Genomic insights into redox-driven microbial processes for carbon decomposition in thawing Arctic soils and permafrost

ABSTRACT Climate change is rapidly transforming Arctic landscapes where increasing soil temperatures speed up permafrost thaw. This exposes large carbon stocks to microbial decomposition, possibly worsening climate change by releasing more greenhouse gases. Understanding how microbes break down soil carbon, especially under the anaerobic conditions of thawing permafrost, is important to determine future changes. Here, we studied the microbial community dynamics and soil carbon decomposition potential in permafrost and active layer soils under anaerobic laboratory conditions that simulated an Arctic summer thaw. The microbial and viral compositions in the samples were analyzed based on metagenomes, metagenome-assembled genomes, and metagenomic viral contigs (mVCs). Following the thawing of permafrost, there was a notable shift in microbial community structure, with fermentative Firmicutes and Bacteroidota taking over from Actinobacteria and Proteobacteria over the 60-day incubation period. The increase in iron and sulfate-reducing microbes had a significant role in limiting methane production from thawed permafrost, underscoring the competition within microbial communities. We explored the growth strategies of microbial communities and found that slow growth was the major strategy in both the active layer and permafrost. Our findings challenge the assumption that fast-growing microbes mainly respond to environmental changes like permafrost thaw. Instead, they indicate a common strategy of slow growth among microbial communities, likely due to the thermodynamic constraints of soil substrates and electron acceptors, and the need for microbes to adjust to post-thaw conditions. The mVCs harbored a wide range of auxiliary metabolic genes that may support cell protection from ice formation in virus-infected cells. IMPORTANCE As the Arctic warms, thawing permafrost unlocks carbon, potentially accelerating climate change by releasing greenhouse gases. Our research delves into the underlying biogeochemical processes likely mediated by the soil microbial community in response to the wet and anaerobic conditions, akin to an Arctic summer thaw. We observed a significant shift in the microbial community post-thaw, with fermentative bacteria like Firmicutes and Bacteroidota taking over and switching to different fermentation pathways. The dominance of iron and sulfate-reducing bacteria likely constrained methane production in the thawing permafrost. Slow-growing microbes outweighed fast-growing ones, even after thaw, upending the expectation that rapid microbial responses to dominate after permafrost thaws. This research highlights the nuanced and complex interactions within Arctic soil microbial communities and underscores the challenges in predicting microbial response to environmental change.

T emperatures in the Arctic permafrost zone have risen 0.6°C per decade over the last 30 years, twice as fast as the global average (1).As a consequence, the nearsurface permafrost has been thawing and exposing substantial quantities of organic carbon (C) to accelerated decomposition by soil microbes.It is estimated that the permafrost C stock, which contains about 50% of the global soil C (2, 3), will decline 30%-40% by the end of the 21st century (4).This anticipated decline is largely due to stimulated microbial decomposition of the organic C currently stored in permafrost (5).Under aerobic conditions, organic C can be mineralized by soil microorganisms to produce carbon dioxide (CO 2 ), whereas under anaerobic conditions methane (CH 4 ), production can be the dominant C emission pathway in addition to C loss as CO 2 via anaerobic respiration/fermentation (6).Methane, accounting for approximately 2.3% of the expected C loss from thawed permafrost, may be oxidized by CH 4 -oxidizing bacteria before reaching the atmosphere.In addition, sporadic emissions of another potent greenhouse gas (GHG), nitrous oxide, can occur from thawing permafrost due to denitrification (7)(8)(9).Predicting C loss from the Arctic regions, however, has a high level of uncertainty (i.e., ranging from a net C sink to a strong source) (4,10,11), in part due to the poorly understood and complex microbial functions that drive soil C-cycling during permafrost thaw (12).
Describing soil C dynamics in a warming climate requires understanding how soil microorganisms degrade organic C upon permafrost thaw (13,14).Post-thaw soil moisture varies with topography, leading to the development of both well-drained and inundated areas within the same location (5).While the majority of studies have primarily focused on aerobic processes (15)(16)(17)(18) shedding light on the consequences of aerobic respiration following permafrost thaw and drainage, it is equally important to consider anaerobic microbial processes in inundated areas.In inundated areas, anaerobic microbial processes predominantly drive organic C decomposition CO 2 via respiration and fermentation, encompassing various redox reactions (19)(20)(21)(22).The dominance of specific reactions in inundated areas is dictated by prevailing redox conditions and the availability of alternate electron acceptors, such as C, iron, and sulfur (23,24).The anaerobic fermentation of complex organic compounds in Arctic soils yields simple organic acids, for example, formic or acetic acid, where iron (Fe 3+ ) or sulfate (SO 4 2− ) reducers can rapidly utilize most produced acids (such as acetate) (24)(25)(26)(27).Consequently, both experiments (24) and models (28) have indicated that the rate-limiting step for methane CH 4 production typically lies in the generation and availability of substrates through fermentation and microbial biomass.In addition.the relative abundance of methanogens, specifically acetoclastic and hydrogenotrophic types, is influenced by environmental factors like temperature, nutrient levels, and soil pH (23).This sensitivity can lead to instances where abundant substrates are not fully utilized, particularly in acidic soils where acetoclastic methanogenesis is hindered (29,30).Such limitations could restrict CH 4 emissions despite of increased substrate availability due to warm ing.Methane emissions in soils are also controlled by microbial CH 4 oxidation (8,31) converting an estimated 20%-60% of CH 4 emissions into CO 2 in Arctic systems (32)(33)(34).In-depth investigations into the distribution and functional roles of post-thaw microbial communities in inundated soils (23) are necessary to decipher pathways of anaerobic C decomposition after permafrost thaw.
Viruses play vital roles in soil ecosystems, influencing microbial dynamics through lysis, regulating microbial physiology, and affecting biogeochemical cycles via auxili ary metabolic genes (AMGs) (35).Although their significance in Arctic soil systems is less explored, in thawing permafrost soils, metagenomic data enable the tracking of virus-host abundance patterns, revealing intricate dynamics among microbial host lineages (36).Moreover, in thawed permafrost, viruses can impact soil C cycling through mechanisms like host lysis and enzyme production, leading to shifts in viral commun ities following thaw (35)(36)(37).Notably, viral abundances show promise in improving predictions of climate-relevant C chemistry measurements, particularly in methane production pathways and consumption, emphasizing their role in shaping C cycling dynamics (36).While our understanding of soil viruses and their role in soil biochemical processes is expanding rapidly, the interactions between microbial and viral communi ties in decomposing organic C, particularly under anaerobic conditions, are inadequately understood.
In this study, we determined the microbial community structure and metabolic potential under anaerobic conditions of the active layer and permafrost samples from a low-centered polygon from a coastal Arctic tundra site by expanding upon previously published process-based measurements of the major C loss pathways in these soils (38).At this location, near-surface permafrost is presently thawing, leading to the deepen ing of the active layer.Prior studies have demonstrated differences in the microbial communities found in the active layer and permafrost in terms of their community structure and function (39), as well as their sensitivity to C loss and thaw trajecto ries (40).The low-centered polygon studied here is representative of predominantly anaerobic and saturated Arctic landscape features (41).When incubated under anaerobic conditions for 60 days at 4°C (similar to the summer thaw period and the average summer soil temperature), active layer (mineral) soils produced 4-5 times more CO 2 and >10 times more CH 4 than permafrost (38).Rapid production of CO 2 was detected for the active layer during incubation while there was a substantial lag in CH 4 production, reported in Roy Chowdhury et al. (38).
We examined the microbial community composition and their functional potential in active layer and permafrost soils, pre-and post-thaw, under anaerobic conditions.We aimed to pinpoint shifts in microbial metabolic pathways and growth strategies governing anaerobic C decomposition and CO 2 /CH 4 ratios.We hypothesized that upon thaw populations with diverse gene pools for C decomposition, mainly the dominant Actinobacteria, would dominate both soil layers.Given the observed lag time for CH 4 production during the thaw and the prominence of Fe 3+ reduction (Fe 2+ concentration increased ~4 µmol/g dry soil per day in active layer soil) at this location (38), the methanogenic archaea are likely to be constrained and contribute less to C decomposi tion compared to iron reducers.At this permafrost location, total sulfate concentrations were up to 16 times higher than total iron concentrations (40).Recent reports sugges ted that mineral active layer soils produce 2.3 times less CH 4 under SO 4 2− -reducing conditions than under Fe 3+ -reducing conditions (42).We hypothesized that Fe 3+ -reducing microbes would constitute a larger portion of the post-thaw soil layers than SO 4 2− -reduc ing bacteria.Recognizing the intricate dynamics of microbial interactions, we utilized both gene-centric and genome-centric analyses to identify populations potentially competing for C substrates and electron donors after thaw.

Site description, soil sampling and greenhouse gas emission data
We used soil samples from a published experiment (38) where site description, sample collection, microcosm incubation set-up, CO 2 and CH 4 measurements, and changes in the biogeochemistry have been described in detail.Briefly, four intact frozen soil cores up to 0.91 m deep and 7.6 cm in diameter were collected in sterile liners from the ridge (two cores) and center (two cores) of a 24 m diameter low-centered polygon (LCP) within the wet acidic tundra of the Barrow Environmental Observatory (BEO) in Utqiaġvik, AK (71.282 o N, 156.610 o W), in April of 2012 and 2013.This is a continuous permafrost region with an average seasonal thaw depth of active layer ranging from approximately 34-53 cm (38,43).Active layer soil temperatures fluctuated between −22°C and 5°C over the course of a year (43).The active (Act) and permafrost (Perm) soil layers from each soil core were visually assessed, mixed thoroughly for each layer, and incubated without oxygen at a temperature of +4°C in the absence of light for 2 months creating conditions similar to the summer thaw period and the average summer soil temperature.Anoxic incubations were set up in triplicate, with each replicate consisting of 15 g of wet soil in a 60 mL sterile serum bottle, resulting in 24 microcosms (i.e., four cores × two layers [Act and Perm] × three replicates).Following the conclusion of the incubation (60 days), 2 g of soil was collected from each replicate and combined for microbial analysis.Pre-incubation soils were analyzed as a reference.In total, 16 soil samples were analyzed (four cores × two layers × two treatments (reference and incubated soil) = 16 samples).Due to damage during shipping, a post-incubation Act replicate was lost, leaving only 15 samples for DNA extractions instead of the performed (16).CO 2 and CH 4 surface emissions are compiled from the Dengel et al. data set containing flux tower measurements from the study site (19) and compared to soil chemistry and fluxes reported in Roy Chowdhury et al. (38).

DNA extraction and metagenome sequencing
Total DNA was extracted from 2 g of soil using an AllPrep DNA/RNA kit (Qiagen, Valencia, CA, USA) as described previously (40).The metagenomic shotgun sequencing libraries were prepared and sequenced at the Department of Energy's Joint Genome Institute (JGI) using an Illumina HiSeq 2000.Sequence quality control with BBMap (44) and assembly with SPAdes (45) (min contig length n = 200 bp) were performed by JGI.Both assembled and unassembled sequences were uploaded to the Integrated Microbial Genomes and Metagenomes (IMG/M) database (46) which provided sequenc ing statistics (Table S1), gene identification, and annotations.Metagenomic sequence coverage and diversity (N d ) of each metagenome were assessed using Nonpareil (v.3.3.3)(47).Sequencing depth (coverage) per sample ranged between 63% and 92% (Table S2) of the predicted diversity.

Gene-centric taxonomic and functional potential analysis
The phylogenetic distribution of the metagenomes was determined using both the assembled and unassembled sequences with 60+% identity IMG/M database (46), and the translated proteins from all detected coding regions of each metagenome were annotated by IMG/M.Microbial pathways involved in fermentation, CH 4 production, Fe 3+ and SO 4 2− reduction, C and nitrogen cycles were identified.The genes involved in C and nitrogen cycling were identified by grouping KEGG annotations into subsets according to Mackelprang et al. (12).

Metagenome-assembled prokaryotic genomes (MAGs) and metagenomic viral contigs (mVCs)
Quality filtered sequences from the JGI Genome Portal were downloaded and a combined assembly of all 15 samples was performed using MEGAHIT v1.1.3 (48) resulting in 1,824,395 contigs of ≥1 kbp.670 contigs were longer than 50 kbp, and the longest contig was 374,306 bp.The average GC content of this assembly was 52.83%.Two binning tools, MaxBin2 v2.2.5 (49) and MetaBAT2 v2.12.1 (50), were used and output bins were further dereplicated and aggregated with DASTool v1.1.10 (51).The CheckM v1.0.11 program (52) was used to determine the completeness and contamination of bins.Contaminated bins were further refined to improve bin quality (53).Protein-cod ing genes associated with a genome bin were manually checked to confirm that they had similar phylogenetic affiliation and guanine-cytosine content.The bins were also screened for 16S ribosomal RNA (rRNA) genes, where available, or phylogenetic marker genes for primary taxonomic annotation.Kaiju (v 1.6.2) (54) and Genome Taxonomy Database (GTDB-Tk v2.1.0,database release R207_v2) (55) were used to assign taxonomy.Genomes were annotated using Distilled and Refined Annotation of Metabolism (DRAM) (56), Metabolic v4.0 (57), Department of Energy Systems Biology Knowledgebase (KBase) (58), and Rapid Annotations using Subsystems Technology (RAST) (59) servers.MAG abundance was calculated by the CoverM (v1.1.5)pipeline (60).Translated methyl-coen zyme M reductase A (mcrA) functional gene sequence identities were confirmed with BlastP (61) to the online NCBI database (62).Growth Rate InDex (GRiD; v 1.3 in multiplex mode) (63) was used to estimate growth rates of those MAGs having >75% completeness and <10% contamination.The GRiD algorithm estimates bacterial growth rates based on the principle that a rapidly dividing bacterium at any given time will have more copies of DNA close to the origin of replication (ori) in comparison to the terminus (ter) region (63).MAGs were identified as slow growers when the ori/ter was <2.5 (64).We filtered all GRiD estimates with the following stringency to reduce the likelihood of false growth estimates ( 63): (i) strain heterogeneity <0.3; (ii) coverage >0.2; (iii) GRiD values < 10; and (iv) if GRiD values were greater than 3, then genomes with sizes less than 4 Mb.In addition, we also used Compute PTR (CoPTR) (65) to estimate microbial growth rates from MAGs with >90% completeness.Both GRiD and CoPTR values were validated for two or more replicates of each treatment (i.e., eight values at least for each MAG to obtain reasonable comparisons among treatments).As a result, 73 MAGs passed the quality filtering with GRiD method (1,812 data points) and were used for comparative analysis.Via CoPTR method only 18 MAGs passed the quality filtering (122 data points).GRiD results were used in comparisons between different treatments, and CoPTR estimates are reported in supplementary files.
The combined metagenome assembly was screened for DNA viral sequences using VirSorter2 v 2.2.3 (66) and quality checked with CheckV (67).Viral sequences were filtered to retain sequences greater than or equal to 10 kbp in length following MIUViG standards (68).vConTACT2 v0.9.8 (69) was used for the clustering of viruses belonging to the viral reference taxonomy databases.Viral taxonomy was assigned via geNomad (70) application implemented in NMDC Edge (https://nmdc-edge.org).DRAM-v annotations (56) were used to assess viral versus non-viral genes, where mVCs containing more than 18% of non-viral genes were assumed to be misclassified and subsequently discarded.Potential mVC hosts were determined with iPHoP using default parameters (71).Relative mVC abundance was calculated as described for MAGs.Genes from mVCs were predicted with Prodigal v2.6 (72) and annotated to the PFAM 35.0 database (73) with DIAMOND v2.0.0 (74) and matched with DRAM-v annotations.Conserved regions and active sites in the protein sequences were analyzed using PROSITE (75).To assess protein structural similarity, we predicted quaternary structures using SWISS-MODEL with a Global Model Quality Estimation (GMQE) score above 0.5 (76).Ligand-binding sites were predicted with the PrankWeb (77) server with models generated by SWISS-MODEL.

Statistical analysis
For gene-centric analysis, de novo annotations obtained from IMG/M were formatted in Excel to include relevant metadata for subsequent analyses using R (version 3.6.1)(78).We conducted normalization procedures to the retrieved metagenomes to remove the technical bias and make relatively fair comparisons.We divided each gene (KO) count with the count of the rpoB gene and then calculated the relative abundance (40).Ordination based on the relative abundance of phylogenetic clades and KEGG genes was performed using non-metric multidimensional scaling (NMDS), with a Bray-Curtis distance measure, using the package vegan (version 2.5) (79).Phylogenetic clades and genes were considered enriched or decreased after incubation if the relative abundance was significantly different between pre-and post-incubation.Results were defined to be significant at P < 0.05.The t-tests and Tukey's HSD tests were run to determine if there was a significant change in relative abundance due to incubation conditions and soil layer.To determine differences in biogeochemical pathways the relative abundances of select KEGG genes involved were compared before and after incubation.

Bacteroidota and Firmicutes outgrow dominant Actinobacteria and Proteobac teria in the thawed permafrost
Upon thaw, microbial diversity and composition altered in permafrost but remained unchanged in the active layer.Comparison of pre-incubation samples from active layer mineral soils (Act Ref ) and permafrost (Perm Ref ) revealed no significant differences in either sequence coverage or sequence-based alpha-diversity (N d ), as estimated by Nonpareil (Tukey's HSD: P = 0.15 and P = 0.08, respectively) (Fig. S1; Table S2).While anaerobic incubation did not markedly change the alpha diversity in both soils, there was a 6% decline in N d Perm Inc (Fig. S1B).
We observed shifts in microbial community composition associated with soil layers at the end of the incubation period.In permafrost microbial community composition at the phylum level significantly changed after incubation (ADONIS: r = 0.92, P = 0.03; ANOSIM: r = 1.00,P = 0.04); however, no significant changes were detected for the active layer (ADONIS: r = 0.37, P = 0.55; ANOSIM: r = −0.04,P = 0.57) (Fig. 1C).Specifically, Firmicutes became the most abundant phylum in Perm Inc , followed by Bacteroidota (Fig. 1A; Table S3).In addition, microbial metabolic guilds showed a similar pattern as taxonomic composition (Fig. 1B and D).
To gain a more comprehensive understanding of these microbial communities, we reconstructed 231 MAGs from 20 phyla, including 37 high-and 194 medium-quality draft following MIMAG standards (80) (Table S4).MAG sizes were between 0.5 and 6.4 Mb with GC contents ranging from 30% to 72%.In all, 81 MAGs contained a 16S rRNA gene.Each MAG constituted, on average, 0.064% (0.007%-0.805%) of the total reads in its respective metagenome, in total 12.7%-22.4% of each metagenome (Fig. S2A).Similar to the read-based findings, the composition of the MAGs was significantly different between Act Ref and Perm Ref (ADONIS: r = 0.8, P = 0.03; ANOSIM: r = 1, P = 0.03) (Fig. S2B).The Actinobacteriota MAGs were significantly more abundant in Perm Ref relative to Act Ref and in contrast, MAGs from all other phyla were significantly more abundant in Act Ref (Fig. 2; Fig. S2A).
Incubation significantly changed the composition of the MAGs in both the active layer (ADONIS: r = 0.61, P = 0.03; ANOSIM: r = 0.5, P = 0.05) and permafrost (ADONIS: r = 0.76, P = 0.03; ANOSIM: r = 0.9, P = 0.02), and these changes were more pronounced in permafrost (Fig. 2; Fig. S2B).Major increases in abundance were observed for three Bacteroidota and one Firmicutes MAGs in Perm Inc (Fig. 2; Fig. S2A).The abundance of most Actinobacteriota MAGs decreased in permafrost soils after thaw (Fig. 2).These shifts mirrored the trends observed in our read-based analysis, further confirming the transformative effects of thawing on microbial communities.

Slow-growing MAGs remain abundant after thaw
Using GRiD values, we classified growth strategies: values below 2.5 were indicative of slow-growing bacteria (64).Both soil layers contained a similar number of MAGs with slow and fast growth strategies.They had average GRiD values of Act Ref = 3.10 ± .09(n = 3, samples) and Perm Ref = 3.44 ± .08 (n = 4, samples) (Fig. 3B).GRiD estimates were similar for most phyla in Act Ref and Perm Ref , except that the GRiD was higher for Chloroflexota and Thermoproteota in Perm Ref (Fig. S3).The GRiD values and relative abundances for the MAGs were negatively correlated in both the active layer and permafrost, especially following incubation (Fig. 3C) indicating that slow-grow ing populations were abundant both in reference and incubated samples.The CoPTR and GRiD estimations were significantly correlated, except for the active layers after incubation (Fig. S4).
At the phylum level, post-incubation growth rates varied significantly (Fig. S3).Specifically, the GRiD values for Acidobacteriota, Actinobacteriota, and Chloroflexota increased after permafrost incubation (Fig. S3).In the active layer, the GRiD values for Actinobacteriota increased, but Nitrospirota decreased after incubation (Fig. S3).When grouped taxonomically, the MAG relative abundances remained negatively correlated to GRiD values (Fig. S5).These results show that despite the average increase in estimated growth rates (Fig. 3B), slower-growing MAGs were still two times more abundant than fast-growing MAGs (Fig. 3A).Observed trends across phyla indicate that, during seasonal thaw, putative fast-growing organisms may not consistently leverage the post-thaw environment.

Thaw increased carbohydrate decomposition potential
At this location, both CO 2 and CH 4 surface fluxes gradually decreased throughout the summer (Fig. 4A) (19).While both the active layer and permafrost had a comparable amount of total organic C (t-test = 0.6, P = 0.564), permafrost contained less total nitrogen (t-test = 2.5, P = 0.034) and higher pH (t-test = −41.5,P < 0.001) (38).Over the incubation period, Roy Chowdhury et al. (38) reported higher CO 2 and CH 4 fluxes from active layer soils in comparison to permafrost (Fig. 4B).The relative abundances of key genes involved in C and nitrogen cycling were higher in Perm Ref than in Act Ref (Fig. 4C).Higher relative abundances of substrate acquisition genes in Perm Ref may result from an increased need to generate C and energy reserves compared to the active layer microbial community (81).Following incubation, significant increases were observed in genes encoding acquisition pathways for trehalose, cellobiose, and sucrose, as well as nitrogen fixation genes in both Act Inc and Perm Inc (Fig. 4C), indicating increased C and nitrogen cycling activity upon thaw.Sugar transporter genes increased and NADH dehydrogenase I and hydrocarbon degradation genes decreased only in Perm Inc (Fig. 4C).The relative abundances of fermentation pathway genes remained similar, but a major change post-incubation was the significant decrease in the functional potential for pyruvate fermentation to formate and acetyl CoA in Act Inc (Fig. 4D).By contrast, genes involved in pyruvate fermentation to formate significantly increased in Perm Inc (Fig. 4D).Predominant Bacteroidota MAGs, which are highly abundant in thawed permafrost, contain a substantial number of CAZy family genes including GH2, GH3, GH13, GH29, and CE4 (Table S5).These enzymes target various structural carbohydrates, such as glucose, xylose, chitin, and trehalose.A similar, though less diverse, carbohydrate degradation capability was detected in the dominant Firmicutes MAG but was absent in the Actinobacteriota MAG.Instead of having a variety of carbohydrate degradation genes, the Actinobacteriota MAG possessed genes associated with pyruvate oxidation.In addition, Bacteriodata had genes like lactate dehydrogenase, acetate kinase, and acetyl-CoA synthetase, indicating their potential to utilize lactate and acetate.By contrast, Firmicutes MAG contained genes enabling the conversion of pyruvate to formate and acetyl coenzyme A. In combination with gene-centric analysis results, these observations demonstrate that thaw gives rise to microbial populations that are capable of targeting complex C reserves to drive their activities and potentially a smaller pool of volatile fatty acids in both soil layers.
The Methyl-coenzyme M Reductase genes (mcr) exhibited a higher presence in Perm Ref compared to Act Ref .Post-incubation, their abundance saw an uptick, albeit not a significant one (Fig. 4C; Fig. S6C).In addition, genes associated with acetogene sis, as evident in MAGs, experienced a marked rise in both abundance and estimated growth rate after permafrost incubation (Fig. 5A; Table S5).Of the 13 archaeal MAGs analyzed, four were identified as methanogens, specifically falling under the categories of Methanosarcina, Methanobacterium_B, and the Methanomassiliicoccales order (Table S4).The relative abundance of the methanogenic MAGs was similar between Act Ref (0.21% ± 0.03%) and Perm Ref (0.14% ± 0.03%); however, they were significantly more enriched in AL Inc compared with Perm Inc (Fig. 5D), which might explain the previously reported higher CH 4 production in AL Inc compared with Perm Inc (38).Methanobacte rium_B bin_metabat2.485had a complete methanogenesis pathway (Table S6), while bin_metabat2.910had 96.1% similarity to MAG 20100900_E2S Fen_5 from Stordalen Mire (8).Both Methanobacterium_B bins contained a single copy of the mcr gene cluster.Methanosarcinia bin_metabat2.102and Methanomassiliicoccales bin_metabat2.644had methyltransferase genes and a Wood-Ljungdahl pathway, providing H 2 for methylotro phic methanogens (83).Methanobacterium_B bin_metabat2.485(t-test, P = 0.05) and bin_metabat2.910(t-test, P = 0.04) were significantly more abundant in Act compared to Perm after incubation.Out of the six Bathyarchaeota MAGs examined, we did not detect the mcr gene cluster.However, we did identify other genes associated with methane metabolism, hinting at potential alternative functions for these genes within the Bathyarchaeota MAGs.
We identified 12 MAGs that contained decaheme c-type cytochrome family genes (mtrBC) that can potentially utilize iron/manganese reduction for energy production.Their relative abundances were similar between Act Ref and Act Inc , but higher in Perm Inc (Fig. 5B; Table S5).The Rhodoferax bin_metabat2.271,related to Rhodoferax ferrireducens T118, contained the mtr gene family, cytochrome c-552, and three copies of cytochrome c551 peroxidase genes.Its relative abundance in permafrost increased 19-fold from 0.05% ± 0.01% to 0.96% ± 0.55% after incubation (Fig. S10).By contrast, the Geomonas bin_maxbin2.039was significantly enriched in both the active layer and permafrost following incubation (Table S4).Similar to isolated members of Geomonas (86), this MAG can potentially utilize nitrate, iron, and a wide range of substrates as electron donors.In addition, 25 MAGs that contained iron oxidase Cyc gene were detected, and they showed a similar pattern as the FeRB MAGs (Fig. 5B).The relative abundances of FeRB and Fe 2+ oxidizing bacteria were positively correlated in the active layer (Fig. S11).These findings show that both the active layer and permafrost harbored diverse FeRB and Fe 2+ oxidizing bacteria and align with prior research suggesting the centrality of iron chemistry in C decomposition at this location (82).
As for sulfate reducers, more than 10 potential SO 4 2− -reducing genera, where both the Act Ref and Perm Ref had a similar composition (Fig. S9C).The relative abundance of these genera significantly increased by 41% after permafrost incubation (Fig. S9C).The Desulfosporosinus genus, being the dominant sulfate reducer in both soil layers, was further enriched by 246% (Fig. S9C).Sulfate reduction MAGs were comparably abundant between Act Ref and Perm Ref , but they significantly increased upon permafrost thaw (Fig. 5H; Table S5).We recovered six MAGs with genes for thiosulfate disproportionation (belonging to Acidobacteriota, Chloroflexota, Desulfobacterota, and Firmicutes) that were capable of dissimilatory sulfate reduction.These MAGs significantly increased following the thaw in permafrost (Fig. 5J).The Desulfosporosinus bin_maxbin2.359specifically held genes enabling full reduction of sulfate to hydrogen sulfide (Table S5).This MAG's abundance increased 28-fold in permafrost post-incubation (Table S4).Post-thaw conditions in permafrost significantly impact both iron-cycling bacteria and sulfate reducers, underscoring the dynamic microbial shifts induced by changing environmental conditions.

Thaw reshaped viral community structure
In total, 186 mVCs were detected in our samples (Table S7), consisting of 3 complete, 50 high-quality, and 133 medium-quality putative viral genomes.For the overall abundance of mVCs, their relative abundances between the Act Ref and Perm Ref were not signifi cantly different (Turkey's HSD: P = 0.14) (Fig. 6A).However, the relative abundance of the mVCs significantly increased in permafrost following incubation (Fig. 6A), exceeding levels in Act Inc .After incubating the permafrost, 61 mVCs showed a significant increase in their levels, while 19 mVCs displayed a significant decrease (Fig. S12).The mVC composition was different between Act Ref and Perm Ref (ADONIS, r = 0.64, P = 0.03), and incubation significantly changed their composition only in Perm Inc (ADONIS, r = 0.68, P = 0.02, Fig. 6B).In all, 55 mVCs could be linked to their potential hosts in nine bacteria and one archaeon (Table S7).mVCs carried AMGs with active sites that potentially encoded carbohydrate-active enzymes (Table S7; Table S8).We predicted quaternary structures for these putative enzymes and found that structural predictions were not always consistent with the sequence-based functional prediction (Table S7) and only considered putative functional annotations of the genes where the predicted structure is consistent with the predicted biological function.Specific AMGs carried on mVCs include the following: the Acetobacterium-associated mVC_00586898 contains a putative gene encoding nitroreductase involved in the production of hydroxylamine; the Azonexus-associated mVC_00718886 contains genes from the AbiE system (81) that can provide resistance and enable stabilization of mobile genetic elements, such as plasmids; the mVC_00839254 associated with Flavobacterium carried a cluster of putative anti-freeze genes (81) that would prevent the formation of ice crystals in their hosts.
mVC-host abundance correlations were tested to investigate potential connections between viruses and their hosts.Significant positive correlations between mVCs (abundance >0.01%) and hosts (abundance >0.1%) were found in all the soils, except for Perm Inc (Fig. S13).Several MAGs were correlated to the relative abundances of specific mVCs.Similarly, mVCs associated with MAGs that were significantly abundant after thaw, namely Geomonas, Flavobacterium, and Acetobacterium, were also more abundant in post-thaw samples (Table S7).

DISCUSSION
Thawing permafrost drastically shifts C substrate complexity and oxidation potentials to support microbial growth niches and alters existing ones, potentially disadvantaging permafrost microbes acclimated to life under sub-zero conditions over millennia (5).Our findings reveal distinct microbial functionalities in the seasonally thawed active layer, contrasting with the reduced diversity found in permafrost.Despite shifts in functional gene composition post-thaw, we observed no potential loss in functionality.Understanding the driving factors behind microbial structural and functional shifts is highly challenging under in situ conditions due to the variable conditions and many covarying and compounding factors.Consequently, laboratory-scale controlled incubations are an indispensable tool for dissecting the nuances of anaerobic C cycling and discovery (12,20,87).Notably, a study by Wilson et al. (88) demonstrated microbial functional consistency in C cycling between field observations and incubation experi ments, underscoring the value of controlled incubations in bridging the gap between laboratory findings and real-world microbial dynamics.At the study location, seasonal surface fluxes result in CO 2 :CH 4 ratios ranging from 3 to 400 (19) with a seasonal average of 40 from the active layer (Fig. 4).Larger ratios originated from delayed methanogenesis at the beginning of the thaw season.In comparison, laboratory-scale incubations studied here had CO 2 :CH 4 ratios of 1-40 in mineral soils of the active layer and CO 2 :CH 4 ratios of 100-290 upon permafrost thaw (Fig. 4).We observed an enrichment of several anaerobic C degradation pathways in permafrost and an uptick in resource acquisition genes in response to thawing, potentially compromising cellular growth efficiency and increasing C decomposition (5,89).Although experimental handling (e.g., sampling and homogeni zation prior to incubation) of soils induces perturbations that limit the extrapolation of all observations back to field measurements, the inferred metabolic functions related to GHG measurements are largely consistent with field observations (40,88).CH 4 production rates reported from these incubations were substantially higher in the active layer than in permafrost (Fig. 4) (38), with a significantly greater accumulation of acetate observed (82).In the active layer, acetate concentrations in the water-extracta ble fraction of organic C reached approximately ~12 µmol g SOC −1 (82), and Methanosar cina spp., which some members can grow with acetate as a sole energy and C source, were the dominant methanogens in both active layer and permafrost.Despite the higher methanogenesis potential in permafrost, as indicated by mcr gene abundance, CH 4 production from thawed permafrost remained minimal (38).These findings suggest that the elevated CH 4 production in the active layer, compared to permafrost, can be attributed to higher microbial cell density and substrate availability (14,38), rather than to genomic constraints on the permafrost microbes.Under strong syntrophic oxidation of acetate or glucose tied to methanogenesis, CO 2 :CH 4 ratios approaching 1 would be expected, where without contributions from the respiration of alternative terminal electron acceptors, methanogenesis would be the sole source of CO 2 production (90).The substantially higher ratios observed in the permafrost incubations, along with changes in the distribution of fermentation pathways and enrichment toward formate production, signal a potential substrate limitation for the predominantly acetoclastic methanogens in permafrost.Incubations studied here artificially separate active layer soils from thawing permafrost.In natural settings, thawing permafrost interacts with compaction from both organic and mineral soils of the active layer, facilitating the transport of dissolved C, nutrients, and potentially microbes into the thawed permafrost via porewater.Freshly transported resources can support resident microbes and impact CO 2 :CH 4 ratios.
We show that changes in redox-driven microbial processes indicate a shift toward CO 2 production by fermentation, followed by Fe 3+ and SO 4 2− reduction.Especially the replacement of globally prevalent permafrost taxa such as Actinobacteria and Proteobac teria (91) with Bacteroidota and Firmicutes shows that under anaerobic and ice-free conditions these latter groups are more adept at thriving.Bacteroidota and Firmicutes were found to prevail upon permafrost thaw because of their metabolic capabilities, mainly tied to the fermentation of complex organic matter, flexibility, and resilience in coping with thaw-induced environmental shifts.In the active layer, an estimated 6% of the available acetate can be utilized for acetoclastic methanogenesis (82), indicating competition with other microbes for acetate utilization.We observed a higher abun dance of FeRBs compared to methanogens, and this was complemented by an increase in genes associated with Fe 3+ reduction in the permafrost layer after thaw.Fe 2+ concen tration increased from 63.5 to 150.3 µmol g −1 dwt soil (38) during incubation where Fe 3+ reduction was predicted to consume more acetate than methanogenesis (82).In addition, we observed an emergence of SO 4 2− reducers, including a significant increase in Desulfosporosinus, which acted as competitors for available C in permafrost following warming.In sparsely vegetated high Arctic permafrost, microbial sulfate reduction was previously found to dominate the anaerobic processes, outcompeting methanogenesis for H 2 and acetate after permafrost thaw (26).The thermodynamic advantages of Fe 3+ and SO 4 2− reduction pathways compared to methanogenesis (92), the emergence of FeRB and sulfate reducers, and acidic soil conditions at the sampling location (38) collectively limit methane production from thawed permafrost.
It is assumed that the active layer microbial communities are more diverse and grow faster than those in the permafrost (5).Our study confirms the greater diversity in active layers compared to permafrost.However, upon thawing, we observed no significant shifts in growth strategies.Both the active layer and permafrost hosted a mix of fast and slow growers, where slower growth strategies we significantly abundant in both layers.To estimate potential growth rates, we employed the peak-to-trough ratio (PTR) method, which offers a snapshot of growth at the time of sampling (63).However, the PTR method's sensitivity, particularly for incomplete and fragmented genomes like MAGs, hinges on the availability of high-quality, long contigs for detecting the origin of replication (65).Therefore, a stringent quality assessment is crucial for reliable and ecologically meaningful estimations (63).PTR methods, in conjunction with further genomic analysis, were used to reveal shifts in community growth rates in response to environmental change in soil and groundwater (93-97) ecosystems.
Our findings indicated a negative correlation between microbial abundances and estimated growth rates.The stringent cutoff values used for strain heterogeneity (<0.3) to avoid errors and the overall lack of representation of closed arctic microbial genomes in databases limit our ability to estimate growth strategies more comprehensively.In both the active layer and permafrost, slower-growing microbes were more prevalent than their faster-growing counterparts, a pattern that persisted post-thaw.It is impor tant to note that in situ growth rates are influenced by both growth yield and the rate of resource acquisition (98), the latter not being captured by the PTR method.The relationship between growth rate and yield is complex and context-dependent.Depending on conditions such as temperature (99), this relationship can be either negative or positive (100).High yields can be achieved through various growth rates, not solely through maximum growth rate, underscoring that while high yield is a consistent, and trait-based strategy, a high growth rate is more a product of aligning any strategy with favorable environmental conditions (101).Slow-growing microbes have been predicted to invest more resources and energy to produce extracellular enzymes to break down extracellular resources (98) and genes for these processes were detected in the MAGs of this study.Consistency of negative correlations between MAG relative abundance and their estimated growth rates in both soil layers and under pre-and post-thaw conditions suggests that slow growth is likely a community-wide strategy tied to the thermodynamic potential of the soils (102).The Arctic microbes at this site are thus likely balancing acclimation to stressful pre and post-thaw conditions at the cost of a lower growth rate (103).
Arctic soils host a diverse virosphere and active virosphere (36) that are impacted by environmental conditions like soil thaw duration (37).Viral populations in these soils have been found to play a role in CH 4 dynamics (36) and may carry genes that aid their hosts in C utilization (104).It has been proposed that similar to marine systems, where viruses carry AMGs that support host energy generation via photosynthesis (105), soil viruses contribute to C cycling via genes involved in C decomposition.This is evidenced by the discovery of a chitosanase enzyme from a soil virus that aids in breaking down complex C resources (106).Here, we describe 186 bacteriophages that have the potential for coding AMGs.Strong host-viral load correlations underpin the importance of virulence in regulating host abundances.However, accessing mVCs via metagenomics has limitations (107).Total DNA extraction and sequencing, without targeted enrichment for viruses, may underestimate a large proportion of the sample viral diversity (108).With this limitation, we are more likely to detect viruses in the lytic replication phase where their hosts can proliferate and reach high densities (109).Consequently, a strong positive correlation between an abundant MAG and their mVC could be expected.Several of the mVCs that we detected that were associated with multiple taxonomically different MAGs carried anti-freeze genes that, if functional, would prevent the formation of ice crystals in their hosts (110).As with other AMGs, anti-freeze genes likely improve viral capacity to produce more progeny under freezing conditions increasing their chance of survival.

Conclusions
In this study, we demonstrate that thaw-induced changes lead to a predominance of metabolic functions involved in CO 2 production through fermentation, along with Fe 3+ and SO 4 2− reduction, collectively constraining methanogenesis.The majority of the active layer and permafrost microbes under post-thaw conditions grow slowly, suggesting that especially permafrost C decomposition may be more gradual than expected under saturated conditions.Potential C decomposition leading to CO 2 via fermentation can limit substrate pool, and cause high CO 2 to CH 4 ratios in Arctic soils under post-thaw anaerobic conditions.Our estimations show that it is crucial to recognize that maximizing growth rates is not always suitable for all arctic microbes and their yield might be decoupled from their observable growth strategy.

FIG 1
FIG 1 Read-based (gene centric) microbial community composition (A) and metabolic pathways (B) in pre-(Ref ) and post-incubation (Inc) samples of the active layer (Act) and permafrost (Perm).NMDS analysis of the relative abundance of microbial phyla (C) and metabolic genes (D) from the metagenomes in the active layer (blue) and permafrost (black) layers.Pre-incubation samples are represented by triangles and after-incubation samples are circles.

FIG 2
FIG 2 Mean effect sizes of thaw on the relative abundance of MAGs (log response ratios, LogRR).Perm Ref vs Act Ref indicates differences between the active layer and permafrost before incubation; Act Inc vs Act Ref indicates LogRR of incubation effects in the active layer; Perm Inc vs Perm Ref indicates LogRR of incubation effects in permafrost.The bar plot on the right shows the mean values of the relative abundance of each MAG in AL Ref and PL Ref .Only significantly changed MAGs between each pair tested by t-test are shown.

FIG 3 8 FIG 4
FIG 3 Cumulative relative abundance of all the MAGs that had validated Growth Rate InDex (GRiD) (A), fast-growing MAGs (i.e., GRiD > 2.5, FG in panel A), and slow-growing MAGs (i.e., GRiD < 2.5, SG in panel A).GRiD of all the MAGs (B) and their associations explored via linear correlations (C) in pre-(Ref ) and post-incubation (Inc) samples of the active layer (Act) and permafrost (Perm).The Pearson correlation coefficient (R) and P-values are shown.* indicates significant changes based on t-test; ***P < 0.001, **P < 0.01, *P < 0.05, ns not significant.Different lowercase letters in B indicate significant differences (P < 0.05) among the four groups.

FIG 5
FIG 5 Relative abundance of MAGs that have complete acetogenesis (A), iron reduction (B) and oxidation (C), methanogenesis (D), anaerobic methane oxidation (E), N 2 fixation (F), nitrate reduction (G), sulfate reduction (H), sulfur oxidation (I), and thiosulfate disproportionation (J) pathways in active layer (Act Ref ), post-incubation in active layer (Act Inc ), permafrost (Perm Ref ), and post-incubation in permafrost (Act Inc ).In addition, for iron oxidation and reduction pathways MAGs with incomplete pathways are also shown.Different lowercase letters in each panel indicate significant differences (P < 0.05) among the four groups, n.s.indicated there were no significant differences among the four groups.

FIG 6
FIG 6 Relative abundance of metagenomic viral contigs (mVCs) (A) and nMDS analysis of their composition (B) in active layer reference (Act Ref ), post-incubation in active layer (Act Inc ), permafrost reference (Perm Ref ), and post-incubation in permafrost (Perm Inc ).Different lowercase letters in panel A indicate significant differences (P < 0.05) among the four groups.
work (proposal: 10.46936/10.25585/60000598)conducted by the U.S. Department of Energy Joint Genome Institute (https://ror.org/04xm1d337),a DOE Office of Science User Facility, is supported by the Office of Science of the U.S. Department of Energy operated under contract number DE-AC0205CH11231 to Lawrence Berkeley National Laboratory.Yaoming Li was supported by the National Natural Science Foundation of China (41871067) and the China Scholarship Fund.Funding in part was provided to Neslihan Taş through by the Office of Biological and Environmental Research in the DOE Office of Science-Early Career Research program.Additional support was provided by the Microbiomes in Transition Initiative LDRD Program at the Pacific Northwest National Laboratory, a multi-program national laboratory operated by Battelle for the DOE under Contract DE-AC06-76RL01830.Neslihan Taş: Funding acquisition, data generation and analysis, writing-original draft, and editing.Yaoming Li: data analysis, writing-original draft, and editing.Yaxin Xue, writing-review & editing.Taniya Roy Chowdhury and David E. Graham: soil sample collection, incubation setup, writing-review & editing, Susannah G. Tringe and Janet K. Jansson: Funding acquisition, writing-review & editing.The author(s) read and approved the final manuscript.