Methane-oxidizing seawater microbial communities from an Arctic shelf

Marine microbial communities can consume dissolved methane before it has a chance to escape to the atmosphere and contribute to greenhouse warming. Seawater over the shallow Arctic shelf is characterized by excess methane compared to atmospheric equilibrium. This methane originates in sediment, permafrost and hydrate. Particularly high concentrations 10 are found beneath sea ice. We studied the structure and methane oxidation potential of the microbial communities from seawater collected close to Utqiagvik, Alaska, in April 2016. The in situ methane concentrations were 16.3 ± 7.2 nmol L -1 , approximately 4.8 times oversaturated relative to atmospheric equilibrium. The group of methane-oxidizing bacteria (MOB) in the natural seawater and incubated seawater was >97% dominated by Methylococcacales (γ-Proteobacteria). Incubations of seawater under a range of methane concentrations led to loss of diversity in the bacterial community. The abundance of 15 MOB was low with maximal fractions of 2.5% at 200 times elevated methane concentration, while sequence reads of nonMOB methylotrophs were four times more abundant than MOB in most incubations. The abundances of MOB as well as non-MOB methylotroph sequences correlated tightly with the rate constant (kox) for methane oxidation, indicating that nonMOB methylotrophs might be coupled to MOB and involved in community methane oxidation. In sea ice, where methane concentrations of 82 ± 35.8 nmol kg -1 were found, Methylobacterium (α-Proteobacteria) was the dominant MOB with a 20 relative abundance of 80%. Total MOB abundances were very low in sea ice, with maximal fractions found at the ice-snow interface (0.1%), while non-MOB-methylotrophs were present in abundances similar to natural seawater communities. The dissimilarities in MOB taxa, methane concentrations and stable isotope ratios between sea ice and water column point toward different methane cycling processes in the two environments.


Introduction 25
Methane (CH 4 ) is the third most abundant greenhouse gas contributing to climate change (IPCC, 2014)exceeded only by water vapor and carbon dioxide. Despite much lower concentrations than carbon dioxide, it has a 28x higher accumulative radiative forcing potential (IPCC, 2014) over a time span of 100 years. In the ocean, the two major sources of methane are ongoing biogenic production by microbes in anoxic sediment (Formolo, 2010;Reeburgh, 2007;Whiticar, 1999) and release of fossil methane from geological storage (summarized by Kvenvolden and Rogers, 2005;Saunois et al., 2016). Other 30 sources include release from permafrost, river runoff, submarine groundwater discharge (Lecher et al., 2016;Overduin et al., 2012) and production from methylated substrates under aerobic conditions (Damm et al., 2010;Karl et al., 2008;Repeta et al., 2016). More than 90% of the methane sourced in the seabed is oxidized within the sediment by anaerobic and aerobic oxidation (Barnes and Goldberg, 1976;Boetius and Wenzhöfer, 2013;Knittel and Boetius, 2009;Reeburgh, 1976). The remaining methane either diffuses into the water at the sediment surface, or is released as bubbles, which completely or 5 partially dissolve while rising through the water column (Leifer and Patro, 2002). Dissolved methane is used as a substrate and oxidized by aerobic methanotrophic bacteria (methane-oxidizing bacteria, MOB) in the water column (Hanson and Hanson, 1996;Murrell, 2010) or diluted by the surrounding water column (e.g. Gentz et al., 2014). As a result of these biological and physical processes, oceanic methane concentrations are frequently at low nanomolar levels, leaving only a small fraction of subseafloor-sourced methane to eventually exchange with the atmosphere (Karl et al., 2008;Reeburgh, 10 2007).
By contrast, in the Subarctic and Arctic shelf areas, shallow water depths and seasonal sea ice cover complicate the picture.
High concentrations of methane have been reported from the entire water column up to the surface around Svalbard (Damm et al., 2005;Mau et al., 2013), the Siberian Shelf (Shakhova et al., 2010) and the Beaufort Sea (Lorenson et al., 2016). In addition, during periods of near 100% sea ice cover, gas exchange from the water column to the atmosphere is restricted 15 (Loose et al., 2011). While under ice free conditions, methane concentrations are frequently found in the range of 15 to 30 nmol L -1 or around 700% supersaturated with regard to atmospheric equilibrium, winter concentrations are often 10 to 100 times higher. Maximal concentrations of 5000 nmol L -1 , oversaturation of 160000%, have been reported from the Siberian Shelf (Lorenson et al., 2016;Shakhova et al., 2010;Zhou et al., 2014).
Along with factors like oxygen and trace metal availability (Crespo-Medina et al., 2014;Sansone et al., 2001;Semrau et al., 20 2010), dissolved methane concentration is an important control on the community of MOB and thus methane oxidation rates (Crespo-Medina et al., 2014;Kessler et al., 2011;Mau et al., 2013). Methane hotspots, promoted by limited gas exchange under sea ice, might thus be candidate regions for accumulation of methane oxidizers. In addition, sea ice, particular the icewater interface, is a hotspot for microbial activity. The ice surface, penetration of light, and constant exchange with the underlying water column favor the development of communities composed of small eukaryotic organisms, microalgae, 25 prokaryotes and viruses; the biomass often being several orders of magnitudes denser than in the underlying water column (Thomas and Dieckmann, 2002).
Methane-oxidizing bacteria use methane as their sole carbon and energy source (Hanson and Hanson, 1996). In the first step, methane is oxidized to methanol catalyzed by the enzyme methane monooxygenase. Since methane monooxygenase is characteristic of nearly all MOB (Knief, 2015), pmoA, the gene encoding for a subunit of the membrane-bound particulate 30 methanemono-oxygenase, has been used as a specific molecular marker for detection and characterization of MOB (Knief, 2015;Lüke and Frenzel, 2011;reviewed by McDonald et al., 2008;Tavormina et al., 2008). Methanol is further metabolized to formaldehyde, from which it is either mineralized to carbon dioxide (CO 2 ), or assimilated into organic compounds and finally biomass (reviewed by Hanson and Hanson, 1996; reviewed by Strong et al., 2015). Different types of MOB are distinguished by their phylogeny and assimilation pathways for formaldehyde. While γ-Proteobacteria or Type I MOB assimilate formaldehyde via the ribulose monophosphate pathway (RuMP), α-Proteobacteria or Type II MOB use the serine pathway (Hanson and Hanson, 1996). Besides these two proteobacterial groups, MOB also occur in the phylum Verrucomicrobia (e.g. Dunfield et al., 2007;Pol et al., 2007).
Methane-derived carbon is also assimilated in non-methane utilizing methylotrophs (non-MOB methylotrophs) or other 5 bacteria in freshwater and temperate marine environments. These non-methane oxidizers are suggested to cross-feed on metabolites produced by the MOB (Hutchens et al., 2003;Jensen et al., 2008;Saidi-Mehrabad et al., 2013).
Knowledge of the microbial communities responsible for methane oxidation in the Arctic and Subarctic is still sparse.
During the last years, the first few studies have determined methane oxidation rates from seawater in these regions to cover a range from 10 -4 up to 3.2 nmol L -1 d -1 (Gentz et al., 2014;Lorenson et al., 2016;Mau et al., 2013Mau et al., , 2017Steinle et al., 2015). 10 In only two of these studies, both performed off Svalbard, oxidation rate measurements were combined with analysis of the microbial community. Steinle et al. (2015) quantified MOB by fluorescence in situ hybridization and microscopy. Low but relatively constant cell-specific oxidation rates were determined from the oxidation rates and MOB abundance, indicating that MOB community size is an important control on the total methane oxidation rate in the system. Mau et al. (2013) analyzed the bacterial community with denaturing gradient gel electrophoresis (DGGE) of the 16S gene and compared 15 patterns of PCR products for pmoA. Different MOB communities were observed in the meltwater layer and deep water in this stratified system, also reflecting the observed differences in methane oxidation rates. Only one of the eleven analyzed DGGE bands was identified as methanotroph (from the genus Methylosphaera) from the deep water in this study, while none were detected in the meltwater, possibly due to the limitations of the method. To our knowledge, no high-throughput sequencing studies of methane-oxidizing bacteria in the Arctic have been published in peer-reviewed literature to date. 20 We studied methane-oxidizing communities from seawater sampled on the Beaufort Sea shelf close to Utqiagvik, Alaska.
Incubation experiments were performed under different methane concentrations to directly compare the bacterial community structure with methane oxidation rates. Seawater incubations, freshly sampled sea water, and sea ice were analyzed for their entire community diversity (16S rDNA) and the presence of MOB (16S rDNA and pmoA) using high-throughput Illumina MiSeq sequencing. The aim of this study was to (1) investigate the response of the entire microbial community to an 25 increase in methane abundance, (2) identify types of MOB involved in the oxidation of methane, (3) test for the presence of MOB in natural seawater and sea ice communities and (4) relate these community features to methane oxidation rates.

Study site
Samples were collected at two sites between 7 April 2016 and 15 April 2016 in the Beaufort Sea (Table 1). Site " Elson 30 Lagoon" (EL) is located north of Utqiagvik, Alaska, (7.4.2016, 71.334° N, -156.363° W). At the time of sampling EL was covered with 1.5 m thick sea ice; at approx. 1.5 m water depth, this left only a narrow layer of water between the sea ice and the sediment. Site "Ice Mass Balance Buoy" (IMB) is located 1 km offshore of Utqiagvik, close to the ice mass balance buoy of the sea ice physics group of University of Alaska, Fairbanks (7.4.2016, 71.373° N, -156.548° W, and9.4.2016, 71.372° N, -156.540° W). This site was characterized by 1 m thick fast ice cover and a water depth of approximately 7 m.

Sampling and instrument deployment
Seawater temperature and salinity records were taken with an YSI Professional Plus probe (YSI, Ohio, USA) and a YSI 600 5 OMS V2 sonde (YSI, Ohio, USA). Water was collected using either a peristaltic pump (Masterflex Environmental Sampler, Cole Parmer, Illinois, USA) or submersible pump (Cyclone, Proactive Environmental Products, Florida, USA) from different water depths. For determination of methane concentration and isotope ratios, water samples were collected as described in Uhlig and Loose (2017). Briefly, in the field, 0.7-0.9 L seawater was transferred bubble-free directly into in foil sample bags (# 22950, Restek, Pennsylvania, USA). On return to the laboratory, a 0.1 L headspace of Ultra-High Purity nitrogen (Air 10 Liquide, Anchorage, AK) was introduced into the bags through the septa, and the samples were equilibrated at 30°C for in situ measurements for at least 6 h.
For DNA extractions, between 1-2 L of seawater were filtered onto Sterivex® filter cartridges (Millipore) with 0.2 µM PES filter membranes directly in the field, or were filled into foldable polypropylene containers and filtered upon return to the laboratory. For nutrient analysis an aliquot of the flow-through of the Sterivex® filters was collected in 15 mL 15 polypropylene tubes (Falcon Brand, Corning, New York, USA) and frozen at -80°C. Seawater was fixed with 2% final concentration formaldehyde (Mallinckrodt Chemicals, Surrey, UK) and stored at 5°C to for later determination of the cell abundance.
Additionally, at site IMB, seawater temperature, salinity and velocities were recorded with an Aquadopp Profiler (Nortek AS, Norway), and a salinity temperature recorder (SBE37SMP, Sea-Bird Scientific, Washington, USA). These were 20 deployed at about 7 m depth on the seafloor between 9 and 15 April.
Sea ice was collected at site IMB only, using a Kovacs Mark II ice corer (Kovacs, Roseburg, Oregon, USA). The ice cores were sectioned into 15 cm and split lengthwise. The outside was cleaned with a sterilized knife to remove microbes possibly transferred from the sampling equipment. The core sections were sealed into custom-made gas-tight tubes (Loose et al., 2011) for determination of methane concentration and isotope ratios. In the laboratory, the gas-tight tubes were flushed with 25 ultrapure nitrogen for several gas volumes (Lorenson and Kvenvolden, 1995). Due to technical limitations, ice core 1 (IC1) was melted within a week at 5°C, while ice core 2 (IC2) was melted within a day, while frequently being mixed, at room temperature. Samples for molecular biology and cell counts were collected from the melted sea ice similar to the procedure described for seawater. In addition, the bottom 2 cm of one ice core was sampled into a sterile sample bag (Whirlpak, Nasco, Fort Atkinson, WI, USA) for molecular biology processing only. Sea ice brine volume fractions were calculated according to 30 Cox and Weeks (1983).

Net methane oxidation/production and determination of isotope fractionation factors
Rates for net methane oxidation/production were determined from the methane mass balance according to Uhlig and Loose (2017). In short, seawater was sampled into multi-layer foil bags. In addition to a headspace of hydrocarbon-free air (Air Liquide, Anchorage, AK), some sampling bags were supplied with a spike of methane. Final dissolved methane concentrations ranged between 3.0 and 4000 nmol L -1 , representing approximately 0.2x (no methane addition, resulting in 5 degassing of in situ methane to the headspace), 2x, 10x and 200x of the in situ methane concentration. Samples were incubated at 0-1°C for 5 to 46 days. Some variation in the incubation period was introduced by logistical constraints. To account for potential diffusive loss of methane, a killed control was prepared for the 200x treatment by adding 0.1M NaOH.
Assuming first order kinetics for oxidation of methane (Reeburgh et al., 1991;Valentine et al., 2001), net oxidation/production rate constants ( ) were determined from the methane mass balance in the incubations (Uhlig and 10 Loose, 2017) as with ( 4 ) , being the total molar mass of methane in the bag at time t i .
The net oxidation/production rate (r ox ) was calculated from the first order constant and the in situ concentration of methane in the water: 15 Isotopic fractionation factors of methane oxidation ( = 12 13 ) were determined as described in Preuss et al. (2013), using the isotope fractionation approach (Coleman et al., 1981).

Methane concentration and stable isotope ratios 25
Methane concentrations and stable isotope ratios were determined with a Picarro G2201-i cavity ring-down spectrometer (Picarro, Santa Clara, California, USA) coupled to a Small Sample Isotope Module (SSIM) as described by Uhlig and Loose (2017). After equilibration, the headspace above the seawater or melted ice was subsampled with a gas tight syringe and 1 to 15 mL was injected into the SSIM. Measurements were performed in fast measurement mode. Dissolved methane concentrations were calculated as described in Magen et al. (2014), with the equilibrium constant according to Yamamoto et al. (1976).

Nutrient analysis and flow cytometry
Phosphate, nitrate and nitrite concentrations were determined using a QuickChem QC8500 automated ion analyzer (Lachat,5 Loveland, Colorado, USA). The total number of prokaryotic cells was counted on a BD Influx TM flow cytometer with BD FACS TM software. Formol-fixed samples were stained with a final concentration of 1x SYBR Green I (Invitrogen, Molecular Probes, Eugene, Oregon, USA) for 20 to 45 min at room temperature in the dark before analysis.

Nucleic acid extraction and sequencing
DNA was extracted with the PowerWater® DNA extraction kit (MoBio, Carlsbad, California, USA). To remove the filter 10 membrane, the Sterivex® cartridge was opened with a pair of sterilized pliers. The filter membrane was cut out along the edge with a scalpel, transferred into the bead tube, and DNA subsequently extracted according to the manufacturer's protocol. A minor modification was made: the tube was vortexed once for 3 minutes, rotated 180°, and then vortexed for another 3 minutes. The DNA was eluted in 80 µL buffer PW6, after incubating the buffer for 1 min on the membrane.
The primer sequences specified above included adapter sequences (xx, yy) to attach Nextera indices and adapters in a second PCR reaction of 6 cycles with 50 ng template DNA (http://web.uri.edu/gsc/next-generation-sequencing/). Amplicons were 30 sequenced with Illumina MiSeq at 2x250 bp read length.

V4-V5 region of 16S rRNA gene
Demultiplexing and adapter removal was performed with Illumina software. V4-V5 sequence quality control and clustering was performed in mothur (Schloss et al., 2009) as follows. Contigs were prepared from forward and reverse reads and culled if they contained ambiguous bases or homopolymers longer than 6 bases. Contaminating sequences observed in kit and filter 5 blanks accounted for 1.4% of all sequences and were removed from all samples. After alignment to the Silva small subunit reference database (v123; Quast et al., 2013), the 408 bp long sequences were preclustered (1% variability allowed) and filtered for chimeras (de novo algorithm) with the UCHIME (Edgar et al., 2011) wrapper in mothur. Sequences identified as chloroplast, mitochondria, Archaea, Eukaryota or unknown were removed and operational taxonomical unit (OTUs were built at a 3% distance level with the opticlust algorithm. OTUs with fewer than 2 reads were removed from further analysis.  (Table 2), were excluded from the analysis.

Identifying potential methane-oxidizing bacteria
To select groups representing methylotrophs and methanotrophs, 16S OTUs were filtered according to their phylogenetic annotation assigned by mothur for containing the string "meth" on family, order and genus level. This filter is expected to find 97% of taxonomically annotated methanotrophs, according to a current review on the diversity of methanotrophs (Knief, 20 2015).
Further, phylogenetic groups potentially involved into methane cycling were identified as differentially more abundant 16S OTUs between incubations (0.2x, 10x, 200x) and in situ samples using DESeq2 (Love et al., 2014). Only OTUs with an adjusted p-value in DESeq2 < 0.05 were kept for further analysis. OTUs identified from spike 0.2x were considered to represent groups favored due to the incubations (the "bottle effect") rather than addition of methane, and removed from 25 further analysis. Treatments EL 0.2x, EL 10x and IMB 2x (Table 2) were not included in this analysis, since no replicate samples were available. The abundance of all candidate 16S OTUs, identified as described above, was determined within every in situ or incubated sample.
Absolute numbers of methanotrophs and methylotrophs were calculated by multiplying the relative 16S sequence abundance with flow cytometric cell counts. The absolute numbers were further corrected for the mean of the 16S gene copy number for 30 the lowest taxonomic rank (class to genus) available in the rrnDB-database (Stoddard et al., 2015).

Particulate methane monooxygenase: pmoA
In addition to 16S genes, the alpha subunit of the particulate methane monooxygenase (pmoA) was used as molecular marker for MOB. Only pmoA forward reads were analyzed. High-quality pmoA reads were retrieved according to the following protocol. Using mothur (Schloss et al., 2009), all reads were trimmed to a length of 225 bp to remove sequence fractions with a mean quality score below 30 (fastqc;Andrews, 2010). In addition, reads were clipped whenever the average quality 5 score over a 50 bp window dropped below 30. Sequences with ambiguous bases and homopolymers larger than 6 bp were culled. Only sequences that translated into uninterrupted protein reading frames (Emboss 6.60 / transseq; Rice et al., 2000) were kept for further analysis. Nucleic acid sequences were aligned to a reference dataset of pmoA sequences (fungene; Fish et al., 2013) and sequences of a length of at least 220 bp were preclustered (1% variability allowed). De novo chimera filtration was run with the UCHIME (Edgar et al., 2011) wrapper in mothur. A similarity of 93% between pmoA sequences 10 was defined to match the 97% cutoff as species definition for the 16S gene (Lüke and Frenzel, 2011). PmoA OTUs were built at a maximal distance of 7% between the furthest neighbors to maximize resolution between OTUs due to the short read length and limited number of unique sequences (Supplementary Table 1). To determine the phylogenetic relationship of pmoA sequences, nucleotide sequences were aligned against selected reference sequences in Mafft 7.017 (Katoh and Standley, 2013) and a neighbor joining tree calculated in Clustal 2.1 (Larkin et al., 2007) with 1000 replications. 15

Water column properties
On 7 April 2016, the narrow layer of water between the sediment and ice in Elson Lagoon (N=1) had a salinity of 21 and a temperature of -1.5°C. Phosphate and nitrate concentrations were 0.74 µM and 4.87 µM, respectively. Methane concentration for Elson Lagoon (N=1) was 53.2 nmol L -1 with a stable isotope ratio of -73.8 ‰ ( Figure 1) and cell density 20 7.7 x 10 4 cells mL -1 . For most days during the sampling period, the water column at station IMB was characterized by temperatures around -1.8°C and salinities of 33.9 to 36.4 ( Figure 1, Supplementary Figure 1, 2). Between 11 April and 13 April warmer water (max. temperature observed -0.9°C) was advected, coinciding with a change in current direction. A lower salinity of 27.5 at the ice-water interface indicates melting of the sea ice. Phosphate concentrations at station IMB were 0.99 ± 0.33 µM (N=9) and nitrate 6.59 ± 4.04 µM (N=9), with neither showing any trends in the depth profiles (data 25 not shown). Nitrite concentrations were below detection (0.3 µM based on technical replicates). Water column methane concentrations at station IMB ranged between 9.2 and 25.3 nmol L -1 (16.3 ± 7.2, N=5) ( Figure 1), with stable isotope signatures between -55.4 and -70.5 ‰ (-60.6 ± 6.3, N=5)). Total prokaryotic cell densities, determined as SYBR Green stained cells with flow cytometry, were 6.9 x 10 4 ± 5.7 x 10 3 cells mL -1 (N=16).
Microbial activity during storage of IC1 at 5°C for one week before analysis might have led to the differences in methane concentrations and isotope ratios. For ice samples, cell counts were performed on IC2 only; they show an increase from 1.0 x 10 4 cells mL -1 in the top layers to 8.2 x 10 5 cells mL -1 in the bottom two cm of the ice core. 10

Net methane oxidation/production and isotope fractionation
The methane oxidation potential of microbial seawater communities at stations EL and IMB was determined from the methane mass balance in incubation experiments (Table 3;

Bacterial community structure
The V4-V5 region of the 16S rRNA gene was sequenced from a total of 10 seawater samples and 7 ice samples (Table 2).
Non-metric multidimensional scaling analysis of the Bray-Curtis diversity revealed high similarity across the in situ water 25 samples analyzed for 16S diversity (Figure 4). Samples from site IMB clustered together repeatedly, and we did not observe any differences in community structure coinciding with water depth or temperature (Figure 4). For the IMB samples, IMB 2 was the only sample slightly different from the other in situ samples, though IMB 1, IMB 2, IMB 4 are all characterized by a colder water mass. Only IMB3 showed some influence of an incoming warm water mass in the YSI profile (Figure 1), though not yet reaching the bottom (Supplementary Figure 1), but this shift is not seen in the community structure. In 30 contrast to the in situ water samples, the community structure of incubated samples is driven by incubation time. While communities in the short-incubation treatments (5-10 days; 0.2x, 2x, 200x short) were similar to the in situ samples, the long incubations (41-46 days; 10x, 200x long) clearly deviated from the in situ samples. In both the long and short incubated clusters samples originating from IMB 1 (0.2x, 10x) and IMB 2 (2x, 200x) are present (Figure 4, Table 2). Microbial communities in ice cores were clearly distinct from those in the water samples and were more distant to each other than were the communities in water samples. 5 In the in situ seawater communities, Proteobacteria were dominant with relative sequence abundances of 59.5% and 65.5% ± 2.5 % for EL (N=1) and IMB (N=9), respectively (Supplementary Figure 4). Within the phylum of Proteobaceteria, α-and γ-Proteobacteria made up the majority. The second most abundant phylum was Bacteriodetes with 23% and 19.6% ± 1.4% for EL and IMB, respectively.
Similar to the seawater, sea ice (N=7) showed a dominance in Proteobacteria (58.9% ± 9.8%), but Bacteriodetes sequences 10 (29.1% ± 11.7%) were slightly more abundant in the ice than in the water. γ-Proteobacteria dominated in all but one sample (IC2 30-46 cm). This one sample, which had clearly visible sediment included into the sea ice structure, was dominated by α-Proteobacteria.
In all incubated samples that were sequenced (N=10), species richness decreased ( Figure 3) and the communities shifted toward higher fractions of γ-Proteobacteria over time. In short incubations (5-10 days; N=5) γ-Proteobacteria dominated 15 with 61.8% ± 2.9% of sequences, while reaching 81.0% ± 11.1% in long-incubation samples (41-46 days; N=4). In particular, one operational taxonomical unit (OTU), from the genus Oleispira, was very abundant in the long-incubation samples, with 50.1 to 76.3%, compared to abundances <0.04% in the in situ samples. The same OTU was only slightly more abundant in the short-incubation treatments (0.5%-1.6%) compared to in situ abundances. In addition to the shift in community structure, total cell densities increased to 1.9 x 10 5 and 3.3 x 10 6 cells mL -1 for short and long incubations, 20 respectively, based on flow cytometric cell counts.

Methanotrophs, methylotrophs and differentially abundant OTUs
Using their 16S taxonomic annotation, we identified six groups of aerobic methanotrophs (MOB) ( Figure 5). With a maximum of 1.76% ± 0.73%, the relative abundance of MOB was low in all samples (Table 4) Table 4). Ice samples showed the largest difference in abundance between non-MOB methylotrophs and MOB, with a ratio of 21:1 between the two groups. Ice samples also had the highest overall relative abundance of methylotrophs (MOB and non-MOB) of all in situ samples (max: 1.63%, IC1 0-16 cm). Only the 200x long incubations had a higher total number of methylotrophs (3.3%), while this sample was in addition dominated by MOB (2.49%). The second highest relative abundance of MOB was found for in situ EL and IMB with 0.24% ± 0.09% (N=10).
Taking into account the total cell number, a strong increase of MOB groups MMG1 (2 to 700 times) and Milano-WF1B-03 5 (25 to 75 times) was observed for the 10x and 200x long-incubation samples compared to in situ conditions (Figure 5b).
Taxonomic groups that became differentially more abundant in the incubated samples than in natural communities were the y-Proteobacteria Oleispira, Colwellia and Glaciecola, as well as Rhodobacteracea (α-Proteobacteria). Except for Oleispira, which became dominant, the other taxa had relative sequence read abundances from 1.1% to 12.6% after the oxidation experiments, compared to abundances <0.25% for in situ samples (Supplementary Figure 5). 10

Particulate methane monooxygenase (pmoA) sequences
A 225 base pair section of the particulate methane monooxgenase gene (pmoA) was sequenced in a total of 15 samples (Table 2). The absolute abundance of pmoA fragments obtained in sequences ranged from 9331 (IMB in situ, 6.5 m depth) to 72781 (IMB 200x long) reads. In general, incubations with higher methane concentration had more pmoA reads than incubations with lower methane concentration or in situ samples. About three times more reads were filtered from the Elson 15 Lagoon in situ sample (33844 reads, N=1) than the IMB in situ samples (11700 ± 1833, N=4).

Methane concentration and stable isotope ratios in seawater and ice 25
Seawater methane concentrations in April 2016 close to Utqiagvik Alaska were supersaturated 250% to 700% compared to atmospheric equilibrium (3.6 nmol L -1 ). The concentration at site EL (52.90 nmol L -1 , N=1, 7.4.2016) was in the range of a study by Lecher et al. (2016) in Elson Lagoon under ice free conditions (3.3-124.0 nmol L -1 ). At site IMB, concentrations were slightly lower (9.5 nmol L -1 -25.2 nmol L -1 : N=5, 15.4.2018) than previously reported from the same area for ice free (Lecher et al., 2016;mean: 40.6 nmol L-1), and ice covered conditions (Zhou et al., 2014; March/April: 37.5 ± 6 nmol L-1). 30 Shallower depths at IMB exhibit lower methane concentrations (Figure 2), and the isotopic signature mirrors this pattern with more positive values toward the surface. This indicates that methane might be biologically oxidized on the way through the water column, after being released from the sediment.
Although both ice cores were sampled within 300 m distance from each other at site IMB, they differ in concentration and isotope signature. These differences could either be caused by spatial variability between the two ice cores or differences in the processing procedure described in section 2.2. Spatial variability as driving difference between the two ice cores is corroborated by the sediment present at 30-46 cm depth in IC1, which was not observed in IC2, indicating that both ice 10 cores have different freezing histories. The same event that led to inclusion of the sediment into IC1 possibly resulted in inclusion of higher methane concentrations into IC1 compared to IC2 during freeze-up. In addition, microbial processes like oxidation of methane or methanogenesis could have taken place in situ or during sample processing and storage. Microbial  (Figure 4a). Another microbial process that may have led to the discrepancies between IC1 and IC2 could be methane production from ice algae-derived organic carbon in IC1. With typical carbon isotopic signatures of -20‰ to -30‰ for ice-derived carbon (e.g. Wang et al., 2014), methane produced from this 20 substrate would be enriched in 13 C (more positive) compared to the initial pool of methane (about -60‰, Figure 2, Figure 6).
Yet, sequences of bacterial taxa that might indicate anoxic conditions (Eronen-Rasimus et al., 2017), which would favor anaerobic methane production, were not significantly more abundant in IC1 than in IC2 (Supplementary Table 2).
Compared to the underlying water column, methane concentrations in the sea ice were two to five times higher. Further, the isotope signatures indicate less oxidized methane (-60.4 to -63.8 ‰) in most of the ice sections compared to the upper water 25 column (-55%). Lorenson and Kvenvolden (1995) report higher methane concentrations in sea ice than in the water column for the Beaufort Sea. They attributed the high methane concentrations in the fast ice to inclusion of sediment-sourced methane during the initial freeze-up over the shallow shelf at <10 m water depth (Lorenson et al., 2016). Methane concentrations in IC2, which are close to water column concentrations reported in previous studies for the Barrow shelf (Lecher et al., 2016;Zhou et al., 2014), suggest the same process for our ice cores. Further, in our study, the lower methane 30 concentrations together with more positive (heavier) isotopic signature in sea water compared to ice, might indicate that the microbial community in the water column is oxidizing more methane during the ice covered period than in the freeze-up period. Higher oxidation rates during ice covered periods compared to ice free conditions were previously reported for the Beaufort Sea. Due to reduced sea-air gas exchange, higher methane concentrations can build up under sea ice cover, which might lead to higher oxidation rates (Lorenson and Kvenvolden, 1995).

Methane cycling at different methane concentrations
Net methane oxidation/production rates were determined from water sampled at station IMB 1 and IMB 2 on 7 and 9 April 2016. Both days were characterized by the cold water temperatures (≤-1.8°C; Fig. 1). Different water masses have 5 previously been reported to influence the methane oxidation potential of water column microbial communities off Svalbard (Steinle et al., 2015). In this study, we observed a change in current direction and water temperature consistent with advection of a different water mass into the study area ( Supplementary Fig. 1). However, this change occurred on 12 April subsequent to sampling IMB 3, and thus this event would not have influenced the net oxidation potential determined in this study. 10 Net oxidation rates of the long-incubation treatments at 10x (46 days) and 200x (41 days) methane concentration fall into the mid-range of rates published for Arctic and subarctic environments (Damm et al., 2015;Gentz et al., 2014;Lorenson et al., 2016;Mau et al., 2013Mau et al., , 2017Steinle et al., 2015) or marine sites with high oxidation rates at oil spills or gas flares Valentine et al., 2010), as discussed in Uhlig and Loose (2017). The fractionation factors (α ox ) that we observed are higher than previously reported from cold marine environments with a range of α ox from 1.002 to 1.017 15 (Cowen et al., 2002;Damm et al., 2008;Grant and Whiticar, 2002;Heeschen et al., 2004;Keir et al., 2009;Tsunogai et al., 2000). Some of these fractionation factors, which were calculated from in situ data, might however be underestimates due to mixing effects in the water column (Grant and Whiticar, 2002). The fractionation factors in our study seem to be inversely dependent on the methane spike concentration, with higher fractionation in the 50x (1.023, N=6) treatments than in the 200x (1.010, N=2) treatments. The relative and absolute abundances of MOB, as well as the dominant MOB types, differed 20 between both treatments, possibly providing explanations of the differences in fractionation rates. Logistical constraints forced us to stop several incubations already after 5 to 6 days. These short-incubation 2x and 200x treatments did not resolve oxidation of methane. While the 2x treatments did not meet the sensitivity threshold for the method (Uhlig and Loose, 2017), the 200x short treatments were likely just about to leave the lag phase when the experiments were stopped. A lag phase of 6 days was observed for the long-incubation 200x samples, in which the microbial community possibly to shifted toward an 25 abundance of MOB that was large enough to cause detectable methane oxidation. To facilitate comparisons between treatments, incubation duration should be kept constant in future studies.
The increase in methane concentration in treatment IMB 0.2x (10 days incubation) is surprising since experiments were performed under aerobic conditions. Since the seawater was not pre-filtered through a larger pore-size filter, which would exclude larger particles but allow bacterial cells to pass, production of methane in microanoxic zones (de Angelis and Lee, 30 1994;Oremland, 1979) should be considered. Furthermore, several studies suggested pathways for methane production in oxygenated marine systems from methylated compounds or dissolved organic matter (Damm et al., 2010;Florez-Leiva et al., 2010;Karl et al., 2008;Repeta et al., 2016). The methane production rate of 0.06 nmol L -1 day -1 observed in our study is two to six orders of magnitude lower than previously published methane production rates under aerobic conditions (Damm et al., 2010;Karl et al., 2008). In addition to biological processes, we cannot rule out an abiotic effect leading to the increased methane concentrations, since our experimental setup did not include a killed control at the same methane concentration.

Abundances of MOB and non-MOB methylotrophs control the methane oxidation potential
We found a strong linear correlation between the net oxidation rate constant (k ox ) and the relative abundance of 16S MOB 5 sequences (Spearman rank order coefficient ρ s = 0.79, p=0.006) (Figure 7a, Table 5). This strong correlation is confirmed when correlating against the total abundance or DESeq2 normalized abundance of 16S MOB sequences ( Table 5). The correlation to k ox is even stronger for the absolute abundance of pmoA sequences retrieved from the respective datasets (ρ s = 0.86, p=0.006) (Figure 7b). This presentation of a direct and statistically significant linear relationship is the first to our knowledge. It agrees with other qualitative reports of positive correlations between methane oxidation rates and abundance 10 of pmoA or MOB 16S rRNA genes determined using a variety of methods -quantitative PCR, FISH, or sequencing -for marine water column and lake sediments (Crespo-Medina et al., 2014;Deutzmann et al., 2011;e.g. Rahalkar et al., 2009;Steinle et al., 2015). Future application of marine-specific pmoA primers may further improve this correlation (Tavormina et al., 2008).
Cell-specific net oxidation rates in our study (3.2-7.5 fmol cell -1 h -1 ) were relatively constant between treatments. They are 15 two orders of magnitude higher than reported for subarctic sea water (Steinle et al., 2016). Since the cell-specific rates only span a narrow range, the ultimate control on the methane oxidation potential is the number of MOB.
Despite the long incubation time in our experiments and the fact that methane was the only added source of carbon, the relative abundance of MOB determined from 16S reads was low (<2.5%, Table 4). Other studies of natural or man-made gas or oil spills, with dissolved methane concentrations comparably high to our 10x and 200x treatments, reported maximal 20 values of 8 to 34% of MOB (Crespo-Medina et al., 2014;Kessler et al., 2011;Steinle et al., 2015Steinle et al., , 2016. Surprisingly, relative sequence abundances of MOB in the natural seawater communities were higher than in the incubations except for the 200x treatment (Table 4). Inferred absolute MOB numbers were higher in 10x and 200x incubations than in situ ( Figure   5b). In contrast, absolute MOB numbers in 0.2x and 2x incubations were very similar to in situ abundances, indicating that either the provided methane concentration was too low or the incubation time too short to stimulate MOB growth. 25 It is puzzling why the fraction of methane oxidizers in the bacterial community did not increase above the observed low percentages although the cell-specific oxidation rates were high and sufficient methane was available, particularly in the 10x and 200x treatments. Oxygen and methane can be ruled out as limiting factors, since both were abundant. Copper, which is essential for expression of particulate methane monooxygenase, can restrict MOB growth (Avdeeva and Gvozdev, 2017;Zhivotchenko et al., 1995). In the absence of copper, many MOB express a copper-independent soluble methane 30 monooxygenase (Hakemian and Rosenzweig, 2007). Since we did neither determine copper concentrations, nor the expression of particulate and soluble methane monooxygenase, we cannot exclude that copper was limiting in our study.
Further, the low relative abundance of MOB sequences could be due to competition with other bacterial taxa for other macro-or micronutrients. In the absence of other added C substrates, these other taxa could have utilized the initial pool of dissolved organic carbon (DOC). DOC concentration is about 68 µM carbon in the Southern Chukchi Sea (Tanaka et al., 2016), which is in the same range as the amount of consumed methane carbon in the 200x treatments and two orders of magnitude higher than the consumed carbon in the 10x treatments.
As a result of the low MOB abundances, the potential of the microbial community to mitigate release of dissolved methane 5 to the atmosphere by oxidation is small. For example, for methane concentrations in the Laptev Sea area, the rates observed in this study would result in 0.2% consumption during the ice covered period. This supports the results from a previous study for the Beaufort Sea, where 1-2% of dissolved methane was calculated to be oxidized (Lorenson et al., 2016).

Structure of the methane degrading microbial community
This first 16S MiSeq sequencing based study on methane-oxidizing sea water communities in the Arctic provides a broader 10 view on the community structure than approaches with FISH and DGGE. The dominance of γ-Proteobacteria MOB in our natural and incubated seawater samples agrees with previous records of MOB diversity for polar and subpolar waters (Mau et al., 2013;Steinle et al., 2015;Verdugo et al., 2016). In addition, non-methane-utilizing methylotrophs were present in all of our samples. The relative read abundance of non-MOB methylotrophs were, similar to MOB, tightly correlated to k ox , and the same correlation holds for the relative abundance of total methylotrophs (MOB plus non-MOB). In contrast, the 15 correlation between OTUs that were differentially more abundant in the incubated samples and k ox was weak (Table 5). This points toward a possible link between the MOB and non-MOB in this methane-oxidizing microbial community, in which non-MOB methylotrophs might play a role for community methane oxidation, whereas the OTUs that were differentially more abundant, are not directly linked to methane oxidation.
Methylophilacaea, the most abundant non-MOB methylotroph in our experiments, have been found to be abundant in 20 sediment methane-oxidizing communities in lakes and marine systems (Beck et al., 2013;Redmond et al., 2010). Possible cooperative behavior between methanotrophs (Methylococcacaea) and non-MOB methylotrophs (Methylophilaceae) was suggested (Beck et al., 2013), in which the latter cross-feeds on intermediate metabolic products of the MOB, i.e. methanol, and can even positively alter the metabolism of the MOB toward methane assimilation (Krause et al., 2017).
To test if the non-methane MOB could be supported by the intermediate substrates produced by MOB, we calculated a 25 budget between the methane carbon assimilated by the growing microbial population (C CH 4 ,assim ), and the cell carbon gained during growth (C cell−growth ) (Figure 8). We assumed (i) a cellular carbon content of 150 fg for exponentially growing bacterial cells (Vrede et al., 2002) and (ii) that about 1/3 of consumed CH 4 -carbon is assimilated, with the remaining 2/3 respired to CO 2 (Bastviken et al., 2003;Roslev et al., 1997). C CH 4 ,assim exceeds MOB-C cell−growth by a factor of 9 to 17, indicating that some of the C CH 4 ,assim was available for secondary consumption by non-MOB. The entire methylotrophic 30 community (MOB + non-MOB-methylotroph) growth can also be explained solely by C CH 4 ,assim , supporting the possible link of non-MOB methylotrophs to methane consumption. In contrast, only about 0.1% of the total community growth could be supported by C CH 4 ,assim in the 10x treatment and 15% in the 200x treatment. The remaining cell growth, e.g. of the differentially more abundant OTUs, must have been supported by other carbon sources, such as initially available DOC.

MOB and methylotrophs in sea ice
The two sea ice cores analyzed in this study give a first insight into the possible role of methane oxidizers in sea ice. In 5 contrast to seawater samples, MOB found in sea ice samples were mostly α-Proteobacteria. The relative sequence read abundance of MOB in the ice was very low (maximal 0.1%), pointing to an overall low contribution of methane oxidation inside sea ice. The highest relative abundances of MOB were found in the top-most ice sections in both ice cores (Figure 5a).  (Kirst et al., 1991).

Conclusions 15
We studied the structure and methane oxidation potential of microbial communities from Arctic seawater and sea ice. The natural seawater community had relative sequence abundances of MOB of 0.24% ± 0.09% and was dominated by γ- Higher methane concentrations in the sea ice compared to the underlying water and an offset in stable isotope ratios suggest 25 that either fractionation and solute concentration occurred during freeze-up, or different microbial processes took place within the ice and water. Possible causes explaining this observation include (i) microbial production of methane, even within the ice (Damm et al., 2015), and (ii) microbial oxidation in the water column and at lower rates in sea ice. To address these hypotheses, future studies should directly compare both sea ice and water, particularly during ice freeze-up, and involve investigation of the microbial processes. 30 Wood, A. M. and Joye, S. B.: The rise and fall of methanotrophy following a deepwater oil-well blowout, Nat. Geosci., 7(6), 423-427, doi:10.1038/ngeo2156, 2014 Pathways of methane in seawater: Plume spreading in an Arctic shelf environment (SW-Spitsbergen), Cont. Shelf Res., 25(12-13), 1453-1472, doi:10.1016/j.csr.2005.03.003, 2005. Damm, E., Kiene, R. P., Schwarz, J., Falck, E. and Dieckmann, G.: Methane cycling in Arctic shelf water and its relationship 5 with phytoplankton biomass and DMSP, Mar. Chem., 109(1-2), 45-59, doi:10.1016/j.marchem.2007.12.003, 2008. Damm, E., Helmke, E., Thoms, S., Schauer, U., Nöthig, E., Bakker, K. and Kiene, R. P.: Methane production in aerobic oligotrophic surface water in the central Arctic Ocean, Biogeosciences, 7 (3)

density (c), cell abundance (d), methane concentration (a) and stable isotope ratios (b). Error bars on cell numbers (d) represent the standard deviation on
5 two technical replicates. Temperature and salinity were determined with an YSI hand held (circle) and YSI sonde (triangle). Salinity for the YSI hand held on 15 April was determined with in the laboratory, thus in situ temperature is missing. Salinity for the YSI sonde on 9 and 15 April is missing due to freezing of the sensor. Methane data is only available for EL at 7 April and for IMB 4 at 15 April.       ice core 2 in situ CH 4 , T/S, DNA, cell counts 1 Station abbreviations are Elson Lagoon (EL) and ice mass balance buoy (IMB) 2 Parameters: in situ concentration and δ 13 CH 4 (in situ CH 4 ), net oxidation/production rate (ox rate), temperature and salinity (T/S), collection of biomass for DNA extraction (DNA), cell counts, nutrients 3 No complete depth profile available 5    1 Levels: ρ S < 0.8 very strong, 0.6 < ρ S < 0.8 strong 2 normalized to total abundance of reads using the DESeq2 package 3 MOB cell density was calculated from relative abundance and flow cytometry cell counts, weighted for copy number of 16S 5 for respective OTUs 4 Significance levels: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 'ns ' 1 5 Candidate OTUs are OTUs that were differentially more abundant in 10x and 200x incubated samples