Spatial Heterogeneity in Methane Biogeochemistry and Prokaryotic Community Structure in Sub‐Arctic Waterbodies in Northern Canada

ABSTRACT

Arctic warming is leading to permafrost thawing which modifies, in cascade, hydrosystems at diverse levels. This study aimed to compare prokaryotic community structure and methane (CH₄) dynamics across 16 sub‐Arctic waterbodies, and to assess how these features are shaped by permafrost thaw. The sampled waterbodies, located in an ice‐rich discontinuous permafrost region (Southwestern Yukon, Canada) differed in size, depth, stratification and degree of thaw influence. Prokaryotic communities were characterised through 16S rRNA gene sequencing and qPCR targeting mcrA (methanogenesis) and pmoA (methanotrophy) genes. Community structures differed significantly between shallow stratified, deep stratified and non‐stratified waterbodies. Methanogens, predominantly represented by the Methanobacterium genus, were mostly detected in shallow non‐stratified waterbodies. Methanotrophs, primarily represented by the Methylacidiphilaceae family, were more abundant in oxic layers whereas bacteria of Crenothrix and Methylobacter genera dominated in anoxic conditions. Our results showed that non‐stratified waterbodies directly affected by permafrost thaw harboured distinct prokaryotic communities, including specific methanogens and methanotrophs. The two sites with the highest CH₄ emissions were affected by permafrost thaw, with fluxes reaching up to 1.7 × 10⁻¹ mg m⁻² s⁻¹. Future investigations should address gaps in CH₄‐related processes in thaw‐affected systems, as they are hotspots for methane emissions and harbour different microbial communities.

Prokaryotic communities and methane dynamics differed markedly across 16 sub‐Arctic waterbodies. Maximum depth and permafrost thaw emerged as key drivers, with non‐stratified thaw‐affected sites hosting distinct communities and elevated methane fluxes, highlighting their potential role as emerging hotspots of methane emissions.

1. Introduction

High‐latitude regions are warming up to four times faster than the global average on Earth (Rantanen et al. 2022). Warming affects the distribution of aquatic ecosystems (Smith et al. 2005; Shugar et al. 2020) and influences their functioning by favouring autotrophic and heterotrophic production in these regions (Saros et al. 2023). Moreover, high latitudes host the largest number of lakes in the world as well as the most extensive lake surface areas (Verpoorter et al. 2014). Permafrost thawing that is widely and rapidly occurring in this area is reshaping the hydrological landscape. The thawing of ice‐rich permafrost (containing a high volume of ground‐ice, i.e., > 50%) leads to the formation of many thermokarst ponds and lakes (i.e., thermokarst process; Payette et al. 2004; Smith et al. 2007; Kallistova et al. 2019; Bouchard et al. 2020). Permafrost contains high stocks of carbon (i.e., 1500 Pg; Schuur et al. 2015) and other compounds (e.g., nitrogen, nutrients) that can be released (mainly under organic and inorganic dissolved forms) in large amounts into these aquatic ecosystems (Vonk et al. 2015). These molecules can then fuel greenhouse gas emissions, which can reinforce climate warming through a positive feedback loop. Such inputs can occur directly into thermokarst lakes actively developing by thawing but can also occur by hydrological transfers into non‐thermokarst aquatic ecosystems. Yedoma permafrost (ice‐rich late Pleistocene permafrost), although covering only ~14% of the total permafrost area, holds approximately 30% of its carbon stocks, estimated at around 400 Gt of carbon (Strauss et al. 2017, 2021), and is highly susceptible to climate change due to its high ground‐ice content (Schuur et al. 2015).

Freshwater ecosystems in sub‐Arctic and Arctic regions are significant sources of CH₄ to the atmosphere (Bastviken et al. 2011; Tan and Zhuang 2015; Saunois et al. 2020) due to the processes described above. Methane has a global warming potential 27 times greater than carbon dioxide over a 100‐year time horizon (IPCC 2023). CH₄ emissions from aquatic ecosystems in northern regions are influenced by physicochemical and biological factors, such as temperature, solar radiation, and primary productivity (Wik et al. 2014; Yvon‐Durocher et al. 2014; DelSontro et al. 2016; Dellagnezze et al. 2023). Morphological features of the aquatic ecosystems may also play a key role, with small and shallow ones often exhibiting higher CH₄ concentrations and emissions (Downing 2010; West et al. 2016; Wik et al. 2016). Moreover, CH₄ emissions result from the balance between methanogenesis and methanotrophy. Methanogenesis in aquatic environments primarily occurs in sediments, where anaerobic and reduced conditions, and abundant organic matter prevail, promoting CH₄ production (Borrel et al. 2011; Sanches et al. 2019). Methane produced at the bottom of these ecosystems is then transported through the water column to reach the atmosphere via rapid (ebullition) or slow (diffusion) transport. Methane can also be produced in oxic conditions through various mechanisms including anoxic microniches, oxygen‐tolerant methanogens, cyanobacterial production and alternative metabolic pathways (Grossart et al. 2011; Bogard et al. 2014; Günthel et al. 2019; Bižić et al. 2020) or horizontally transported from the littoral zones to the pelagic areas of the waterbody (Encinas Fernández et al. 2016). The water column production, though documented for nearly 50 years (Scranton and Farrington 1977), can represent a substantial fraction of total CH₄ emissions. Donis et al. (2017) estimated that oxic CH₄ production could contribute up to 90% of diffusive CH₄ emissions to the atmosphere, whereas Günthel et al. (2019) demonstrated that for lakes larger than 1 km², oxic water column production can be the dominant source of atmospheric emissions. Although the conventional vertical model of CH₄ cycle focuses on oxidation at the oxic–anoxic interface (Hanson and Hanson 1996), methanotrophy can also take place throughout the water column via anaerobic oxidation of methane (AOM), a process documented in several northern lakes (Martinez‐Cruz et al. 2017; Rissanen et al. 2018; Cabrol et al. 2020; Thalasso et al. 2020). A large portion of the CH₄ produced (30%–99%) can be oxidised by methanotrophs before reaching the atmosphere, making them key players in mitigating the emissions of this greenhouse gas and, therefore, its impact on climate (Bastviken et al. 2008). Thus, characterising water column methanogenic and methanotrophic communities is essential for understanding the CH₄ cycle, regardless of the dominant production compartment.

Although the majority of studies on microbial methane cycling in high‐latitude regions have focused on ponds (Negandhi et al. 2013; Crevecoeur et al. 2015, 2016, 2017; Vigneron et al. 2019) or on lake sediments (He et al. 2012, 2015; de Jong et al. 2018; Emerson et al. 2021; Meisel et al. 2023), far fewer have examined the water column of northern lakes (Cabrol et al. 2020; Thalasso et al. 2020; Cadieux et al. 2022). Many questions remain thus unanswered. Our study complements these previous investigations by providing new insights into microbial communities and CH₄ dynamics across a wider range of waterbody types. It also contributes to understanding the role of prokaryotes (Bacteria and Archaea) in methane cycling within freshwater ecosystems, particularly in Yedoma regions, which remain understudied.

The aim of this study was to investigate the relationships between microbial community structure and methane dynamics in ponds, thermokarst meltwaters and lakes located in the ice‐rich permafrost region of Southwestern Yukon (Canada). This region is the southernmost limit of permafrost and it is therefore expected to continue to change dramatically in the future. Specifically, we characterised prokaryotic communities, identified methanogens, methanotrophs and methane‐producing processes, quantified methane concentrations and fluxes, and characterised water chemistry and permafrost degradation stages in the water column of 16 waterbodies. The sampled waterbodies differed in terms of size, depth, stratification occurrence and permafrost degradation stage. Prokaryotic communities were assessed by sequencing their 16S rDNA (metabarcoding) and by quantifying functional prokaryote groups (methanogenesis and methanotrophy) using qPCR (targeting mcrA and pmoA, respectively) in combination with in situ measurements of physicochemical parameters and CH₄ concentrations and fluxes. Carbon and deuterium isotopes of CH₄ were also analysed to investigate the processes driving these fluxes.

2. Material and Methods

2.1. Site Description and Sampling Strategy

The study area, covering 827 km², is located in the Yukon Territory (Canada) between 62.164360° N and 62.611760° N and between 140.998120° W and 140.683200° W near the border with Alaska (USA). This area is part of the Yukon River watershed (Brabets et al. 2000) and was only recently recognised as part of the Yedoma permafrost domain (Strauss et al. 2017; Fortier et al. 2018). It is characterised by extensive discontinuous ice‐rich permafrost of late Pleistocene and Holocene age with an average thickness of 80 m and multiple thermokarst lakes (Fortier et al. 2018). Sampling was conducted from August 16 to 24, 2022. Water samples for microbiological and physico‐chemical analyses were collected from 16 waterbodies (12 lakes, 1 pond and 3 thermokarst meltwaters) (Figures 1 and S1, Table 1). Since there is no universally agreed‐upon definition, the largest and/or deeper waterbodies were named herein ‘lakes’ (L01–L17) and the smaller and shallower ones, ‘ponds’ (P01, P05, P06, P16) for convenience. Most of the ‘ponds’ were visibly the result of thermokarst (Table 1, Figure S1). P16, located in the degradation zone of L15, could be fed by thermokarst meltwater (from P05) and groundwater (Figure S1, Table S1).

FIGURE 1.

Location of the 16 waterbodies (map creation with Google Earth Pro, version 7.3), (for more details of the geomorphological features of each waterbody see Figure S1).

TABLE 1.

Classification, limnological properties and number of samples collected for each waterbody.

Category	Site	Geographic coordinates	Sampling date	Surface (m2)	Maximum DEPTH (m)	T° (°C)	DO (%)	Conductivity (μS/cm)	pH	Microbiological sample number
Shallow Non‐Stratified	L01*	62.260270° N 140.683960° W	18 August	132	1	5.9–6.2	83.9–84.2	261–261.9	7.31–7.39	5
	L05	62.164710° N 140.691540° W	20 August	79,454	3	7.6–8.7	98.7–104.5	226.1–228	7.91–8.04	4
	L08	62.576910° N 140.957720° W	21 August	5891	2.3	7.9–9.2	86.5–88.8	263.3–1251	7.15–7.51	3
	L09	62.611760° N 140.991040° W	21 August	803	0.8	13.7–14.4	22–24	247.8–251.4	6.89–6.95	3
	L10	62.607920° N 140.990590° W	21 August	211,317	1.3	14.8–17.1	88.2–88.3	44.6–45.3	6.87–6.88	3
	L12	62.242120° N 140.688810° W	22 August	98	0.3	12.7	56.5	196.3	6.48	2
	L13	62.253500° N 140.683200° W	22 August	244	0.5	14.3	116.8	81.7	7.22	2
	L14*	62.253830° N 140.684630° W	22 August	148	0.8	12.4	84.5	774.0	7.57	1
	P01*	62.260050° N 140.684100° W	16 August	0.5	0.3	1.8	76.6	191.7	6.86	3
	P05*	62.536530° N 140.997860° W	23 August	0.8	0.4	4	1	103.3	6.07	2
	P06*	62.191220° N 140.681470° W	24 August	0.5	0.3	3.5	72.2	64.9	6.47	2
	P16*	62.536500° N 140.998120° W	23 August	1	0.3	12.1	0	64.5	6.01	3
Shallow Stratified	L04	62.164360° N 140.688350° W	19 August	1695	2	10.8–14.2	1.4–71.5	475.9–667	6.58–7.60	8
	L17	62.210660° N 140.697780° W	24 August	984	4	2.8–13.8	0–61.4	132.1–1307	6.60–7.27	4
Deep Stratified	L11	62.241830° N 140.692050° W	22 August	51,269	7.3	6.9–15.1	0–86.6	47.1–53.6	6.17–7.18	10
	L15	62.535340° N 140.993790° W	23 August	19,183	11	3.5–15.5	0–89.4	68.1–109.9	5.99–7.31	8
									Total	63

Note: Waterbodies marked with an asterisk showed some geomorphological features suggesting permafrost thaw impact (see also Figure S1, Table S1).

2.2. Environmental Variables and Morphological Features

Physico‐chemical parameters were measured at the center of each waterbody throughout the water column. Temperature (°C), dissolved oxygen (DO, %), specific conductivity (μS cm⁻¹), pH and redox potential (ORP, mV) were measured using a multi‐parametric probe (ProDSS Multiparameter Sonde, YSI, USA). The entire water column was first screened with the probe to assess stratification in real time. When the water column of shallow waterbodies showed homogeneous conditions (oxygen concentration, redox potential, temperature and conductivity along depth), a single depth was selected for both water sampling and measurement recording. Dissolved organic carbon (DOC, mg L⁻¹), total dissolved nitrogen (TDN, mg L⁻¹), NH₄ ⁺, NO₃ ⁻ and NO₂ ⁻ (mg L⁻¹) were analysed as detailed in Barret et al. (2022). Dissolved organic nitrogen (DON) was calculated as ([TDN] − ([NH₄ ⁺] + [NO₃ ⁻] + [NO₂ ⁻])). The maximum depth of the lakes was measured using a portable bathymetric sounder (Echotest II, Plastimo). The lake surface area was determined from SPOT‐6 satellite image of August 12, 2022. Waterbodies were classified as permafrost thaw‐impacted based on visual indicators including submerged ‘drunken’ trees, shoreline collapse, ice‐wedge polygon erosion forming subsidence areas and presence of thermokarst meltwater inputs (Kokelj and Jorgenson 2013) (Table S1).

2.3. Methane Fluxes

CH₄ emissions (hereafter referred to as CH₄ fluxes) were calculated at the surface of each waterbody using a static floating chamber connected in a closed‐loop configuration to a portable greenhouse gas analyzer (LI‐COR Li‐7810 Trace Gas Analyzer, Li‐COR Biosciences). The method, adapted from Gerardo‐Nieto et al. (2017), was as follows: a chamber with a volume of 4.24 L and a surface area of 453 cm² was placed at the center of each waterbody for 3 to 5 min. Fluxes were determined using Equation (1), where F is the methane flux (in mg m⁻² s⁻¹), ΔC is the slope of the linear regression applied to the time series of methane concentration, V _c is the chamber volume, A _c is the chamber area in contact with water and M _g is the molecular mass of methane (16.04 g mol⁻¹).

$$F=\mathrm{\Delta }{C}^{*}{\left(V_c/A_c\right)}^{*}{M_g}^{*}1000$$ (1)

Once placed at the water surface, CH₄ concentration in the chamber was measured at 1 Hz. The first 30 s of measurements were discarded due to disturbances associated with chamber deployment. All fluxes had an R ² (coefficient of determination from the linear regression between concentration and time during 1–3 min) greater than 0.82. All measurements were performed in triplicates. A custom interactive application was developed to support this analysis and streamline data processing (Szylit et al. 2025b).

2.4. Dissolved Methane Concentrations

Dissolved CH₄ concentrations were assessed at multiple depths along the water column, employing a headspace injection method. A vial of 150 mL was completely filled with the water sample. Then, a headspace was created in the vial by removing 60 mL of water and simultaneously replacing it with 60 mL of air using two syringes. After shaking the vial for 3 min, the gas phase was manually injected into the LI‐COR analyzer in an open‐loop configuration, using air as a standard reference. The injection volume was typically 1 mL, but was manually reduced to 0.3 or 0.1 mL for samples with exceedingly high CH₄ concentrations to remain within the analyzer's detection range. The actual injection volume was considered in the calculation (Equation 2). Methane concentration (C _g , in mol L⁻¹) in the headspace was calculated using Equation (2), where I is the integrated CH₄ peak area following the injection (in ppb s), Q is the instrument flow rate (in mL s⁻¹), V _inject is the volume of gas phase injected into the instrument (in mL), C _base is the methane concentration of the reference air before injection (in ppb), C _conv is the conversion factor for CH₄ (10⁹ to convert ppb to molar fraction) and V _ig is the molar volume of an ideal gas at the sample temperature (in L mol⁻¹).

(2)

Henry's law constant (H, in mol L⁻¹ atm⁻¹) for CH₄ was calculated according to Equation (3), where P is the standard atmospheric pressure (1.013 bar), R is the universal gas constant (0.082 L atm mol⁻¹ K⁻¹), T is the water temperature (in Kelvin) and T _std is the standard temperature (298.15 K). β _Methane is the methane temperature adjustment factor, corresponding to 1700 K, K _methane is the Henry's constant for methane solubility in water at 298.15 K and corresponds to 0.0014 mol L⁻¹ atm⁻¹ (Sander 2023).

$$H=1/\left(P^{*}{K}^{*}{T}^{*}{K_{\text{methane}}}^{*}e\left[{{\beta }_{\text{Methane}}}^{*}\left(\left(1/T\right)-\left(1/T_{\mathrm{std}}\right)\right)\right]\right)$$ (3)

From the CH₄ concentration in the gas phase and the Henry constant, the concentration in the liquid phase was calculated using Equation (4), where C _l is the CH₄ concentration in the liquid phase (in mol L⁻¹), C _g is the CH₄ concentration in the gas phase calculated using Equation (2), V _g and V _l are the respective volumes of the gas and liquid phases in the vial (60 and 90 mL, respectively) and H is the Henry constant calculated using Equation (3).

$$C_l=\left(\left({C_g}^{*}{V}_g\right)+\left({\left(C_g/H\right)}^{*}{V}_l\right)/V_l\right)$$ (4)

An interactive application was developed to process, visualise and ensure the reproducibility of the data (Szylit et al. 2025a).

2.5. Methane Isotopes

Water samples were collected across all 16 waterbodies at various depths depending on the waterbody (detailed sampling protocol in ‘Sampling and DNA extraction’ section). The stable isotopic signatures of CH₄ (δ¹³C‐CH₄ and δD‐CH₄) were analysed to investigate the processes involved in CH₄ cycling. Both isotopic signatures of CH₄ were analysed using a ThermoScientific Precon concentration unit interfaced with a ThermoScientific Delta V Plus isotope ratio mass spectrometer. For further details on the analytical procedure, see Barret et al. (2022).

3. Microbial Community

3.1. Sampling and DNA Extraction

Water samples were collected from all 16 waterbodies (Table 1, Figures 1 and S1). In deep and large waterbodies, sampling was performed from an inflatable boat using a 2.2 L Van Dorn bottle. In smaller waterbodies, water was collected from the shore using a clean plastic beaker attached to a 4 m rod or held by hand in very small aquatic ecosystems. In shallow (less than 1 m deep) waterbodies, water was sampled at 10 cm depth. In stratified systems, multiple depths were sampled along the water column. At each depth, between one and three replicates were collected, resulting in a total of 63 samples (Table 1). Sampled water was immediately prefiltered through 100 and 30 μm filters (Polycarbonate membranes, Milipore Isopore) to remove large particles and then filtered through a 0.2‐μm polycarbonate membrane (Whatman, Nucleopore) until filter‐clogging. The filtered volume ranged from 50 to 2500 mL (average 184 mL/sample). The 0.2‐μm filters were fixed in 2 mL of 100% ethanol and then frozen at −80°C until DNA extraction. DNA was extracted from the 0.2‐μm filters using the DNeasy PowerLyzer PowerSoil Kit (Qiagen, Germany). Because of the low DNA content in our samples, a modified protocol was set up to optimise DNA extraction (Supplementary Material 1). The whole extraction process was performed under sterile conditions on ice. Extracted DNA was quantified by fluorescence (Qubit 3.0 Fluorometer, Invitrogen) and stored at −20°C until amplification.

3.2. Quantitative PCR

The copy numbers of genes encoding for 16S rRNA of Bacteria and Archaea, the pmoA gene (encoding the beta subunit of particulate CH₄ monooxygenase, a phylogenetic marker for methanotrophic bacteria) and the mcrA gene (encoding the alpha subunit of methyl coenzyme M reductase, a marker for methanogens and ANMEs) were quantified with primers and amplification conditions described in Table S2, using SsoAdvanced Universal SYBR Green Supermix (BioRad, United States) and a QuantStudio 3 Real‐Time PCR Detection System (Applied Biosystems, United States). For each gene, a standard was produced by ligation of a specific version of the gene of concern onto the synthetic plasmid pEX‐A128 (Eurofins Genomics, Belgium). Standard curves were prepared by dissolving and serially diluting plasmids. The standard curves obtained from the genes described in Table S2 ranged from 0.2 × 10¹ to 0.2 × 10⁷ copies μL⁻¹ for all genes, except for bacterial 16S rRNA, which extended up to 10⁸ copies μL⁻¹. The R ² values of the standard curves were above 0.90 and PCR efficiencies ranged between 88% and 111%. Melting curve analysis and electrophoresis gels were run to confirm amplification specificity and amplicon size obtained from lake samples, respectively.

3.3. 16S rRNA Gene Amplification and Sequencing

Amplification and sequencing were carried out on the same DNA samples as for qPCR. Microbial community composition was determined by high‐throughput sequencing of the gene encoding the small subunit of 16S rRNA in prokaryotes (Bacteria and Archaea) using primers U515F and Pk926R (411 bp) targeting the V4–V5 region (Walters et al. 2015). Unique 10 bp molecular barcodes were added to the forward and reverse primers for each sample to tag the amplicons and differentiate them after sequencing. PCR was performed using the PCR mix and thermocycling procedure described in detail in Table S3. The size of the amplicons was verified on a 1% agarose gel. Amplicons from 3 to 5 PCR reactions of the same sample were pooled (to obtain enough material) and purified using the QIAquick PCR Purification Kit (Qiagen, Germany) before being paired‐end sequenced on an Illumina MiSeq 2 × 300 platform (GENEWIZ, Germany). Alongside but separately, mock communities (Microbial Community DNA Standard, Zymo Research) were amplified and sequenced in triplicate according to Berube et al. (2022) to validate the accuracy and reliability of sequencing (Figure S2).

3.4. Bioinformatic Pipeline

The 7,664,066 raw sequences were processed using the QIIME2 pipeline version 2023.9 (Bolyen et al. 2019). Reads were imported and demultiplexed using cutadapt version 4.5 (Martin 2011). The DADA2 plugin version 1.26.0 (Callahan et al. 2016) was used for denoising, paired‐read merging (with a minimum overlap of 15 base pairs and no mismatches allowed), and chimera removal. Using the same plugin, forward and reverse reads were trimmed to remove primers, truncated based on quality scores (with a threshold of Q38 for forward reads and Q35 for reverse reads), and subsequently processed to generate ASVs. A supplementary step was performed to remove ASVs identified as reverse complements. In our dataset, reads were present in both forward and reverse orientations. Without reorientation, identical biological sequences were interpreted as distinct ASVs depending on their strand orientation, which artificially inflated ASV richness. We used the q2‐readorient tool (https://gitlab.com/DeemTeam/q2_readorient) to detect and merge these reverse‐complement ASV pairs. The ASVs were then taxonomically assigned using VSEARCH version 2.22.1 (Rognes et al. 2016) with the SILVA SSU database version 138.1 (Quast et al. 2013). All taxonomic classifications and nomenclature presented throughout this study are based on this SILVA database version. Singleton ASVs and ASVs affiliated with chloroplasts and mitochondria were removed. For the analysis of the total prokaryotic community, ASVs being less than 0.005% of total reads were removed (Bokulich et al. 2013). The dataset was then rarefied to 23,000 sequences per sample to optimise the balance between sequencing effort and sample retention, resulting in the removal of six samples and leaving a total of 63 samples and 1,449,000 reads for downstream analyses.

3.5. Functional Group Classification

Because of their low abundance, the filtering step that excluded ASVs being less than 0.005% of total reads was not applied prior to rarefaction for functional group analyses. This allowed the retention of rare ASVs that could be functionally relevant. As for the total prokaryotic community, rarefaction was performed at a depth of 23,000 sequences per sample, leading to the exclusion of five samples with insufficient sequencing depth. Methanogens and methanotrophs were identified within the total communities based on the ASV taxonomic affiliation, according to Adam et al. (2017) and Seppey et al. (2023). Methanogenic ASVs were those affiliated with the class Methanonatronarchaeia; the orders Methanobacteriales, Methanococcales, Methanopyrales, Methanofastidiosales, Methanocellales, Methanomicrobiales, Methanomassiliicoccales and Methanomethyliales; families Methanosaetaceae, Methermicoccaceae and Thermogymnomonas; and the genera Methanimicrococcus, Methanococcoides, Methanohalobium, Methanohalophilus, Methanolobus, Methanomethylovorans, Methanosalsum, Methanosarcina, Methanoliparia and Candidatus Methanomethylicus. Methanotroph ASVs were those affiliated with the class ‘ANME‐1’; the families ‘ANME‐2a‐2b’, ‘ANME‐2c’, Methanoperedenaceae, Methylacidiphilaceae, Methylococcaceae, Methylohalobiaceae, Methylomonadaceae and Methylomirabilaceae; and the genera Methylobacterium‐Methylorubrum, Methylocapsa, Methylocella, Methylocystis, Methyloferula, Methylosinus, Methylovirgula, Methyloceanibacter and Archaeoglobus. Samples with < 20 reads of methanogens and/or methanotrophs were removed and were not further considered. Removing very low read samples resulted in 27 samples and 28 ASVs (being 0.11% of total reads) for methanogens and 50 samples, 127 ASVs (1.33% of total reads) for methanotrophs.

3.6. Data Analysis

All statistical analyses were performed using R version 4.3.3 (R Core Team 2013). The dataset was processed using the phyloseq package version 1.46.0 (McMurdie and Holmes 2013). Data manipulation was performed using the dplyr package version 1.1.4 (Wickham et al. 2023) and tidyr version 1.3.1 (Wickham et al. 2024). Graphs were created using the ggplot2 package version 3.5.1 (Wickham 2016).

Alpha diversity indices (Richness, Shannon and Simpson (1‐D)) were calculated using the rarefied prokaryote dataset using the estimate_richness function within the phyloseq package for each sample. Bray‐Curtis distance (Bray and Curtis 1957) was used to study beta diversity in the rarefied prokaryotic community. A Principal Coordinate Analysis (PCoA) was performed to visualise the distances between samples.

Several analyses were performed to evaluate the effects of environmental variables on microbial communities. A permutational multivariate analysis of variance (PERMANOVA; Anderson 2001) was conducted to test for significant differences in microbial community composition between waterbodies (Table 1) using the adonis2 function from the vegan package (Oksanen et al. 2022; version 2.6.4) and the pairwise.adonis function (pairwiseAdonis package, version 0.4.1) for pairwise post hoc tests. To further explore the influence of environmental variables on the microbial community composition, the first step was to determine whether the community response was linear or unimodal. A Detrended Correspondence Analysis (DCA) revealed the length of the first axis exceeded four, confirming the unimodal nature of the data (Leps and Smilauer 2003). Canonical Correspondence Analysis (CCA) was performed. Forward selection was applied using the ordiR2step function (vegan package) to identify environmental variables significantly affecting microbial community composition across all waterbodies. The selected variables were then assessed for collinearity using the vif.cca() function (vegan package) to calculate Variance Inflation Factors (VIF), and those with VIF > 5 (indicating multicollinearity) were iteratively removed. The indval function from the indicspecies package (version 1.7.15) was used to identify taxa significantly associated (p < 0.05) with the sample groupings identified in the PCoA, with an indicator value threshold of 0.8. Indicator taxa were defined as those combining high specificity and fidelity to a given group, meaning they were both frequent and relatively exclusive to it.

For methanotrophic and methanogenic communities, PCoAs were performed, based on Bray‐Curtis distance, and PERMANOVA tests were conducted to test for differences in microbial community composition between waterbodies. The indval function was also used to identify ASVs significantly associated (p < 0.005) with the identified groups, using a lower indicator value threshold of 0.5.

4. Results

4.1. Environmental Variables

The sampled waterbodies differed in several geomorphological and physicochemical features (Table 1). Surface areas ranged from 0.3 to 211,317 m². Maximum depth ranged from 0.3 to 11 m. The deepest waterbodies, L11 and L15, reached depths of 7.3 and 11 m, respectively. Deeper waterbodies generally had larger surface areas, except L10, which exhibited the largest surface area and a maximum depth of only 1.3 m (Figure 1, Table 1). Four waterbodies (P16, P01, P05 and P06) had a surface area between 0.5 and 1 m².

Four waterbodies (L04, L11, L15 and L17) were characterised by thermal stratification, with a mean surface temperature of 14.7°C ± 0.8°C decreasing with depth to 6.4°C ± 3.2°C at the bottom (Table 1, Figure 2A). These waterbodies also showed oxygen stratification, with oxygen saturation dropping from 77% ± 14.5% at the surface to hypoxic or anoxic conditions (0.35% ± 0.7%) in the deeper layers (Figure 2B). A waterbody was considered ‘stratified’ if it showed both temperature and oxygen stratification, with dissolved oxygen saturation falling below 1.4% in the deepest layer (Table 1). Among these waterbodies, only L11 showed a decrease in DON and DOC concentrations with depth (Figure S3). Based on maximum depth and stratification, the waterbodies were classified into three categories: deep stratified waterbodies (L11 and L15), shallow stratified waterbodies (L04 and L17), and shallow non‐stratified waterbodies (L01, L05, L08, L09, L10, L12, L13, L14, P16, P01, P05, P06) (Table 1). Within non‐stratified waterbodies, the lowest temperatures were recorded in the three ponds P01, P05 and P06, with values ranging from 1.8°C to 4°C (Table 1, Figure 2A). P16 and P05 were anoxic, whereas the water column of L13 was supersaturated with oxygen (116.8%) (Figure 2B). The pH also varied between waterbodies, with a clear stratification in deep stratified systems, where pH reached a minimum at the oxycline. In contrast, shallow stratified waterbodies exhibited less marked pH variations. The lowest pH values were measured at P16 (6.01) and P05 (6.07), whereas L05 showed the highest pH, reaching 8.04 (Table 1, Figure 2C).

FIGURE 2.

(A) Temperature (°C), (B) dissolved oxygen (% saturation) and (C) pH profiles across the 16 waterbodies.

Permafrost thaw actively impacted several waterbodies. Active thawing produced thermokarst meltwater at P01, P05 and P06, with meltwater from P01 and P05 supplying L01 and P16, respectively. Five waterbodies (L01, L04, L14, L17 and P16) showed active development from ongoing permafrost degradation (Table S1).

In the PCA carried out with the physico‐chemical characteristics of waterbodies (Figure S4), the bottom waters of L17 and L04, as well as ponds P05 and P16, were oriented along vectors representing CH₄, DOC and DON, indicating that these samples shared a chemical profile associated with elevated concentrations of these compounds.

4.2. Prokaryotic Community Structure

The observed mean prokaryotic richness of all waterbodies was 314 ± 110 ASVs, with the lowest richness found at the surface of L13 (143 ASVs) and the highest in the anoxic layer of L15 (553 ASVs). ASV richness was significantly higher in deep waterbodies, with 381 ± 97 ASVs compared to 288 ± 105 ASVs in shallow waterbodies (Wilcoxon test, p = 0.003). Among stratified waterbodies, only L15 showed higher ASV richness in anoxic layers compared to oxic ones, with 492 ± 79 ASVs versus 255 ± 32 ASVs, respectively (Wilcoxon test, p = 0.04) (Figure S5). In contrast, Shannon evenness and Simpson 1‐D indices did not differ significantly between shallow and deep waterbodies. However, Simpson diversity in L11 was higher in oxic layers (0.98 ± 0.004) than in anoxic ones (0.93 ± 0.04) (Wilcoxon test, p = 0.01).

Metabarcoding analysis showed that Bacteria represented as much as 97.2% ± 5.1% of all reads of the prokaryotic community. qPCR analysis confirmed the dominance of Bacteria in the prokaryotic community, with 16S bacterial gene copies ranging from 1.1 × 10⁴ to 3.1 × 10⁷ copies mL⁻¹, accounting for 97.8% ± 0.9% of the total (Bacteria + Archaea) 16S gene copies (Figure S6).

Proteobacteria (26.8%), Actinobacteriota (20.3%) and Bacteroidota (19.3%) were overall the most abundant bacterial phyla. Their relative abundances varied across waterbodies and within the water column of each waterbody. The three waterbody categories, characterised by stratification and depth, were reflected in microbial community structure, with distinct assemblages in each (Figure 3). The deep stratified waterbodies (L11 and L15) were characterised by abundant Planctomycetota, especially in anoxic layers, reaching 46.3% of the prokaryotic reads at the bottom of L11 and 31.2% below the oxycline of L15. This phylum was also relatively abundant in L10, reaching 8.4% of the total community, while it represented only 0.4% ± 0.3% in the other waterbodies. Cyanobacteria were also present at the surface of the deep stratified waterbodies (20.2% and 9.2% in L11 and L15, respectively) and their abundance decreased with depth (Figure 3A). Cyanobacteria were also abundant in the anoxic zone at the bottom of L04, accounting for 18.4% of the community (Figure 3B). Actinobacteriota were predominantly present in the epilimnion of L04 and L17, with a mean relative abundance of 42.9% ± 4.8%, but showed low abundance under anoxic conditions, averaging 1.2% ± 0.6% in the fully anoxic, non‐stratified waterbodies (P16 and P05), and even lower values in the bottom layers of the stratified waterbodies L04, L15 and L17 (Figure 3). ASVs affiliated with Patescibacteria were mostly present at the non‐stratified waterbodies L01, L08, L14, P16, P01, P05 and P06 (mean 17.9% ± 9.6%), reaching a maximum in P16 (31.9%, Figure 3C). Archaea reached their highest abundance at the bottom of the shallow stratified waterbody L17 representing 25.1% of the total reads and were dominated by the phylum Nanoarchaeota (22.8%) (Figure 3B). qPCR results confirmed this observation, with 1.86 × 10⁶ archaeal 16S rRNA coding gene copies mL⁻¹ measured in bottom waters of L17, representing 5.6% of the total 16S rDNA copy number (Figure S6).

FIGURE 3.

Mean relative abundance of replicates at each depth for the 16 most abundant prokaryotic phyla, representing 98.99% of total reads. Waterbodies are grouped as (A) deep stratified, (B) shallow stratified and (C) shallow non‐stratified. Grey bars indicate oxic (light grey) and anoxic (dark grey) layers. Asterisks indicate waterbodies that were both non‐stratified and impacted by permafrost thaw. The ‘Other’ category includes phyla representing < 0.24% of the community. Detailed taxonomic composition of the seven dominant phyla is shown in Figure S8.

Variability in prokaryotic community composition across lakes and along depth gradients was visualised using a PCoA. The first two axes of the PCoA explained 21.8% and 16.9% of the total variance, respectively (Figure 4A). The PERMANOVA confirmed that the prokaryote communities were significantly different between waterbodies and also between waterbody categories (i.e., shallow non‐stratified, shallow stratified and deep stratified, p < 0.01, Table S4A,B). The deep and shallow communities were clearly separated along the first axis of the PCoA. This observation can be partly explained by the high number of ASVs (191 ASVs) shared exclusively between the two deep stratified waterbodies L11 and L15 (Figure S7).

FIGURE 4.

Principal Coordinate Analysis (PCoA) plot of the (A) prokaryote (C) methanogens (D) methanotrophs communities based on Bray–Curtis distance. Samples with fewer than 20 reads were removed (see M + M), resulting in 27 and 50 samples for methanogens and methanotrophs, respectively. (B) Canonical correspondence analysis of microbial community composition with eight environmental variables after selection by forward analysis (see M + M) including Redox Potential (ORP), dissolved oxygen (DO), dissolved organic nitrogen (DON), Conductivity (Cond) and CH₄ concentration (CH₄, log‐transformed), Temperature, (Temp), surface area (Surface Area) and maximal depth (Max Depth). Two samples from L11 and P06 were excluded due to missing environmental data. Asterisks indicate waterbodies that were both non‐stratified and impacted by permafrost thaw.

The waterbodies that were both non‐stratified and impacted by permafrost thaw (L01, L14, P01, P05, P06 and P16) were also distinct from the other waterbodies (PERMANOVA, p < 0.01, Table S4A). Among the 63 ASVs most indicative of this group, 28 were affiliated with the phylum Patescibacteria (Table S6D). In contrast, although the stratified waterbodies L04 and L17 also showed signs of permafrost thaw, they did not cluster with these non‐stratified sites, reflecting distinct community compositions.

The environmental variables explaining most of the variance in prokaryotic communities were identified by CCA (Figure 4B). The CCA model was significant (p = 0.001, Table S5A) and the constrained variance accounted for 42.7% of the total variance (Table S5B). However, the constrained variance was almost equally shared between the first 5 axes (from 20.6% to 10.7%), with the first 2 axes explaining 37.5% of the constrained variance (i.e., 16.6% of the total variance, Table S5C). All tested variables had a significant contribution (p = 0.001, Table S5D). Maximum depth discriminated the deep stratified waterbodies, while dissolved CH₄ concentration specifically characterised the bottom of L17. DON concentration was associated with P16 and P05 (Figure 4B). Finally, DOC was also tested and found significant (p = 0.001), but was not retained due to collinearity with DON.

4.3. Quantification and Community Structure of Methanogens and Methanotrophs

Methanogens represented a minor fraction of the prokaryotic community, as shown by both qPCR and metabarcoding data, with strong variations across waterbodies and depths. Based on qPCR analysis, the abundance of the mcrA gene ranged from 8.0 to 1.1 × 10⁴ copies mL⁻¹ representing between 0.0002% (in L11 at 3 m depth) and 0.03% (in L17 bottom) of total prokaryotic 16S rRNA gene copies (Figure S9C,D). In stratified waterbodies, the mcrA gene was detected throughout the water column in L11, and only at the surface and bottom layers in L04, L15 and L17 (Figure S9C,D). Metabarcoding analyses resulted in higher relative abundances, with methanogen‐affiliated reads accounting for 0.1%–1.1% of total reads. These reads were predominantly detected in ponds P06 and P01 (Figure 5C). Methanogen communities were primarily composed of archaea from Methanobacterium (68%), Methanofastidiosales (15.5%) and Methanosarcina (11.2%) genera (Figure 5C). Methanogens were detected at the surface of L04, dominated by ASVs affiliated with Methanofastidiosales (81%), and at the bottom of L17, primarily composed of Methanoregula (50%) and Methanofastidiosales (43%). In non‐stratified waterbodies, this functional group was mostly dominated by Methanobacterium, reaching up to 100% of the community in P16 (Figure 5C). In contrast, less than 20 methanogen reads were detected in deep stratified L11 and L15, indicating very low relative abundance. The structure of methanogen communities across waterbodies was assessed using PCoA. The first two axes explained 49.6% and 15.9% of the total variance, respectively. Samples from permafrost‐impacted waterbodies (L01, P01, P05, P06 and P16) formed a tight and distinct cluster, significantly separated from the others (PERMANOVA, p = 0.001, Table S4A; Figure 4C). Only one ASV was identified as an indicator of this cluster, affiliated with the genus ‘Rice Cluster II’ (Table S6B). Oxygen availability (oxic vs. anoxic) had no significant effect on methanogen community structure (PERMANOVA, p = 0.13).

FIGURE 5.

Composition of the eight most abundant methanogen and methanotroph taxa—based on the lowest assigned taxonomic rank, with prefixes (e.g., F_, O_) indicating higher‐level classifications, in (A) deep stratified, (B) shallow stratified and (C) shallow non‐stratified waterbodies. Percentages on the x‐axis refer to the total prokaryotic community. ‘Other’ includes taxa representing < 0.16% (methanogens) and < 0.14% (methanotrophs). Black dots indicate samples with fewer than 20 reads, excluded from the analysis (see M + M). Asterisks indicate waterbodies that were both non‐stratified and impacted by permafrost thaw.

Methanotrophs also represented a minor component of the prokaryotic community. Based on qPCR analysis, the abundance of the pmoA gene ranged from 23.0 to 1.0 × 10⁵ copies mL⁻¹, representing between 0.16% in the top of L05 and 2.98% in the top of L17 (Figure S9A,B). In stratified waterbodies, pmoA was consistently detected throughout the water column and reached its maximum abundance at the surface of L17, accounting for 3% of the prokaryotic community. Relative abundances based on metabarcoding were 1.7% on average, exceptionally peaking at 9.4% at 2 m depth in L15 (Figure 5A). Methanotroph communities were primarily composed of Methylacidiphilaceae (57.1%), Methylobacter (12.8%) and Methylomonadaceae (12.6%) (Figure 5A–C). Methanotrophs were detected in the epilimnion of deep stratified waterbodies, with relative abundances reaching 4.1% in L11 and 9.4% in L15, both dominated by members of Methylacidiphilaceae. In the anoxic layer of L15, Methylobacter was also detected, though at lower abundance (1.3%). In shallow stratified waterbodies, methanotrophs were more abundant at the anoxic bottom: 0.4% in L04 and 1.0% in L17. These communities were dominated by bacteria of Crenothrix in L04 and Methylobacter in L17, both from Methylomonadaceae. In non‐stratified waterbodies, methanotrophs were mainly represented by unclassified members of Methylomonadaceae. The methanotroph community in L10 differed from those of the other non‐stratified waterbodies and was similar to the communities found in the oxic surface layer of L15, with dominance of representatives of Methylacidiphilaceae. The structure of methanotroph communities was assessed using PCoA. The first two axes explained 24.3% and 14.6% of the total variance, respectively (Figure 4D). Communities differed significantly between waterbodies and between deep and shallow samples (PERMANOVA, p = 0.001, Table S4A,B). L10 samples clustered near the oxic layer of L15 in the ordination space. As observed for the total community, permafrost‐impacted non‐stratified waterbodies (L01, L14, P01, P05 and P06) formed a distinct group from the others(PERMANOVA, p = 0.001, Table S4A; Figure 4D), indicating distinct community compositions. The seven most indicative ASVs for this group were affiliated with Candidatus Methanoperedens, Crenothrix, Methylobacter and other members of the Methylomonadaceae family (Table S6C). In contrast to methanogens, oxygen availability significantly shaped methanotroph community structure (PERMANOVA, p = 0.004). IndVal analysis identified one ASV affiliated with Methylacidiphilaceae as indicative of oxic conditions and nine ASVs indicative of anoxic conditions, affiliated with Methylobacter, Crenothrix, other Methylomonadaceae and Methylococcaceae (Table S6A).

4.4. Methane Concentration and Fluxes

In stratified waterbodies, CH₄ accumulated at the bottom of the water column, with concentration profiles showing different patterns between deep and shallow waterbodies. The highest dissolved CH₄ concentrations were found in the anoxic layers of shallow stratified waterbodies (7.0 × 10⁻⁴ mol L⁻¹ at the bottom of L04 and 2.8 × 10⁻³ mol L⁻¹ at the bottom of L17), more than five times higher than in the anoxic water of deep stratified waterbodies (1.6 × 10⁻⁵ and 1.4 × 10⁻⁴ mol L⁻¹ at the bottom of L11 and L15, respectively) (Figure 6A). In the four stratified waterbodies, the high CH₄ concentrations observed at the bottom, combined with δ¹³C‐CH₄ values ranging from −67‰ to −78‰ and δD‐CH₄ values from −389‰ to −397‰ (Figure 6A,B,D). Moreover, CH₄ and dissolved oxygen concentrations showed an inverse vertical distribution from bottom to surface, with CH₄ concentrations decreasing markedly at depths where oxygen levels increased (Spearman's ρ = −0.52, p = 0.0017). In L04 and L17, methane oxidation occurred progressively upward from the bottom of the water column, as shown by a gradual decrease in CH₄ concentrations and corresponding isotopic enrichment; this pattern was particularly pronounced in L04. In L11 and L15, although CH₄ concentrations began to decrease slightly upward from the bottom layer, a sharper decline occurred higher up, at the oxycline, with concentrations dropping to 1.4 × 10⁻⁷ mol L⁻¹ at 4 m in L11 and 1.2 × 10⁻⁷ mol L⁻¹ at 2.5 m in L15. This was accompanied by δD‐CH₄ enrichments of 22‰ and 16‰, respectively. Above the oxycline in L15, CH₄ concentrations increased again towards the surface, reaching 4 × 10⁻⁶ mol L⁻¹ at 1 m depth. The isotopic signatures of CH₄ at this depth (δ¹³C‐CH₄ = −34‰, δD‐CH₄ = −165‰) suggest a methanogenic pathway distinct from that at the bottom of the water column.

FIGURE 6.

Vertical depth profiles of dissolved oxygen (%), mean dissolved CH₄ (mol L⁻¹; logarithmic scale) and stable isotopes of carbon (δ¹³C‐CH₄) and hydrogen (δD‐CH₄) (‰) in (A) Deep Stratified and (B) Shallow Stratified waterbodies. (C) Mean values of the same variables in shallow non‐stratified waterbodies. Points indicate measurements at different depths, bars represent the average values. (D) Carbon and hydrogen isotope ratios of methane (δ¹³C‐CH₄ vs. δD‐CH₄). Shaded fields indicate the expected isotopic signatures for different CH₄ origins (adapted from Whiticar 1999) (E) Methane fluxes (mg m⁻² s⁻¹, log scale). The shaded area shows the range reported for aquatic ecosystems in Kuhn et al. (2021). No measurements were available for P06 and P01. Asterisks indicate waterbodies that were both non‐stratified and impacted by permafrost thaw.

In the shallow non‐stratified waterbodies, dissolved CH₄ concentrations ranged from 8.8 × 10⁻⁷ to 1.3 × 10⁻⁴ mol L⁻¹, reaching their highest values in the anoxic waterbodies P05 and P16 (Figure 6C). In these waterbodies, δ¹³C‐CH₄ values around −69‰ suggest methane production via the acetoclastic or methylotrophic pathway, with a strongly depleted δD‐CH₄ signature of −421‰ in P16, further supporting this interpretation (δD‐CH₄ data were not available for P05). CH₄ fluxes varied from 1.6 × 10⁻³ to 15 × 10⁻⁵ mg m⁻² s⁻¹ across all waterbodies, except for P05, which had exceptionally high fluxes, averaging 1.7 × 10⁻¹ ± 1.4 × 10⁻¹ mg m⁻² s⁻¹ (Figure 6E). Like the dissolved CH₄ concentrations, the highest CH₄ fluxes were observed in the fully anoxic waterbodies with the smallest surface areas, P05 and P16. CH₄ fluxes were significantly positively correlated with dissolved CH₄ concentrations at the surface (Spearman correlation, ρ = 0.6, p = 8.5 × 10⁻⁴) (Figure S10A) and negatively correlated with surface DO saturation at the surface (Spearman correlation, ρ = −0.7, p = 7.8 × 10⁻⁶) (Figure S10B).

5. Discussion

Distinct spatial patterns emerged among the water column of waterbodies, reflecting variations in community structure and CH₄ dynamics (Figure 7). Maximal water depth emerged as the primary factor structuring prokaryotic communities across all waterbodies, overriding the effects of stratification and associated physicochemical gradients. Within shallow, non‐stratified waterbodies, those directly impacted by permafrost thaw exhibited unique prokaryotic community structures and notably elevated CH₄ fluxes, highlighting their potential significance as methane sources (Figure 7). These findings are discussed below in relation to inter‐waterbody variability, vertical gradients within stratified lakes, functional groups involved in CH₄ cycling and the impact of permafrost thaw based on geomorphological indicators and microbial signatures.

FIGURE 7.

Synthetic representation of prokaryotic communities and CH₄ dynamics across the studied waterbodies. Red arrows indicate CH₄ fluxes (mg m⁻² s⁻¹). CH₄ concentrations ([CH₄]) are illustrated by vertical profiles in stratified waterbodies and by the size of the [CH₄] label in non‐stratified systems. Anaerobic oxidation of methane (AOM) and oxic methane production are also represented, with certain taxa (Crenothrix, Methylobacter) capable of both processes depending on oxygen availability. Asterisks indicate waterbodies that were both non‐stratified and impacted by permafrost thaw.

5.1. Processes Driving Methane Concentration and Fluxes

The conventional vertical model of CH₄ concentration in stratified lakes is based on higher concentrations at the bottom of the waterbody due to production occurring in anoxic sediments, and subsequent transport (diffusion or ebullition) through the water column, with decreasing concentrations towards the surface due to methanotrophic activity (e.g., Thottathil and Prairie 2021). In the stratified waterbodies of our study (L04, L11, L15 and L17), methane concentration profiles generally followed this model (Figure 6A,B). CH₄ concentrations at the bottom of these stratified waterbodies were within the same range as those reported above the sediments of other Arctic lakes (Juutinen et al. 2009; Cadieux et al. 2017; Savvichev et al. 2020). The low δ¹³C‐CH₄ and δD‐CH₄ suggested a predominant methylotrophic or acetoclastic pathway (Whiticar 1999). Methane oxidation was most pronounced at the oxic/anoxic interface in L11 and L15, as shown by a decrease in CH₄ concentration and an enrichment in δD‐CH₄ and δ¹³C‐CH₄. This layer combined low oxygen concentrations with enough CH₄ to promote methanotrophic activity (Mau et al. 2013; Blees et al. 2014; Martinez‐Cruz et al. 2015). Methane concentrations showed a less pronounced decline from the hypolimnion upward in all four stratified waterbodies. While this pattern may reflect physical processes such as upward diffusion, it does not exclude the occurrence of anaerobic methane oxidation (AOM). In the hypolimnion, methane‐oxidising bacteria (MOB) affiliated with Crenothrix and Methylobacter, both from the Methylomonadaceae family (Figure 5A,B), were dominant, supporting the potential for AOM, as both genera are known to thrive under low‐oxygen conditions (Mayr et al. 2020). These bacteria have already been identified as key players of AOM in situ in Arctic lakes (Cabrol et al. 2020) and in lake sediment incubation experiments (Martinez‐Cruz et al. 2017). Anaerobic methanotrophic archaea (‘ANME’), known to oxidise CH₄ through the reverse methanogenesis pathway (Timmers et al. 2017) and bacteria performing CH₄ oxidation via nitrite dismutation, such as Methylomirabilis oxyfera (Ettwig et al. 2012), were not detected in our dataset. Besides AOM, the enhanced oxidation activity in anoxic waters could also result from a coupling between methanotrophy and photosynthesis (Milucka et al. 2015; Perez‐Coronel and Michael Beman 2022), the latter being possibly carried out by members of Cyanobium, identified as the dominant phototrophic cyanobacteria in anoxic conditions. AOM has been reported to remove a variable fraction of CH₄ in the anoxic zone, ranging from 3% to 99.9% depending on the ecosystem (Segarra et al. 2015; Cabrol et al. 2020; Gao et al. 2022). It represents a potentially significant sink for CH₄ before it reaches the oxic layer, but its quantitative importance in aquatic CH₄ cycling at a global scale remains poorly understood.

The epilimnion of L15 was characterised by increased CH₄ concentration coupled with δ²H and δ¹³C depletion, which suggests a CH₄ input into this layer (Whiticar 1999). Methane production in oxic conditions in surface waters has been observed in aquatic and marine ecosystems worldwide (e.g., Bižić et al. 2020) and represents a significant source of CH₄ emissions from these environments (Günthel et al. 2019; Perez‐Coronel and Michael Beman 2022); yet the origin of this CH₄ is still unknown. Several studies have identified cyanobacteria as key players in methane production in oxic conditions (Khatun et al. 2019; Bižić et al. 2020). However, the only cyanobacteria detected in L15 oxic water were affiliated with Cyanobium, and to date, no study has documented the ability of this genus to produce CH₄. Furthermore, although Cyanobium was also detected in the epilimnion of L17, L04 and L11, no concurrent increase in CH₄ concentration was observed in these lakes, which does not support a systematic role of this taxon in aerobic CH₄ production. Another hypothesis explaining oxic CH₄ production is the presence of organic matter particles and/or zooplankton faecal pellets, creating reduced anoxic micro‐niches suitable for methanogens (Ditchfield et al. 2012). However, most of the aggregates have been removed during the 30 μm prefiltration of our samples, preventing their detection. Oxic CH₄ production could also result directly from methanogen activity, as shown in a German lake (Grossart et al. 2011), with some methanogens being able to tolerate oxygen exposure (Jarrell 1985). The presence of the mcrA gene detected by qPCR throughout the water column, including in the surface layer, supports this hypothesis (Figure S9), though no methanogens were detected at this depth by metabarcoding. The genus Methylotenera was detected in the epilimnion of L15. Although its members are generally considered methylotrophs, a recent study combining metagenomics and proteomics showed that one ecotype of this genus produces CH₄ under oxic conditions from methylphosphonate (Li et al. 2025) Finally, this increase in CH₄ concentration at the surface may also result from its transport from the littoral zone (Encinas Fernández et al. 2016). Overall, the origin of CH₄ in the oxic epilimnion of L15 remains uncertain and potentially results from a combination of the above mechanisms, including both local production and lateral transport.

The measured CH₄ atmospheric fluxes were generally consistent with those reported for various northern freshwater systems, ranging from ponds to large lakes (Natchimuthu et al. 2014; Rasilo et al. 2015; Sepulveda‐Jauregui et al. 2015; Kuhn et al. 2018, 2021) at the exception of the extremely high values observed at P05 (Figure 6E). As expected, CH₄ fluxes were correlated with methane concentration in the surface layer (Stumm and Morgan 1996) (Figure S10A). CH₄ fluxes were negatively correlated with dissolved oxygen concentration at the surface (Figure S10B). This observation suggests that anoxic surface waters reflect what happens in the whole water column of fully anoxic ecosystems, where methanogenesis is favoured in detriment of methanotrophy. In several large‐scale studies (Holgerson and Raymond 2016; West et al. 2016; Deemer and Holgerson 2021), methane fluxes were negatively correlated with lake maximum depth and surface area and positively correlated with surface temperature, patterns that we did not observe in our ecosystems. However, the two waterbodies where the highest methane fluxes were measured (i.e., permafrost meltwater P05 and pond P16) were both shallow and very small, confirming the importance of small ecosystems (< 0.001 km²), recognised as important contributors to methane emissions from inland waterbodies (Rosentreter et al. 2021).

5.2. Inter‐Waterbodies and Vertical Prokaryotic Community Variability

As previously observed in various Arctic freshwater ecosystems (Crevecoeur et al. 2015; Zakharova et al. 2022; Meisel et al. 2023), alpha diversity indices of prokaryotes indicated a relatively high taxonomic richness and evenness across all waterbodies (Figure S5). The strong seasonality in Arctic and subarctic waterbodies (e.g., ice cover, temperature fluctuations, shifts in physicochemical parameters) enhances alpha‐diversity variations, with higher alpha diversity in summer, that is, during our sampling period (Zhang et al. 2019; Fang et al. 2023). Prokaryotic communities were dominated by the phyla Proteobacteria, Bacteroidota and Actinobacteriota; however, their relative abundances varied considerably between waterbodies (Figure 3). These phyla are commonly reported among the most abundant groups in freshwater ecosystems (Zwart et al. 2002; Newton et al. 2011). Prokaryotic communities also significantly differed between deep and shallow waterbodies, regardless of stratification (Figure 4A). The distinctive phylum of the two deep (stratified) waterbodies (L11 and L15) was Planctomycetota. This phylum is commonly found in various aquatic ecosystems (Fuerst and Sagulenko 2011; Newton et al. 2011). In this study, it was dominated by ASVs of the genus ‘CL500‐3’ of the class Phycisphaerae (Figure S7), a ubiquitous genus found in various aquatic environments (Andrei et al. 2019; Messina‐Pacheco 2019). The presence of this genus throughout the water column of the two deep stratified waterbodies can be explained by its wide range of metabolism and ecology, being either aerobic, facultative anaerobic or strictly fermentative (Lenferink et al. 2024).

Among the stratified waterbodies, only L15 exhibited higher richness and diversity indices in the hypolimnion compared to the epilimnion, a pattern previously reported in other Arctic ecosystems (Schütte et al. 2016; Marois et al. 2022). In this lake, the bottom layers contained higher concentrations of organic compounds such as DON and DOC. This greater substrate availability may result from the downward transport of organic matter from the upper layer and from interactions at the sediment–water interface (Forsberg 1989). In addition, the prokaryotic community structure also varied along the water column depth. Actinobacteriota were less abundant at the bottom of L04, L15 and L17, than at the surface, as previously observed in several aquatic ecosystems (Taipale et al. 2009; Crevecoeur et al. 2015; Block et al. 2021). Actinobacteriota were dominated throughout the water column in stratified waterbodies by members of the ‘hgcI clade’ (Sporichthyaceae), which were also abundant across both oxic and anoxic non‐stratified waterbodies (Figure S8). This clade, previously known as ‘acI’, is abundant across diverse freshwater ecosystems (Warnecke et al. 2004), and its genetic profile indicates an ability to assimilate carbohydrates and nitrogen‐rich organic compounds (Ghylin et al. 2014). Its occurrence across environments ranging from hypoxic to oxic conditions (Liu et al. 2015), along with genomic evidence from a representative ‘acI‐B1’ lineage indicating a facultative aerobic metabolism (Garcia et al. 2013), suggests a potential tolerance to low oxygen concentrations that could explain its persistence in anoxic waters. Cyanobacteria, which are key primary producers in general and in high‐latitude freshwater systems in particular (Vincent 2002; Vincent and Quesada 2012; Makhalanyane et al. 2015), also exhibited a stratified abundance profile. In our dataset, they were mostly found near the surface of deep stratified waterbodies (L11, L15). Interestingly, cyanobacteria were also detected in the anoxic bottom waters of L04, where the community was dominated by Cyanobium and Gastranaerophilales. Both of these genera were present throughout the water column but were most abundant in the bottom layer. The autotrophic bacteria of the Cyanobium genus, although typically associated with oxic environments, have also been reported in anoxic conditions, likely due to post‐bloom aggregation and sedimentation (Nwosu et al. 2021; Rogers et al. 2021). The non‐photosynthetic cyanobacteria of the Gastranaerophilales order, in contrast, are heterotrophic obligate fermenters adapted to anoxic conditions (Soo et al. 2014, 2017), observed in animal gut microbiomes, faecal samples (Di Rienzi et al. 2013) and freshwater lake sediments (Monchamp et al. 2019). These vertical patterns suggest that multiple prokaryotic lineages may occupy distinct ecological niches along the water column, including unexpected taxa detected in deep anoxic layers.

5.3. Spatial Patterns of Methane‐Cycling Prokaryotes

Methanogens and methanotrophs' relative abundances differed between qPCR and metabarcoding estimates, partly because of the targeted genes that differed between the two approaches. While the primers used for metabarcoding targeted the gene coding for the 16S rRNA, those used for qPCR targeted three genes: 16S rDNA, mcrA and pmoA. The variation in copies per genome, up to 37 copies for 16S rDNA (Pan et al. 2023), 3 for pmoA (Stolyar et al. 1999; Tchawa Yimga et al. 2003), and 1 for mcrA, could lead to an underestimation of the relative abundance of methanogens and methanotrophs in qPCR analyses. Adjusting 16S counts based on gene copy number per organism is not recommended, as this number can vary significantly among genomes of the same taxa (Gao and Wu 2023). Additionally, (Wilkins et al. 2015) found that 16S rDNA and mcrA gene sequencing yielded different taxonomic compositions of methanogens, indicating that each gene provides only partial coverage of their diversity. In contrast, for methanotrophs, this discrepancy appears less pronounced, as pmoA and 16S rDNA‐based phylogenies are globally similar (Costello and Lidstrom 1999).

As commonly reported, methanogens (based on qPCR detection) were present in the water column of all our waterbodies but at low relative abundance. Methanogens occurred at very low abundances (< 20 sequences over 23,000 sequences) in deep waterbodies but were most often observed at significant abundances (> 20 sequences) in shallow ones (Figure 5). In addition, two distinct patterns were observed between stratified and non‐stratified shallow waterbodies. In stratified waterbodies, Methanofastidiosales prevailed. Members of this group, known as strictly methylotrophic methanogens (Nobu et al. 2016), were detected at the surface of L04 and at the bottom of L17. This observation suggests a vertical succession of phylogenetically distinct Methanofastidiosales lineages with varying metabolic capabilities, including both aerobic and anaerobic conditions, potentially associated with other functional traits. In contrast, non‐stratified waterbodies were dominated by Methanobacterium members, which are predominantly hydrogenotrophic (Demirel and Scherer 2008), although some lineages exhibit methylotrophic metabolism (Borrel 2011; Kallistova et al. 2023). Representatives of the Methanobacterium genus were abundant in both oxic and anoxic/hypoxic waterbodies (P05 and P16). The higher methanogen abundance in these two shallow waterbodies reflects both sediment proximity and favourable growth conditions, including continuous organic matter inputs from permafrost thaw, and high DOC and DON concentrations. This suggests that some Methanobacterium lineages could have the ability to grow in oxic environments, despite their strict anaerobic classification (Boone 2015). However, although Methanobacterium spp. were the most abundant methanogen in non‐stratified waterbodies, the isotopic signatures indicated acetoclastic or methylotrophic production rather than the hydrogenotrophic pathway typically associated with this genus. This discrepancy could be explained by the significant differences in microbial communities between sediments and the water column, both in composition and assembly processes (Zeng et al. 2019; Espín et al. 2021). The isotopic signatures reflected CH₄ produced in sediments by methanogenic communities distinct from those sampled in the water column. The water column methanobacteria may therefore have contributed minimally to total CH₄ production, with most CH₄ originating from sediment‐based acetoclastic or methylotrophic methanogens.

Methanotrophs were generally more abundant than methanogens in all waterbodies, except in P16 and at the surface of L04 (Figure 5). In deep stratified waterbodies, their distribution followed the vertical oxygen gradient. In the oxic epilimnion, especially in L15, methanotrophs reached up to 9.4% of the prokaryotic community and were dominated by the members of the family Methyloacidiphilaceae. Although typically associated with acidic environments (Sharp et al. 2014; Kalyuzhnaya et al. 2019; Dedysh et al. 2021), this family was present here across a pH range close to neutrality (6.56 < pH < 7.31). In contrast, the anoxic hypolimnion of stratified lakes L04, L15 and L17 was dominated by methylobacteria of Methylomonadaceae, with Crenothrix genera prevailing in L04 and Methylobacter genera in L17 and L15. These observations suggest that methanogen and methanotroph assemblages were structured not only by oxygen availability but also by environmental gradients occurring across waterbodies and within the water column of stratified systems. This spatial distribution likely reflects distinct metabolic strategies and niche partitioning, which in turn may influence net methane emissions by shaping the balance between production and oxidation processes.

Together, these results highlight contrasting community structures of methanogens and methanotrophs across waterbodies and oxygen conditions. For example, Methanofastidiosales, Methanobacterium and Methylobacter were detected in both oxic and anoxic layers, whereas Methylacidiphilaceae, typically associated with acidic environments, were mainly found in near‐neutral pH waters. The occurrence of identical taxa under contrasting conditions may indicate physiological versatility or the presence of ecotypes adapted to specific environmental niches. This microdiversity can involve differentiation in metabolic pathways (Neuenschwander et al. 2018) and ecological traits, resulting in niche partitioning among closely related lineages (Chase and Martiny 2018). This fine‐scale microdiversity is particularly relevant under fluctuating environmental conditions, as it enhances community stability by enabling the coexistence of closely related ecotypes occupying slightly different niches (García‐García et al. 2019). ASVs with nearly identical 16S rDNA gene sequences can have markedly different genomes and ecological functions (e.g., Jaspers and Overmann 2004). Genome‐resolved metagenomic methods are therefore essential to uncover cryptic ecotypes and to better characterise their functional roles in methane cycling and microbial ecosystem processes.

5.4. Challenges to Link Permafrost Thaw With Microbial Communities and Methane Cycling

Identifying particular microbial communities and CH₄ cycles in waterbodies affected by permafrost thaw is challenging. Although the study site is located within the extensive discontinuous ice‐rich permafrost zone (Fortier et al. 2018), identifying waterbodies impacted by permafrost thaw is not straightforward. Thaw‐impacted waterbodies often exhibit morphometric (e.g., depth, surface area) and physico‐chemical (e.g., DOC, total nitrogen, total phosphorus) characteristics overlapping with those of waterbodies considered unaffected by thaw (Grosse et al. 2013; Arsenault et al. 2022), as observed in our study (Figure S4). Consequently, we relied on geomorphological indicators of permafrost degradation, including submerged ‘drunken’ trees, shoreline collapse and ice‐wedge polygon erosion (Kokelj and Jorgenson 2013). Based on these criteria, we classified eight waterbodies as thaw‐affected: two stratified lakes (L04, L17) and six non‐stratified systems, including two lakes (L01, L14), three thermokarst meltwater (P01, P05, P06) and one small pond (P16). These lakes and ponds may represent different development stages of thermokarst lakes as observed elsewhere in the Arctic.

CH₄ fluxes varied notably among thaw‐affected waterbodies (Figure 6E). Highest fluxes were recorded in shallow, anoxic ponds (P05, P16), whereas lake L01 exhibited among the lowest fluxes, despite thaw impact. The high flux observed in P05 and P16 was likely due to a combination of factors: both were shallow and anoxic, conditions that promote methanogenesis while limiting methane oxidation. They were also located in an area of active permafrost thaw (Figure S1, Table S1), characterised by the presence of ice‐wedge polygons. Fluxes at P05 were nearly two orders of magnitude greater than those at P16, likely because of a continuous input of permafrost meltwater supplying organic matter and methane (Paytan et al. 2015; Dabrowski et al. 2020; Olid et al. 2021). In contrast, stratified thaw‐affected lakes (L04, L17) had lower fluxes due to methane oxidation along vertical gradients, underscoring the critical role of lake depth and oxygen gradients in regulating emissions. These observations highlight the importance of integrating local thaw intensity into predictive models of methane fluxes, beyond morphometric parameters alone.

Among these thaw‐affected waterbodies, only the bottom waters of L17 and L04, along with P16, P05 and P01, were clearly separated in the PCA space derived from physico‐chemical analyses, associated with elevated CH₄, DOC and DON concentrations. However, L01 and L14, despite clear thaw indicators, did not separate from unaffected sites, reflecting the overlap in physico‐chemical characteristics between thaw‐affected and unaffected waterbodies. This suggests that additional factors, such as groundwater inputs (A. Séjourné, pers. comm.), may characterise these particular systems. For example, thermokarst lakes, although receiving permafrost meltwater input due to permafrost thawing, also receive groundwater input via the active layer, which depends on the topography. Permafrost features (soil type, ground‐ice and organic matter content) are horizontally and vertically heterogeneous, varying at depth and within an area. This pattern can influence waterbodies' characteristics and promote heterogeneity among these ecosystems. Furthermore, the influence of thaw in complex stratified lakes such as L04 and L17 appeared confined primarily to bottom layers, and was potentially masked in other layers by intrinsic (e.g., chemical gradients, stratification) and extrinsic factors (e.g., hydrological input, surrounding soil characteristics, climate).

The differentiation between waterbodies that were both non‐stratified and impacted by permafrost thaw and all other sites (i.e., ecosystems non‐affected by permafrost thaw and stratified thaw‐affected systems) was reflected in prokaryotic community structure. Indicator taxa included a significant proportion affiliated with Patescibacteria (Figure 3). These organisms, known to be predominantly parasitic of prokaryotes (e.g., Lemos et al. 2019), could indirectly influence methane cycling through parasitizing methanogens, potentially leading to physical damage (e.g., cell wall deformation) and metabolic impairment (e.g., reduced ribosomal activity) (Kuroda et al. 2022). However, Patescibacteria‐methanogen interactions are not entirely resolved, putative syntrophy being suggested (Khamespanah et al. 2024). Although Patescibacteria have previously been detected in thermokarst lakes (Vigneron et al. 2020), they have also been observed in a wide range of environments (Harris et al. 2004), meaning that their presence alone cannot reliably indicate permafrost thaw. The communities of waterbodies that were both non‐stratified and affected by permafrost thaw did not form a clear cluster in the CCA (Figure 4B). Among the thaw‐affected sites, only the microbial communities of L01, P01 and P05 were clearly separated, primarily in association with elevated concentrations of DOC and DON. This highlights the potential role of carbon and nitrogen‐rich organic matter, likely derived from ice‐rich permafrost thaw, as a key environmental driver shaping microbial communities in these systems.

Methane‐cycling communities were also impacted by permafrost thaw, with a clear differentiation between non‐stratified thaw‐affected waterbodies and other systems (Figure 4C,D). Methanogenic communities in thaw‐affected sites showed high relative abundances of Methanobacterium, consistent with substrate‐rich conditions (Sakai et al. 2009). The single ASV indicator identified belonged to ‘Rice Cluster II’ from Methanomicrobiales (Table S6B). To date, no genome of this lineage has been reconstructed impairing our understanding of its metabolism. Methanotrophs indicative of thaw conditions included methane oxidizers such as members of Methylobacter and Crenothrix genera (both from the Methylomonadaceae family) as well as Candidatus Methanoperedens (Table S6C), a nitrate or iron‐reducing anaerobic methanotroph (Shen et al. 2023).

In conclusion, our results showed that, despite the presence of vertical gradients within stratified waterbodies, maximum depth was the parameter that mostly structured prokaryotic communities. This likely reflects complex combinations of physico‐chemical conditions associated with water column depth. We also observed that waterbodies combining no stratification and permafrost thaw impacts harboured distinct prokaryotic, methanogenic and methanotrophic communities. These findings open several avenues for future research aimed at deepening our understanding of CH₄ cycling and microbial dynamics in permafrost‐affected aquatic systems, especially in the context of global warming. Building on these results, future investigations should address key gaps in our understanding of CH₄ related processes in thaw‐affected aquatic ecosystems, specifically shallow and stratified ones. Indeed, the latter contribute greatly to methane emissions on a global scale and will likely gain importance with global warming because they will increase in numbers due to permafrost thaw. A more integrative approach, including additional compartments (e.g., particulate aggregates, sediment, benthic interface samples), may help to better understand the methane cycle by identifying precisely the promoters of methane emissions both at the bottom and the surface of these waterbodies. Although we coupled CH₄ isotopic analyses with 16S rRNA profiling, 16S data provide limited metabolic resolution. The use of metagenomics would help distinguish metabolic pathways, appreciating symbioses involved in CH₄ production, identifying alternative methanogenic pathways and elucidating the possible presence of diverse lineages with different oxygen metabolism for example. Furthermore, transcriptomic approaches would provide crucial insights into the actual microbial activity, allowing us to move beyond community composition to understand which metabolic pathways are actively expressed and directly link microbial functions to measured methane fluxes. This would be particularly valuable for quantifying the relative contributions of different methanogenic and methanotrophic populations to overall CH4 cycling. Lastly, time‐series approaches, which are difficult to achieve in such environments, would help capture the temporal dynamics of prokaryotic communities and integrate seasonal methane fluxes.

Author Contributions

Conceptualization: U.C., M.B., L.C., L.J. Methodology: A.Sz. Investigation: A.Sz. Data curation: A.Sz., L.G., F.B., S.O. Writing – original draft: A.Sz. Writing – review and editing: A.Sz., U.C., M.B., L.C., L.G., A.Se., F.B., S.O., L.J. Supervision: U.C., M.B., L.C., L.J. Project administration: A.Se. Funding acquisition: U.C., A.Se., L.J. All authors have read and agreed to the published version of the manuscript.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1: emi70210‐sup‐0001‐Supinfo.docx.

Szylit, A. , Christaki U., Barret M., et al. 2025. “Spatial Heterogeneity in Methane Biogeochemistry and Prokaryotic Community Structure in Sub‐Arctic Waterbodies in Northern Canada.” Environmental Microbiology 27, no. 12: e70210. 10.1111/1462-2920.70210.

Data Availability Statement

Sequence data have been deposited in the European Nucleotide Archive under project accession number PRJEB90091. The rest of the data that support the findings of this study are available on request from the corresponding author.