The ice-free margin of southwest Greenland is characterized by a low continental climate and is strongly influenced by the high-pressure system over the Greenland Ice Sheet (Bennike, 2000). In Kangerlussuaq, the average annual temperature is −5°C from 2008 to present, the maximum temperature measured was 15.4°C in June of 2018, and the minimum temperature measured was −44.9°C in February of 2015. The average rainfall is 243 mm. Teardrop lake (informal name) is located ∼5 km from the terminal moraine of the Russell Glacier (Figure 1). It is a relatively small lake, with a surface area of 0.97 ha, volume of 34,000 m³ and maximum depth of 5.25 m. Teardrop is a moderately brackish, mesotrophic lake, with mean conductivity of 2,740 μS cm⁻¹and mean dissolved organic carbon of 97.6 mg L⁻¹ (Cadieux et al., 2016). Under open-water conditions in 2013, the pH was alkaline and buffered, ranging from 8.98 to 9.3. Currently, Teardrop lake is a hydrologically closed basin, with no direct inflow or outflow channel. Due to continuous permafrost, groundwater seepage is assumed to be limited. While many lakes in the Kangerlussuaq region are cold monomictic (Anderson et al., 2001), Teardrop lake is a dimictic lake, exhibiting ice cover from late October to early June (Cadieux et al., 2016).

Water samples and measurements were taken at two depth profiles located in the deepest basin of Teardrop lake (Figure 1C) in August of 2015. Profiles were similar between the two locations, noted East and West (Supplementary Figures S1,S2). Herein, we will present findings from the East sampling station located in the center of the basin at the maximum depth point. Dissolved oxygen (D. O., mg L⁻¹), pH, specific conductivity (μS cm⁻¹) and oxidation reduction potential (ORP, mV) were measured using a YSI 6039 Data Sonde (Yellow Springs Inc., Yellow Springs, OH, United States) deployed at 0.5 m depth intervals down to maximum depth. Photosynthetically active radiation (PAR) was measured using a Li-Cor LI193 Spherical Quantum Sensor. Water samples were taken with a submersible pump at 0.5 m–1 m intervals throughout the water column at each location, with the exception for methane, where only one sample was obtained from the well-mixed epilimnion and the remainder of the water column was sampled at 0.5 m intervals.

For methane, 1 L of water at each interval was immediately stripped in the field using a headspace-equilibrium technique to extract methane gas from water (Westendorp, 1985). Headspace gas was displaced into a Cali-5-Bond bag using surficial water (Cadieux et al., 2016, 2017). For chlorophyll, 250 ml of water was filtered onto a 0.7 nm glass filter, and the filter was transferred to an amber bottle and remained frozen until analysis. Additionally, 1 L of water was collected into sterile HDPE Nalgene bottles and filtered through 0.22 μm Isopore Membrane Filters (Millipore SiGMa, Burlington, MA) for microbial analysis. Filters were frozen until analysis.

Chlorophyll-a concentrations were determined using acetone pigment extraction in aqueous acetone solution and spectrophotometric determination of chlorophyll a with pheophytin correction at Indiana University (APHA, 2012). Concentrations of dissolved aqueous methane were measured on a Gas Chromatography Flame Ionization Detector (GC-FID) at Indiana University. Stable C and H isotopes of methane were analyzed using continuous-flow gas chromatography-oxidation/pyrolysis isotope ratio mass spectrometry (GC-ox/pyr-IRMS) with a laboratory-built gas pre-concentrator interfaced with a ThermoFinnigan GCC and Delta Plus XP IRMS at the Stable Isotope Research Facility (SIRF) at Indiana University in Bloomington Indiana (Cadieux et al., 2016). Carbon isotope values are referenced to Vienna Pee Dee Belemnite standard (VPDB) and hydrogen isotope values are referenced to Vienna Standard Mean Ocean Water (VSMOW) by comparison with pure methane where the isotopic composition has been measured by off-line combustion. Instrument uncertainty of δ¹³C and δ²H from the GC-C-IRMS was calculated as 0.4‰ and 10‰ respectively, which is one standard deviation for repeated measurements of the gas standards analyzed during sample runs.

Genomic DNA was extracted using the PowerWater DNA isolation kit (MOBIO laboratories, Carlsbard, CA). Amplification of the 16S rRNA genes was performed using the forward primer 515f and a barcoded reverse primer 806r (Caporaso et al., 2012, http://www.earthmicrobiome.org). Each reaction was 25 μl in volume with 5 μl 5′ HF Buffer (New England BioLabs), 1.5 μl of 50 mM MgCl₂, 0.5 ml 10 mM dNTPs, 0.5 μl of each 10 mM primer (Integrated DNA Technologies, Coraville, IA) 0.25 μl of 2,000 U/ml Phusion DNA Polymerase (New England Biolabs), 10 ng–20 ng of DNA, and 15.75 μl PCR water. Amplification was performed using an initial incubation step at 94°C for 3 min followed by 35 cycles of 94°C for 45 s, 50°C for 1 min, 72°C for 1.5 min and a final extension step at 72°C for 10 min. PCR amplicons were cleaned using the PCR purification kit (QIAgen, Valencia, CA). Amplicons were pooled in an equimolar mixture and the pool was cleaned an additional time using the Agencourt AMPure XP (Beckman Coulter, Brea, CA). Sequences were generated on the MiSeq Illumina platform using MiSeq Reagent Kit v2 (300 cycles, PE 300 bp, Illumina Inc., San Diego, CA) and custom-made sequencing primers (http://www.earthmicrobiome.org).

After quality control amplicon sequence variants (ASVs) were generated using the DADA2 (Callahan et al., 2016) subcommand from within QIIME2 (Bolyen et al., 2019) release 2018.2 with the parameters “--p-trunc-len-f 200- --p-trunc-len-r 95”. A portion of the Mothur MiSeq SOP (Kozich et al., 2013) was then followed to align reads to the RDP training set v.9 and remove reads identified as anything other than bacteria or archaea. Remaining reads were imported back into QIIME2 and chimeras were removed using the “vsearch uchime-denovo” subcommand. ASVs were classified using the “classify-sklearn” command in QIIME2 against release 138 of the Silva SSU database (Quast et al., 2013) and a rooted tree was generated using the phylogeny subpackage with default parameters. Shannon entropy and Faith pd diversity measures were generated within QIIME 2. We used these sequence data to assess changes in microbial community composition based on differences in relative abundances with depth in Teardrop lake. Recent studies have pointed out that using differences in relative abundances without determining changes in total abundance through e.g., qPCR can be misleading when inferring changes in microbial community composition in the environment (Aitchison 1982; Gloor et al., 2017; Knight et al., 2018; Morton et al., 2019). An increase in the relative abundance of e.g., methanotrophs can be associated with low absolute numbers of methanotrophs at a particular location suggesting that the relevance of this function may still be limited. Furthermore, changes in the relative abundance of one taxa results in compensatory changes in at least one other taxa as the relative abundances of the microbial taxa are not independent from one another (Aitchison 1982; Gloor et al., 2017). Many different underlying scenarios for changes in the microbial community lead to the same relative abundances of taxa (Morton et al., 2019). However, in this study we assessed the changes in microbial community composition in combination with changes in multiple geochemical parameters. This approach has allowed us to investigate linkages between inferred microbial functional groups and corresponding geochemical processes.

Phytoplankton samples were concentrated via sedimentation and settling using an Utermoehl settling chamber. Taxa present were identified, and plankton enumeration was done using a nanoplankton counting chamber at 200x magnification ensuring a minimum of 200 organisms were counted for each sample depth. Colony forming phytoplankton were enumerated and individual cells were counted from a minimum of 10 organisms to generate an average cell count. This multiplication factor was used to generate cells per milliliter (APHA, 2012).

The traditional conceptual model of methane cycling in northern lakes involves methane production in anoxic sediments that is transported vertically up the water column by ebullition and diffusion, with oxidative removal along the transport path that generates a concentration profile of dissolved methane that decreases up the water column (Wik et al., 2016; Thottathil and Prairie, 2021). Our results for dissolved methane in Teardrop lake exhibited this general pattern - methane concentrations decreased from a maximum just above the sediment/water interface to a minimum in the upper water column. Oxidative removal of methane along the path was also likely. Dissolved oxygen was clinograde, reaching hypoxic conditions at the sediment-water interface. Oxidation of methane by aerobic methanotrophic bacteria was likely occurring throughout the water column as evidenced by the inverse relationship between dissolved oxygen and dissolved methane concentrations throughout, the change in methane isotopic signature, and the presence of methanotrophs throughout the water column (Figure 5). Dissolved oxygen concentrations were sufficient to support methanotrophy throughout the water column which is consistent with an increase in relative abundance of putative methanotrophs we observed below 3.5 m (Figure 5). Additionally, a strong positive correlation was observed between δ¹³C and δ²H of dissolved methane (n = 12, R = 0.93, p < 0.001; Supplementary Figure S12), characteristic of residual methane following aerobic methanotrophic activity (Whiticar and Faber, 1985; Happell et al., 1994; Whiticar, 1999; Bastviken et al., 2002; Alstad and Whiticar, 2011; Wik et al., 2020). The average slope of enrichment of ¹³C and ²H of dissolved methane, calculated as δ²H/δ¹³C (5.2) is similar to the slope of projected oxidation predicted by previous experimental studies, which range from 5.9 to 13 (Coleman and Risatti, 1981; Whiticar, 1999; Kinnaman et al., 2007). These results are consistent with previous findings from Teardrop lake, other Greenlandic lakes in this area (Cadieux et al., 2016), and other Arctic freshwater environments (Wik et al., 2020).

Putative methanotrophic communities were present throughout the water column of Teardrop lake and were similar to other lacustrine systems. Gammaproteobacteria, the dominant putative methanotrophs observed here, were also the most abundant taxa in high oxidation zones in several other lacustrine systems (Carini et al., 2005; Blees et al., 2014; Oswald et al., 2015, 2016; Zigah et al., 2015). In neighboring Potentilla lake, where microbial communities were collected under ice in late winter, we detected high relative abundances of Gammaproteobacteria known to be type I methanotrophs, including Crenothrix and Methylobacter, which dominated between 4 m and 5 m depth (Schütte et al., 2016). In Teardrop lake, Methylovulum was the dominant type I methanotroph, and exhibited the highest relative abundance at 4 m. Beijerinckiaceae (Alphaproteobacteria) including Methylobacterium were present in low relative abundance in Teardrop lake. However, Tamas et al. (2014) reported that members of Beijerinckiaceae lost their capabilities of methanotrophy and are not methanotrophs. Others report that Methylobacterium is methylotrophic and not methanotrophic (Dedysh et al., 2004; Theisen and Murrell, 2005). Most other studies of Arctic lakes described methanotroph community composition mainly in lake sediments rather than from the water column (He et al., 2012, 2022). Consistent with our results, He et al. (2012) found the methanotrophic community in an Arctic lake in Alaska to be dominated by members of the Gammaproteobacteria. In contrast, studies from boreal lake systems found that Gammaproteobacteria, in particular Candidatus Methylumidiphilus within the order Methylococcales, had high relative abundances throughout the lake, but Alphaproteobacteria known to be type II methanotrophs had high relative abundance in surface water where methane concentrations were comparatively low (0 μM–1 μM) (Martin et al., 2021). In comparison, methane concentrations in Teardrop lake were elevated throughout the water column (2 μM–27 μM), consistent with the ubiquitous presence of type I methanotrophs, which have been suggested to have lower methane affinity compared to type II methanotrophs (Martin et al., 2021).

Coexisting with a traditional methane cycle functioning throughout the water column, we found evidence of methane cycling centered between 3 and 5 m depth that may be decoupled from the full water column cycle. Extreme isotope enrichment of ¹³C and ²H of dissolved methane centered at 4 m depth may result from methanogenesis coupled tightly to methanotrophy in what approaches closed-system characteristics. Microbial oxidation of methane under these conditions leads to a systematic enrichment in both ¹³C and ²H of the residual un-oxidized methane pool (Whiticar and Faber, 1985; Happell et al., 1994; Whiticar, 1999; Bastviken et al., 2002; Alstad and Whiticar, 2011; Wik et al., 2020). At the time of sampling in August 2015, dissolved methane in Teardrop lake at 3 m–4 m depth was isotopically enriched by as much as 36‰ for ¹³C and 279‰ for ²H relative to bottom waters at 5 m depth. The lake was weakly thermally stratified and likely approaching holomixis. However, a slight increase in temperature was observed at 4 m depth, potentially supporting increased microbial processes leading to a thermal discontinuity or temporary biogenic meromixis. Oxidation-reduction potential was also elevated within this zone by 15 mV. A subsurface maximum in the isotopic enrichment of ¹³C and ²H in dissolved methane is an unusual pattern relative to other local lakes that we first observed in Teardrop lake in 2013 (Cadieux et al., 2016). However, this pattern has recently been reported in a German lake (Lake Stechlin; Hartmann et al., 2018), as well as group of northern freshwater lakes in the Canadian Shield (i.e., Morency Lake, Croche Lake, Cromwell Lake, Gelil Lake), with δ¹³C mid-depth values as high as +1.7‰ and is attributed to localized elevated methanotrophic activity (Thottathil et al., 2019).

Photosynthesis in the euphotic zone of lakes is known to fuel aerobic methane oxidation as the addition of dissolved oxygen can stimulate methane oxidation and rates of oxygen production can be of the same magnitude as methane oxidation (Oswald et al., 2015). In Teardrop lake, the entire water column was euphotic, with photosynthesis likely occurring at all depths, including 3.5 m–4.5 m depth. Additionally, we observed concentrated layers of potential microbial growth suspended throughout the lake at mid-depths, supported by increased phytoplankton counts at 3 m (Figure 3). These suspended microbial communities could contribute to enhanced oxygenic photosynthesis. At the major methane isotope excursion at 4 m, there was no observable increase in dissolved oxygen, as oxygen is saturated and not a limiting factor. However, there was a 10% increase in ORP to +134 mV between 4.5 m and 3.5 m confirming sufficient concentration of dissolved oxygen to meet terminal electron acceptor demand from aerobic methanotrophy.

Corresponding with the extreme isotope enrichment in ¹³C and ²H of methane at 4 m was an increase in the relative abundance of putative methanotrophs. The putative methanotrophic community was dominated by type I methanotrophs belonging to the Gammaproteobacteria. Despite being microaerophilic, incubation studies suggest that Gamma-MOB also have a high tolerance for dissolved oxygen (Reis et al., 2020). Methylovulum had the highest relative abundance at 4 m in Teardrop lake. The change in relative abundances of putative methanotrophs with depth was consistent with changes in putative methylotrophic bacteria. Methylotrophic bacteria such as Methylotenera and Methylobacterium were present throughout the water column, but showed a significant increase at 4 m consistent with the increase in methanotrophs (i.e., relative abundance of 0.002 at 3 m to 0.312 at 4 m for Methylotenera; Supplementary Figure S8). Recent studies have shown that methylotrophs can co-feed with methanotrophs (Krause et al., 2016). The presence of methylotrophs can lead to a change in gene expression and the production of a different enzyme that results in the release of methanol that methylotrophs can then metabolize (Krause et al., 2016). Metagenomic analysis of methanotrophs in Arctic lake sediments showed that methanotrophs also have the capability to produce methanol (He et al., 2022). This suggests that methylotrophs in Teardrop lake have the capability to co-feed with methanotrophs and that the increase in relative abundance of methanotrophs at 4 m may help account for the increase of methylotrophs at that depth.

In Teardrop lake, methane concentrations increased at the interval corresponding with extreme isotope enrichment of ¹³C and ²H, which is not typical of classic methanotrophy. Production and accumulation of methane in oxic waters, known as the “methane paradox,” has been documented in other lakes, globally (Keppler et al., 2006; Bogard et al., 2014; Tang et al., 2014, 2016; Günthel et al., 2019; Bizic et al., 2020; Bizic 2021; Wang et al., 2021) and could lead to a mid-water-depth increase in methane concentrations. Putative methanogens were identified both above and below the zone of ¹³C and ²H isotope enrichment of methane at 3 m and 5 m, though relative abundances were overall low. Methanogenesis in oxic water has been demonstrated to be driven by acetoclastic production (Bogard et al., 2014). We only detected two ASVs classified as Methanosaeta known to be capable of acetoclastic production with increased relative abundance at 5 m (Fournier and Gogarten 2008). All other putative methanogens we detected carry out hydrogenotrophic and/or methylotrophic methanogenesis (Vanwonterghem et al., 2016). Isotopic composition of methane above and below the interval of isotopic excursion ranged between −37‰ and −38‰ for δ¹³C and −100‰ to −227‰ for δ²H, which is closer to the isotopic range from acetoclastic methanogenesis than to that from hydrogenotrophic methanogenesis (Whiticar 1999). Additionally, accumulation of methane in oxygenated freshwater systems, though lower than archaeal methanogenic production rates, has been associated with the presence of phytoplankton including cyanobacteria, which readily convert fixed inorganic carbon into methane under both light and dark conditions (Bizic et al., 2020, 2021; Hartmann et al., 2020; Bizic 2021). Potential pathways include association of archaeal methanogens with cyanobacteria (Berg et al., 2014), demethylation of methylophosphates (Yao et al., 2016) under phosphorus starvation (Bizic et al., 2020), and a connection to the photosynthetic electron transfer chain in cyanobacteria (Bizic et al., 2020, 2021). In Teardrop lake, phytoplankton were present throughout the water column, with highest abundance especially of the cyanobacteria, Anabaena, at the surface, but a marked increase of Aphanocapsa at 4 m and other phytoplankton at 3 m depth. If methanogenesis in the oxic water column is associated with algal activity, the increased abundance of more labile organic matter could potentially promote methanogenesis. Microbial degradation of methylphosphate to methane has been shown to result in minimal fractionation (Taenzer et al., 2020), but isotopic fractionation from other oxic methane production pathways has not been reported (Bizic et al., 2020; Wang et al., 2021). Therefore, we are unable to use isotopic signatures of methane to confirm if oxic methane production is occurring in Teardrop Lake.

In addition to mid-depth methanogenesis, methane in surface intervals may also derive from different source areas than the anoxic sediments of deeper waters. One possibility is lateral transport of methane from littoral sediments, which has been reported in other lakes, and can amount to 30%–40% of methane emitted during the summer stratification (Günthel et al., 2020). Though in lower relative abundance, we detected putative methanogens throughout the water column (Figure 5). Teardrop lake is a hydrologically closed basin, with no inputs from other water bodies. Lateral methane source areas are limited to shallow littoral sediments. Carbon isotopic composition of porewater methane collected from littoral sediments in 2013 ranged from −48.6 to −36.5‰ (Cadieux et al., 2016), within the range of epilimnetic waters in Teardrop lake. The dissolved methane in profundal sediments of Teardrop lake was more depleted in ¹³C and enriched in ²H relative to the methane in littoral sediments, suggesting different microbial production pathways (Cadieux et al., 2016). Partial oxidation of porewater methane or more isotopically enriched methanogenic substrates in littoral sediments can also contribute to differences in the isotopic composition of dissolved methane between littoral and profundal sediments in small Arctic lakes (Thompson et al., 2018).

Biogeochemical cycles in lakes are often coupled, and driven by underlying microbial community interactions (Wetzel, 2001). In particular, methane and nitrogen cycling are known to be coupled (Bodelier and Steenbergh, 2014; Zhu et al., 2016; Jing et al., 2020). Under anaerobic conditions, methane oxidation occurs using alternative electron acceptors such as nitrate and nitrite instead of oxygen (Valentine, 2002; Oswald et al., 2017). Nitrogen concentrations have been found to both inhibit and stimulate methanotrophic activity and methane oxidation. For example, ammonium-based fertilization has been found to stimulate the growth and activity of methane oxidizers in rice patties (Bodelier et al., 2000; Krüger and Frenzel, 2003), however, ammonium concentrations > 4 mM have also been found to repress methane oxidation in littoral lake sediments (Bosse et al., 1993). Ammonium concentrations in Teardrop lake are 3 orders of magnitude below the threshold for inhibition and unlikely to influence methane oxidation. Methanotrophic bacteria have also been suggested to co-oxidize ammonia when nitrite is present (Harrits and Hanson, 1980). Nitrite concentrations in Teardrop lake exceeded nitrate concentrations, which has also been reported in other local Greenlandic lakes (Cadieux et al., 2016, 2017; Schütte et al., 2016). In Teardrop lake, nitrite concentrations increased slightly at 4 m where extreme isotope enrichment of ¹³C and ²H was observed associated with a drop in the relative abundance of putative ammonia oxidizers (Figure 6A), but concentrations remained overall stable throughout the water column. Additionally, the presence of nitrate has been suggested to stimulate methane oxidation rates under oxic conditions, with Methylotenera being capable of denitrification (van Grinsven et al., 2020). Relative proportions of putative denitrifiers were overall constant in Teardrop lake, though the putative methylotroph Methylotenera increased at 4 m–2.5% (Supplementary Figure S9). This increase in the relative abundance of Methylotenera was mostly driven by one ASV. Despite not finding clear linkages between methane and nitrogen biogeochemical cycles in Teardrop lake, it is possible that the extreme isotope enrichment of ¹³C and ²H in dissolved methane resulted from alternative co-cycling of nutrients.

Arctic lakes are likely to represent a growing net source of methane to the atmosphere with rapid warming of northern landscapes (Wik et al., 2016; Li and Xue, 2021). Measurements of global atmospheric δ¹³C-CH₄ suggest a rapidly increasing microbial source (Lan et al., 2021) that may account for up to 85% of the recent growth in atmospheric methane (Tollefson, 2022). Identification and inclusion of the key biogeochemical processes in climate models is a significant challenge (Lan et al., 2021). In the last 10 years, the methane paradox has been unfolding in the literature, demonstrating that methane in oxic water is not exclusively from transport from anoxic sources, and oxic methane production may account for 18%–90% of surface emissions (Keppler et al., 2006; Grossart et al., 2011; Bogard et al., 2014; Tang et al., 2014, 2016; Günthel et al., 2019, 2020). We suggest that, in Teardrop lake, methane in the water column may derive from a combination of sources: anoxic sediments, lateral transport from littoral areas and oxic methanogenesis in the mid-water column. Additionally, there is evidence for tight carbon recycling within a limited depth interval of the water column, leading to extreme isotope enrichment in ¹³C and ²H of the dissolved methane pool. Our work from Teardrop lake highlights the challenge in predicting potential methane emissions from small, shallow water bodies. Most conceptual models of greenhouse gas emissions from northern landscapes utilize large lakes and assume methane emissions from conventional anoxic sources (IPCC 2013; Saunois et al., 2016). Feedbacks involving small lakes have been lacking (Rasilo et al., 2015). As permafrost thaw rapidly expands with warming climate, smaller thermokarst lakes are increasing in number and area in the Arctic (Abnizova et al., 2012; Gao et al., 2013). It is important to integrate methane cycling of small Arctic lakes into the conceptual modeling framework for greenhouse gas emissions from the terrestrial biosphere.