Methane-related Changes in Prokaryotes along Geochemical Profiles in Sediments of Lake Kinneret (israel)

Microbial methane oxidation is the primary control on the emission of the greenhouse gas methane into the atmosphere. In terrestrial environments, aerobic methan-otrophic bacteria are largely responsible for this process. In marine sediments, a coupling of anaerobic oxidation of methane (AOM) with sulfate reduction, often carried out by a consortium of anaerobic methanotrophic archaea (ANME) and sulfate-reducing bacteria, consumes almost all methane produced within those sediments. Motivated by recent evidence for AOM with iron(III) in Lake Kinneret sediments, the goal of the present study was to link the geochemi-cal gradients in the lake porewater to the microbial community structure. Screening of archaeal 16S rRNA gene sequences revealed a shift from hydrogenotrophic to acetoclas-tic methanogens with depth. The observed changes in microbial community structure suggest possible direct and indirect mechanisms for the AOM coupled to iron reduction in deep sediments. The percentage of members of the Nitrospi-rales order increased with depth, suggesting their involvement in iron reduction together with Geobacter genus and " reverse methanogenesis ". An indirect mechanism through sulfate and ANME seems less probable due to the absence of ANME sequences. This is despite the abundant sequences related to sulfate-reducing bacteria (Deltaproteobacteria) together with the occurrence of dsrA in the deep sediment that could indicate the production of sulfate (disproportionation) from S 0 for sulfate-driven AOM. The presence of the functional gene pmoA in the deep anoxic sediment together with sequences related to Methylococcales suggests the existence of a second unexpected indirect pathway – aerobic methane oxidation pathway in an anaerobic environment.


Introduction
Chemical profiles in the porewater of aquatic sediments reflect the sequence of microbially mediated redox reactions that are driven by the availability of both electron donors and of suitable electron acceptors.The latter are depleted in the order of decreasing chemical potential, beginning with oxygen and proceeding through nitrate, manganese and iron oxides, and then sulfate.Below the main zone of sulfate reduction, the fermentation of organic carbon leads to the formation of methane (CH 4 ) through methanogenesis (Froelich et al., 1979).
The produced methane is isotopically depleted in 13 C, with values of ∼ −50 to −110 ‰ (Schoell, 1988), and the residual dissolved inorganic carbon (DIC) pool is enriched by an isotopic fractionation factor (ε) of 50 to 70 ‰ (see, e.g., Borowski et al., 2000;Whiticar, 1999).When the produced methane diffuses into a zone with a suitable electron acceptor, it can be consumed by microbial oxidation (methanotrophy), the main process by which the important greenhouse methane is prevented from escaping into the atmosphere.While in the terrestrial environment, aerobic methanotrophy is the dominant process (Chistoserdova et al., 2005), in anaerobic marine sediments, archaea are found to consume the majority of upward-diffusing methane coupled to sulfate reduction (Knittel and Boetius, 2009;Thauer, 2010;Valentine, 2002).
Evidence from lipids and from fluorescence in situ hybridization (FISH) showed that a consortium of archaea and sulfate-reducing bacteria are involved in this anaerobic methane oxidation (AOM; Boetius et al., 2000;Hinrichs et al., 1999;Orphan et al., 2001).To date, three groups of anaerobic methanotrophic archaea (ANME), named ANME-1, ANME-2, and ANME-3, are known to perform sulfatedriven AOM (Niemann et al., 2006;Orphan et al., 2002).Recently, Milucka et al. (2012) demonstrated AOM mediated solely by archaea, where the archaea were shown to oxidize the methane and reduce the sulfate to elemental sulfur.Disproportionating bacteria, also involved in this mechanism, oxidize and reduce this elemental sulfur to sulfate and sulfide, respectively.The carbon isotopic fractionation factor (ε) for this methanotrophic process was shown to be in the range of 4-30 ‰ (Kinnaman et al., 2007;Whiticar, 1999).Enrichment cultures of ANME from different environments showed a carbon isotopic fractionation of 12-39 ‰ (Holler et al., 2009).
Other electron acceptors were recently shown to drive AOM.Nitrite-driven AOM through oxygenic bacteria was observed in two different freshwater ecosystems in the Netherlands (in canal sediments) (Ettwig et al., 2009;Raghoebarsing et al., 2006) and also in peatlands (Zhu et al., 2012).Beal et al. (2009) showed the potential of manganeseand iron-driven AOM in marine sediments, and Egger et al. (2015) showed it in brackish costal sediments.In our recent study (Sivan et al., 2011), we provided in situ geochemical evidence for AOM coupled to microbial iron reduction below the main methanogenesis zone in Lake Kinneret (LK) sediments, where dissolved sulfate and nitrate are absent.However, the mechanism that is responsible for this process was not investigated.The goal of the present study is to explore the possible microbial communities that may be involved in the iron-driven AOM in LK sediments.This is accomplished by combining chemical and isotope analyses of porewater samples along a depth profile from LK sediments with molecular biological techniques.By using specific functional genes related to dissimilatory sulfate reductase (dsr) for sulfate reducers (Klein et al. 2001), methylcoenzyme M reductase (mcr) for methanogens and anaerobic methanotrophs (Hallam et al., 2003) and particulate methane monooxygenase (pMMO) for aerobic methanotrophs (Mc-Donald et al., 2008), we pinpoint the most likely candidate for this relatively unknown process.

Study site
Located in northern Israel, Lake Kinneret (LK, Fig. 1) is a warm monomictic subtropical lake.Typical concentrations of major electron acceptors in the water column during the mixed period are 35-50 µM nitrate and 600 µM sulfate (Adler et al., 2011;Nishri et al., 2000;Serruya et al., 1974).In the spring, the newly formed epilimnion is characterized by increasing temperatures and enhanced phytoplankton development, while in the hypolimnion heterotrophic microorganisms gradually deplete oxygen and then nitrate.Organic matter degradation by bacterial iron and manganese reduction takes place below the thermocline in the summer; at the end of the stratification period, sulfate reduction starts in the bottom water.In the upper part of the sediment, sulfate reduction is the dominant microbial process year round, and below depth of 5 cm this process is mostly replaced by methanogenesis (Adler et al., 2011;Eckert and Conrad, 2007).Total iron content (Fe (tot) ) in the sediment increases with depth till 18 cm and then remains uniform around 550 µmol g −1 dry weight (Eckert, 2000).Mn(II) concentration is about 13 µmol g −1 dry weight in the sediment (Serruya, 1971).The total carbon (C tot ) in the solid phase shows a decrease from 14 µmol g −1 dry weight in the top part of the sediment to 8 µmol g −1 dry weight in the deep part of the sediment (∼ 30 cm; Eckert, 2000).

Sampling
Several sediment cores were collected from the center of the lake (Station A, Fig. 1) at a water depth of ∼ 42 m (maximum lake water column depth) at different times using Perspex tubes, measuring 55 cm long by 5 cm in diameter, with a gravity corer.The cores were stored in the dark at 4 • C until they were sliced (on the same day or the day after).Core sampling for the microbial community study took place in December 2009.Over a dozen porewater chemical and isotope profiles were conducted seasonally (every 3-4 months) from 2007 to 2013.The slight seasonal changes allowed the use of typical geochemical profiles (Adler et al., 2011) in order to sample for the microbiology communities in the different electron acceptors zones and to find correlations between the microbial sampling and the geochemical profiles.

Geochemical analyses
Cores were divided into 2 cm slices from top to bottom under a constant flow of N 2 using a slicing device (Eckert, 2000).About 1.5 mL of each sediment slice was transferred into N 2 - flushed crimp bottles containing 5 mL of 1.5 N NaOH for the headspace measurements of CH 4 and δ 13 C CH 4 (after Sivan et al., 2011).CH 4 concentrations were measured on a SHI-MADZU gas chromatograph (GC 8IF) equipped with a flame ionization detector (FID) at a precision of 2 µmol L −1 .
Porewater was extracted immediately from each slice by centrifugation at 27 000 g at 4 • C in a N 2 atmosphere, and the supernatant was filtered through 0.45 µm filters.A 1 mL subsample was fixed with ferrozine and analyzed for dissolved Fe(II) (Stookey, 1970).Four millimeters of subsample was poured into an acidified vial (1 mL of 0.5 N nitric acid) to measure dissolved Mn(II).The sample was analyzed via an inductively coupled plasma mass spectrometer (ICP-MS, Elan DRC II, Perkin Elmer) at a precision of ±10 %.For sulfide profiles, 1 mL of subsample was added to zinc acetate, and hydrogen sulfide concentrations were determined by the methylene blue method, where the MDL (method detection limit) is 1 µM (Cline, 1969).For sulfate measurements, 5 mL of the subsamples was analyzed with a Dionex DX500 highpressure liquid chromatograph with a precision of ±3 %.Iron isotope analysis was done by acidifying the subsamples with 10 % HCl for 1 week (to dissolve any precipitated iron) and then purifying them through anion exchange chromatography (Borrok et al., 2007).δ 56 Fe was measured on a Neptune multi-collector ICP-MS in high-resolution mode according to standard methods and standardized against isotopic reference material (IRMM-014) with a precision of ±0.1 ‰ (John and Adkins, 2010).Total lipids were extracted using the Bligh-Dyer procedure (Bligh and Dyer, 1959) with solvent mixture of 2 : 1 : 0.8 (methanol : dichloromethane : buffer).δ 13 C of the total lipid extraction (TLE) was measured on an elemental analyzer-isotopic ratio mass spectrometer (EA-IRMS) with a precision of 0.1 %.

DNA extraction and quantitative PCR (qPCR) amplification from sediment samples
Part of the sediment slices were kept frozen at −20 • C for the microbial work.Samples from three different depths (0-3 cm representing the sulfate reduction zone, 6-9 cm representing the methanogenesis zone and 29-32 cm representing the deep AOM zone) were defrosted.Those depth zones were chosen based on the geochemical profiles sampled at different times, showing quasi-steady state and stabilization of the sediments (Adler et al., 2011;Sivan et al., 2011).Therefore, microbial community sampling of the three depth zones at a single point in time should represent the different microbial habitats as they pertain to the respective geochemical zones.Total genomic DNA was extracted from the sediment samples using the MO BIO PowerSoil DNA Isolation Kit (MO BIO Laboratories, Solana Beach, CA).Genomic DNA was eluted using 60 µL of elution buffer and stored at −20 • C. Concentrations of DNA were determined via UVvisible spectrophotometry (ND-1000 NanoDrop Technologies, Wilmington, DE; sulfate reduction zone 22, methanogenesis zone 35.8 and AOM zone 14 ng µL −1 ).Quantification of functional genes was performed using mcrA primers (Luton et al., 2002), Forward GGTGGTGTMGGATTCACACARTAYGCWACAG and Reversed TTCATTGCRTAGTTWGGRTAGTT , dsrA primers, DSR1F ACSCACTGGAAGCACG (Wagner et al., 1998) and RH3-dsr-R GGTGGAGCCGTGCATGTT (Ben-Dov et al., 2007) and pmoA primers, a189F GGNGACTGGGACTTCTGG and mb661R CCGGMGCAACGTCYTTACC (Yan et al., 2006) with an ABI PRISM 7000 Sequence Detection System (Applied Biosystems).The qPCR reaction consisted of 10 µL of ABsolute Blue SYBR Green ROX, 150 nM each of forward and reverse primers and 5 µL of each DNA template.Thermal cycling conditions were as follows: 15 min at 95 • C for enzyme activation, followed by 40 rounds of 15 s at 95 • C for denaturation and 1 min at 60 • C for annealing/extension.To verify that the used primer pair produced only a single specific product, a dissociation protocol was added after thermocycling to determine dissociation of the PCR products from 60 to 95 • C. Standards for the calibration curves for quantification were made using pGEM-T Easy plasmid cloned with 1.9 kb dsrA and amplicons of the mcrA and pmoA encoding for functional genes at a known concen-I.Bar Or et al.: Methane-related changes in prokaryotes along geochemical profiles tration with six serial dilution points (in steps of 10-fold).All runs included a no-template control.Plasmid standards and environmental samples were simultaneously assayed in triplicates.The ABI PRISM 7000 Sequence Detection System and SDS Software were used for data analysis.QC (quality control) values were exported into a Microsoft Excel worksheet for further statistical analysis.

Sequence analysis
454 tag-encoded amplicon pyrosequencing was performed by the Research and Testing Laboratory (Lubbock, Texas, USA) as previously described (Dowd et al., 2008).The bacterial and archaeal 16S rRNA gene primers that were used are 28F GAGTTTGATCNTGGCTCAG and 519R GTNTTACNGCGGCKGCTG , and Arch349F GYGCASCAGKCGMGAAW and Arch806R GGACTACVSGGGTATCTAAT , respectively.
Data analysis was made using two different methodologies (MOTHUR and SILVA ngs).The initial trimming of the sequences was made by MOTHUR v1.33 (Schloss et al., 2009) and generated around 300 bp of sequences.The trimmed sequences were taken to further analysis by MOTHUR (Schloss et al., 2011) using SILVA NR v119 database.MOTHUR 454 pipeline filtration and denoising sequences remove from the analysis that were < 150 bp, when they contained homopolymers longer than 8 bp, ambiguous bases, more than one mismatch to barcode sequences or more than two mismatches to the forward primer sequence.We further removed sequences that did not align in the same nucleotide position on the reference database.
Most of the bacterial sequences were in the same region; however, the archaeal sequences were spaced between two different regions.We used one position window that was dominant in the top sample and implied it to the other archaeal analysis in order to better compare the data set between other samples of the same position window.Identical sequences were grouped and then were aligned against SILVA.nr_v119.Chimeras were removed using MOTHUR CHIMERA UCHIME (Edgar et al., 2011).The lengths of the remaining sequences were around 200 bp.A further screening step (pre-cluster) was applied to reduce sequencing noise by clustering reads differing by only 1 base every 100 bases (Huse et al., 2010).In addition, the rest of the sequences were classified in order to remove eukaryote, mitochondria and chloroplast classified sequences.The remaining sequences were used to generate a distance matrix and clustering into operational taxonomic units (OTUs) defined at 97 % cutoff using the average neighbor algorithm.The OTUs were classified using the SILVA.nr_v119database with a confidence threshold of 80 %.Alpha diversity was calculated by MOTHUR using the remaining sequences at 97 % similarity.The beta diversity was calculated using comparable data; the number of sequences per sample was made equal through subsampling (Tables 1 and 2).
The second program which we used for analysis was SILVA ngs pipeline (Quast et al., 2013).The trimmed sequences from MOTHUR were Aligner against the SILVA SSU rRNA seed.Sequences shorter than 50 aligned nucleotides and with more than 2 % of ambiguities or 2 % of homopolymers, respectively, were removed.The sequences that were not aligned as being putative contaminations/artifacts were removed.Then sequences were clustered to OTUs with 98 % similarity and classified by local nucleotide BLAST search against SILVA database v119.To filter out low identity and artificial BLAST hits, hits for which the function (% sequence identity + % alignment coverage)/2 did not exceed the value of 93 % were discarded.Sequences with weak, low scores were classified as "No Relatives" and were disregarded thereafter.Standard deviation betweena the percentages of the two classification showed that at the phylum level there are small differences between the two pipelines (Table S1 in the Supplement).In the order classification, the standard deviations increased but the sequences still showed close similarity.Estimates of phylotype richness, diversity coverage and similarity were calculated according to the abundance-based coverage estimate (ACE), Chao's estimator (Chao, 1984;Chao and Ma, 1993), the Shannon diversity index, and Good's coverage (Good, 1953) in MOTHUR (Tables 1 and 2).The distribution and abundance matrix of the OTUs was normalized to the sample with the smallest number of reads by randomly resampling the MOTHUR data set for statistical comparisons.Raw sequencing data were deposited in the MG-RAST (metagenomics.anl.gov)archive.

Microbial community structure related to the environment conditions
To estimate community similarity among samples, we used PC-ORD 6 software.Subsample OTU data from MOTHUR was used to create a distance matrix based on the Sørensen (Bray-Curtis) dissimilarities of the OTU composition of the samples.The data were normalized to percentages before the analysis.Community relationships were visualized using principal coordinate analysis (PCoA) based on this distance matrix.Different environmental variables were added as well (CH 4 , δ 13 C CH 4 , Fe(II), δ 56 Fe(II), Mn(II), H 2 S, SO 2− 4 , dissolved organic carbon (DOC), δ 13 C TLE , NH + 4 and PO 3− 4 ) in order to identify potential explanatory variables.The environmental vectors were applied and projected to the PCoA ordination.In addition, Venn diagrams for graphical descriptions of unshared and shared OTUs among the three samples were constructed using MOTHUR.

Porewater profiles
This study focused on microbial community shifts along the porewater profiles of electron acceptor gradients related to the methane cycle.To characterize shifts in main electron acceptors with depth, over a dozen porewater profiles were performed seasonally from station A (Fig. 1).The shown profiles (Fig. 2) are representatives of the slight seasonal trends.Sulfate and sulfide profiles show depletion from ∼ 100 µM at the top part of the sediment, to below detection limits within the upper 15 cm (Fig. 2a).This typical concave curvature profile of the sulfate profile in the porewater indicates intensive sulfate reduction in the upper few centimeters of the sediment throughout the year.A typical LK methane concentration profile (Fig. 2b) shows increase from 250 µM at the water-sediment interface to a maximum of about 1.25 mM in the depth range of 7 to 15 cm, and then a decrease below 15 cm depth.The profile of δ 13 C CH 4 (Fig. 2b) shows a decrease from −60 ‰ at a depth of 1 cm to about −65 ‰ at a depth of 7 cm and then an increase in the deeper sediments to a maximum value of −53.5 ‰ at a depth of 25 cm.Also, the profile of δ 13 C TLE (Fig. 2c) shows a decrease in this deepest part of the sediment.
Manganese and iron oxides are the most probable electron acceptors in the deeper part of the sediment where methane is decreased.Dissolved Mn(II) concentration (Fig. 2d) increased from 5 µM at the top of the sediment to a plateau of about 23 µM from depths of 23 to 36 cm.The dissolved Fe(II) concentration profile also showed an increase with depth (Fig. 2d), although following a different pattern.In the upper 15 cm, dissolved Fe(II) concentrations were below the detection limit, and gradually increased below 15 cm to about 90 µM at a depth of 36 cm.The δ 56 Fe profile (Fig. 2c) showed a decrease with depth from 0.5 ‰ in the upper part to −1.7 to −2.3 % in the deeper part of the sediment.

Sediment microbial communities
To study the sediment depth-dependent microbial community shifts, 16S rRNA gene sequences of the prokaryotic community from three different depth zones (0-3, 6-9 and 29-32 cm, see arrows in Fig. 2a) were extracted and analyzed.As mentioned above, these three zones represent the main sulfate reduction zone, the methanogenesis zone and the deep iron-driven AOM zone, respectively.The diverse bacterial and archaeal communities at these depth zones varied in composition and richness throughout the sediment.A total of 13227 and 4881 bacterial and archaeal sequences, representing 3852 and 705 OTUs (cut-off value of 98 %), respectively, were identified by the SILVA database project.The normalized Shannon and Chao index were used as proxies for diversity and richness, respectively (Tables 1 and 2).The upper layer showed lower diversity (Shannon index, 3.91 and 1.8 for bacteria and archaea) and richness (Chao index,925.7 and 83.4 for bacteria and archaea) than the other depths.Moreover, the bacterial coverage of the upper layer was the highest (91 %), while that of the deeper layers was about 80 %.The archaeal coverage was almost the same (∼ 88 %) in all the sampled layers.Most of our sequences were affiliated with uncultured microorganisms mainly from various sediment environments.
Although the bacterial OTUs were distributed over 43 phyla, we present here only the phyla that were over 1 % sequences of the 454 library.The most abundant phyla were Proteobacteria, Chloroflexi, Nitrospirae, Bacteroidetes, Firmicutes and Chlorobi observed in at least one of the libraries (Fig. 3).
The communities of microorganisms in the top layer (0-3 cm) of the sediment had less sequence overlap with those of the deeper layers (70 and 33 shared bacterial sequences with the middle and bottom layers, respectively) and of the deeper layers (Figs.S1 and S2 in the Supplement).The most dominant (∼ 48 %) phylum in this bacterial community was Proteobacteria; class Gammaproteobacteria (91 %) and the genera Acinetobacter (28 %) and Pseudomonas (14 %; Fig. 3).The second most abundant (38 %) phylum was Firmicutes; class Clostridia (21 %) and Bacilli (17 %), family Clostridiaceae (15 %) and genus Bacillus (14 %).Further sequences were related to other phyla and orders (Fig. 3 and Table S1).

Depth comparison of microbial communities
An ordination plot from the 454 pyrosequencing by MOTHUR subsample OTUs data set was derived from principal coordinate analysis (PCoA) of the bacterial and archaeal.The PCoA display the similarities and differences between the bacterial and archaeal communities varied with depth (Fig. 5a and b, respectively).The PCoA of bacteria and archaea show very similar separation among the communities of different layers related to the different environment conditions of each layer.The two deeper layers are on the same point along the x axis, showing relatively more similarity than the upper layer.The vectors of the sulfate and sulfide are correlative with communities of the top layer, while methane, δ 13 C TLE and δ 56 Fe(II) were more associated with the communities of the middle layer.The communities in the bottom layer were more correlated to Fe(II), Mn(II) and NH + 4 .The Venn diagram shows also that more OTUs from the deeper layers were shared than between the upper layer and the deeper layers (Figs.S1 and S2).

Discussion
In this study of LK sediments, we investigated changes in the microbial diversity associated with porewater geochemistry and the transition of the dominant electron acceptors with depth (0-40 cm).The geochemical porewater profiles in LK (Fig. 2) suggest that the sediment can be broadly divided into three different regions of microbial processes: the upper 5 cm which is dominated by sulfate reduction, the methanogenesis zone between 5 and 17 cm, and the deep sediments, dominated by anaerobic oxidation of methane coupled to iron reduction.This division of the sediment and the deep irondriven AOM processes were confirmed by in situ profiles of methane, δ 13 C CH 4 , sulfate, ferrous iron and subsequent use of a numerical mass conservation model (based on the geochemical profiles of DIC and δ 13 C DIC ; Adler et al., 2011) and a set of geochemical incubation experiments conducted in our previous work (Sivan et al., 2011).
The mechanisms that enable the process of AOM via iron reduction can be characterized by the existence of specific prokaryotic populations at each depth and their resemblance to similar environments with a distinct characterization.There are only a few studies that analyzed microbial communities relative to geochemical zones in freshwater sediments (Deutzmann and Schink, 2011;Koizumi et al., 2004;Ye et al., 2009).Therefore, this study can deepen knowledge of microbial community shifts under different electron acceptors conditions, especially those related to the methane cycle.It is also the first study in LK that divides the sediment into different layers and contributes significantly to the understanding of the diversity of microbial ecology in different zones.Pipeline of SILVA ngs was used for the alpha diversity to get more classified sequences and better coverage of the classification and population.However, for beta diversity the data from the MOTHUR pipeline were applied to get more statistical results without the bias of different sample size.The changes between the pipeline results for the alpha diversity were not large at the phyla level (standard deviation of ∼ 0.5 %) but increased with higher taxonomy levels.

Sediment geochemical conditions
In the upper layer of sediment (0-3 cm), sulfate concentrations were the highest and decreased with depth as particulate organic matter from the water column accumulates, making organic carbon degradation available (Hadas and Pinkas, 1992).Adler et al. (2011) calculated that the bulk of bacterial sulfate reduction occurs in the top 1 cm of sediment, a finding that was supported by the microbial work of Hadas and Pinkas (1992).Using sulfide and particulate organic carbon measurements, Eckert and Conrad (2007) showed that sulfate reduction accounts for most of the mineralization of organic matter in the upper 15 cm of sediment.Methane concentrations in the upper layer were relatively low and probably occurred from upward diffusion as methanogens are outcompeted by sulfate reducers there (Lovley and Klug, 1983).
The depletion in methane concentrations, decline of δ 13 C TLE values and the heavier isotopic values of δ 13 C CH 4 in the upper layer (Fig. 2) compared to the methanogenic zone may be explained by AOM via sulfate reduction, although no ANME sequences were found using specific primers and fluorescence in situ hybridization methods or by our phylogenetic alignment of 16S rRNA gene libraries at any depth in the sediment.
Below the sulfate reduction zone, in the middle layer of sediment (6-9 cm), methane reached its maximum concentrations (Fig. 2b).The low values of δ 13 C CH 4 in the methanogenesis zone are typical due to the large carbon isotope fractionation that occurs during methane production, thereby leaving the produced methane isotopically light and the DIC isotopically heavy.Below the maximum methane production zone (around 20 cm), methane concentrations began to decrease while δ 13 C CH 4 values increased (Fig. 2b), an outcome that may be due to the AOM process, which leaves residual methane isotopically heavier.Depleted δ 13 C TLE (Fig. 2c) in the deeper part of the sediment also supports the presence of AOM with the production of light biomass from methane oxidation.In addition, although all the suitable electron acceptors at this depth were depleted, both Fe(II) and Mn(II) increased (Fig. 2d).Also, δ 56 Fe values (Fig. 2c) of dissolved iron in the deeper part of the sediment were isotopically negative, similar to sediments with active dissimilatory bacterial iron reduction (Severmann et al., 2006) and are an indication of active iron reduction in the deep sediments and not just diffusion.The increase in Fe(II) concentrations below a depth of 15 cm was probably supported by the absence of sulfide.The potential of highly reactive Fe(III) oxides to drive AOM in LK was also supported by our set of mesocosm incubation studies (Sivan et al., 2011).Since manganese oxide concentrations were very low (∼ 0.04 %) throughout the LK sediment column (Serruya et al., 1974), iron oxide seemed to play a bigger role in this AOM process.

Concept for methane cycle mechanisms
A few possible mechanisms for the novel process of irondriven AOM in the deep part of the sediment can be suggested based on the geochemical profiles and the microbial communities.A possible direct process could be through new, currently unknown bacteria/archaea that reduce iron and utilize methane.Indirect processes could include (1) reduction of Fe(III) oxides through the oxidation of sulfide (in pyrite or FeS minerals) to sulfur intermediates, and then disproportionation to sulfide and sulfate (such as in Holmkvist et al., 2011;Sivan et al., 2014), and/or sulfate-driven AOM but without ANME; (2) reduction of Fe(III) by utilizing H 2 (Lovley, 1991), which could consume most of produced H 2 in the deep sediment.Thus creating a low concentration of H 2 which drives methanogenesis backwards ("reverse methanogenesis", Hallam et al., 2004;Hoehler et al., 1994) through methanogens in the deep sediment; (3) an oxygenic methane oxidation pathway in an anaerobic environment as described by Ettwig et al. (2010), when methane is oxidized by oxygen that is released from iron oxides.
Changes in the microbial population shown in the present study may support the deep iron-driven AOM mechanism (presented by the sample from 29-32 cm).The bacterial diversity in this depth was lower than in the methanogenesis zone (6-9 cm) but higher than in the upper layer (0-3 cm) of the sediment.The changes in bacterial diversity with depth could be related to the availability of different electron acceptors and of organic matter (Nam et al., 2008).However, there is an overlap between bacterial and archaeal sequences in the middle and the bottom layers (PCoA and Venn diagram), indicating that the environmental conditions are affecting the different processes (sulfate reduction, methanogenesis and AOM).Alpha diversity indexes (Tables 1 and 2) were compared to previous studies conducted only in the upper layer of the sediment of LK using different methods (Schwarz et al., 2007a;Wobus et al., 2003), and showed a similarity in the upper parts of the sediment.However, the deeper layers in this study showed higher diversity and richness than the previous studies previous studies (Nusslein et al., 2001;Schwarz et al., 2007aSchwarz et al., , 2007b)), but diversity and richness indexes were closer to other study sites using 454 sequencing (Hollister et al., 2010;Röske et al., 2012).
The direct mechanism of iron reduction coupled with methane oxidation could be performed by a single microorganism, as methanogens have the ability to reduce iron (Van Bodegom et al., 2004;Bond et al., 2002) although not with methane as the electron donor.This direct mechanism could be performed by unique methanotrophy-like MBG-D (Thermoplasmata class) or by MCG/MBG-B (MCG may represent a new archaeal phylum (Lloyd et al., 2013) or a sister lineage to Thaumarchaeota (Meng et al., 2014).However, MBG-B is still classified as Thaumarchaeota; Marlow et al., 2014), to which some of our sequences were similar.About 5 % of the archaeal sequences in the middle and deep layers were affiliated with MBG-D; however, MCG and MBG-B showed much lower representation and thus are not discussed.Members of the MBG-D have been shown to exist in a variety of freshwater and marine environments (Beal et al., 2009;Borrel et al., 2012), and it is the most widely encountered, uncultured lineage in freshwater lake sediments.Even though their metabolism is unknown, hypotheses about their functionalities are based on the environments in which they were found.Methanogenesis has been suggested, as MBG-D was found in deep lake sediments with high methane concentrations (Borrel et al., 2012), and it was also hypothesized to be involved in AOM, as it was found in AOM zones (Schubert et al., 2011) and in marine seep sediment (Beal et al., 2009).In other environments, in which methane concentrations were low, the utilization of waste products, intermediates, or dead cells by MBG-D was also suggested (Smith et al., 1975).Recently, a single cell genomics study showed that members of MBG-D were capable of exogenous protein degradation in cold anoxic environments (Lloyd et al., 2013).
Another option is a consortium of iron-reducing microbes together with methanotrophs or with methanogens capable of reverse methanogenesis.The well-known iron reducers are Geobacter genus; however, only about 1 % of the sequences in the middle and bottom layers or the cores were affiliated with the Geobacter genus members.In the deep layer, the most abundant class was Sva0485.Sequences of the SVa485 order were similar to those found in different aquatic environments, but the metabolic functions of members of this order are not clear.Pelobacter carbinolicus (Lovley et al., 1995), a member of SVa485, is capable of Fe(III) and sulfur reduction.Desulfuromonadales, which was found in our sequences in the deeper parts of the sediment, was shown as S 0 respiring (Pjevac et al., 2014) and may also reduce Fe(III) and Mn(IV) in marine surface sediments (Lovley, 2006).Schwarz et al. (2007a) showed that in the upper part of LK sediment, Deltaproteobacteria was one of the dominant classes, and that most of them were affiliated with acetateoxidizing sulfate-reducing bacteria, that outcompete the acetoclastic methanogens.Even though Deltaproteobacteria are best known for their sulfate reduction metabolism, they can shift their metabolism in response to depleted sulfate concentrations (Plugge et al., 2011).
In addition, our results suggest that Desulfuromonadales of the Deltaproteobacteria class could be also involved in indirect mechanisms of disproportionation of sulfur together with other Deltaproteobacteria sulfate reducers.The dsrA functional gene shows that sulfate reducers are present at the same level of the core where mcrA functional genes (Fig. 6) of methanogens or methanotrophs are found in the deeper part of the sediment.Additionally, the presence of Deltaproteobacteria at the deepest sediment depth and the observed accumulation of acetate with depth (data not shown) could indicate their exploitation of a different metabolic path in the deep sediment than in the upper part of the sediment.
Other options for functioning iron reducers include members of the Nitrospirae phylum that increased with depth.Indeed, sequences from our samples were classified as the Nitrospirales order in Nitrospirae and were most abundant in the bottom layer of the sediment (7 %).Part of our sequences were similar to those found in a previous study (Schwarz et al., 2007b).Using SIP-RNA, Schwarz et al. (2007a) showed not only that Nitrospirae were present, but also that they were functionally bioactive.Although Nitrospirae is a known nitrate oxidizer (Ehrich et al., 1995), the conditions of this environment suggest that it utilized another metabolic pathway.Nitrospirae also include the iron-reducing candidates such as Magnetobacterium bavaricum (Spring et al., 1993) and sulfur reducers (Sonne-Hansen and Ahring, 1999).Part of our sequences were aligned against an uncultured clone (98 %) from freshwater sediment which enhanced degradation of phenanthrene and pyrene by amorphous ferric hydroxide (Yan et al., 2012).
The reverse methanogenesis pathway could be carried out by the dominant deep layer methanogens Methanosaeta.Methanosaeta are acetoclastic methanogens, which are able to grow only on acetate (Jetten et al., 1990).Methanosaeta were also shown to be the dominant active methanogens in the upper layer in previous studies of LK (Schwarz et al., 2007a, b) and in other mesotrophic to eutrophic freshwater lakes (Glissman et al., 2004;Koizumi et al., 2003).Yamada et al. (2014) showed that Methanosaeta species also have the ability to reduce ferrihydrite with H 2 as the electron source.Methanosaeta concilii was the most similar cultured acetoclastic methanogen (96 %) to our sequences which was observed also in the previous study.Methanosaeta could perform a different metabolic process or may have been inhibited.The other methanogens that could perform the reverse methanogenesis are from the Methanomicrobiales order.In the Methanolinea genus of Methanomicrobiales, a hydrogenotrophic methanogen was observed at a constant percentage for all depths.The percentage of members of the Methanoregula genus, which also contains hydrogenotrophic methanogens, decreased with depth in the core.qPCR analysis of the mcrA functional gene of methanogens/anaerobic methanotrophs shows that methanogens were more abundant in the middle layer than in the upper and bottom layers (Fig. 6).ANME were not found in the sediment using specific primers and fluorescence in situ hybridization methods.An additional reason for not finding any ANME sequences in our samples is likely because 454 sequences related to methanogens were found in high percentages and they are probably dominant in the qPCR analysis of mcrA gene observed in the deep layer.
The additional indirect mechanism of anaerobic methane oxidation via an oxygenic pathway was shown clearly by the pmoA functional gene.The pmoA concentration in the deep part of the sediment was higher than in the upper part, indicating an oxygenic pathway (Fig. 6).However, pMMO is a homologue enzyme of ammonia monooxygenase and might be sequenced together with pMMO (Tavormina et al., 2011).The source of pMMO could be Methylococcales (∼ 1 %), of Gammaproteobacteria, which were observed in the middle and bottom sediment layers.Methylococcales are aerobic methanotrophs, which some members were present within the sediments and overlying water column from dysoxic, methane-rich vent and seep systems (Tavormina et al., 2008).The most abundant order in the deep layer was HOC36 (5 %), which is an uncultured Gammaproteobacteria.However, when compared to the National Center for Biotechnology Information database, it was found to be closely related to uncultured LK clones (99 %) and to cultured Methylocaldum sp.(94 %; Bodrossy et al., 1997), which is a Thermophilic methanotroph isolated from landfill cover soil.However, the bias of pmoA with ammonia monooxygenase could be caused by Thaumarchaeota, Betaproteobacteria or Nitrospirae, which were found in all depth.
Thaumarchaeota comprise not only all known archaeal ammonia oxidizers, but also several clusters of environmental sequences representing microorganisms with unknown energy metabolisms (Pester et al., 2011).Members of Thaumarchaeota phylum could have monooxygenase-like enzymes that are able to capture methane due to the enzymes' phylogenetic proximities to methane monooxygenases, but that activity requires the necessary downstream metabolic pathway.If they had that ability, they would be good candidates for group of methanotrophic archaea.This indirect pathway can be similar to that found in the NC10 phylum (Zhu et al., 2012), which produces oxygen via the reduction of nitrite and the oxidation of methane, but with iron oxides.Also many bacterial ammonia oxidizer sequences were found in our environment.Ammonium profiles show increase with depth mainly due to decomposition of the organic matter, and theoretically the large amount of ammonium could be oxidized by ferric iron minerals and produce nitrite (Clement et al., 2005).Maybe ammonia monooxygenase can function for ammonium uptake for iron reduction and for methane uptake for oxidation, but that is only speculation.However, no nitrate or nitrite was detected in the deep sediment; moreover, no NC10 phylum (Ettwig et al., 2010) was observed in our sequences.
To summarize, this study attempted to find a correlation between the performed geochemical and microbial profiles in lake sediments.The geochemical data suggest three main depth-related zones where electron acceptor activities take place in the sediment: sulfate reduction, methanogenesis and a novel, deep iron-driven AOM.The prokaryotic analysis provided clues regarding the microorganisms that may be involved in this novel process and the metabolic paths that occur throughout the microbial assemblage.For AOM via iron reduction to occur, a number of potential pathways and their combinations have been suggested.Orders that become enriched (Sva0485, Methanosarcinales and Nitrospirales) with depth can be assumed to participate in the AOM process either directly or indirectly.A possible direct process could be through new, currently unknown bacteria/archaea that reduce iron and utilize methane, which may be carried out by a MBG-D as a methanotroph in concert with an ironreducer-like Geobacter.Possible indirect processes could be Fe(III) reduction by sulfide, oxidation of the sulfide to elemental sulfur and other sulfur intermediates and then disproportionation to sulfide and sulfate, and sulfate-driven AOM.However, this process is less likely because ANME were not found.It could also be via reduction of Fe(III) by utilizing H 2 , creating a low concentration of H 2 and driving reverse methanogenesis.Fe(III) reduction processes could be carried out by Nitrospirae and/or Deltaproteobacteria.Members of both groups can reduce iron while in concert with methanogenic Methanosarcinales/Methanomicrobiales, or can reduce iron with sulfur minerals, creating sulfate (see above) that Deltaproteobacteria can utilize while in concert with a MBG-D as a methanotrophic archaea (Schubert et al., 2011).An oxidation of methane coupled to an iron reduction pathway as described by Ettwig et al. (2010) could occur through Thaumarchaeota or Nitrospirae with a monooxygenase enzyme that can utilize methane while using iron oxides to generate the oxygen needed to oxidize the methane.Further research involving even larger samples of the microbial community and characterizations of more diverse functional genes will provide a better indication of the composition of microbial communities at different depths.Also, microbial and geochemical experiments have the potential to provide more insights into the mechanism of this novel iron-driven AOM.
The Supplement related to this article is available online at doi:10.5194/bg-12-2847-2015-supplement.

Figure 1 .
Figure 1.Map of Lake Kinneret, located in the north of Israel.Numbers correspond to altitude measured in meters from the level of Lake Kinneret.Cores were taken from station A (taken from Hambright et al., 2004).

Figure 2 .
Figure 2. Geochemical porewater profiles in LK sediment.(a) Profile of SO −2 4 (black circles) and sulfide (gray diamonds) in the porewater (b).Headspace measurements of methane (gray circles) and δ 13 C CH 4 (black squares) in the sediments (c) δ 13 C of total lipids extraction (TLE; black diamonds) from the sediment, and δ 56 Fe (gray circles) of the dissolved iron in the porewater.(d) Electron acceptor profiles of dissolved Fe(II) (gray triangles) and Mn (II) (black squares) in the porewater.Black arrows indicate the sampled sections for 16S rRNA gene analysis.The presented methane profile was taken 2 weeks prior to the sampling for the microbial communities on December 2009.Sampling for Fe(II), Mn(II) and SO −2 4 profiles was carried out between 2007 and 2011 and for sulfide between 2007 and 2013.The δ 13 C CH 4 and δ 56 Fe profiles were performed 4 months prior to the microbial sampling on August 2009.

Figure 4 .
Figure 4. Classification of archaeal sequences using SILVA ngs pipeline.Phyla and order distributions of sequences of the 454 sequencing above 1 % at the different depths in the communities.

Figure 5 .
Figure5.Principal coordinate analysis (PCoA) of the (a) bacterial and (b) archaeal community matrix (based on operational taxonomic units) and vector fitting of the environmental geochemical conditions in the different depth.The bacterial community from the upper layer is labeled with a red triangle, the middle layer is labeled with a green triangle and bottom layer is labeled with a blue triangle.

Figure 6 .
Figure 6.Profiles of functional genes from the three different depths.Green bars represent the pmoA functional gene of aerobic methanotroph, red bars represent the dsrA functional gene of sulfate reducers and blue bars represent the mcrA functional gene from methanogens and anaerobic methanotrophs.

Table 1 .
Bacterial sequences used for classification in SILVA ngs and for subsample in MOTHUR for the alpha diversity.

Table 2 .
Archaeal sequences used for classification in SILVA ngs and for subsample in MOTHUR for the alpha diversity.