Sediment core samples were obtained in Goldstream Valley, Alaska during prior fieldwork at Big Trail Lake site 1 (referred to BTL in this paper) in March 2023 at GPS coordinates 64.91890 N, 147.81996 W as we previously reported in Smallwood et al. (2025). Sampling equipment was sterilized using 10% bleach followed by 70% ethanol or autoclaved prior to sub-sampling. In March 2023, we conducted a CH₄ emission survey using a handheld optical spectrometer (Quanta3) around the littoral boundary of Big Trail Lake (BTL) and a small feeder pond to the north, concurrent with core sampling (Figure 1a).

Samples from the BTL sediment core were shipped to the Institute of Arctic and Alpine Research (INSTAAR), University of Colorado. Samples were stored in an insulated incubation chamber built at INSTAAR kept at 4 °C to ensure biological fidelity with naturally occurring temperature conditions at the time of sample acquisition. Soil aliquots were separated from the BTL core at 100 cm depth below ground surface. The aliquots were used for incubations targeting temperature-dependent production rates of CH₄ and CO₂ specific to the BTL site. Four ~5 g aliquots (wet mass) were separated from homogenized soil via a surface-sterilized chisel and trowel. Aliquots were placed in a 60 mL jar with 5 mL of Milli-Q water (MilliporeSigma, Burlington, Massachusetts) and homogenized. Jars were then sealed and sparged with nitrogen gas to create an anoxic environment. Samples were then incubated in the dark at temperatures of −30 °C, −4 °C, 4 °C, and 10 °C. Samples were measured at days 0, 7, 15, 22, and 60. It is expected that little microbial activity occurs at −30 °C, increasing starting at −4 °C, and further at 4 °C and 10 °C. At BTL, large step increases in biodiversity of microbial populations have been observed (Smallwood et al., 2025). At all timepoints besides day 60, samples were analyzed for CH₄ and CO₂ concentration vis gas-chromatography flame ionization detection/thermal conductivity detection (GC-FID/TCD). Using a gas-tight syringe, 1 mL samples were injected into a GC-FID/TCD (SRI 8610C). Syringes were filled with 1 mL of nitrogen gas and injected prior to pulling CH₄ to ensure bottles maintained constant pressure. The GC-FID/TCD was calibrated against a 1% CH₄/CO₂ certified reference gas mixture (Scott Specialty Gases, Cat. No. 22561; now part of Air Liquide, Radnor, PA, USA) injected in variable amounts. Error for measurements was determined via the nearest calibration point’s first standard deviation.

To understand carbon isotopic mass balance, soil and headspace carbon-isotopic composition was determined. Soil organic-matter carbon-isotope composition was measured via combustion isotope-ratio mass spectrometry (c-IRMS). Samples were introduced for combustion to a FlashSmart Elemental Analyzer coupled to DELTA Q IRMS (Thermo Fisher Scientific, Waltham, Massachusetts). Carbon-isotope composition was calibrated to Vienna Pee Dee Belemnite (VPDB) via a suite of calibrated secondary references. At day 60, headspace gas was pulled from incubation headspace via a gas-tight syringe, diluted with nitrogen gas in another syringe, and slowly injected into a cavity-ring down spectrometer G2202-i Isotopic Analyzer (Picarro Inc., Santa Cruz, California). Samples were again calibrated to VPDB via a suite of calibrated secondary gas references.

Core subsamples were extracted from the BTL core at depths of 50 cm, 200 cm, and 400 cm. At each depth, three replicate samples were collected. Each core subsample was sectioned using a sterile 5 mL syringe with syringe tip excised, then the syringe plunger was used to push the core into individual 15 mL conical tubes for each replicate. To prepare incubation vials, 2 g of soil were weighed and transferred from the conical tube to a labeled 20 mL screw cap amber vials containing a thin-film holder (Gerstel GmbH, Mulheim, Germany). The pH of each triplicate sample was collected by combining 0.5 g of soil in 1 mL of DI water and measured with the Mettler Toledo S220 SevenCompact™ Benchtop pH/ISE Meter (Mettler Toledo, Columbus, Ohio). Values between triplicates were averaged for the pH of each depth. Both the 15 mL conical and incubation vials were stored at −20 °C until the incubation experiment.

Soil incubation vials contained 20 mm x 4.65 mm Thin-Film-Solid-Phase-Micro-Extraction (TF-SPME) coated with 90 μm mixture of polydimethylsiloxane (PDMS) and a copolymer of divinylbenzene and vinyl pyrrolidinone (HLB) (Gerstel GmbH, Mulheim, Germany). Before sampling, all TF-SPME’s were conditioned using a TurboMatrix 220 tube conditioner (Perkin-Elmer, Shelton, Connecticut) under inert gas flow at 250 °C for 60 min. The sample vials were incubated for a total of 23 days at varying temperatures with intermittent volatile organic compound (VOC) and microbial sampling. The samples were held at 5 °C for the first 7 days, 12 °C for next 8 days, and −4 °C for the final 8 days. These incubation temperatures approximate a range of thawed and re-frozen conditions expected in permafrost soils across spring and summer seasons, capturing microbial activity across a realistic thermal gradient. Incubation tubes were kept in the Lab Armor® 6 L Bead Bath (Lab Armor, Plano, TX) for the first two temperature intervals, and in a Danby Chest Freezer (Danby, Ontario, CA) for the last. Microbial sampling was performed the day following a change in temperature, and VOCs were sampled every 2–5 days.

VOC were passively collected on TF-SPMEs in anaerobic bioreactors (Figure 1b). TF-SPME harvest and replacement was performed in a BACTRON600 Anaerobic Chamber (Sheldon Manufacturing, Cornelius, Oregon) containing a VOC scrubber fan to prevent exposure to oxygen and contaminating VOCs. To sample for VOC’s, TF-SPME of the incubation vial were collected and replaced at pre-determined time points. To replace the TF-SPME, the sampled TF-SPME was removed with sterile tweezers and placed into an empty stainless steel thermal desorption unit (TDU) tube (Camsco, Houston, Texas), then immediately sealed with brass compression caps fitted with PTFE ferrules (Camsco, Houston, Texas) to prevent contamination. The TDU tubes containing sampled TF-SPME were stored at −20 °C until GCxGC-TOFMS analysis.

VOC collection and analysis were performed on a methanogen species to find overlap between the pure culture methanogen and the soil samples VOC signatures. A thin film holder (Gerstel GmbH, Mulheim, Germany) was stapled to butyl rubber stopper and all supplies, besides the TF-SPME, were autoclaved before transferring to the anaerobic chamber. A fresh stock culture was created from an existing culture of Methanosarcina acetivorans strain C2A in late exponential phase by taking 0.5 mL existing culture and measuring optical density at 600 nm on the Thermo Scientific™ Invitrogen™ Nanodrop™ One Spectrophotometer with WiFi and Qubit™ 4 Fluorometer (Thermo Fisher Scientific, Waltham, Massachusetts). The M. acetivorans culture was then diluted to an OD of 0.2 in a high salt medium with 50% methanol added at 0.5% of the total culture volume. Using a serological pipette, 5 mL M. acetivorans stock culture was dispensed into 12 hungate tubes (Chemglass Life Sciences, Vineland, NJ); 6 of which were autoclaved. 5 mL of a defined high-salt (HS) media (Benedict et al., 2012) with 0.5% methanol was dispensed into an empty tube (media blank). Three culture tubes of each treatment (non-autoclaved and autoclaved; 6 total) and media blank were fitted with a TF-SPME for VOC capture and subsequently left undisturbed for the duration of the experiment. The remaining cultures were used to monitor growth, so optical density was measured daily by extracting 300 μL M. acetivorans culture with sterile 1 mL syringe and 23G needle, then combining with 300 μL HS medium in semi-microcuvette. All 13 samples were incubated in a biobag held in 37 °C anaerobic chamber incubator. After 12 days of incubation, each TF-SPME was removed with sterile tweezers and placed into TDU tubes that were then stored in 4 °C until GCxGC-TOFMS analysis.

Following a change in incubation temperature, soil was sampled to determine the microbial community via metagenesis. In the anaerobic chamber, a food-grade plastic straw (Up&Up, Brooklyn, New York) was inserted into the sample to extract ~0.5 g. The straw containing soil was cut and inserted into a sterile 2 mL microcentrifuge tube. The microcentrifuge tubes were stored at −20 °C until sent for sequencing.

For shotgun metagenomics, DNA was extracted with ZymoBIOMICS®-96 MagBead DNA Kits (Zymo Research, Irvine, CA). Extracted genomic DNA was subjected to shotgun metagenomic sequencing. The Illumina DNA Prep Kit (Illumina, San Diego, CA) was used to prep sequencing libraries with ~100 ng DNA. All libraries were quantified with TapeStation® (Agilent Technologies, Santa Clara, CA) and then pooled to equal abundance. The final pool was quantified using qPCR. All metagenomic libraries were sequenced on the Illumina NovaSeq® X platform (2 × 150 bp paired-end configuration). All samples were pooled and sequenced in a single run, eliminating concerns about platform-specific batch effects. No batch correction was required, as library preparation and sequencing were performed under consistent conditions.

To remove low quality features raw sequence reads were trimmed with Trimmomatic-0.33 (Sihi et al., 2019); quality trimming by sliding window with 6 bp window size and a quality cutoff of 20 and reads with size lower than 70 bp were removed. Host-derived reads were removed with Kraken2 and sdust was used to detect and remove low-diversity reads (Wood et al., 2019). The DIAMOND sequencer aligner genetically identified antimicrobial resistance and virulence factor against NCBI reference databases (Buchfink et al., 2014). Bacteria and archaea were identified using the GTDB species representative database (RS207). Sourmash (Brown and Irber, 2016) profiled microbial composition, and GenBank databases, also provided by Sourmash, were used to identify virus, protozoa, and fungi. Using BWA-MEM (Li et al., 2024), reads were mapped back to the genomes identified by Sourmash, and microbial abundance was determined based on the count of map reads. The resulting taxonomy and abundance information were further analyzed: (1) alpha- and beta-diversity analyses (Supplementary Figure 2); (2) microbial composition bar plots using QIIME (Morris et al., 2013); (3) abundance heatmaps with hierarchical clustering (based on Bray–Curtis dissimilarity); and (4) biomarker discovery with LEfSe (Monod, 1949) with default settings (p > 0.05 and LDA effect size >2). Functional profiling was also performed on assembled reads using EggNOG Mapper (v2.1.12) on genes called from the assembled reads by Prodigal (v 2.6.2) to identify the presence of KOs (KEGG Orthology groups) (Kanehisa et al., 2023; Cantalapiedra et al., 2021).

To evaluate patterns in microbial community composition across environmental gradients, we performed Principal Component Analysis (PCA) on standardized genus-level relative abundance data. Abundances were first log-transformed (if applicable) and then scaled using z-score normalization. PCA was conducted using the scikit-bio package in Python (v0.6.3), and the first two principal components were visualized to assess sample clustering by depth and incubation temperature (Rideout et al., 2023). Sample metadata (depth and temperature) was extracted from column names to facilitate grouping and interpretation.

The VOC samples were thermally desorbed using a Gerstel Thermal Desorption Unit 3.5 + (Gerstel GmbH, Mülheim, Germany) integrated into a LECO Pegasus BT4D comprehensive two-dimensional gas chromatograph with time-of-flight mass spectrometer (GCxGC-TOFMS) system (LECO, St. Joseph, MI) and ramped at 60 °C/min from 35 °C to 250 °C and held for 5 min (Smallwood et al., 2025). Desorbed samples were refocused on a Gerstel CIS 4 Cryogenic Inlet, held at −50 °C during TDU desorption and ramped at 12 °C/s to 300 °C to inject the desorbed sample as a single bolus into the GCxGC-TOFMS equipped with a 15 meter, 0.25 mm ID DB-WAX primary column and 2 meter, 0.25 mm ID DB-1 secondary column (Both Agilent, California, United States). The GCxGC-TOFMS system had an initial temperature of 35 °C, ramping at 10 °C/min to a maximum of 230 °C with a 5-min hold. The secondary column was ramped at the same rate, but with a + 5 °C offset from the primary column. The system uses a LECO quad jet thermal modulator design operated with liquid nitrogen cooled nitrogen as the cold jet and heated nitrogen as the hot jet. The modulation period was 4 s, using liquid nitrogen-cooled nitrogen for the cold jet and heated nitrogen for the hot jet. Mass spectra were collected at a rate of 200 spectra/s from 20 mu to 550mu at -70 eV. Both the transfer line and ion source temperatures were set to 250 °C, with a detector voltage offset of -30 V.

Top-down quantitative analysis of metabolites in the hit lists was performed using ChromaTOF TILE software (v1.2.6) (LECO Corp. Michigan, United States). Each individual VOC feature identified from the TILES analysis was compared to the NIH PubChem database (https://cactus.nci.nih.gov/chemical/structureaccessedJune2024) to check for alternative chemical synonyms. Using a custom script, the TILE feature table for BTL, containing unidentified and identified VOCs, was further processed by ClassyFire, an automated tool for taxonomic classification of chemicals (Djoumbou Feunang et al., 2016). Using the chemical classification appended TILE feature table, each taxonomic classification was binned to the category [i.e., 200 cm (5 °C)] with the highest average peak area for that individual feature. For handling of the pure culture TILE feature table, the average instrument blank and media blank signal were subtracted from the averaged peak areas of both the autoclaved and non-autoclaved cultures to isolate the pure culture signal. Once this analysis was performed on the autoclaved and non-autoclaved cultures, the signals from the autoclaved sample were subtracted from the living sample. In the analysis, any rows containing zero were removed. A heat map was created using all remaining rows for non-autoclaved (37 °C), autoclaved (37 °C), and dormant (2 °C) samples using the mean retention time for dimension 1 (RT1) and mean retention time for dimension 2 (RT2) as row labels.

To simulate subsurface CH₄ dynamics during permafrost thaw, we used the PFLOTRAN reactive transport simulator with enhancements to the HYDRATE mode originally developed for marine systems (Hammond et al., 2014). CH₄ was treated as a trace gas in a sequentially coupled flow and reactive transport model solving conservation equations for energy, water, and air components. Mass and energy balances accounted for phase interactions, diffusion, Darcy flow, heat transfer, and phase change. Microbial processes—including aerobic respiration, anaerobic methanogenesis, and CH₄ oxidation—were modeled using temperature-dependent Monod kinetics (Monod, 1949). These simulations were further refined using the Dual Arrhenius Michaelis–Menten Greenhouse Gas (DAMM-GHG) framework to capture temperature-sensitive microbial competition and inhibition effects, following the formulation of Sihi et al. (2019). While the underlying reaction kinetics are consistent with Sihi et al. (2019) model parameters were adapted from Malinverno (2010), which describes CH₄ generation and transport in marine sediments saturated with microbially derived gas. The model domain was one-dimensional (1 m × 1 m × 5 m) with 1 cm vertical resolution. Simulations began with a 130-day spin-up to equilibrate temperature, water content, and CH₄ concentrations to boundary conditions derived from the NOAA-CIRES 20th Century Reanalysis (V2) dataset, which spans 1891 to 2011. The temperature boundary condition used for this simulation corresponds to the year 2011. Following spin-up, we simulated CH₄ transport and fluxes over a 60-day summer period under multiple surface temperature scenarios (0 °C to 4 °C) to evaluate CH₄ production and its delayed emergence at the ground surface. Model results were benchmarked against CH₄ fluxes observed in laboratory incubations of BTL 100 cm depth soils, allowing evaluation of subsurface production and transport lags under warming conditions.

Elevated CH₄ emissions were observed during a March 2023 survey of the littoral boundary of Big Trail Lake (BTL) and an adjacent feeder pond to the north (Figure 1a). CH₄ mole fractions reached up to 12 ppm at 10 cm above ground level (background ~2 ppm), with spatially heterogeneous fluxes evident across the site. Regions highlighted in blue and dark yellow in Figure 1a correspond to high-emission zones, with vertical concentration differences of up to 3,500 ppb between 10 cm and 1.5 m aboveground—indicative of strong surface emissions. The highest CH₄ levels were recorded near the southeast margin of the lake, coinciding with a sediment core collected for permafrost thaw analysis (Figure 1a, red dot). These results confirmed the presence of localized CH₄ hotspots in both the northern and southern littoral zones, reinforcing prior observations of sustained CH₄ enrichment during both cold and growing seasons at BTL.

To assess microbial community dynamics with depth and temperature, we incubated BTL soil core subsamples in anaerobic micro-bioreactors (Figure 1b). Incubation temperatures progressed from frozen conditions to 5 °C, 12 °C, and −4 °C in one-week intervals. Metagenomic sequencing was conducted at each interval, alongside GC × GC–MS to analyze volatile organic compound (VOC) profiles. All replicates exhibited similar alpha and beta diversity initially, with the exception of replicate 3 from the 400 cm sample, aligning with its lower alpha diversity (Supplementary Figure 2). Microbial composition shifted following temperature increases from −20 °C to 5 °C, stabilizing at 12 °C, with some community restructuring observed at −4 °C (Figure 2). At 50 cm, Arthrobacter dominated, followed by Pseudomonas, Massilia, and Caulobacter. Deeper samples (>50 cm) showed expected transitions toward facultative and obligate anaerobes. For instance, Bradyrhizobium (facultative anaerobe) was more prevalent at 200 cm, while Arthrobacter (obligate aerobe) declined. The 400 cm samples showed the highest microbial diversity, with notable enrichment of Candidatus Nitrotoga at −4 °C—indicative of active nitrogen cycling under cooler conditions. Methanosarcina, a key methanogen, was found across depths but showed differential responses to temperature. At 50 cm, its abundance dropped from 2.8 to 0.4% after 5 °C incubation. At 200 cm, it increased slightly at 5 °C, then declined sharply at 12 °C, but rose again to 2% at −4 °C. At 400 cm, Methanosarcina abundance was initially high (~2%) but declined rapidly with warming and remained low even at −4 °C. Deeper soils harbored syntrophic communities supportive of methanogenesis, including Candidatus Nitrotoga, Bradyrhizobium, Phenylobacterium, and Limosilactobacillus. Stacked bar plots of gene abundance confirmed that Methanosarcina was the most abundant CH₄-cycling taxon, particularly at 200 cm (Figure 3). Methanotrophs also increased in relative abundance, especially in 200 and 400 cm soils following incubation at 5 °C, with methanotrophic diversity rising over time.

Principal Component Analysis revealed clear separation of microbial communities based on both sediment depth and incubation temperature (Supplementary Figure 3). The first two principal components accounted for a substantial portion of the variance in the dataset, with samples clustering distinctly by depth (50 cm, 200 cm, and 400 cm) and further stratified by temperature treatments (−20 °C to 12 °C). Shallow samples incubated at higher temperatures tended to group separately from deeper or colder samples, indicating that both environmental variables strongly influence community composition.

Anaerobic incubations of sediment samples from 50 cm, 200 cm, and 400 cm depths revealed distinct shifts in methanogenic and methanotrophic taxa across a stepped temperature regime (Figure 3). Samples were incubated sequentially at 0 °C, 5 °C, 12 °C, and −4 °C, each for one week. At 50 cm, Methanobacterium and Methanosarcina were the dominant methanogens at 0 °C and 5 °C, indicating adaptation to cold, near-surface conditions (Figure 3). A notable increase in Methanosaeta and Methanoregula was observed at 12 °C, suggesting these taxa are more responsive to moderate warming. Upon cooling to −4 °C, the overall abundance of methanogens declined, although Methanobacterium remained detectable, indicating some cold tolerance. At 200 cm, methanogenic diversity increased with temperature. Methanoculleus and Methanoregula were particularly enriched at 12 °C but were less abundant at 0 °C and nearly absent at −4 °C. These patterns suggest a greater metabolic activation of methanogens at intermediate depths under warming. At 400 cm, the methanogenic community showed the greatest richness at 12 °C, with Methanoculleus, Methanobacterium, and Methanosarcina as dominant taxa. These groups persisted, though at reduced abundance, during the return to −4 °C, suggesting a more metabolically flexible or resilient methanogen community at depth.

Methanotrophs were most abundant at 50 cm and primarily active at lower temperatures (Figure 3). Type I methanotrophs, including Methylobacter and Methylomonas, were enriched at 0 °C and 5 °C but declined sharply by 12 °C. At −4 °C, methanotrophic taxa were nearly undetectable. Type II methanotrophs (Methylocystis) were observed sporadically at 200 cm, with peak presence at 5 °C; however, they did not persist at other temperatures. At 400 cm, methanotrophs were largely absent across all temperature treatments, indicating that CH₄ oxidation activity is limited in deeper anaerobic sediments.

Methanogenic taxa increased in diversity and abundance with depth and temperature, peaking at 12 °C, particularly in sediments from 200 cm and 400 cm. In contrast, methanotrophs were confined to shallow sediments and favored lower temperatures, with peak abundance at 0 °C and 5 °C in the 50 cm horizon. Cooling to −4 °C led to a general decline in both methanogen and methanotroph abundance, though several methanogenic taxa persisted, particularly in deeper samples. These results indicate a stratified and temperature-responsive community, with deeper sediments favoring methanogenic activity under warming and shallow sediments supporting CH₄ oxidation under cold, anaerobic conditions.

Methane and CO₂ production were tracked over 60-day incubations across temperature treatments (Figure 4; Supplementary Figure 1). Methane accumulation (log-scale) was highest at 10 °C, whereas emissions at −4 °C plateaued between days 22 and 60 (Supplementary Figure 1). Methane fluxes, calculated from CH₄ mass, showed a strong positive correlation with temperature, while CO₂ emissions were less variable across treatments. A transient CO₂ spike during the initial 7 days was attributed to residual oxygen, which diminished by day 15, after which CO₂ production steadily increased—except in −4 °C and −30 °C treatments.

The temperature sensitivity of CH₄ production was quantified using Q₁₀ coefficients. Q₁₀ values were ~5 at days 7 and 15 but rose to 14 by day 22, suggesting that microbial responses to temperature intensify over time (Figure 4b). The average Q₁₀ across all time points was 7.9, considerably higher than values used in many boreal wetland models. Carbon isotope measurements supported these findings (Figure 4a). Soil organic matter had δ¹³C values consistent with C₃ vegetation (−25.60 ± 0.70‰). Methane δ¹³C values after 60 days ranged from −64.6‰ to −38.9‰ from −4 °C to 10 °C, showing a highly linear response to temperature (R² = 0.997; Figure 4a). Methane was undetectable in −30 °C treatments at day 60 due to low production.

Metagenomic profiles revealed temperature-dependent changes in the relative abundance of key microbial groups with functional roles in fermentation, metal reduction, methanogenesis, and CH₄ oxidation, suggesting possible shifts in metabolic potential across thermal regimes (Supplementary Figure 4). Community composition at the endpoint of the 60-day mesoscale incubation under four temperature treatments (−30 °C, −4 °C, 4 °C, and 10 °C) reflected these trends, with specific groups showing distinct responses to thermal inputs. Anaerobic fermenters such as Clostridium were highly abundant under freezing conditions (2.8% at −30 °C) but declined significantly with warming. Other well-characterized fermenters such as Bacteroides were consistently present across treatments and may play a central role in sustaining anaerobic carbon turnover. In contrast, less abundant taxa such as Citrifermentans and Paludibacter exhibited increases at 4 °C (0.98 and 0.23%, respectively). Although these groups were low in relative abundance, they are functionally specialized anaerobes known to contribute to fermentative degradation of organic matter (Thapa et al., 2025; Qiu et al., 2014). Their increased representation at moderate warming suggests a transition in fermentative community composition and highlights the potential ecological relevance of functionally responsive, yet numerically minor taxa. Metal-reducing bacteria (e.g., Geobacter, Geomonas, and Geotalea) were also enriched at 4 °C, with Geomonas reaching 2.23% and Geobacter peaking at 0.80% (ANOVA, p < 0.01), indicating stimulation of dissimilatory metal reduction pathways that could support downstream syntrophic methanogenesis.

Methanogenic archaea also showed temperature-dependent shifts. Methanosarcina abundance tripled from 0.44% at −30 °C to 1.27% at 10 °C (ANOVA, p < 0.05), while more specialized taxa like Methanoregula and Methanothrix declined. In parallel, methanotroph-like populations such as Rhodoferax and Polaromonas also responded strongly to warming. Rhodoferax peaked at 6.45% at 4 °C, while Polaromonas, typically considered psychrotolerant, rose steadily with temperature from 0.15 to 0.64%, further contributing to potential CH₄ oxidation capacity in warmer conditions. Importantly, two key CH₄ oxidation genes—xoxF and mmoX—were detected in the metagenomic data with clear temperature-associated trends (R01146 and R01142, respectively in Supplementary Table 2). Both genes were enriched at 4 °C and 10 °C, indicating increased genetic potential for CH₄ and methanol oxidation under warming conditions.

Contigs annotated with these CH₄ oxidation genes were taxonomically linked to non-canonical but increasingly abundant microbial genera, including Amycolatopsis, Geobacter, Citrifermentans, and Geomonas. These genera not only increased in abundance with temperature but also carried functional markers (xoxF, mmoX) associated with CH₄ oxidation, reinforcing the notion that CH₄ oxidation potential increased with temperature and involved a wider phylogenetic range than canonical aerobic methanotrophs.

We quantified the relative abundance of functional marker genes associated with anaerobic methane-oxidizing archaea (ANMEs), including multiheme c-type cytochromes (e.g., mtrA–H, omcX, omcI), hydrogenase complexes (echA–F), formate dehydrogenases (fdhA–F), and Rnf complex genes (rnfA–H) (Supplementary Table 3) (Ouboter et al., 2024). Across the four temperature conditions (−30 °C, −4 °C, 4 °C, 10 °C), the majority of these genes exhibited a distinct temperature-dependent pattern. Gene counts for mtrC, omcI, echA, fdhA, fdhD, and rnfC were consistently highest at 4 °C, with several showing multi-fold increases relative to colder or warmer treatments. For instance, fdhA increased from 27 hits at −30 °C to 114 at 4 °C before declining to 60 at 10 °C. Similarly, mtrC rose from 8 at −30 °C to 31 at 4 °C. In total, more than 30 ANME-associated genes peaked in abundance at 4 °C, while fewer genes showed maximal counts at −30 °C or 10 °C.

We detected 2,846 unique VOCs across all incubated samples (Figure 5) during the 1-week temperature steps. Despite all samples producing VOCs, their expression patterns differed significantly by temperature and depth. Samples incubated at 12 °C exhibited the greatest number of highly expressed VOCs across all depths. Unexpectedly, VOC abundances at −4 °C exceeded that at 5 °C in some cases (e.g., 50 and 400 cm), suggesting non-linear temperature dependencies. Hierarchical clustering of VOC profiles revealed consistent temperature-driven grouping. At 200 cm, 2,418 unique VOCs were identified across temperature treatments. A compositional shift in compound classes—particularly between 5 °C and −4 °C—was evident (Figure 6). Organic salts dominated at 12 °C, with organometallics, organic oxygen compounds, lipids, hydrocarbons, and benzenoids also abundant.

The heatmap comparison of VOCs in Figure 5 reveals clear depth- and temperature-dependent patterns in VOC expression across the permafrost soil profiles. At 50 cm depth, the thawing treatments (5 °C and 12 °C) show marked enrichment of specific VOC features, with several compounds showing elevated production and release at 12 °C, indicating strong microbial or chemical activity in the more active surface layer during warming. Upon re-freezing (−4 °C), a subset of these compounds remains elevated, suggesting a legacy effect or persistent volatile production even under cold conditions. At 200 cm, VOC expression at 12 °C also shows moderate enrichment, although the signal is less pronounced than at 50 cm, likely reflecting slower metabolic responses in deeper, less biologically active soil layers.

Notably, some VOCs become more prominent only after re-freezing at this depth, suggesting possible chemical transformation or stress-induced release. The 400 cm layer exhibits relatively low VOC diversity and intensity overall, with limited response to warming. However, several compounds increase specifically under −4 °C conditions, suggesting cryoactive processes or latent metabolic shifts triggered by rapid freeze. It is worth considering whether the low VOC abundance observed at this depth reflects an intrinsically less active microbial community or simply the limited duration of warming applied in these experiments. With sustained warming, deeper communities may have the potential to proliferate and increase VOC production over time, leading to compositional changes. Overall, the data demonstrate that temperature shifts strongly influence VOC dynamics, particularly in shallower soils. These volatile signatures may serve as early indicators of biogeochemical reactivation and microbial community shifts in thawing permafrost, with potential implications for forecasting CH₄ flux and monitoring permafrost-affected ecosystems.

Chemical classification of VOCs revealed distinct patterns across temperature treatments and soil depths, reflecting both thermal sensitivity and microbial source specificity in metabolite production (Figure 6). The distribution of chemical super classes highlights the functional stratification of microbial metabolism and stress responses in thawing permafrost environments.

At 200 cm and 12 °C VOC profiles were enriched in organometallic compounds, benzenoids, organic oxygen compounds, and hydrocarbons. Lipids and lipid-like molecules, along with organoheterocyclic compounds, were also abundant (Figure 6). This chemically diverse profile is consistent with active methanogenic metabolism and secondary metabolite production, suggesting organometallic VOCs as potential biosignatures of archaeal metabolic activity. At 5 °C, the VOC profile at 200 cm shifted toward a less diverse set, dominated by organic salts and organometallic compounds, indicative of reduced but ongoing microbial metabolism. At −4 °C, a further shift occurred, with organic nitrogen compounds and benzenoids becoming prominent. These changes suggest the emergence of stress-induced metabolites, likely associated with cryo-preservation mechanisms or slowed anabolism. VOC diversity was highest in surface soil incubations. At 12 °C, the 50 cm profile was also dominated by organic salts, benzenoids, and organometallic compounds, along with substantial contributions from organic nitrogen and oxygen compounds. These signatures suggest broad microbial activation. At 5 °C and 50 cm depth, lipids and lipid-like molecules were the dominant class, followed by organic oxygen compounds and benzenoids, pointing to membrane turnover and biosynthesis under moderate cold stress. Re-freezing at −4 °C shifted the VOC profile again, with organic nitrogen compounds and organic salts becoming most abundant. Notably, phenylpropanoids and polyketides appeared uniquely under this condition, implying potential roles in microbial stress defense or signaling. In the deeper 400 cm samples, overall VOC diversity was lower. At −4 °C and 400 cm depth, hydrocarbons and organic salts were the dominant classes, possibly linked to cryoactive processes or microbial persistence. At 5 °C and 400 cm, lipids and benzenoids were present, indicating minimal metabolic activity, potentially tied to membrane remodeling or low-level aromatic degradation. VOC presence in instrument and vial blanks was minimal, limited primarily to lipids, organic oxygen compounds, and benzenoids. This low background confirms the biological origin of VOC signals observed in experimental treatments.

Sixteen KEGG-annotated volatile organic compounds (VOCs) were identified as overlapping between BTL soil incubations and pure cultures of M. acetivorans (Supplementary Figure 5; Table 1). Columns represent experimental conditions for BTL soils at 50 cm, 200 cm, and 400 cm incubated at −4 °C, 5 °C, and 12 °C, alongside M. acetivorans cultures grown at 2 °C and 37 °C. Green cells indicate presence, red cells absence. Notable shared compounds across environments include benzaldehyde, acetophenone, o-xylene, and pentanoic acid (Table 1). Overlap was highest at 200 cm and 12 °C, suggesting this thermal regime supports VOC profiles associated with acetoclastic methanogenesis. Unique M. acetivorans compounds (e.g., dimethyl sulfide, squalene, furfural) may serve as archaeal-specific biomarkers. These unique methanogen-associated VOC profiles underscore the potential of volatile signatures to trace archaeal activity and infer methanogenic pathways in permafrost-affected soils.

The overlap between BTL and M. acetivorans VOCs was most pronounced at 200 cm depth and 12 °C—conditions previously shown to favor acetoclastic methanogenesis—where 13 of the 16 compounds were shared with M. acetivorans grown at 37 °C, and 10 were shared with cultures grown at 2 °C. Several VOC classes were consistently detected across both systems. Aromatic compounds such as acetophenone, benzaldehyde, and o-xylene were ubiquitous, appearing in nearly all BTL incubations as well as both M. acetivorans temperature treatments. Long-chain alkanes and aldehydes, including tridecane, decanal, and dodecanal, were also present in both soil and culture samples, potentially reflecting shared processes such as lipid degradation or membrane remodeling. Phenol, detected in surface and 200 cm soil incubations and in the 2 °C M. acetivorans culture, may represent a byproduct of aromatic compound metabolism or microbial stress responses.

VOCs shared between M. acetivorans and BTL soil incubations were not limited to warm conditions. At 200 cm and 5 °C, ten overlapping compounds were identified, including butylated hydroxytoluene, benzaldehyde, and pentanoic acid, indicating that methanogenic VOC signatures persist under moderate cold stress. In contrast, compound overlap diminished substantially at −4 °C, where only six VOCs were shared between the BTL 200 cm incubations and either of the M. acetivorans culture temperatures. The greatest divergence occurred in the deepest BTL samples at 400 cm, where overlap with M. acetivorans VOCs was minimal. At this depth and −4 °C, only six shared compounds were detected, including benzaldehyde, butylated hydroxytoluene, and o-xylene, suggesting limited but persistent microbial metabolic activity under cryogenic conditions.

To evaluate whether observed CH₄ fluxes from warming thermokarst soils could be explained by coupled hydro-thermo-biogeochemical processes, we employed a 1D reactive transport model using a modified version of PFLOTRAN flow and reactive transport simulator. The model simulated CH₄ production, transport, and flux at the soil-atmosphere interface under different temperature boundary conditions over a 60-day summer period, following 130 days of spin-up to equilibrate thermal and hydrologic conditions (Supplementary Figures 6–10; Figure 7).

Our investigation into microbial community dynamics across a vertical profile of anaerobic permafrost soil at a thermokarst hotspot reveals stratified and thermally responsive assemblages with implications for broader carbon cycling processes. At 50 cm, microbial communities were dominated by facultative anaerobes and metabolically flexible taxa, including Pseudomonas, Arthrobacter, and Bradyrhizobium, with relative abundances increasing at elevated temperatures. These taxa likely participate in organic degradation and nitrogen cycling, generating intermediates like acetate and lactate that fuel methanogenesis (Morris et al., 2013).

While methanogenesis is traditionally considered an obligately anaerobic process, recent findings have demonstrated the potential for methane production under oxic conditions in a range of ecosystems (Ernst et al., 2022; Günthel et al., 2019). Although our study focuses on anoxic soils, we acknowledge that oxic methanogenesis may also occur at the lake margin or within microoxic niches, and merits further investigation in future work. The prominence of Pseudomonas, a known degrader of complex carbon compounds, aligns with observations from active layer soils in Stordalen Mire, Sweden, where aerobic heterotrophs have been linked to enhanced carbon turnover during thaw (Hodgkins et al., 2014). Similarly, Arthrobacter—often associated with seasonal thaw layers in interior Alaska—reflects resilience in fluctuating redox and temperature regimes (Haan and Drown, 2021; Liu et al., 2023).

At 200 cm, the microbial community shifted toward increased abundance of Bradyrhizobium and detectable levels of Methanosarcina. At 400 cm—within historically stable, deep permafrost—warming induced a marked community shift, with Ferrigenium and Sideroxydans emerging in higher abundance. The activation of iron-oxidizing bacteria such as Rhodoferax, Geobacter, Gallionella, and Sideroxydans indicates that alternative electron acceptors like iron may play an underappreciated role in subsurface microbial metabolism, particularly under warming conditions (Li et al., 2024; Levar et al., 2017; Patzner et al., 2022; Berns-Herrboldt et al., 2025). These shifts suggest that microbial community structure in anaerobic permafrost is likely shaped by temperature as well as depth-related gradients in redox potential, moisture availability, substrate quality and in situ thermal conditions–all of which co-vary with depth and collectively influence microbial functional potential.

The substantial proportion of unclassified taxa—exceeding 50% relative abundance in some warm treatments at 400 cm—underscores the cryptic diversity of deep permafrost microbiomes. Similar patterns have been reported in metagenomic studies from thaw gradients in Alaska and Siberia, where a significant fraction of sequences remain unassignable to known lineages (Burkert et al., 2019; Woodcroft et al., 2018). Together, these findings emphasize the need for depth-resolved microbial assessments as a foundation for understanding permafrost carbon flux dynamics. The observed thermal responsiveness of microbial communities across depths sets the stage for exploring the functional strategies and ecological niches of key microbial groups driving carbon turnover in these dynamic environments.

The 60-day anaerobic incubation of 100 cm depth thermokarst soils from BTL also revealed a clear temperature dependence in CH₄ production and microbial functional potential. Methane mass and flux increased markedly with incubation temperature, with the highest accumulation observed at 10 °C, while flux plateaued under subzero conditions, particularly at −4 °C (Supplementary Figure 1). CO₂ production, in contrast, was relatively stable across temperatures following an initial spike during the first 7 days—likely due to residual oxygen—which subsided by day 15. After this transition to fully anoxic conditions, CO₂ production increased moderately at 4 °C and 10 °C but remained low at −4 °C and −30 °C.

These gas trends were mirrored in the metagenomic profiles, which showed strong temperature-linked increases in key methanogenesis genes. Specifically, the abundance of methyl-coenzyme M reductase (EC:2.8.4.1), acetate kinase (EC:2.7.2.1), and formylmethanofuran dehydrogenase (EC:1.2.99.5) rose consistently from −30 °C to 10 °C, reflecting enhanced hydrogenotrophic and acetoclastic methanogenesis with ecosystem warming (Supplementary Table 1). Methylotrophic pathways (e.g., methanol co-methyltransferase, EC:2.1.1.90) showed peak gene abundance at 4 °C, suggesting cold-adapted activity of these pathways at intermediate temperatures.

In contrast, aerobic CH₄ oxidation potential was minimal across all temperatures. Methane monooxygenase (EC:1.14.18.3), the canonical marker for aerobic methanotrophs, was detected at extremely low abundance throughout the incubation. Metagenomic data indicate that methanotrophic capacity is minimal under cold, anaerobic conditions, as shown by the consistently low abundance of methane monooxygenase genes across all temperatures. This suggests limited microbial CH₄ oxidation during winter. Combined with physical and process-based modeling, these findings support the hypothesis that CH₄ produced at depth may bypass oxidation and escape more readily through preferential pathways in frozen or partially thawed soils.

The temperature sensitivity of CH₄ production, quantified via Q₁₀ coefficients, further supports the role of microbial amplification with soil warming. Q₁₀ values reached ~14 by Day 22—well above typical ecosystem model values—indicating a sharp increase in CH₄ production per degree of soil warming (Figure 2). However, by Day 60, Q₁₀ values declined, a pattern not consistent with elevated CH₄ oxidation, given the genomic data. This decline may instead reflect substrate depletion, saturation effects, or temperature-induced shifts in methanogenic community composition.

Collectively, these findings suggest that soil warming leads to both elevated methanogenic potential and increased CH₄ flux in deep permafrost soils, while CH₄ oxidation remains largely inactive under anoxic, cold conditions. The resulting imbalance between CH₄ production and oxidation may contribute to greater CH₄ emissions during winter and early thaw periods—especially where structural thawing or preferential gas pathways allow CH₄ to escape before microbial mitigation can occur (Walter Anthony et al., 2024). In addition, thawing and rewetting events can induce mechanical disruption of ice-rich soil matrices, leading to the sudden release of CO₂ and CH₄ that had accumulated over prior time periods. These emissions may reflect past microbial activity rather than active metabolism at the time of thaw, complicating efforts to distinguish immediate biological responses from physical gas release. However, a key limitation of this 60-day incubation experiment is the potential accumulation of CH₄ in sealed microcosms, which may not capture the full range of diffusion, ebullition, or oxidation processes occurring under natural field conditions. The buildup of headspace CH₄ could artificially enhance local CH₄ concentrations, potentially triggering methanotroph activation that would otherwise be spatially restricted or redox-limited in situ. While CH₄ oxidation genes exhibited temperature- and CH₄-dependent increases, this result could partly reflect incubation artifacts, and the potential for transient or localized methanotrophic blooms cannot be excluded. Additional studies—both laboratory-based and field-deployed—are needed to constrain the feedbacks between CH₄ production, transport, and oxidation in geographically and geochemically diverse thermokarst environments.

The heatmap illustrates temperature- and depth-dependent patterns of VOC expression, revealing distinct chemical signatures across the permafrost soil profile (Figure 5). At 50 cm, warming to 12 °C induces a broad upregulation of VOCs, consistent with microbial activation and increased metabolic flux in the biologically active surface layer. Several VOCs remain elevated even after refreezing (−4 °C), suggesting persistence of volatiles or residual metabolic signals from prior thaw-induced activity. In contrast, the 200 cm depth—a more anaerobic and colder zone—exhibits a narrower but more diagnostic suite of VOCs, which become more pronounced under both warming and freezing conditions. These depth- and temperature-specific patterns point to microbial consortia, particularly methanogenic lineages such as Methanosarcinales, as being metabolically responsive to thermal perturbation. Their volatile byproducts likely reflect early stages of anaerobic carbon turnover. Given the low molecular weight and diffusivity of VOCs, their early appearance may precede detectable gas emissions and offer a window into real-time microbial activity. The distinct and consistent volatiles observed at 200 cm support the potential use of VOCs as biosensors of microbial reactivation, although concurrent CH₄ and VOC measurements in field settings will be essential for validation.

Organometallic compounds, benzenoids, and organic oxygen compounds were particularly enriched in high-temperature methanogen incubations (e.g., 200_12 °C), supporting active acetoclastic methanogenesis by Methanosarcina spp. These VOC classes have putative associations with cofactor biosynthesis, central metabolic intermediates, and increased biosynthetic activity during methanogenic growth (Nazaries et al., 2013; Thauer, 2011). In contrast, the elevated abundance of organic salts, organic nitrogen compounds, and phenylpropanoids under frozen or freeze–thaw conditions (e.g., 50 cm at −4 °C, 200 cm at −4 °C) likely reflects microbial turnover and cryo-adaptation processes, including protein denaturation, osmoprotectant accumulation, and nitrogen mineralization. While our monoculture studies inform potential links, broader community overlap and environmental complexity limit specificity. However, these VOC trends (Table 1) are consistent with findings from permafrost soils, where freeze–thaw cycling leads to nitrogenous compound accumulation and the release of stress-induced metabolites (Schuur et al., 2015; Mackelprang et al., 2011; Waldrop et al., 2023).

Benzenoids, lipid-like compounds, and hydrocarbons were consistently detected across depths and temperatures and may represent a core VOC group tied to both microbial maintenance and interspecies signaling. Especially when found in specific ratios or combinations, these metabolites may function as biosignatures of incipient methanogenic activity—prior to detectable CH₄ production. While it is true that CH₄ exhibits minimal sorption to soil components compared to many VOCs, the observed delay in surface CH₄ flux is more likely driven by slower diffusive transport through frozen or saturated soils and by microbial oxidation in the upper soil column. In contrast, VOCs—though more prone to sorption—are often produced rapidly during early microbial metabolic transitions. While we did not directly measure the temporal release of VOCs relative to CH₄ in this study, their detection in incubation headspace and supporting literature suggest they may emerge earlier under certain conditions, making them promising candidates for early microbial activity indicators. This early-warning capacity is particularly significant for CH₄ release forecasting, as it enhances our ability to detect and track hotspot emergence before CH₄ fluxes accelerate. Since VOCs likely diffuse more rapidly than CH₄ under partially thawed conditions and are readily detectable in incubation headspace, they represent a powerful, non-invasive tool for monitoring microbial biokinetics in thaw-sensitive landscapes. In other words, VOCs represent a novel or emerging diagnostic tool compared to traditional gas flux or genomic methods. Future work should aim to integrate these VOC fingerprints with metagenomic and transcriptomic datasets to strengthen causal links to methanogenic pathways. Field validation under natural thaw scenarios will be key for deploying VOC-based biosensors in boreal and Arctic observatories.

While our findings point to microbial origins for many VOCs, particularly those linked to acetoclastic methanogenesis, we acknowledge that VOC profiles in soil are shaped by both biological activity and the physicochemical properties of organic matter. Depth-resolved variation in VOC composition may reflect differences not only in microbial metabolism but also in the quality and quantity of organic carbon substrates. We have not yet fully disentangled the influence of organic matter composition on VOC production, but future work using advanced techniques such as Fourier Transform Ion Cyclotron Resonance Mass Spectrometry (FTICR-MS) or pyrolysis-GC/MS could help resolve the biotic and abiotic sources of key VOCs across thaw gradients.