collected at 118 m (Table 3). Unless cellular material was present in the form of very small, free living cells (<0.45 lm) that were not captured on our filters and, thus, overlooked in our genetic assays, the PCR data indicate that the cell density of known aerobic methanotrophs must be very low and the rates of aerobic methane oxidation at and around 122 m depth must be insignificant. Taken together, the available evidence leads us to conclude that some methane oxidation takes place below the depth of oxygen penetration and is, therefore, anaerobic.
Within the depth interval of putative anaerobic methane oxidation in Lake Matano, both sulfate and nitrate occur at exceedingly low concentrations, unlike Fe and Mn (Figs 3  and 4). Methane oxidation coupled to the reduction of Fe or Mn (hydr)oxides is energetically favorable under standard conditions: Reactions 1 and 2 are also favorable over a broad range of environmental conditions, including those encountered within the zone of CH 4 oxidation in Lake Matano (see Appendix 3, Fig. 10). Given the slow oxidation kinetics of CH 4 under most environmental conditions, CH 4 oxidation coupled to Mn or Fe reduction would likely need to be biologically catalyzed. Indeed, sterilized controls failed to elicit methane oxidation activity in incubations of marine sediments, purportedly occurring via the reduction of Fe and Mn (Beal et al., 2009).
The minimum energy yield required for organisms to conserve energy for growth from a chemical reaction is thought to be around 15 kJ mol )1 , based on ATP generation using four protons per ATP molecule (Caldwell et al., 2008). Given the chemical conditions present at 120 m in Lake Matano, under standard state (25°C and concentration = activity), reactions 1 and 2 yield free energies of )166.8 and )610.9 kJ mol )1 , respectively, well in excess of the minimum microbial energy requirement. Accordingly, both reactions could support microbial growth down to sub-nanomolar CH 4 concentrations under circum-neutral pH conditions, indicating that the pathway could operate in a broad range of sedimentary and soil environments at low CH 4 concentrations (see Appendix 3, Fig. 10). In Lake Matano, AOM coupled to Fe(III) or Mn(III ⁄ VI) reduction provides a thermodynamically feasible alternative to highly active aerobic methanotrophs working at the exceedingly low O 2 concentrations encountered near 122 m depth. The following discussion considers the rates of CH 4 oxidation and Fe(III) and Mn(III ⁄ IV) supply in the pycnocline region. The sharp Mn(II) and Fe(II) concentration gradients in Lake Matano that are maintained by oxidation drive large fluxes of Mn and Fe near the persistent pycnocline and should generate an abundant supply of both authigenic Fe and Mn (hydr)oxides in the vicinity of point II (Figs 3 and 6B). The Fe(II) concentration gradient between 120 and 110 m (0.003 mol m )4 ) translates to volume-specific Fe oxidation rates of 5 · 10 )8 mol L )1 day )1 , assuming that oxidation occurs over a 10 m depth interval and a vertical diffusivity of 0.16 m 2 day )1 (Katsev et al., 2010). Thus, to generate sufficient authigenic Fe(III) to oxidize 5 · 10 )7 mol L )1 day )1 CH 4 requires that every Fe atom be recycled eighty times within the pycnocline region. Fe and Mn are known to be recycled hundreds of times in bioturbated marine sediments where the activity of macrobenthic fauna mixes reduced Fe and Mn into the oxic layers of the sediment (Canfield et al., 1993). In Lake Matano, such recycling could be achieved through the upward flux of Fe(II) and Mn(II), their oxidation at the oxycline, and a return flux of particulate Fe(III) and Mn(III ⁄ IV). Alternatively, photosynthetic Fe oxidation (photoferrotrophy) could regenerate Fe(III) anaerobically. Such a tightly coupled Fe-CH 4 cycle, whereby photogenic Fe (hydr)oxides are rapidly reduced during CH 4 oxidation, could account for the apparent lack of particulate Fe(III) in the vicinity of the persistent pycnocline. Similar arguments could be made for Mn cycling. However, to our knowledge, phototrophic Mn oxidation has never been documented. Another possibility is that short-lived events, such as heavy rainfalls, which were missed by our discrete sampling intervals, deliver pulses of particulate material from the catchment soils to the water column. These events could supply Fe(III) and Mn(VI) intermittently to the chemocline and fuel episodic methane oxidation. A preliminary examination of the chemistry of Fe in the lateritic catchment soils reveals that these soils contain up to 25 wt.% ascorbate-extractable Fe (A. Mucci and S. A. Crowe, unpublished data), and a significant portion of the Fe washed into the lake would be available for microbial reduction.
We can also estimate the total rate of Fe supply to the lake based on a sedimentary Fe content of 20 wt.% (Crowe et al., 2004) and a deepwater sedimentation rate of 0.08 cm year )1 (Crowe et al., 2004). Sediment water contents of 90% and a particle density of 2.97 g cm )3 (S. A. Crowe, unpublished data) translate to an Fe burial rate of 8.5 · 10 )3 mol m 2 day )1 , a factor of 5 greater than the upward flux of CH 4 (1.6 · 10 )3 mol m )2 day )1 , see next section). These Fe burial rates reveal that AOM coupled to Fe-reduction could account for a maximum of 66% of total methane oxidation, based on the stoichiometry of reaction 1. The amount of Fe available for CH 4 oxidation would also depend on its reactivity towards microbial reduction and on competition with microbes using Fe(III) as an electron acceptor in organic matter oxidation. Due to the sparseness of the sediment data, these estimates should be taken as preliminary. They also assume no recycling of Fe(III), which could increase the amount of CH 4 oxidized via Fe(III).
The rates we calculate for AOM in Lake Matano (5 · 10 )7 mol L )1 days )1 , See Appendix 1 for calculations) are lower than the rates of alleged Fe-and Mn-dependent AOM observed in marine sediment incubations (Beal et al., 2009). Thus, the Fe-and Mn-reducing, methane-oxidizing micro-organisms in Lake Matano appear to operate at rates consistent with the observed physiological capabilities of other such putative organisms. A lack of phylogenetic affiliation between our clone sequences and the ANME does not rule out AOM, and the phylogenetic distribution of methaneoxidizing Archaea may extend beyond known ANME. Furthermore, the micro-organisms involved in NO 3 ) -dependent AOM are not closely related to the ANME (Raghoebarsing et al., 2006) and may be exclusively bacterial (Ettwig et al., 2008). A similar explanation was put forth to account for decreases in the relative abundance of ANME as a function of incubation time in laboratory experiments with marine sediments (Beal et al., 2009). Further resolving the rates and pathways of CH 4 oxidation in Lake Matano will require measurements of gross CH 4 oxidation rates, gross rates of anoxygenic photosynthesis, organism culturing, and better vertical sampling resolution.

Balancing carbon budgets
The concentration profiles of CH 4 observed in 2006 and 2007 were identical within the precision of our sampling and analyses. Hence, the CH 4 cycle in Lake Matano is at steady state over the time scale of our observations. At steady state, the rate at which methane is supplied from the sediments and deep waters equals the rate at which it is oxidized in the vicinity of the persistent pycnocline. The latter can be estimated from the fluxes of methane between 200 and 100 m depth. Using an average vertical diffusion coefficient of 0.16 m 2 day )1 , estimated from the Brunt-Väisälä stability frequency (Crowe et al., 2008b;Katsev et al., 2010), and a methane concentration gradient of 9.7 · 10 )3 mol m )4 between 200 and 100 m depth, our calculation yields a methane oxidation rate of 1.6 · 10 )3 mol m )2 day )1 . The rates of methane production and methane oxidation in Lake Matano are relatively high and comparable to those observed in the Florida Everglades (King et al., 1990) and freshwater tidal swamps (Megonigal & Schlesinger, 1997). At this rate, methane concentrations in the deep waters would reach the present levels within about 1000 years following a hypothetical mixing event (i.e., overturn). This suggests that the lake has not experienced complete mixing in the last millennium, a time frame consistent with the 14 C CH4 measurements.
The steady-state organic matter degradation rate in Lake Matano was estimated at 3 · 10 )3 mol m )2 day )1 , based on the ammonium concentration gradient (Fig. 4A), an organic matter C:N ratio of 10, and assuming steady-state. The settling organic matter C:N ratio of 10 was calculated from the ratio of the difference between the RCO 2 in the surface and bottom water and the NH 4 + concentrations in the bottom water. This C:N ratio is consistent with the measured C:N ratio of 12 determined from the elemental analysis of POM recovered from 32 m depth. A comparison of the rates of methanogenesis using the stoichiometries of acetate disproportionation and hydrogenotrophic methanogenesis reveals that 50 or 100%, respectively, of carbon degradation occurs through methanogenesis. Some intermediate value is most likely, given our isotopic data (calculated isotope separation values) and the methanogen ecology. Clearly, methanogenesis and methanotrophy play a key role in Lake Matano's carbon cycle. Integrated primary production rates in the surface waters (3.8 · 10 )3 mol C m )2 day )1 ) (Crowe et al., 2008a) indicate that autochthonous primary production can supply sufficient carbon for the observed rates of organic matter degradation and methanogenesis. The agreement between the computed and measured C:N ratios and a C:N ratio of 49 for allochthonous leaf litter further suggests that respiration and methanogenesis in Lake Matano are largely fueled by autochthonous POM. The difference between the total organic carbon degradation and primary production rates reveals that roughly 78% of the organic matter generated by primary production is degraded, and the rest is buried in the sediment at an approximate rate of 8.4 · 10 )4 mol m )2 day )1 . Lake Matano sediments contain between 5 and 8 wt.% organic carbon, which, given a deep water sedimentation rate of 0.08 cm year )1 (Crowe et al., 2004), translates to organic carbon accumulation rates of between 8.5 and 13 · 10 )4 mol C m )2 day )1 , consistent with the above estimate.
The minimum fraction of organic matter degraded through microbial Fe reduction can be estimated by considering the flux of Fe(II) through the pycnocline. Again, we assume that steady state conditions and a diffusion coefficient of 0.16 m 2 day )1 apply. The upward flux of Fe(II) is driven by the concentration gradient sustained by Fe(II) oxidation reactions at or near the pycnocline. At steady state, it equals the rate of Fe(II) production via bacterial Fe(III) reduction in the bottom waters and sediments. However, only a fraction of the total Fe(II) produced in the bottom waters and sediments is consumed through oxidation within the pycnocline. Fe(II) is also likely removed by precipitation as siderite (FeCO 3 ), with which the deep waters of Lake Matano are supersaturated, and ⁄ or via sorption to sedimentary particles. Fe(II) production rates calculated from the oxidative Fe(II) flux underestimate the total Fe(II) production rates, as they do not include Fe(II) sequestered by precipitation ⁄ adsorption. A minimum estimate for the carbon degradation rate by Fe reduction, calculated from the flux of Fe(II) at the pycnocline, is 2.2 · 10 )4 mol m )2 day )1 , less than 10% of the estimated total organic C degradation rate (3 · 10 )3 mol m )2 day )1 ) and only 7-14% of the C degraded by methanogenesis (1.6 · 10 )3 mol m )2 day )1 ), depending on the pathway considered.

Methane in Fe-rich environments and the Archean Ocean
The prominence of methane in the carbon cycle of Lake Matano demonstrates that methanogenesis and methanotrophy can be viable, and even major, metabolic pathways in ferruginous ecosystems. Although ferruginous ecosystems are rare on the modern earth, Fe-rich, S-poor conditions have dominated ocean chemistry throughout much of Earth's history, particularly during the Archean Eon (Canfield et al., 2006). Geological evidence for the early evolution of methanogens and their activity in ferruginous Archean Oceans (Canfield et al., 2006) is provided by findings of isotopically light carbon in Archean sedimentary kerogens (Hayes, 1983), which is diagnostic of Archaeal methanogenesis and coeval methanotrophy, and CH 4 trapped in fluid inclusions (Ueno et al., 2006). It has been proposed that the lack of C isotopic evidence for methanogenesis and methanotrophy in the carbonate minerals preserved in banded iron formations (BIF) could be a consequence of anaerobic oxidation of CH 4 coupled to Fe reduction (Konhauser et al., 2005). Based on the equilibrium carbon isotope fractionation factor for the siderite-HCO 3 ) (aq) system (10 3 lna = 0.5 ± 0.2&) (Jimenez-Lopez & Romanek, 2004), the d 13 C RCO2 ()7& to )10&) in the anoxic Lake Matano water column indicates that methanogenesis followed by methane oxidation under ferruginous conditions could yield carbonate minerals with a C isotopic composition of )6& to )10&. Such values are well within the range of )5& to )15& observed for carbonate minerals in BIFs (Becker & Clayton, 1972;Baur et al., 1985;Konhauser et al., 2005). Hence, the absence of a light d 13 C signature in BIF carbonates does not preclude methanogenesis during BIF diagenesis but may be indicative of anaerobic methane oxidation, possibly coupled to Fe reduction as previously suggested (Konhauser et al., 2005). Though the evidence is mounting, the tenability of such hypotheses clearly rests on a conclusive demonstration of Fe-and Mn-dependent AOM.

CONCLUSIONS
Biogenic CH 4 accumulates in Lake Matano's deep waters to concentrations exceeding 1.4 mmol L )1 , amongst the highest recorded in natural waters and comparable to concentrations in lakes receiving volcanic gas inputs, such as Lake Kivu. The sum of the partial pressures of dissolved CO 2 and CH 4 in the bottom waters of Lake Matano does not exceed 6% of the in situ hydrostatic pressure. Thus, ebullition at depth is unlikely, and the risk of gas-driven limnic eruptions is low. Despite the abundance of Fe(III) solid phases in Lake Matano, methanogens effectively compete for substrates and are responsible for recycling up to 50% of the C fixed during primary production. Pelagic methanogenesis was confirmed by the presence of methanogens from the R-10 ⁄ Fen cluster and of the family Methanosaetaceae within the water column. Given the isotopic compositions of CH 4 and CO 2 and the methanogen ecology, methanogenesis in Lake Matano may occur via both disproportionation of acetate and reduction of CO 2 using H 2 as an electron donor. Upward diffusing CH 4 is oxidized in the vicinity of the persistent pycnocline, and some methane oxidation occurs, seemingly in the absence of O 2 , SO 4 2) , and NO 3 ) but in the presence of Fe and Mn (hydr)oxides. This suggests AOM coupled to Fe or Mn reduction in Lake Matano, but positive identification requires further studies. 3.5 · 10 )6 to 1.0 · 10 )4 mol L )1 [O 2 ] 1 0 )9 to 10 )4 mol L )1 Cell density 10 )4 to 10 )6 cells mL )1 C content of methanotroph cell 6.23 · 10 )14 mol C cell )1 van Bodegom et al., (2001) For typical parameter values (Table 4), competition for oxygen between aerobic methanotrophy and Fe(II) oxidation is illustrated graphically in Fig. 9A. Irrespective of cell density, methanotrophic oxygen consumption plateaus when the oxygen concentration exceeds 1 lmol L )1 , whereas abiotic O 2 consumption by Fe(II) is log-linear at all O 2 concentrations. The percentage of O 2 consumption by methanotrophs vs. total O 2 consumption by methanotrophy and Fe(II) oxidation as a function of ambient O 2 concentration is shown in Fig. 9B for a range of Fe(II) concentrations and a cell density of 10 5 cells mL )1 . At low O 2 concentrations, a higher percentage of oxygen is consumed by methanotrophs than by abiotic Fe(II) oxidation. At methanotroph cell densities of 10 5 cells mL )1 , typical of microaerophilic zones in methane-rich chemoclines (Eller et al., 2005;Lehours et al., 2005), and Fe concentrations of 35 lmol L )1 , methanotrophy consumes approximately 50% of the available O 2 at O 2 concentrations less than 1 lmol L )1 .
Using the Monod expression for aerobic methanotrophy (equation 6), cell densities up to 10 5 cells mL )1 , and oxygen concentrations up to 5 nmol L )1 (the O 2 concentration at 120 m depth was estimated at 4.7 nmol L )1 by fitting a logarithmic function (R 2 = 0.92) to the O 2 concentration profile between 105 and 110 m and extrapolating to 120 m depth), we estimate that the rate of aerobic methane oxidation is at most 3.1 · 10 )11 mol L )1 min )1 . Note that using the measured Fe(II) concentrations and the solubility limit of Fe(III) at pH 7.21 gives an estimated pe of 5.58 which, at equilibrium, translates to O 2 concentrations of 1.4 · 10 )26 nmol L )1 . Hence, unless O 2 is rapidly supplied to 120 m depth, it is likely present at concentrations much lower than the rough estimates gleaned from the O 2 concentration profile. Table 5 Free energy yield (kJ mol )1 ) for reaction 5 assuming that concentrations are equal to activities (i.e., ideal solution) and a pH of 7.0 CH 4(aq) (M)  Table 6 Free energy yield (kJ mol )1 ) for reaction 6 assuming that concentrations are equal to activities (i.e., ideal solution) and a pH of 7.0 APPENDIX 4 (BACTERIAL PHYLOGENY) Figure 11 shows the phylogeny of non-chimeric clones sequenced from PCR products generated using primers to target known aerobic methanotrophs. These clones cluster within the b subdivision of the proteobacteria, whereas members of Type I and Type II methanotrophs all belong to the c and a subdivisions, respectively.