In April 2018, soil core samples were drilled from completely frozen active-layer permafrost in Ny Ålesund, Svalbard (Figure 1). These two drill sites were near Bayelva Monitoring Site (Boike et al., 2018). Bayelva Permafrost Site 1 (BPF1) (N 78° 55.237′ E 011° 50.495′) is closest to the Bayelva Monitoring Site, 21 m above sea level. BPF2 (∼84 m from BPF1, N 78° 55.261′ E 011° 50.294′) is near a summer glacial melt riverbank at an elevation of 20 m above sea level (Figure 1A). Three boreholes were drilled at each of the two sites with a SIPRE auger drill (NSF, United States) with ∼0.6 m snow cover that was shoveled away prior to drilling (Figures 1B,C). Depths of core samples were limited by the ability to recover intact material. Core depths for BPF1 were 21, 20, and 58 cm, and core depths from BPF2 were 30, 21, and 21 cm below the surface (Figure 1D). Freshly cored samples were removed from the drill and kept inside presterilized polycarbonate core liners (Jon’s Machine Shop, United States) used during drilling. Core liners were capped and stored inside a sterile lined cooler to maintain frozen temperature.

Cores were removed from the core liner and sliced into 2-cm-depth intervals using a sterile geological sampling hammer and chisel at the King’s Bay AS Marine Laboratory (Ny Ålesund, Svalbard). A portion of the sample was weighed and then dried in a 60°C oven for 24 h to determine gravimetric water content. Bulk density of the cores was estimated by measuring volume and dry mass of one intact 2.8-cm core puck with a diameter of 8 cm collected from BPF1; the resulting uniform bulk density was applied to all core samples for bulk density.

To study colony variation and CFU/ml, we tested various soil dilutions and three different agar types on the BPF1 0–2 cm. Following previously published methods (Vishnivetskaya et al., 2000), we tested R2A, TSA, and 1/2-strength TSA with a Master Soil Mixture (MSM) made from 10 g soil and 100 ml of 1 × phosphate-buffered saline (PBS, pH 7.2). The following four final soil suspensions ratios were plated: MSM:1XPBS; 0.1:0.9 (1 ml plated), 0.5:0.5 (1 ml plated), 1:0 (1 ml plated), and 3:0 (3 ml plated) for each of the three media types, in triplicate. CFUs/ml were counted after growth at 4°C for 3 weeks. Full-strength TSA yielded a lawn of colonies after just a few days, and 1/2 TSA yielded only a few colonies (Supplementary Figure 1). R2A agar was chosen for our study, since it had the greatest colony diversity and allowed for slower growing colonies to form. The MSM:PBS dilutions listed above were made for eight depth intervals between BPF1 and BPF2 core sites (Supplementary Figure 2). From BPF1, the depth intervals were 0–12 cm, 12–24 cm, 24–36 cm, 36–48 cm, and 48–58 cm. From BPF2, the depth intervals were 0–12 cm, 12–20 cm, and 20–30 cm. An MSM from these eight depth intervals were plated with the same soil suspension scheme and incubated at 4°C for 3 weeks. After CFU/ml and colony variants were measured on the experimental plates, we streaked for isolation on R2A. Of the hundreds of isolates obtained on plates, 10 were selected for continued analysis based on their morphological diversity. Isolates were grown on R2A for 4 weeks at 4°C, after choosing R2A over TSA and 1/2-strength TSA because the diluted medium allowed for isolation of oligotrophic microbes (Supplementary Figures 1, 2). The exact depth intervals they came from are listed in Table 1.

Total carbon, nitrogen, carbon isotopic signature (δ¹³C), and nitrogen isotopic signature (δ¹⁵N) were determined on completely dried soil that was ground using a mortar and pestle into a fine powder. Large stones were removed, and the final particle size of the fine powder was not measured before analysis. To quantify the organic fraction of carbon, 1 ml of 1 N HCl was added to 5.0 g of soil and oven dried at 45°C for 56 h to volatilize inorganic carbon (Harris et al., 2001). The δ¹³C and percentage of inorganic carbon were calculated through mass balance for samples with > 0.7% inorganic carbon (Komada et al., 2008). Total carbon, nitrogen, δ¹³C, and δ¹⁵N were determined on finely ground soil samples (30 mg for BPF1 and 40 mg for BPF2) using a Costech ECS4010 Elemental Analyzer coupled to a Thermo-Finnigan Delta+XL mass spectrometer via a Thermo-Finnigan Conflo III device. Helium was used as the carrier gas, and the oxidation furnace was operated at 1,050°C and the reduction furnace at 650°C. These measurements were performed at the Stable Isotope Laboratory at the University of Tennessee, Knoxville, United States.

Extractable dissolved organic carbon (DOC), phosphate, and inorganic nitrogen were measured on each of the 2-cm intervals of core samples (Supplementary Figure 3). Extracts were prepared by combining 5.0 g soil with 20 ml of 0.5 M potassium sulfate and shaking at room temperature for 4 h. Samples were then filtered through a Whatman GF/B glass microfiber filter (1.0 μm pore size) using a vacuum extraction manifold. Filtrate was collected and frozen at –20°C for 12 h. DOC was quantified by reacting extracts with a 0.42 M potassium persulfate (Doyle et al., 2004) to oxidize soil organic carbon to CO₂. A series of potassium hydrogen phthalate standards were included in each analysis to use for calculation of persulfate-oxidized organic carbon. Samples and standards were reacted overnight (80°C) in sealed glass vials with rubber septa for headspace gas sampling, which was conducted after samples had cooled to room temperature using an infrared CO₂ gas analyzer. Sample extracts were carried through three colorimetric assays to measure extractable phosphate (PO₄^3–), nitrate (NO₃^–), and ammonium (NH₄⁺) concentrations. The sum of nitrate and ammonium concentrations is used to calculate total inorganic nitrogen, and the sum of organic nitrogen and inorganic nitrogen is equal to total nitrogen. Inorganic phosphate was measured using the Malachite Green assay (D’Angelo et al., 2001). Nitrate was determined using a vanadium (III) chloride reagent (Doane and Horwáth, 2003), and ammonium was quantified using the Berthelot reaction (Rhine et al., 1998). All nitrogen and phosphorous assays were conducted using protocols modified for a 96-well microplate reader (Synergy H1 Hybrid Reader, Biotek Inc., Winooski, VT, United States).

Additional measurements for electrical conductivity, pH, and labile carbon were performed on a separate core sample from each site that was divided as follows: BPF1, 0–12 cm, 12–24 cm, 24–36 cm, 36–48 cm, and 48–58 cm; BPF2, 0–12 cm, 12–20 cm, and 20–30 cm (Supplementary Figure 3). Electrical conductivity and pH were measured on these samples using calibrated bench-top meters (15 g soil in 45 ml deionized water). Additionally, pH, labile carbon, and electrical conductivity measurements were performed on eight soil sample increments (BPF1, 0–12 cm, 12–24 cm, 24–36 cm, 36–48 cm, 48–58 cm; BPF2, 0–12 cm, 12–20 cm, and 20–30 cm) using calibrated bench-top meters (15 g dry soil in 45 ml deionized water). Permanganate oxidizable carbon (POXC), used to analyze total labile carbon, was extracted and quantified according to the method of Weil et al. (2003). Briefly, 2.5 g was reacted with 20 ml 0.02 M potassium permanganate (KMnO₄) + 0.1 M calcium chloride solution by shaking at 120 rpm for 2 min. After settling, extracts were diluted 1:100 with milliQ water then measured for absorbance at 550 nm wavelength using a microplate spectrophotometer. Standards of a known concentration of KMnO₄ were included with each plate and used to determine the moles of KMnO₄ oxidized upon reaction with soil organic carbon. Assuming 9,000 mg carbon oxidized per mole KMnO₄ (Weil et al., 2003), the amount of POXC was corrected for soil water content and reported in μg POXC g dry soil^–1.

For DNA extraction from isolates, a pellet was formed with low-speed centrifuging (7,000 × g) of isolates that had been grown on R2A at 4°C for 4 weeks, and DNA was extracted with a Qiagen DNA Power Soil Kit (Qiagen, Germany). To amplify the 16S rRNA gene of the isolates, a PCR Master Mix was prepared as follows: 0.25 μl of Speedstar Taq polymerase (TaKaRa Bio, United States), 4 μl of 2.5 mM deoxyribonucleotide triphosphates (dNTPs), 5 μl of 10 × Fast Buffer 1 (TaKaRa Bio, United States), 10 μl of 27F primer (5′-AGAGTTTGATYMTGGCTCAG-3′) 10 μl of 1492R primers (5′-TACGGYTACCTTGTTACACTT-3′) (Frank et al., 2008) (Eurofins Genomics, United States), 29.25 μl of dH₂O, and 2 μl of DNA, totaling 200 μl of volume per sample. Samples underwent PCR thermocycling in a BioRad T100 ThermoCycler (BioRad, United States) for 95°C for 1 min, 95°C for 5 s, and 65°C for 20 s. The last two steps were repeated 34 times. After the thermocycler, samples were dyed with 6 × TriTrack DNA Loading Dye (Thermo Fisher Scientific, United States). Visualization of the PCR product was compared to a GeneRuler 1 kb Plus DNA Ladder (Thermo Fisher Scientific, United States). The samples were then placed into a 1.5% agarose gel with Midori green DNA stain at 90 V for 45 min with BioRad PowerPac Basic (BioRad, United States) for PCR product verification.

Amplified 16S rRNA gene (27F, 1492R) PCR product was cleaned with Qiagen PCR Clean Up Kit (Qiagen, Germany) and Sanger sequenced at the Sequencing Core Facility at the University of Tennessee, Knoxville. Sequences were viewed, and forward and reverse reads were combined from the chromatogram in 4Peaks (v.1.7.2), and DECIPHER v2.17.1 (Wright et al., 2012) was used to check for chimeras. The combined sequences were analyzed in nucleotide BLAST (v2.11.0), and the closest related organisms were downloaded for comparison. Output sequences were classified with SILVA Sina (v1.2.11). SINA Alignment (v1.2.11) was also used to compare these isolates to 16S rRNA genes cataloged in the SSU database and to make a RAxML tree. All sequenced isolates have been deposited on National Center for Biotechnology Information (NCBI) GenBank accession numbers MZ773212–MZ773221.

DNA extractions from the 10 cultured isolates were sequenced with an Illumina MiSeq V3, 600 cycles (2 × 300) at The University of Tennessee, Knoxville Center for Environmental Biotechnology. Whole genome sequences were retrieved from Illumina BaseSpace and were assembled with SPAdes v3.13.0 (Bankevich et al., 2012) on KBase (link in “Data Availability” section) (Arkin et al., 2018). Prokka v. 1.14.6 was used for annotations (Seemann, 2014). All whole genomes are available on NCBI accession number PRJNA649544.

DNA was extracted from the longest core sample from each site (BPF1, 58 cm; BPF2, 30 cm) using the Qiagen DNA PowerSoil DNeasy DNA Extraction kit (Qiagen, Germany) with 0.5 g of starting material at King’s Bay AS Marine Laboratory, no more than 3 h after removal from the ground. DNA was quantified on each 2 cm soil sample using a Qubit™ 4 Fluorometer (Thermo Fisher Scientific, United States). To get a total of 10 ng/μl for each metagenome, BPF1 samples were pooled into two groups (0–30 cm and 30–58 cm), and all BPF2 samples were pooled (0–30 cm). These three samples were sequenced on an Illumina MiSeq with a V3 flow cell, using a 600 Nextera cycle kit with 275–300 bases paired-end reads. Data were downloaded from Illumina’s BaseSpace platform and analyzed with KBase (Arkin et al., 2018) with default program settings. Forward and reverse fastq reads were assembled with 98% of the two reads surviving, and adapters were trimmed with Trimmomatic v0.36 (Bolger et al., 2014); then, all three metagenomes were assembled separately with MetaSPAdes v3.14.1 (Nurk et al., 2017) and annotated with Prokka v. 1.14.6 (Seemann, 2014). Metagenome libraries of soil samples and whole genomes of isolates were compared with read mapping in terms of “reads per kilobase per read library” through Bowtie2 (Langmead and Salzberg, 2012) and in-house python scripts.

The same protocol was used for cell counts of the bulk soil intervals and the cultures. The soil intervals used a dilution of 1:20 (1 × PBS: Soil suspension); the cultures were diluted based on visual opacity and spectrophotometer reading but did not exceed a 1:40 dilution (1 × PBS:culture broth) (see Supplementary Data File 1). 5 × SYBR gold stain was added and filtered on a vacuum Hoefer box on 0.2-μm Millipore round filters. Filters were adhered to microscope slides with Vecta Sheild©, and a cover slip was applied. A microscope used was a Zeiss Axio Imager M2 Epifluorescence Microscope (Oberkochen, Germany). Cells were counted in 30 random fields of view at 10 × magnification in the singular grid hemocytometer for eyepiece PL 10 × /23 in 23 mm × 23 mm. Total cell counts were calculated by the following equation:

where xcells¯ is the average of the cells counted, sample filtered is the total sample used in the dilution, A_filter is the area of the Millipore 0.2 μm filter used, A_grid is the area of the hemocytometer grid, and dilution factor is the initial dilution factor of the sample:1 × PBS.

The organic carbon of BPF1 ranged from 1 to 3.5% (Figure 2A), while inorganic carbon ranged between 0 and 0.5% (except for 1 point at 1.5%). BPF2 had a slightly decreasing trend with depth for organic carbon (2.0–1.0%) and slight increase in inorganic carbon from ∼0 to 1.5% with depth (Figure 2B and Supplementary Data File 1). Total nitrogen decreased with depth in BPF2 and ranged from 0.05 to 0.10%. Total nitrogen was nearly twice as high in BPF1, ranging from 0.05 to 0.25% (Figure 2C). The δ¹³C of organic carbon ranged from –26 to –25‰ with the exception of two points (Figure 2D). The inorganic δ¹³C values for BPF1 varied between –30% and –20‰, while the inorganic δ¹³C of BPF2 contains more ¹³C with depth (–20 to ∼10‰) (Figure 2E). δ¹⁵N in BPF1 and BPF2 decreased with depth (2.5–1.5‰) in the top 30 cm only (Figure 2F). Carbon to nitrogen ratios (C/N) in BPF1 ranged from ∼13 to 18, and those of BPF2 increased with depth from ∼19 to 50 (Figure 2G). The water content per weight of the samples decreased with depth at both sites, except for BPF1 30–32 cm, which was mostly ice (Figure 2I). Dissolved organic carbon (DOC) was similar in both cores, ranging from 0.3 to 3.2 mg/g_{dry soil}. Inorganic nitrogen (sum of the measured NH₄⁺ and NO₃^–) ranged from 0.05 to 0.2 μg/g_{dry soil} and was similar between the two cores (Figure 3). BPF2 had a lower concentration of PO₄^3– at each depth compared to BPF1. However, both had negative values that were below the limit of detection of 0.0015 μg/g_{dry soil}. NH₄⁺ was highest at the surface (0.07 μg/g_{dry soil} for BPF1 and 0.18 μg/g_{dry soil} for BPF2) and consistently close to 0.05 μg/g_{dry soil} for both cores below the surface. NO₃^– was similar for both cores, varying from 0 to 0.04 μg/g_{dry soil} (Figure 3). The measured electrical conductivity was < 0.05 mS in BPF1 and ranged from 0.025 to 0.17 mS in BPF2 (Supplementary Figure 3). The pH range of BPF1 and BPF2 were both between 7.4 and 8.1 (Supplementary Figure 3). Labile carbon (mg POXC/g_{dry soil}) was lower in BPF2 (200–250), while BPF1 had greater POXC and higher variability with depth (320–825) (Supplementary Figure 3).

Ten cultures were chosen from the plated 0.5 ml soil suspension on R2A based on preliminary culture tests (see section “Materials and Methods,” Supplementary Figures 1, 2). Eight cultured isolates originated from BPF1 soil (Table 1) and two from BPF2 soil. All were Gram-negative rods of Pseudomonas sp. (Figure 4 and Supplementary Figure 6). Isolates of Pseudomonas sp. have been previously found in soil from Thuringian Basin, Council Alaska, Livingston Island Antarctica, and other Antarctica soils (Table 1 and references therein; Figure 4). The 16S rRNA gene sequence of each isolate was ≥ 99% similar to a previously cultured 16S rRNA genes on NCBI (Figure 4 and Table 1).

Despite being < 100 m apart, BPF1 and BPF2 differed in the δ¹³C signatures of inorganic carbon and percentages of inorganic carbon, organic carbon, and total nitrogen. The organic carbon δ¹³C of both cores showed signatures originating from plant material (–25 to –35‰) (Mu et al., 2014; Figure 2D). At BPF1, the δ¹³C of inorganic carbon was similar to the δ¹³C of organic carbon at all depths. This likely indicates that the inorganic carbon was produced from organic matter through microbial degradation, since heterotrophy expresses little to no kinetic isotopic effects. In BPF2, the inorganic carbon was ¹³C-enriched ranging from -20 to 14‰ with depth. Three possibilities to explain this are the following: (1) inorganic carbon becomes ¹³C-enriched due to microbial carbon fixation, which can leave behind ¹³C-enriched inorganic carbon due to kinetic fractionation (Boschker and Middelburg, 2002); (2) a large source of ¹³C-enriched inorganic carbon diffuses upward in the core, mixing with ¹³C-depleted heterotrophically produced inorganic carbon as it diffuses downward; and (3) both processes are occurring. Evidence that less microbial heterotrophy may be occurring in BPF2, which would allow autotrophy to have a greater effect on the δ¹³C values for inorganic carbon, comes from the fact that there are lower amounts of labile organic matter, compared to BPF1. C/N ratios can indicate how labile carbon compounds are for microbial activity in the soil, with lower ratios (13–18) indicating a greater availability for metabolic use (Sinsabaugh et al., 2009). BPF1 had significantly more organic carbon (p < 0.015) differences by comparable depths between sites and a statistically lower C/N ratio (p < 10^–4), largely driven by the statistically higher nitrogen content (p < 10^–5), suggesting that there may be greater activity of microbial carbon degradation in BPF1 due to higher carbon lability (Figures 2A,G). POXC values, a proxy for organic matter lability (Weil et al., 2003), were also higher in BPF1 (582 mg POXC/g_{dry soil} ± 109 vs. 225 mg POXC/g_{dry soil} ± 31 for BPF2) (Supplementary Figure 3). Microbial biomass was also greater in BPF1, shown by higher DNA concentrations (Figure 2H, 87% of samples at the same depth were higher in BPF1) and cell counts in the comparable soil increment depths (∼80,000 cells/g soil vs. ∼26,000 cells/g soil, Supplementary Figure 5B and Supplementary Data File 1) than BPF2. However, carbonates could have come from the wide range of carbonate deposits that are found in Svalbard (Koevoets et al., 2016). These could contribute to the proglacial active layer after glacial scouring and would be likely to have values ¹³C-enriched by at least 20‰ relative to organic matter (Kraimer and Monger, 2009), so the possibility of the δ¹³C trends in inorganic carbon could also reflect a carbonate source from below mixing with a heterotrophic source above. Given the evidence for a decrease in heterotrophy in BPF2, and the prevalence of ¹³C-enriched carbonate deposits in the area, both processes likely contribute to the observed stable carbon isotope trends.

Nitrogen is often a limiting nutrient in Arctic environments, and this can reduce the compounds that are available for microbial metabolism (Manzoni et al., 2012; Kou et al., 2020). The inorganic nitrogen at these two sites (NH₄⁺ and NO₃^–) makes up a small portion of the total nitrogen present (<10^–8%, Figures 2C, 3), which means the limited nitrogen that is present here is organic nitrogen, most likely in the form of plant material and input from animals (Solheim et al., 1996). NH₄⁺ and NO₃^– are highest in the surface (0.05 and 0.04 μg/g_{dry soil}, respectively). At both sites and the lowest values of NH₄⁺ (0.02 μg/g_{dry soil}) are higher than the highest values of NO₃^– (0.04 μg/g_{dry soil}; Figure 3). This could be an indication of NH₄⁺ oxidation or nitrate absorption by plants, since the δ¹⁵N signature of plant organic matter is 2‰–3‰ (Gauthier et al., 2013). This may occur in the upper few centimeters at both sites, as NH₄⁺ decreases and NO₃^– slightly increases with depth. Therefore, the δ¹⁵N signature that is decreasing with depth at both locations likely originates from plant processes and degraded proteins.

Phosphatase generally has the highest activity among the enzymes commonly measured in soils because most soil microbes are capable of creating extracellular phosphatases, and when phosphorus is low, phosphatase enzyme expression is increased (Lee Taylor and Sinsabaugh, 2015). Lastly, the low/below limit of detection values for phosphate in the soil profile (Figure 3) and the increased activities for phosphatase (Figure 5A) further emphasize how these organisms in these soils are limited in nutrients for microbial activity. Even though BPF1 has more labile carbon and higher microbial biomass than BPF2, it only had higher activities in two enzymes (AG and BG) and only in the upper section. Even with less heterotrophy in BPF2, the reason the microbial community maintains similar rates of carbon-degrading enzymes may be to access the nitrogen and phosphorous in organic compounds.

In both Bayelva active layer cores, the highest measured enzymatic activity was for leucine aminopeptidase, and that activity was generally higher at 25°C than at 15 or 5°C. Leucine aminopeptidases are critical cell maintenance enzymes that drive peptide turnover (Matsui et al., 2006) and are often used as an indicator of total peptidase potential in an environment due to the non-specificity of the enzyme (Sinsabaugh et al., 2008). There are many environments that report LAP activity to be the highest among this tested suite of enzymes (Steen et al., 2015). Bacteria have been shown to use leucine, along with other amino acids, as an alternative source of carbon and nitrogen in energy-limited environments (Díaz-Pérez et al., 2016). Nitrogen demand may explain why LAP is nearly two orders of magnitude higher in activity than the other six tested enzymes for both the bulk soil and cultured isolates (Figures 5A,B).

Characterizing environmental microbes is challenging, as many organisms resist common culturing methods (Lloyd et al., 2018). Our isolates belong to a genus that has previously been found in active layers (Figure 4 and Table 1). Pseudomonas species are Gram-negative, aerobic, bacilli organisms (Ramos and Levesque, 2006). This genus contains over 140 species, and most are saprophytic (Ramos and Levesque, 2006) and common in soil. Some Pseudomonas sp. can be psychrophilic, while others are mesophilic (Vishnivetskaya et al., 2000; Sonjak et al., 2006; Finster et al., 2009). The isolates we obtained from the Bayelva sites are most closely related to species that have been found from similar cold regions such as Alaska, Antarctica, and Tibetan plateau (Table 1).

A wide range of enzyme classes, including peptidases and carbohydrate lyases, have been shown to be cold adapted (Feller and Gerday, 2003; Kim et al., 2021). One of the best studied enzymes that has a lower K_m (and therefore greater catalytic efficiency) at cold temperatures is chitobiase, or N-acetyl-β-D-glucosaminidase, obtained from an Arthobacter sp. culture (Kim et al., 2021). NAG is also often used as an indicator of nitrogen mineralization within soils (Salmon et al., 2018). This enzyme showed higher V_max values at 15°C, rather than 25°C in four depths of BPF1 and in five of the Pseudomonas sp. cultures. This implies that psychrophily in this class of enzymes may be widespread among active-layer microbes, since multiple isolates and the natural population had this property. The only other enzyme that appeared to be cold adapted was xylosidase, which had its highest V_max at 15°C in three soil samples and in the same five Pseudomonas sp. isolates that had psychrophilic N-acetyl-B-D glucosaminidase. One depth of BPF1 had the highest activity for this enzyme at 5°C. Xylosidase has been found to be cold adapted in a Duganella sp. isolate from Antarctic soil (Kim et al., 2021). Both enzymes are good evidence that the cold adaptations observed in pure-culture experiments are upheld in natural microbial populations. This suggests that these enzymes are well adapted to a cold environment but have the flexibility to continue functioning as temperatures warm.

The read mapping between the metagenomes and the whole genomes of the isolates showed that each isolate was present in the environment (Supplementary Figure 8). However, the low reads per kilobase per million mapped reads (RPKM) values suggests that these organisms were not the dominant community members; instead, they were the best at growing on the general medium used. There were more genes encoding LAP than other enzymes within both the metagenomes and isolates’ whole genomes (Table 3). The second highest gene counts and activity were in PHOS across the whole genomes and the metagenomes. The whole genomes had more gene counts for three of the four carbohydrate-degrading enzymes, (AG, BG, and XYL) than the metagenomes. However, enzymatic activity was often higher in bulk soil than in cultured isolates. This could indicate that microbes in the natural communities have enzymes with higher V_max values or that the cultured Pseudomonas sp. produced lower amounts of enzymes when grown in culture, perhaps because they were not nutrient limited. Additionally, gene presence may not always mean a higher rate of enzymes exported for compound utilization. For instance, leucine can be cleaved by multiple types of enzymes, which may inflate the number of genes for LAP (Steen et al., 2015). This is reflected in the lack of correlation between gene dosage per genome or per metagenome and the measured activity for that enzyme class.

Dissolved organic carbon (DOC) is a potential source of carbon and energy for heterotrophic organisms (Kaplan and Newbold, 2000; Fischer et al., 2002). Previous studies have shown high spatial variability in permafrost DOC values. For instance, active layers in the Eight Mile Lake soil in Alaska, United States, ranged from 0.063 to 0.46 mg/g (Waldrop et al., 2010). This location does not reach maximum DOC, 46.38 mg/g, until the bottom of the active layer at 35–45 cm. Comparatively, the DOC in the two Svalbard active-layer sites was highest at the surface (2 mg/g_soil, Figure 3).

Nitrogen is the limiting nutrient for plant growth in arctic tundra, and an increase in nitrogen will increase primary production (Beermann et al., 2017). When the active layer thaws, it provides the rooting zone, which is a region for plant roots and soil microbes to compete for nitrogen (Salmon et al., 2018). Salmon et al. (2018) found that C/N decreased, %N decreased, and %C decreased with depth. The C/N of organic matter is fairly constant with depth at both BPF1 and BPF2, and the range of values from 20 to 28 is lower than the C/N in active layer soil of Eight Mile Lake, Alaska (21–46) (Salmon et al., 2018). This indicates that the increase in nitrogen limitation with depth at the Alaskan site may not apply to our Bayelva, Svalbard active layer samples.

These Svalbard active layer soils differ from Taylor Valley in Antarctic McMurdo Dry Valley soils in DOC, water content, inorganic nitrogen, and labile carbon (Zeglin et al., 2009). Moisture from soils near McMurdo Dry Valley had 20 times the DOC, two times the water content, and three orders of magnitude higher inorganic nitrogen than the Svalbard active layer samples. The McMurdo Dry Valley has a higher inorganic nitrogen contribution from NO₃^–, while Svalbard’s inorganic nitrogen contribution is mainly from NH₄⁺. Additionally, we found the Svalbard active layer to have a higher amount of organic nitrogen. Labile carbon (POXC) in Svalbard was up to a hundred times higher than the Antarctic active layer, even though the DOC was up to 20 times lower. These measurements could indicate that Arctic and Antarctic sites may have different outcomes for carbon degradation as permafrost thaws and microbial activity increases. Our work suggests that Svalbard will have higher activity, compared to McMurdo Dry Valley soils, due to the higher amounts of organic nitrogen and labile carbon based on POXC and C/N ratios.