Introduction

Permafrost is primarily distributed across the Arctic region, underlying ~25% of the land area in the Northern Hemisphere and hosting an estimated carbon stock of 1500–1700 Pg1,2. Permafrost thaw due to climate change leads to the thickening of the active layer, which releases the frozen carbon and increases the availability and mobility of soil organic carbon (SOC)3,4. This ongoing effect accelerates the potential of microbial organic matter decomposition and is thought to increase emissions of greenhouse gases, CH4 and CO22,5, which may result in a positive feedback effect into the climate system1. Depending on ice richness and soil drainage, permafrost thaw can result in water-saturated soils due to the abrupt collapse of ice wedges6,7, or drier soils due to better drainage and evapotranspiration8,9. While drier and aerated soils may host a burst of CO2 via the microbial degradation of ancient carbon1,10, anoxic conditions in wetter soils may hamper this process and contribute to SOC accumulation11. Although well-drained Arctic soils are well known CH4 sinks12, anoxia in water-saturated soils favors anaerobic methanogenesis and the expansion of anoxic conditions after permafrost thaw may enhance the proportion of future CH4 emissions13.

CH4 has ~34 times higher warming potential compared to CO2 on a 100-year timescale14. Biogenic CH4 is primarily produced by methanogenic archaea carrying out anaerobic methanogenesis, despite some bacteria and eukaryotes producing CH4 using different mechanisms15,16,17. Anaerobic methanogenic archaea constitute the major archaeal communities in permafrost soils18,19. Typical methanogens are a diverse archaeal group occurring within 8 validly described orders, Methanobacteriales, Methanococcales, Methanopyrales, Methanocellales, Methanomicrobiales, Methanonatronaechaeales, Methanosarciniales and Methanomassiliicoccales20,21. Some uncultured lineages, e.g., Ca. Bathyarchaeota, Ca. Methanofastidiosa and Ca. Verstraetearchaeota, are speculated to also carry out methanogenesis based on the coding potential of their metagenome-assembled genomes21. The evidence was recently found in Methanosuratincola (formerly affiliated with Ca. Verstraetearchaeota) within Thermoproteota22,23, and even in Korarchaeia24. Methanogens consist of three main metabolic groups: acetoclastic, hydrogenotrophic and methylotrophic methanogens, utilizing acetate, H2/CO2 and formate, and methanol/methylamines/methyl-sulfides/ethanol as their substrates, respectively25. These pathways are the major players in methanogenesis in soils, despite recent findings on two novel methanogenesis pathways namely methoxydotrophy26 and alkylotrophy27.

In permafrost environments, 20%-60% of the microbial-source CH4 can be consumed by methanotrophs before emitting to the atmosphere28,29. Methanotrophs consist of aerobic methane oxidizing bacteria (MOB) within Alphaproteobacteria, Gammaproteobacteria and Verrucomicrobia, as well as anaerobic bacteria Methylomirabilales and anaerobic archaeal methanotroph (ANME) groups within Methanosarcinales25,30. Aerobic MOB of Alpha- and Gamma-proteobacteria are classified as type II and type I methanotrophs, respectively, and they are the major methane oxidizers occurring in most environments30. While these MOB mostly oxidize the microbial-source CH4 in soils and sediments31, some groups are able to utilize atmospheric CH432. These atmospheric MOB (atmMOB) have a high-affinity particulate methane monooxygenase and can utilize CH4 in a low-CH4-concentration environment, e.g., the atmosphere33. They have been detected in many different soils34,35,36,37; however, so far, only one strain, Methylocapsa gorgona MG08, has been isolated which was also detected in Arctic regions32. The atmMOB are gaining increasing attentions as they are responsible for the mitigation of atmospheric CH4, which can help to counteract climate change.

The diversity and abundance of methane-cycling microbiomes are expected to vary across the Arctic due to the heterogeneity in the structure and physicochemical conditions of permafrost soil regions38. In soils, both methanogens and methanotrophs are important in determining CH4 fluxes39,40,41. It is therefore pivotal to characterize these microbiomes across space and time, for a fundamental understanding of the key players in CH4 dynamics in permafrost soils. However, due to the reduced accessibility of the Arctic permafrost region for sampling, the diversity of these microbiomes and the dominant microbes in these soils are poorly characterized, particularly across horizons and on a broader, pan-Arctic scale.

Arctic wetlands and water-logged soils are known hotspots of CH4 emissions42,43. This is due to the wet conditions after permafrost thaw that contribute to the increase of methanogen abundances and the shift in their community compositions44,45. A substantial number of studies have particularly addressed the consumption of atmospheric CH4 in well-drained Arctic soils12,46,47,48,49,50, but surprisingly little attention has been paid to the microbiomes driving this phenomenon. Only a few studies mentioned atmMOB as the main drivers of CH4 consumption in Arctic soils37,51,52. Moreover, studies comparing the abundance and composition of methane-cycling microbiomes across soil moisture gradients following permafrost thaw remain scarce.

This study aims to reveal the pan-Arctic distributions of methanogens and methanotrophs in intact permafrost soils and their future development in response to different water conditions after permafrost thaw. We analyzed the microbiomes of intact permafrost soil samples from eight locations across the pan-Arctic over a course of eight years (Fig. 1a) from different horizons (Fig. 1b). The relative abundance and distribution patterns of phylotypes associated with methanogens and methanotrophs across space and horizons are shown. To understand the impact of permafrost degradation on those methane-cycling microbiomes in times of climate change, we selected three hydrologically different degraded permafrost sites and their corresponding intact site in Fairbanks, Alaska. The relative abundance and community composition of methanogens and methanotrophs were compared among the degraded sites and between the degraded sites and the intact site.

Fig. 1: Sampling sites and horizons in the pan-Arctic.

a Sampling locations including (from west to east) Herschel Island and Beaufort coast (Canada), Disko Island and Zackenberg (Greenland), as well as Tazovskiy Peninsula, AriMas, Logata and Cherskiy (Russia). b Schematic diagram of soil horizons including organic layer, topsoil, subsoil, subsoil cryoturbated organic matter (cryoOM) and permafrost. The base map in (a) is from https://www.grida.no/resources/5234 (Hugo Ahlenius and UNEP/GRID-Arendal, 2016). Dark blue refers to continuous permafrost >90% area coverage; medium blue refers to discontinuous/sporadic 10–90% coverage; light blue refers to isolated patches; white over land refers to no permafrost.

Materials and methods

Studying sites and soil sampling

Over the course of 8 years, soils samples were taken from 8 different intact Arctic sites located in the continuous permafrost domain (Fig. 1a), including Herschel Island (2016) and Beaufort Coast (2016) in Canada, Disko Island (2017) and Zackenberg (2010) in Greenland, and Logata (2011), AriMas (2011), Tazovskiy (2012) and Cherskiy (2010) in Siberia, Russia. For each sampling, soil samples were taken from different horizons, including organic layer, topsoil, subsoil, cryoturbated organic matter (cryoOM) and permafrost (Fig. 1b). The sampling campaigns were conducted in summer (July or August) at the peak of the growing season within different projects. The detailed descriptions of these sites and sampling protocols are available in previous studies19,53,54,55. The number of samples taken at each layer in each site is shown in Supplementary Fig. S2 and Supplementary Fig. S3. In total, 621 samples were collected.

A ninth intact permafrost site is located in the city of Fairbanks, in the Interior Alaska, USA (Fig. 2a). This Intact permafrost site (64° 51′ 56.1′′ N, 147° 51′ 18.9′′ W) is close to Smith Lake with a shallow permafrost table and no major degradation features (Fig. 2b), like other intact sites from the pan-Arctic (Fig. 1a). To study the impact of different water conditions on microbiomes after permafrost thaw, two degraded permafrost sites in the vicinity of the Intact site were selected. The Wet degraded permafrost site (64° 52′ 02.4′′ N, 147° 51′ 17.3′′ W) is located on a low hill at the bottom of the depression with degraded permafrost soils and high-water contents (Fig. 2b). The Dry degraded permafrost site (64° 51′ 33.5′′ N, 147° 51′ 18.0′′ W) is at the shoulder of a north-facing slope with well-aerated and degraded permafrost soils and low water contents (Fig. 2b). A detailed description of these three sites is available in a previous study56. Additionally, a third degraded permafrost site was investigated on a mid-slope position (64° 51’ 41” N, 147° 51’ 33” W) with soil development and moisture conditions being in between those of Dry and Wet sites (Fig. 2b), in the following referred to as Intermediate-Wet site.

Fig. 2: Sampling sites and soil profiles in Alaska.

a Sampling locations in Fairbanks, Alaska. b Soil profiles of the four sites with different water conditions. Int. Intact, Inter-Wet Intermediate-Wet. This figure is modified from a previous study56.

Soil samples in Alaska were taken in late August to early September in 2019 and 2021. In 2019, each three soil pits (150 × 100 cm) were dug with a depth of either ~100 cm for degraded (Dry and Intermediate-Wet) sites or until reaching the permafrost surface for the non-degraded (Intact) site. In 2021, an additional site with a higher moisture content (Wet site) was sampled. Only one pit was dug at each site and the same approach was used for sampling the profile as in 2019. In compensation, we additionally sampled soils from 6 satellite pits at the depths of top- and sub-soil layers (~5 cm or ~50 cm, respectively) to account for the large, small-scale heterogeneity of the soils in this area. The number of samples taken from each different depth is shown in Supplementary Fig. S4.

Soil physicochemical properties

The soil water content and pH were measured for samples from the 8 pan-Arctic sites (Fig. 1a). Water content was measured as water to fresh soil weight ratio by drying the soil at 105 °C until weight stabilized. The pH values were measured using a soil-water-suspension with a fixed 5:2 ratio (water to soil). The water content of Disko samples and the pH of Zackenberg samples were not available due to different managements and measurements of the involving projects.

Soil samples taken in Alaska were analyzed on a variety of soil physicochemical parameters, including soil moisture, pH, soil organic carbon (SOC), total nitrogen (TN), base saturation (BS), dissolved organic carbon (DOC) and dissolved total nitrogen (DTN), micro- and macro-aggregate (MiA and MaA) proportions, and organic carbon and nitrogen in MiA and MaA. Data were adapted from a previous study56.

Microbiome analysis

Soil samples from the 8 pan-Arctic sites (Fig. 1a) were processed with DNA extraction and 16S rRNA gene amplicon sequencing using the primer pair 515F/806R. Illumina sequencing with pair-end 100 bp was performed with Cherskiy and Zackenberg samples collected in 2010, while the other samples were sequenced with a later pair-end 250 bp Illumina platform. The sequencing data of the pan-Arctic samples were generated and published by previous studies19,53,54,55. To make all sites comparable, forward sequences were all trimmed to 100 bp and then used for downstream analysis. The UNOISE algorithm57 was used to denoise and error-correct the sequencing data, and zero-radius Operational Taxonomic Units (ZOTUs) were generated. The taxonomy of each ZOTU was assigned against the SILVA 138.2 database using a Naive Bayesian Classifier algorithm implemented in dada2 pipeline58 in R v3.6.359. Based on 621 analyzed datasets, a subset of ZOTUs associated with methanogens and methanotrophs (Supplementary Table S1) were used for downstream analyses. The core members that existed in all sites except for Tazovskiy (due to absence or low abundances of both methanogens and methanotrophs) were identified. The representative sequences of associated ZOTUs were further used as enquiry for NCBI BLASTN against the 16S rRNA gene sequence (bacteria and archaea) database to verify the taxonomy identification (Supplementary Data 1).

Alaska soil samples (Fig. 2a) taken in 2019 were transported on ice and kept at −20 °C before the DNA extraction. DNA was co-extracted with the RNA using the RNeasy PowerSoil total RNA kit and the RNeasy PowerSoil DNA elution kit (Qiagen, Hilden, Germany) from ~2 g of homogenized soil samples. In 2021, ~3 g of each soil was mixed with 2 volumes of LifeGuard Soil Preservation Solution (Qiagen, Hilden, Germany) directly in the field. All treated soils were kept at 4 °C before the DNA extraction. DNA was also co-extracted with the RNA using the same kits. The LifeGuard solution was removed by centrifugation and ~2 g of the treated soil was used for the extraction. The extracted DNA samples were used for 16S rRNA amplicon sequencing using Illumina platform (pair-end 250 bp) using the primer pair 515F/806R. The sequencing data of the Alaska samples was generated for this study and has not been published elsewhere. These data were analyzed separately and the phylotypes of methanogens and methanotrophs were inferred as amplicon-sequencing-variants (ASVs). However, ASVs and ZOTUs both provide species-level resolved phylotypes57,58 and are thus comparable with each other.

The dada2 pipeline was used to process the Alaska data. Sequences failing to meet the filter criteria (maxEE = 2, truncQ = 2, maxN = 0) were removed. Those filtered sequences were de-replicated, the ASVs were deduced, and the paired-end sequences were merged. Afterwards the chimeric sequences were removed. The sequence of each ASV was assigned to taxonomy using the SILVA 138.2 database as described above. ASVs associated with methanogens and methanotrophs were identified with the same criteria (Supplementary Table S1). Those ASV sequences were also used as query to verify the taxonomy identification on NCBI as described above (Supplementary Data 1). The Beijerinckiaceae ASVs were additionally used as query against 16S rRNA gene of M. gorgona MG08 with a local BLASTN to verify their relationships (Supplementary Data 1). To link the methanotroph ASVs in Alaska to ZOTUs in the pan-Arctic, a local BLASTN was run using ASVs as the enquiry and ZOTUs as the database. Only ASVs showing a 100% identity to a ZOTU are considered as the potential same phylotype.

Statistics

The downstream analyses were done with R v3.6.359. The normality of abundance data distributions was checked before linear regressions or t tests using the Shapiro-Wilk test. If not normally distributed, the data were log10 transformed. A minimum value was added to avoid zeros when necessary, before the transformation. The coefficients and significance level of the correlations were determined by Pearson’s correlation analysis using vegan v2.5.6 package60. The significant correlations were further confirmed by non-parametric Spearman’s correlation analysis using vegan. Pairwise t tests were performed to compare the means of methanogen and methanotroph relative abundances between each two sites or horizons. Spearman’s correlation was also used to check the correlations between methanogen and methanotroph abundances and environmental parameters for Alaska samples. For all multiple comparisons, P values were adjusted by the false discovery rate (FDR) method. All of the plots were generated using ggplot2 v3.3.3 package61.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Results and discussion

Functional guild abundances across the pan-Arctic and soil horizons

Methanogen relative abundance (in total microbiome) varied strongly between locations, from the highest found in Herschel Island (~0.5%) and Zackenberg (~0.75%) to almost undetected in Disko Island and Tazovskiy (Fig. 3a). The methanogen distribution along the vertical profiles was often location-specific, sometimes increasing with depth, and sometimes having the highest abundance in the organic layers (Fig.3c and Supplementary Fig. S2). In mineral top- and subsoils, methanogens were lowly abundant likely due to the lack of substrates. Nevertheless, their relative abundance and soil water content were significantly positively correlated in organic layer, cryoOM and permafrost (Supplementary Fig. S1a) in which the methanogens were also abundant (Fig. 3c). This indicates that the hydrologic status is one key driving factor for methanogen abundance by limiting oxygen diffusion into the soil profile in layers where they habitat. pH only showed a significant and negative correlation with methanogen relative abundances in the permafrost layer (Supplementary Fig. S1c).

Fig. 3: Relative abundances of methanogens and methanotrophs in the pan-Arctic.

a, b Relative abundances of methanogens (a) and methanotrophs (b) in different sites. c, d Relative abundances of methanogens (c) and methanotrophs (d) in different horizons. Abundances are shown as mean + standard error. The bars with no common letter are significantly different (P < 0.05) characterized by pairwise t tests with P values adjusted by the false discovery rate method.

Methanotroph relative abundance (in total microbiome) followed the methanogen pattern, with the highest (~0.32%) and lowest (almost undetected) abundances found in Zackenberg and Tazovskiy, respectively (Fig. 3b). Along the soil profiles, methanotrophs were less abundant in the organic layer and topsoil compared to the other deeper layers (Fig. 3d), but this pattern again varied between sites (Supplementary Fig. S3). Compared to methanogens, water content only showed a weak but still significant correlation with methanotroph abundance in the organic layer (Supplementary Fig. S1b). Like methanogens, methanotroph abundance was negatively and significantly correlated with pH only in the permafrost layer (Supplementary Fig. S1d). Overall, methanotroph abundances showed a strong and positive correlation with methanogen abundances in all layers except permafrost (Supplementary Fig. S1e). This suggests that these methanotrophs in the active layers likely rely on the CH4 produced by the methanogens. While methanogen abundances could be mainly driven by abiotic factors such as water (Supplementary Fig. S1a), methanotroph abundances could be driven by the resource availability in pan-Arctic permafrost affected soils.

Functional guild compositions across the pan-Arctic and soil horizons

Methanogens conducting hydrogenotrophic methanogenesis dominated the methane producing community (Fig. 4a). We considered phylotypes found in all locations (excluding Tazovskiy) as members of the pan-arctic core methanogenic microbiome. One member was associated with the uncultured candidatus family Methanoflorentaceae (former Rice-cluster-II) which had been found widespread in thawing permafrost44,62. The other phylotype was associated with the H2-dependent methylotrophic order Methanomassilicoccales. Another one was closely related with Methanosarcina which can use multiple methanogenesis pathways. The most abundant phylotype, however, was closely related to Methanobacterium which can perform methanogenesis from both H2/CO2 and methanol/H2 pathways.

Fig. 4: Distributions of methanogen and methanotroph phylotypes in different sites and horizons in the pan-Arctic.

a, b The relative abundance of each methanogen (a) and methanotroph (b) phylotype. Numbers in brackets indicate different phylotypes that were assigned to that genus as shown in Supplementary Data 1. RA relative abundance, NA not assigned at genus level, MOB methane oxidizing bacteria, ANME anaerobic methanotroph, * core methanogen or methanotroph members that were found in all sites (excluding Tazovskiy).

The pan-Arctic methanotrophs were dominated by type I MOB (Fig. 4b). Only two phylotypes of the Ca. Methanoperedens clade (formerly ANME-2d) were found and one was notably abundant across many locations, especially in the subsoil (Fig. 4b), suggesting a potential vital role of anaerobic methane oxidation in mitigating methane emissions in anoxic deeper layers of permafrost soils. Similarly, only two type II MOB phylotypes related to Methylocella were found and one was found in all the sites (Fig. 4b). They are known as facultative methanotrophs that can grow on low-molecular-weight organic compounds30. Among the detected type I MOB phylotypes, nine were related to the obligate methanotroph Methylobacter and seven were found as the core methanotrophs (Fig. 4b). Taken together, these Methylobacter-like phylotypes accounted for in average 77% of the total methanotrophic community. This proportion varied across studied locations, ranging from min. 28% (Tazovskiy, Russia) to max. 98% (Disko, Greenland). This finding is remarkable, since it suggests that the microbial CH4 filter in permafrost-affected soils across the Arctic is of a strikingly low diversity and dominated by one single genus. Further, NCBI BLASTN results suggested that these Methylobacter-like phylotypes were all closely related to M. tundripaludum SV96 (95–100% identity, Supplementary Data 1) which was first isolated from Svalbard peatlands63. This hints that the pan-Arctic microbial CH4 filter might even be dominated by one single species, although the species-level identity is not robust based a short fragment (~250 bp) of 16S rRNA gene. M. tundripaludum was also found as dominant methanotrophs in other locations on Svalbard64 and in the Lena Delta in Siberia65, which supports our finding. Moreover, M. tundripaludum was found as an active methane oxidizer in high Arctic wetlands using stable isotope probing66. A recent study showed physiological adaptions of M. tundripaludum to varying temperatures by adjusting their central metabolism, protein biosynthesis, cell walls and storage67. This might explain the prevalence and dominance of M. tundripaludum across heterogeneous Arctic regions.

Responses to different water status after permafrost degradation – case study in Alaska

The relative abundance of methanogens in the Intact site was <0.01% in both 2019 and 2021 (Supplementary Fig. S4a), which is among the lowest relative abundances detected in the other intact sites across the pan-Arctic (Fig. 3a). Accordingly, only one phylotype related with Methanobacterium was found in the Intact site in 2021 (Fig. 5a). However, 4 phylotypes of the order Methanomassiliicoccales were mainly found under Dry and Intermediate-Wet conditions (Fig. 5a), but still only accounting for up to 0.03% of the total microbiome (Supplementary Fig. S4a). They were only detected in the deeper soil layers (>30 cm below the surface) (Fig. 5a). The low abundance in the Intact site of Alaska might be due to the shallower sampling depth as more methanogens were found in the layers lower than 40–60 cm in the other sites (Fig. 5a). The deeper layers in the Intact site were frozen permafrost in which microbes should not be active despite their existence.

Fig. 5

Distributions of methanogen and methanotroph phylotypes in different sites and depths in the Alaska study. a, b The relative abundance of each methanogen (a) and methanotroph (b) phylotype in 2019 and 2021. Numbers in brackets indicate different phylotypes that were assigned to that genus as shown in Supplementary Data 1. RA relative abundance, NA not assigned at genus level, MOB methane oxidizing bacteria, ANME anaerobic methanotroph.

The methanotrophs were less diverse in the Intact site compared to the other pan-Arctic sites (Figs. 4b and 5b). Only Methylobacter-like phylotypes were detected in both 2019 and 2021. Their relative abundance in the Intact site is comparable to that in Herschel Island and Disko Island (Fig. 3b and Supplementary Fig. S4b). In contrast, the methanotrophic communities in Dry and Wet sites were dominated only by Methylocapsa-like and Methylobacter-like phylotypes, respectively (Fig. 5b). In the Intermediate-Wet site, both groups were found, with additionally Ca. Methylomirabilis phylotypes detected (Fig. 5b). Compared to the Intact site, the relative abundance of methanotroph was higher in the Dry and Intermediate-Wet sites and similar in the Wet site (Supplementary Fig. S4b).

All type I MOB were only found in depth >30 cm below the surface where methanogens were detected, while the type II MOB in the Dry site were mostly found in upper soil layers, with a few found in the deeper layers (Fig. 5a, b). Similar to the pan-Arctic sites, Methylobacter-like phylotypes dominated the MOB community in the Intact and Wet sites (Fig. 5b). This is astonishing as their dominance is consistent across the pan-Arctic permafrost affected soils, even including the Wet degraded permafrost site. We ran a local BLASTN to link Alaska ASVs to the pan-Arctic ZOTUs. Most of the type I MOB ASVs could be linked to a ZOTU with 100% identity (Supplementary Table S2), suggesting that most of these detected Alaska Methylobacter were likely identical to those found at the other pan-Arctic sites.

The distribution pattern of methanotrophs in different Alaska sites is likely due to the different water conditions. In the Dry site, the Methylocapsa-like phylotypes are closely associated with M. palsarum and M. gorgona (>97% identity, Supplementary Data 1). It is known that Methylocapsa hosts atmMOB and that M. gorgona MG08 is so far the only isolated atmMOB32. Due to their high-affinity to CH4, these MOB can utilize CH4 at very low CH4 concentrations, e.g., the atmospheric level[33](https://www.nature.com/arti