2.1 Geochemical Characteristics
Permafrost in the Alazeya River basin from borehole AL1_15 upper layers (FP 2.95-6.0 m) was formed ~10-40 kyr ago, while deeper permafrost sediments (FP 15.0 and 25.1 m) were formed ~0.9 and 1.1 Myr ago (Table S1) [4]. The geological age of permafrost sediments from borehole CH1_17 also increased with depth. The permafrost layer from FCP 1.75-5.8 m likely formed ~10-100 kyr ago, whereas permafrost from MP 11.5-19.6 m was formed about ~105-120 kyr ago (Table S1) [4, 6]. Temperature in borehole AL1_15 varied from -1.8°C in upper layers to -5.9°C in deeper layers, whereas temperatures in borehole CH1_17 were lower and decreased with depth from -4.8°C in FCP sediments to -8.1°C in deeper MP layers. Redox potential measured in MP sediments was +100 – (+150) mV [6] and was lower in FP varying from +40 – (–100) mV in upper layers to –256 mV in deeper layers [27] indicating anaerobic conditions in permafrost.
FP samples from AL1_15 had 0.04-0.11% total dissolved solids and low salinity in the range of 0.1-0.2 ppt (parts per thousand). FCP from CH1_17 formed in a coastal zone had 0.09-0.38% total dissolved solids and salinity 0.3-1.3 ppt increasing with depth. The concentration of total dissolved solids and salinity increased to 1.58% and 6.1 ppt, respectively (Table S1), in the deeper MP layers in the same borehole. Concentration of ions Cl- and Na+ showed a similar trend in both boreholes [15, 30], increasing with depth from 7.5 and 0.89 mmol kg-1 to 25 and 1.76 mmol kg-1, respectively, in freshwater AL1_15 sediments; and from 6.5-55 and 3.7-33.2 mmol kg-1 in freshwater coastal CH1_17 sediments to 80-230 and 68-193.6 mmol kg-1, respectively, in marine permafrost CH1_17 sediments (Table S1). Concentration of K+ was in the range of 0.27-0.66 mmol kg-1 in FP; 0.3-0.6 mmol kg-1 in FCP; and increased from 1.8 to 7.1 mmol kg-1 with depth in MP (Table S1). Ratio of SO42- to Cl- in all samples was below 1 indicating the dominance of Cl- anions in all studied permafrost samples. Cl-, Na+ and K+ are major osmotically active ions. In addition, Cl- is widespread in bacteria and may be involved in the stabilization of membrane potential, regulation of intracellular pH gradients, regulation of key enzymes and salt adaptation [57].
The isotopic signature of methane collected from borehole AL1_15 at a depth of 23 m was -84.9 ± 0.5‰ for d13CCH4 and -316.2 ± 1.5‰ for d2HCH4 that shows a biogenic origin.
Total carbon in FP ranged from 0.762% to 2.663%; in FCP (1.678-2.088%); and in MP (0.214-0.911%). Total nitrogen measured in FP (0.072-0.175%); in FCP (0.149-0.158%); in MP at depth of 11.5-16.9 m (0.022-0.077%) and at 19.6 m (0.108%) was lower than total nitrogen in Siberian tundra soils (0.5-2%) [58]. During the previous study of freshwater permafrost samples collected from the Alazeya River site [59], a low concentration of ammonium (<100 ppm g-1 wet soil) and traces of nitrite and nitrate were detected in the upper sediments followed by a layer (12-24 m), where the concentration of ammonium was 3 times as high as that of the top layer. The same study [59] found high concentrations of nitrite (5-110 ppm g-1 wet soil), nitrate (65-115 ppm g-1 wet soil) and ammonium (110-152 ppm g-1 wet soil) in the Chukochy Cape marine sediments formed 100-150 kyr, while only ammonium was detected in freshwater coastal sediments in concentration of 22-85 ppm g-1 wet soil.
2.2 Microbial Diversity From 16S rRNA Gene Sequencing
In the freshwater permafrost from AL1_15, alpha-diversity determined by the Shannon and Simpson indexes was higher in the shallow and younger permafrost samples and biodiversity decreased with age (Fig. 1 A). A similar trend was observed in CH1_17 (Fig. 1 A).
The decrease in diversity with age is in accordance with previous findings in ancient permafrost [15, 60]. Notably, despite the similar origin and close estimated age, biodiversity in MP_11.5 was higher than in MP_16.9 and MP_19.6 (Fig. 1 A), which may be related to conditions of sedimentation or small-scale heterogeneity connected to sample depth. The non-metric multidimensional scaling (NMDS) analysis of all amplicon sequence variants (ASVs, based on relative abundance) showed that the microbial communities differed between the sediments based on age and salinity (Fig. 1 B). In both boreholes the shallow and younger samples grouped separately from deeper and older permafrost. Despite having different salinities ranging from 0.1 ppt in young FP to 0.3-1.3 ppt in FCP, the FP that formed ~10-40 kyr ago and FCP that formed ~10-40 and 100 kyr grouped in close proximity to each other, separately from deeper, older, and more saline permafrost. Based on 16S rRNA gene relative amplicon abundance from duplicate extractions, Actinobacteriota was the most abundant phylum at 2.95, 6.0 and 15 m in FP, while microbial communities from 25.1 m were dominated by Gammaproteobacteria (Fig. 2).
The increase of Gammaproteobacteria in the deeper FP samples (15 and 25.1 m) could be explained by an increase in salinity. Previous studies of freshwater and saline lakes and ponds showed a similar trend when Actinobacteria were more abundant in freshwater and low salinity waters [61] while Gammaproteobacteria increased with increasing salinity [61, 62]. There was no correlation between abundance of Gammaproteobacteria and salinity in samples from CH1_17.
In CH1_17, the low-salinity top permafrost layers (1.75, 5.4 m) and 11.5 m were dominated by Actinobacteriota and shifted to Chloroflexota in MP at 16.9 and 19.6 m, while 11.5 m sample contained considerable amount of Chloroflexota along with Actinobacteriota. Members of the Chloroflexi phylum are dominant in numerous subseafloor environments [63] that is consistent with our finding of high amounts of Chloroflexota in marine permafrost. ASVs from spore-forming Clostridia (Firmicutes_A phylum) and Bacilli (Firmicutes phylum) classes were, on average, most abundant at 15.0 m in FP (7.41%). These ASVs increased from a mean relative abundance of 2.74% at 1.75 m to 20.69% at 5.4 m in FCP, and then declined to 2.10, 1.18 and 1.57% at 11.5, 16.9 and 19.6 m, respectively, in MP. The highest abundance of the Firmicutes was observed in FCP at 5.4 m with moderate salinity of 1.3 ppt (1.1 mS/cm) and methane concentration of 76 µmol/kg. Previous studies showed a positive correlation between relative abundances of Firmicutes and either salinity gradient of 0.36-6.72 mS/cm in temperate soils [64] or CH4 production in anaerobic digester [65]. Archaea were detected in all the samples except for FP_25.1. Halobacterota and Crenarchaeota contributed a relatively high abundance in FCP at 5.4 m and in MP at 11.5 m (averaging 16.32 and 16.46%), respectively. Within the Halobacterota, ASVs associated with methanogens, such as Methanobacteria, Methanosarcinia, Methanomicrobia, Syntrophoarchaeia and Methanocellia, were most abundant at 6.0 m in FP (averaging 2.09%) and at 5.4 m in FCP (averaging 8.52%). While these samples originated in lake and coastal mixing zones, respectively, and were collected at similar depths, the sample from 5.4 m had 10 times higher salinity, and more than twice the abundance of carbon, nitrogen, and methane (Table S1). The presence of methane at 5.4 m is supported by presence of salt-tolerant methanogens, such as Methanobacteria, Methanocellia, Methanomicrobia [66]. Decrease in abundance of the methanogens in deeper MP samples may be connected to the increase of SO42-. Similar to our study, negative correlation between presence of SO42- and CH4 production in artificial sea water microcosms was reported [23]. Methanogenic archaea were not identified at 15.0 and 25.1 m in FP despite a high concentration of methane in those samples. This methane could therefore be explained by methane compression in deeper layers during epigenetic freezing from the top down [67].
2.3 Identification and Distribution of MAGs
Metagenomic binning from the co-assembly of the seven permafrost samples (two marine permafrost samples failed to yield metagenomes of good quality) resulted in a total of 60 MAGs with ≥ 50% completeness and ≤ 10% contamination, altogether contributing 0.24 to 59.10% of sequencing reads of the respective samples, with the highest proportion of reads in FP_25.1 and lowest in FP_2.95 (Fig. 3 A). The low percentage of reads recovered in MAGs from young permafrost samples may be attributed to higher biodiversity in that sample (Fig. 1 A). Because of different coverages of metagenomes, the co-assembly provided obvious benefits for capturing more of the diversity due to higher read depth, robust assembly, improved MAGs recovery [68], and facilitated comparison across permafrost samples. The reconstructed MAGs comprised phylogenetically diverse members from 2 archaeal and 15 bacterial phyla (Fig. 3 A and Table S2). Of the 60 recovered MAGs, 15 MAGs were retrieved from both boreholes, while 20 MAGs were only identified in AL1_15 and 25 MAGs were exclusively found in CH1_17. Clearly partitioned communities were observed between the two boreholes (Fig. 3 B). MAGs aligning with Asgardarchaeota, Bacteroidota, BMS3Abin14, Desulfobacterota, MBNT15, Myxococcota, Nitrospirota, Patescibacteria and Verrucomicrobiota were only identified in CH1_17, while members of phylum Firmicutes_A were exclusively detected in AL1_15.
The microbial communities also shifted with depth and age. In CH1_17, the number of recovered MAGs was highest in FCP_1.75 (n=22) and lowest in FCP_5.4 (n=4). The majority of the recovered populations from FCP were related to Chloroflexota_A at 1.75 m (40.53%, all quoted percentage values in the sections below refer to the relative abundance from the respective metagenome) and Actinobacteriota at 5.4 m (99.96%), whereas the recovered populations from MP at 16.9 m were mostly related to Chloroflexota and Gammaproteobacteria (34.93 and 29.21%, respectively). In addition, Acidobacteriota, Bacteroidota, Chloroflexota_A, Nitrospirota, Patescibacteria and Verrucomicrobiota MAGs were exclusively observed at 1.75 m, while BMS3Abin14, Desulfobacterota, MBNT15 and Myxococcota were only detected at 16.9 m. Lastly, archaeal MAGs aligning with Asgardarchaeota, which is suggested to be the closest prokaryotic relatives to eukaryotes [69, 70], were also only recovered from MP at 16.9 m. Another analysis of MAGs obtained from discrete depths of marine permafrost and annotated against TIGRFAM and COG [31] showed the presence of Asgardarchaeota, Bacteroidetes, Nitrospirae, and Deltaproteobacteria. Contemporary descendants of the microorganisms, which were exclusively discovered in CH1_17, were found to be highly metabolically flexible organisms that could adapt to resource variability by using different electron donors and acceptors or being involved in syntrophic interactions. For example, Myxococcota from anoxic aquatic environments being a strict anaerobe are capable of using fermentation, nitrate reduction, and dissimilarity sulfate reduction for energy acquisition [71]. The current discovery of MAGs for Desulfobacterota and MBNT15 in deeper MP_16.9 sample are in line with the previously obtained data that obligately anaerobic taxa such as sulfate reducers and candidate lineage MBNT15 thrive in more stable deeper marine sediments [72]. The 2 MAGs belonging to Asgardarchaeota were also discovered in the deeper marine sample. A recent review suggested that many Asgardarchaeaota are involved in syntrophic interactions [73], e.g., the syntrophic exchange of formate and hydrogen was shown between a Lokiarchaeon and a sulfate-reducing Deltaproteobacterium [74].
In freshwater permafrost from AL1_15, the metagenome from 25.1 m generated the highest number of MAGs (n=19), followed by 6.0, 2.95 and 15.0 m (n=18, 8 and 4, respectively). Even though the FP_25.1 metagenome size was approximately half that of the FP_2.95 metagenome, it generated the highest number of MAGs. This could be explained by low biodiversity in sample FP_25.1 based on the 16S rDNA amplicon analysis (Fig. 1 A). While poor MAG recovery from younger permafrost samples could be related to high diversity of the microbial community [75]. A high abundance of the recovered populations was aligned with Chloroflexota_A at 2.95 and 6.0 m (73.73 and 44.27%, respectively), whereas MAGs recovered from 15.0 and 25.1 m were dominated by Actinobacteriota (59.16%) and Gammaproteobacteria (89.57%), respectively. In addition, populations affiliated with Chloroflexota_A and Gemmatimonadota were only identified at 2.95 and 6.0 m, while Firmicutes_A MAGs were exclusively detected at 25.1 m.
Amplicon and metagenome approaches detected similar populations of bacteria; however, amplicon analysis showed their presence in more samples than was detected by metagenomics. The number of MAGs correlated positively with microbial diversity based on amplicon analyses in FCP samples (higher number of MAGs in FCP_1.75 at higher Alpha diversity), and correlated negatively in FP samples (higher number of MAGs in FP_25.1 at lower Alpha diversity). This discrepancy may be connected to biases during isolation of DNA from samples with different salinity, sequencing approaches, depth of sequencing, platform for sequencing technology and bioinformatic approaches [76].
2.4 Metabolic Potential in Permafrost
Recovered MAGs were screened for genes encoding hydrolysis, fermentation, respiration, CO2 and N2 fixation, motility, bacterial secretion, spore formation and stress resistance (Fig. 4-5 and Tables S3-S5) to identify their metabolic potential. Genes coding for certain functions that were not identified in MAGs or in many samples were also searched in un-binned, assembled metagenomic sequences and a summary of potential metabolic pathways identified in permafrost metagenomes presented in Fig. 6.
2.4.1 Aerobic Respiration
Despite the scarcity of oxygen (<0.07 mg O2 kg-1 at redox potential below +200 mV [77]), genes encoding the machinery to reduce oxygen were identified in 30 MAGs, including members of Acidobacteriota, Actinobacteriota, Alphaproteobacteria, Bacteroidota, Chloroflexota, Gammaproteobacteria, Gemmatimonadota, MBNT15 and Nitrospirota. Depending on the sample, 5.29-100% of the recovered populations contained genes coding for cytochrome c oxidase (Fig. 4).
These communities were most abundant at 15.0 and 25.1 m in FP (100 and 99.76%, respectively) and at 16.9 m in MP, respectively (35.69%). The high abundance of genes coding for cytochrome c oxidases could be explained by high identity of this enzyme in diverse organisms [78]. Cytochrome c oxidase catalyzes redox-driven proton pump that take part in generating the proton gradient in both prokaryotes and mitochondria that drives synthesis of ATP [78]. This highlighted the importance of energy acquisition through oxygen reduction in the permafrost populations, and accumulation of these enzymes in the deeper and older samples. In addition, MAGs containing genes encoding both low and high affinity cytochrome c oxidase had the highest relative abundance at 25.1 m in FP (92.49%) and at 16.9 m in MP (30.94%, Fig. 4), suggesting that the recovered populations from the deeper samples were capable of operating under different levels of O2 concentration. This is in line with previous cultivation studies showing that isolates from the Siberian permafrost grew well at atmospheric oxygen concentrations [10, 79].
2.4.2 Carbohydrate Hydrolysis
Chitin and plant-derived materials, such as cellulose and xylan, are the two most abundant types of polysaccharides in the ecosystem [80]. MAGs containing genes encoding cellulases and/or β-glucosidases, which are involved in cellulose degradation, were most abundant in older permafrost at 15.0 and 25.1 m depth in FP (100 and 92.63%, respectively) and at 16.9 m in MP (80.85%) (Fig. 4) with carbon content of 1.896, 2.663 and 0.911%, respectively. These populations were dominated by Actinobacteriota in FP_15.0, Gammaproteobacteria in FP_25.1, and Chloroflexota and Gammaproteobacteria in MP_16.9 (Fig. 5).
MAGs with the capacity of carrying out xylan degradation were identified in all the samples apart from FP_2.95 and FCP_5.4. Approximately 10.99-31.31% of the recovered communities contained xylanase and/or β-xylosidase genes, with the highest at 15.0 m in FP (17.29%) and 16.9 m in MP (31.31%). These potentially xylan degrading microorganisms were dominated by Actinobacteriota and Gammaproteobacteria, respectively. The higher abundance for cellulases over the protein-coding sequences for xylan breakdown shows that the Siberian permafrost populations have potential to degrade cellulose, the most commonly utilized polysaccharide. A recent study of Alaskan permafrost with ages of 19-33 kyr showed that sequences for enzymes targeting structural polysaccharides, such as xylan and cellulose, were less abundant than those targeting smaller molecular weight compounds [81]. However, the same study displayed that microbial community changed along the chronosequence from young permafrost community with potential to target hemicellulose, through increased potential to target starch, to the old permafrost microbial population having enzymes to target recalcitrant substrates like peptidoglycan and cellulose, that is in line with our findings. Populations encoding genes assigned to chitin degradation were most abundant at 2.95 and 15.0 m in FP (57.82 and 59.16%) and at 5.4 m in FCP (37.21%). MAG_99 and MAG_142, members of Actinobacteriota, were the dominating population for chitin degradation at 15.0 m in FP and at 5.4 m in FCP, respectively, whereas the dominant population at 2.95 m in FP belonged to one Chloroflexota_A MAG.
2.4.3 Fermentation
The potential for fermentation was identified in 23 MAGs, including members of an archaeal phylum, Asgardarchaeota. Microorganisms with the capability of producing acetate and/or lactate via fermentation were identified in all the samples. In FP sediments, fermentation-encoding microorganisms contributed a high abundance of the recovered communities at 15.0 and 25.1 m (100 and 91.76%, respectively), while approximately 5.29 and 12.02% of the recovered populations contained genes for fermentation at 2.95 and 6.0 m, respectively (Fig. 4). The most abundant MAGs encoding fermentation pathways were Gammaproteobacteria at 2.95, 6.0 and 25.1 m, whereas these populations primarily belonged to Actinobacteriota at 15.0 m (Fig. 5). Formate fermentation genes were only observed at 6.0 m in FP, whereas butanoate fermentation genes were exclusively identified at 25.1 m. The recent metagenomics study of the alluvial silty loams continuously frozen for 0.01-1.1 Myr showed that genes involved in the synthesis of formate, acetate, and butyrate were more numerically abundant in the older sediments [15]. In addition, MAGs containing genes indicating the capability for fermentative production of ethanol were identified at 6.0 and 25.1 m in FP and at 16.9 m in MP. In FCP and MP, genes encoding fermentation were most enriched in MP_16.9 (71.46%), followed by FCP_5.4 and FCP_1.75 (37.21 and 12.79%, respectively). Chloroflexota and Gammaproteobacteria MAGs appeared to be the most abundant groups with pyruvate metabolic genes at 16.9 m, while the most abundant fermentation encoding microorganisms were Bacteroidota at 1.75 m and Actinobacteriota at 5.4 m. Populations containing genes for formate and butanoate fermentation pathways were identified at 1.75 m in FCP and 16.9 m in MP. Analyses of neighboring samples showed presence of low–molecular-weight organic acids, such as formate, acetate, and propionate, at all depths with the highest concentrations of acetate (36.2 ± 3.3 μg g-1) and formate (12.0 ± 0.7 μg g-1) in FCP at depth of 5.8 m [31]. Lastly, one Myxococcota MAG containing genes encoding pyruvate fermentation to propanoate was exclusively identified in MP at 16.9 m. Overall, our data suggested that fermentation contributed significantly to anaerobic degradation of soil organic carbon in the deeper and older permafrost (Fig. 4 and Table S4), which is in line with previous studies emphasizing its importance in permafrost [82, 83].
2.4.4 Potential Metabolic Interaction Between Populations
Short-chain fatty acids, such as alcohols, hydrogen and CO2 converted by fermentative populations can subsequently be used by other microorganisms. Genes involved in ethanol degradation were detected in all the samples except for FCP_5.4. They were particularly abundant at 25.1 m in FP (82.63%) and at 16.9 m in MP (53.68%, Fig. 4). These populations were dominated by Gammaproteobacteria in FP_25.1 and MP_16.9 (Fig. 5). Hydrogen is an important electron donor for microorganisms in anaerobic conditions. MAGs containing homologs for anaerobic hydrogen oxidation were only identified in FP_25.1, FCP_1.75 and MP_16.9 (0.24, 7.43 and 3.15%, respectively). These populations belonged to Bacteroidota (n=1), Desulfobacterota (n=1) and Firmicutes_A (n=1). However, the NADP-dependent hydrogenase catalyzes reversible oxidation of H2, therefore, the detected hydrogenase could also be involved in hydrogen production.
Five MAGs containing genes coding for form I ribulose-bisphosphate carboxylase (RuBisCO), the key enzyme in CO2 fixation via the Calvin-Benson-Bassham (CBB) cycle and the most abundant form in eukaryotes and bacteria [84], were only identified in FP (Fig. S2 and Table S2). Of these populations, two Alphaproteobacteria MAGs and two Gammaproteobacteria MAGs contained both cbbL and cbbS genes for form I RuBisCO, while the Chloroflexota_A MAG only included the cbbL gene. These microorganisms were most abundant in FP_6.0 (22.84%) and were mainly encoded by Chloroflexota_A (Fig. 4 and 5). In addition, three archaeal MAGs, from Asgardarchaeota and Crenarchaeota, were suggested to hold cbbL genes encoding form III RuBisCO and were only identified in MP at 16.9 m (10.59%). This form is known to be present in many archaea (Jaffe et al., 2019) and reported to be involved in the incorporation of CO2 into ribulose-1,5-bisphosphate (RuBP) from nucleotides like adenosine monophosphate (AMP) [85, 86]. Genes coding for form IV RuBisCO were identified in 3 MAGs, including the two MAGs also containing genes for form I and one archaeal MAG also with genes for form III. Notably, form IV, which is often referred to as a RuBisCO-like protein, appears to be involved in sulfur metabolism, methionine salvage pathway, and D-apiose catabolism [87, 88], rather than the CBB cycle. The microorganisms carrying this form were found at 2.95, 6.0 and 25.1 m in FP and 16.9 m in MP. MAG_40, closely related to Nitrospirota, encoded genes assigned to CO2 fixation via incomplete reductive tricarboxylic acid (TCA) cycle, and was exclusively identified in FCP at 1.75 m (3.38%). Lastly, key genes encoding both anaerobic carbon-monoxide dehydrogenase and acetyl-CoA synthase for the Wood-Ljungdahl pathway were found in 4 MAGs. The Actinobacteriota MAG was only identified in FCP at 1.75 m (4.19%), whereas the Desulfobacterota and Chloroflexota MAGs were detected in MP at 16.9 m (33.64%). Overall, the higher abundance of CO2 fixation potential at 16.9 m in MP (33.64%) suggested that autotrophy was more widespread when this permafrost layer formed relative to the others.
2.4.5 Methane Metabolism
Methane is a potential electron donor for anaerobic respiration, coupled to sulfate, iron, manganese, nitrate, nitrite reduction or denitrification [89-95]. It is also a potent greenhouse gas that contributes to global warming. We screened genes encoding methanogenesis, anaerobic and aerobic methane oxidation from the recovered MAGs as well as un-binned metagenomic sequences. None of the MAGs were suggested to carry genes involved in anaerobic methane oxidation (AOM) or methanogenesis. Within the un-binned metagenome data, the key genes involved in methanogenesis (based upon the presence of the mcrBCDG genes) were identified in all samples except FP_15.0 and FCP_1.75 (Fig. S3). These genes were most enriched in FCP_5.4 (367.46 TPM), corresponding to the 16S rRNA amplicon data showing the highest relative abundance of methanogenic taxa, and correlated with a presence of methane in layer between 4.8 and 6.4 m with the highest concentration of methane 76 mmol kg-1 detected near this depth at 5.6 m. The concurrent presence of genes involved in methanogenesis, ASVs associated with methanogens and methane detected by the static headspace method [67] suggests that methane at 5.6 m depth has a biological origin and methane production may happen in situ at FCP conditions. Overall, the capacity for methanogenesis appeared to be relatively low in the remaining samples (0.006-7.81 TPM) which is in line with the previous studies showing that methanogenesis was extremely limited in intact permafrost [17, 96]. Despite high levels of methane 157.1 and 167.2 mmol kg-1 detected in FP at 15.0 and 25.1 m, respectively [67], the key genes for methanogenesis were not identified at 15.0 m and were negligible at 25.1 m (0.006 TPM). A previous study, which sampled permafrost from the same location at three depths (1.4, 11.8, and 24.8 m), also showed low abundance of methanogens [15]. The low abundance of methanogenesis genes in samples with high concentrations of methane that has biogenic origin (-84.9 ± 0.5‰ for d13CCH4) may suggest that this methane has either surficial or deep-sediment origin and likely accumulated in lithological traps earlier during permafrost formation [67] or represents a consequence of methanogenesis in deeper permafrost layers, similar to observations from organic-rich Antarctic marine sediments where the methanogenesis genes are in low abundance despite the observation of biogenic methane [97]. Low abundance of methanogens in metagenomes from permafrost samples with biogenic methane may also be attributed to biases associated with sample size, DNA extraction, sequencing technology, depth of sequencing, sequence assembly, annotation, and database used for identification [76]. Analysis of replicate metagenomes and ultra-deep sequencing of additional samples from methane containing permafrost layers would likely result in more prominent detection of methanogens.
In contrast to the rare occurrence of anaerobic methane oxidation in permafrost, aerobic methane oxidation typically attenuates methane release [98, 99]. However, MAG_40, encoding the capacity to oxidize methane aerobically, was only found in FCP at 1.75 m (most closely to Nitrospirota, 3.38%, Fig. 4). Within the un-binned metagenomic sequences, genes involved in aerobic methane oxidation (based on the presence of the pmoABC-amoABC) were identified at 2.95 and 6.0 m in FP (1.38 and 4.06 TPM, respectively, Fig. S3) and at all depths in FCP and MP, with highest abundance in FCP_1.75 (15.05 TPM). The generally low abundance of genes for aerobic methane oxidation was not surprising considering the lack of oxygen in permafrost. In agreement with the high abundance of methanogenesis genes at 5.4 m in FCP, genes for aerobic methane oxidation were lowest at this depth (0.77 TPM), indicating the strictly anaerobic microenvironments at this depth. It should be noted that the identified genes encoding aerobic methane oxidation were also suggested to co-oxidize ammonia aerobically [100].
2.4.6 Nitrogen and Sulfur Metabolism
The potential ability to fix nitrogen was not identified in the recovered MAGs. Within the un-binned metagenome data, nitrogenase genes were identified in all samples with the exception of FP_15.0. An overall low abundance of form I and III nifH, nifD and nifK genes was observed across the samples (totaling 0.12-16.95 TPM, Fig. S3 and S4). Genes for the enzymes needed to utilize nitrate or nitrite as a terminal electron acceptor in the process of nitrification were detected in 26 MAGs. Populations that were suggested to reduce nitrate to either nitrite or ammonium as the end products were most abundant at 25.1 m in FP (62.84%) and at 16.9 m in MP (48.46%) that correspond to higher concentration of ammonium in deeper sediments (~300 ppm NH4- g-1 at depth of 12-24 m in FP and 110-152 ppm NH4- g-1 in MP [59]). These populations were dominated by Gammaproteobacteria in both samples. Genes encoding the capacity to reduce nitrite to ammonium were also most enriched at 25.1 m in FP (34.13%) and were identified at 1.75 m in FCP with a relatively lower abundance (5.45%). In addition to the dissimilatory nitrate reduction pathway, three MAGs also contained a complete or near complete set of genes coding for denitrification. These populations represented a small fraction of the recovered communities, which is present in all the samples (0.03-4.74%), except for FP_15.0 and FP_25.1 (40.84 and 58.19%, respectively). The presence of nitrite and nitrate in different FP layers and in deeper MP layers supports a potential of the communities to the denitrification process. No MAGs were suggested to carry out denitrification in FCP_5.4 and nitrate was not detected in FCP layer. Overall, our data suggest that nitrogen metabolism plays an important role in the older permafrost, likely due to nitrate being the most energetically favorable electron acceptor in the absence of oxygen. The generally higher abundance of populations involved in nitrate reduction is consistent with previous reports of their dominance under reducing, high-carbon and low-nitrate conditions [101, 102].
Dissimilatory sulfate reduction by anaerobic microorganisms is a predominant pathway of organic material mineralization in marine sediments [103]. However, none of the MAGs contained the key dsrA and/or dsrB genes for dissimilatory sulfate reduction, suggesting the rare utilization of this pathway by permafrost populations in the studied environments. Populations containing genes for enzymes needed to reduce sulfur via polysulfide were exclusively identified in MP at 16.9 m (51.08%). The presence of hydrogen sulfide in MP samples was identified by the specific odor during core extraction in the field. These populations were encoded by Asgardarchaeota, Crenarchaeota and Chloroflexota. Genes encoding thiosulfate oxidation via the Sox pathway were only identified in FP at 15.0 and 25.1 m (0.34 and 6.35%, respectively). Populations containing these genes exclusively belonged to Alpha- and Gammaproteobacteria, where the Alphaproteobacteria MAGs were only identified in FP_25.1. In agreement with the genomic analysis, an overall low abundance of the dsrAB genes coding for sulfate reduction were detected at 2.95 and 6.0 m in FP, 1.75 and 5.4 m in FCP and 16.9 m in MP (1.57-18.68 TPM) within the un-binned metagenome data. Nevertheless, the possibility of the sulfate reduction was supported by the presence of sulfate in FP (4.4-8.8 mmol kg-1, except sample FP_25.1 where sulfate was ~10 times lower), FCP (1.7-2.7 mmol kg-1), and MP (8.1-36.8 mmol kg-1) samples. The content of sulfate in all studied samples was <0.2% by mass what is considered low sulfate soils [104]. The previous study also discovered the presence of sulfate reduction in marine and deep freshwater permafrost sediments with S2- in concentration of 0.12-0.22 and 0.12-0.35 g kg-1 wet soil, respectively [27]. The phylum Desulfobacterota that comprises sulfate-, sulfur-, and ferric iron-reducing bacteria was solely detected in MP at 16.9 m. In addition, genes encoding the use of polysulfide sulfur were more abundant in MP within the un-binned metagenome data (Fig. S3), suggesting that anaerobic degradation of organic carbon coupled to sulfur reduction was more common in MP. Of the aforementioned genomes that encoded the capacity for fermentation and/or anaerobic respiration, a high abundance of the recovered communities also contained genes encoding the machinery to reduce oxygen at 15.0 and 25.1 m in FP (100 and 96.96%, respectively). Taking these together, we suggest that the recovered communities from the older samples of FP are likely to be facultatively anaerobes, whereas the recovered populations from 5.4 m in FCP to 16.9 m in MP are more likely to be strict anaerobes at the time of freezing (Table S4).
2.4.7 Potential Adaptation in Permafrost
Biofilm formation is considered to be a survival strategy to enable adaptation of microorganisms to extreme environments [105], i.e., bacteria were tightly associated with soil particles in Siberian permafrost [106]. In addition, Psychrobacter arcticus strain 273-4 from Siberian permafrost showed the capability to form biofilms [107]. Genes coding for both surface attachment through flagellar (including chemotaxis) and/or type IV pili and extracellular polymeric substance (EPS) secretion via type II secretion system (T2SS) which are potentially involved in biofilm formation were identified across seven MAGs representing Gammaproteobacteria, MBNT15 and Myxococcota. Biofilm-forming Gammaproteobacteria were most abundant at 25.1 m in FP (73.10%) and decreased in abundance toward the surface (Fig. 4). In FCP and MP, the highest abundance of the biofilm-forming microorganisms was detected in MP_16.9 (33.07%). None of the MAGs from FCP_5.4 contained genes encoding both surface attachment and EPS secretion. Of the 7 MAGs, one Gammaproteobacteria MAG also included genes for type IV secretion system (T4SS) which encodes conjugation machinery and DNA release and uptake systems [108]. Another three Gammaproteobacteria MAGs, MAG_24, MAG_35 and MAG_67 also contained genes coding for type VI secretion system (T6SS, delivering toxins into eukaryotic and prokaryotic cells, [109]), whereas genes encoding type III secretion system (T3SS, injecting effector proteins into eukaryotic cells, [110]) were also detected in MAG_35. In addition, three MAGs included genes for flagellar and/or type IV pili assisted motility and T6SS, where one MAG also contained genes encoding T4SS. These microorganisms were exclusively identified in FP at 25.1 m (3.80%). MAGs containing genes for flagellar or type IV pili mediated motility but lacking genes encoding secretion systems were also identified at 6.0 and 25.1 m in FP (0.12 and 4.89%, respectively), 1.75 m in FCP (8.81%) and 16.9 m in MP (5.12%). In addition, approximately 0.02-14.24% of the recovered MAGs contained genes assigned to type I secretion system (T1SS), T2SS or T6SS at 2.95, 6.0 and 25.1 m in FP and 1.75 m in FCP, with the highest abundance observed in FP_25.1. The high abundance of recovered MAGs that contained genes encoding both surface attachment and bacterial secretion system in FP_25.1 and MP_16.9 indicated that biofilm formation might be an important survival strategy for microbes in the older and deeper perennially frozen sediments, providing exchange of molecules and ions between live microorganisms and liquid brine veins surrounding cells and soil particles [5, 16]. This is generally in agreement with the un-binned metagenome data analysis where genes involved in chemotaxis, flagellar assembly, type IV pili and bacterial secretion systems (type I-III and VI) were more abundant in the oldest permafrost sample (FP_25.1), while genes encoding T4SS were most abundant in MP_16.9 (Fig. S5). The higher abundance of T4SS in marine permafrost could indicate that horizontal gene transfer may play a role in allowing microorganisms to adapt to changes in their environment. Overall, our data support the findings from a previous metagenome study of Alaska permafrost showing that chemotaxis and bacterial secretion system pathways were enriched in older permafrost up to 33 kyr [60], but also pointed to the presence of bacterial populations that were potentially involved in biofilm formation.
Sporulation is a widely utilized strategy for microorganisms to survive in extreme environmental conditions [111]. MAG_104, most closely related to Firmicutes_A, was suggested to carry out multiple stages of spore formation and was exclusively found in FP at 25.1 m (0.24%, Fig. 4 and 5). The low relative abundance of spore-forming populations in the older permafrost contradicted a previous study from the same location showing that the older layers in freshwater permafrost was dominated by spore-forming bacteria [15], but was consistent with other studies of Siberian and Antarctica permafrost [112, 113] as well as permafrost from Svalbard, Norway [114]. Our un-binned metagenome data analysis also showed that metagenomes from 15.0 and 25.1 m had a lower abundance of these genes compared to the top layer samples in FP (Fig. S5), suggesting that spores were not the most prominent survival strategy in the older permafrost. The discrepancy between this study and the previous study [15] of freshwater permafrost suggested that the permafrost sampled at this site may be highly heterogeneous, providing large amounts of micro niches with different environmental characteristics, even from the same borehole.
Since permafrost is frozen, genes encoding cold shock protein (CSPs) were prevalent across the recovered MAGs (38 of 60 MAGs). CSPs are a loosely defined group of DNA binding proteins that are commonly found in cold-adapted microorganisms and were most enriched at 15.0 m in FP (100%), followed by 25.1, 6.0 and 2.95 m (99.76, 67.74 and 62.95%, respectively, Fig. 4). In FCP and MP, CSP genes were highest in FCP_5.4 (100%, respectively) and lowest in FCP_1.75 m (25.46%). The overall high abundance of CSP genes across the samples indicates that microbial communities are well-adapted to stresses associated with freezing temperatures, nutrient starvation and growth deprivation [115].