Soils used in the analyses were sampled in October and November 2011 from active layers overlaying permafrost in six pits (0.40 m × 0.40 m) dug under turves of D. antarctica. Three pits were dug at Polynesia Point on Signy Island (60.7107°S, 45.5849°W; altitude 35–45 m a.s.l.) and three were dug on Walton Terraces on the north-western side of Léonie Island (67.5984°S, 68.3561°W; altitude 50–60 m a.s.l.). The three pits dug at each island were separated by a mean distance of 311 m. Soils were sampled by hammering five sterile tubes (50 ml capacity) horizontally into the vertical faces of each pit at depths of 2, 4, and 8 cm, generating 15 soil samples from each pit. Further details of sampling are provided by Cox et al. (2016, 2019) and Newsham et al. (2018). Soil was frozen at −20 or −80°C within a few hours of sampling and was transported back to the United Kingdom at these temperatures, where it was thawed prior to the analyses described below.

Soil pH, moisture and C and N concentrations were measured in bulked soils from each of the three depths in the three pits dug at each island, generating 18 soil samples in total. Soil pH was measured by adding approximately the same volume of deionized water to each soil sample to generate slurries and recording pH with a glass electrode (pH 21, Hanna Instruments, Leighton Buzzard, United Kingdom). Soil moisture was measured gravimetrically after drying c. 1 g of fresh soil at 105°C for 17 h. Sub-samples (c. 2.5 mg) of air-dried soil were weighed into foil capsules and were analyzed for total C and N (Carlo Erba NA1500 elemental analyser, CE Instruments, Wigan, United Kingdom). Details of samples are provided by Robinson et al. (2020).

Owing to the expense of these determinations, analyses were made on the five replicate soils sampled from depths of 2, 4, and 8 cm in one of the three pits dug at each island, generating 30 samples in total. The soils were submitted to the NERC (now NEIF) Radiocarbon Laboratory, where soil C was converted to carbon dioxide (CO₂) using either bomb combustion in a high-pressure oxygen atmosphere, or an elemental combustion system (Costech ECS 4010, Italy). Sample CO₂ was cryogenically purified under vacuum and split into aliquots. The δ¹³C (relative to the Vienna PDB standard) was determined on one aliquot of sample CO₂ by isotope ratio mass spectrometry (Delta V, Thermo-Fisher, Germany). A second aliquot was converted to graphite using Fe:Zn reduction (Slota et al., 1987) and analyzed for ¹⁴C content using accelerator mass spectrometry at the Scottish Universities Environmental Research Centre (East Kilbride, United Kingdom). Following convention (Stuiver and Polach, 1977), ¹⁴C results were normalized to δ¹³C = −25‰ and expressed as % modern and conventional radiocarbon ages (in years BP, where 0 BP = AD1950). Mean residence time (MRT) of C in soil was determined from ¹⁴C concentration using a bomb-¹⁴C soil carbon turnover model (the Meathop Model; Harkness et al., 1986), as described by Horrocks et al. (2020). For samples that showed no clear evidence of bomb-¹⁴C incorporation (i.e., < 97%Modern), the conventional radiocarbon age was used as the MRT (Bol et al., 1999). Sample details, including unique publication codes for each sample, are provided by Robinson et al. (2020).

As described by Cox et al. (2016, 2019), both nucleotides were extracted simultaneously from soil samples (5 mg) using RNA PowerSoil Total RNA Isolation and DNA Elution Accessory kits (MoBio Laboratories, Carlsbad, CA, United States). Extracted RNA was treated with a Turbo DNA-free kit (Life Technologies, Carlsbad, CA, United States), checked for contaminant DNA using PCR, and reverse transcribed using AccuScript High-Fidelity Reverse Transcriptase (Agilent, Santa Clara, CA, United States) and random non-amers. Internal transcribed spacer (ITS) regions in the extracted DNA and cDNA were amplified in triplicate PCR reactions using ITS1F (Gardes and Bruns, 1993) and ITS4 (White et al., 1990) primers. The ITS4 primer was modified with the Roche 454 A adapter and a 10 bp barcode specific to each sample, and the ITS1F primer was modified with the 454 B adaptor. The triplicate PCR products were pooled and purified using AMPure XP bead purification (Beckman Coulter, Inc., Brea, CA, United States) and quantified using a Qubit dsDNA HS Assay (Life Technologies, Carlsbad, CA, United States) before normalization to consistent concentration. The purified and normalized PCR products were run on the 454 Roche Titanium FLX platform at the Liverpool Centre for Genomic Research.

The DNA and RNA sequences were processed using the QIIME pipeline (Caporaso et al., 2010). Sequences were filtered to remove reads that were of low quality, < 300 bp or > 1,200 bp, and were split according to barcodes. The remaining sequences were denoised (Reeder and Knight, 2010), and checked for potential chimeras (Edgar et al., 2011). After deleting potential chimeras, the resulting libraries from Signy Island and Léonie Island consisted of 2,313–11,401 and 1,994–15,482 DNA reads and 2,478–11,840 and 5,250–10,943 RNA reads, respectively (Supplementary Figure S1). The ITS2 regions of the remaining sequences were extracted using ITSx (Bengtsson-Palme et al., 2013) and grouped into operational taxonomic units (OTUs) at 97% sequence similarity using USEARCH 6.1 (Edgar, 2010). The ITS sequence of the isolate (see below) showed 100% similarity to a representative OTU in the DNA and RNA libraries, enabling its abundance in field soil to be determined (see Robinson et al., 2020). DNA and RNA sequences have been deposited in the NCBI Sequence Read Archive (accession codes SRP068654 and PRJNA518063, respectively).

The fungus was isolated in 2012 from surface sterilized organic matter (OM). Briefly, decomposing OM picked from Léonie Island soil was washed 30 times in sterile water, blotted dry on sterile filter paper and pressed into soil extract medium (SEM) before incubation at 7°C for up to 16 weeks (Newsham et al., 2018). An isolate of the fungus was subsequently maintained at c. 15°C on half strength potato dextrose agar (PDA) medium. The isolate was also inoculated onto oat agar (OA) and Azure B agar media (Newsham et al., 2018). Phialides and conidia formed on the latter medium at the inoculation point after 6 weeks of growth at c. 15°C. Squash mounts in water were examined at ×400 and ×1,000 magnifications using light microscopy (Olympus BX51, Olympus Life Sciences, Tokyo, Japan), with images of structures being captured using Cell P software (Olympus Life Sciences).

DNA was extracted from hyphae of the isolate growing on sterile cellophane overlaying half strength PDA medium using a commercially available kit (Extract-n-Amp Plant PCR, Sigma-Aldrich, St Louis, United States). Following standard protocols, the ITS region of rDNA was amplified with the ITS1F (5′-CTTGGTCATTTAGAGGAAGTAA-3′)/ITS4 (5′-TCCTCCGCTTATTGATATGC-3′) primer pair, and a portion of the large subunit (LSU) gene was amplified with the JS-1 (5′-CGCTGAACTTAAGCATAT-3′)/LR5 (5′-TCCTGAGGGAAACTTCG-3′) primer pair. The amplicons were bidirectionally sequenced at a commercial facility, using the same primers. The sequences were manually trimmed and consensus sequences were generated prior to deposition in GenBank (accession codes MW033386 and MW033387). The ITS sequences were subjected to blastn searches in UNITE v. 8.2¹ to determine if the fungus had been sampled and sequenced previously, and to provide a starting point for a taxonomic placement. In order to better place the fungus taxonomically, an exemplar sequence of a voucher LSU sequence was downloaded from NCBI GenBank from each of the classes Orbiliomycetes and Pezizomycetes (as outgroups), and sequences for exemplars from each order within the class Leotiomycetes. Where possible, sequences from type material were chosen to add rigor to the taxonomic placement, with other sequences being derived from material deposited in the Centraalbureau voor Schimmelcultures (CBS) culture collection, the majority of which are reported by Vu et al. (2019), or from vouchered specimens. Sequences were aligned using MAFFT 7.450 (Katoh and Standley, 2013) in the GENEIOUS R11 package (Biomatters, Ltd., Auckland, New Zealand). Ambiguous columns were excised where alignment ambiguities existed around regions containing gaps. A phylogenetic analysis was undertaken to place the test LSU sequence against the voucher sequences, and to identify the likely order to which it belonged. An exemplar sequence from each family in the order was downloaded and a new round of phylogenetic analyses was conducted to place the test LSU sequence to family. An exemplar from each genus of the likely family and closely related families was downloaded to further refine the taxonomic placement. A similar strategy was used to provide context to closely matching ITS sequences in order to produce an ITS phylogeny. Phylogenetic analyses were conducted using approximate maximum likelihood in the package FastTrees 2.1.11 (Price et al., 2010) in the GENEIOUS package using a General Time Reversible (GTR) model, four rate categories per site and the Gamma20 likelihood optimized. The final phylogeny was produced in MrBayes 3.2.3 (Ronquist et al., 2012) using a GTR model with a gamma distribution. The analysis used four heated chains and four replicates. Twenty million generations were run, with sampling every thousandth generation. Convergence, stability and stationarity were determined using Tracer 1.7.1 (Rambaut et al., 2018).

The isolate was grown on half strength PDA medium amended to pH values of 3.9, 4.2, 4.5, 5.1, 5.6, and 6.1 with citric acid/phosphate buffer (Stoll and Blanchard, 1990). The pH values of the media were measured with a Jencons pH meter (Jencons Scientific Ltd., Leighton Buzzard, United Kingdom) prior to autoclaving. Media at pH 3.9 and pH 4.2–6.1 were solidified with 2 and 1.5% technical agar (Oxoid no. 1, Thermo Fisher Scientific Ltd., Basingstoke, United Kingdom), respectively. The medium at each pH was poured into six replicate 90 mm diameter non-vented Petri dishes and was inoculated with plugs (7 mm diameter) of the fungus bored from the margins of actively growing colonies on half strength PDA medium. The dishes were incubated at 15°C in the dark for 5 weeks. Two measurements of each colony’s diameter were recorded weekly at c. 90° to each other and radial extension rates (mm d^–1) were subsequently calculated.

Plugs (7 mm diameter) bored from the margins of the isolate growing on half strength PDA medium were inoculated onto SEM in 0.2 μm-vented sterile cell culture flasks (Thermo Fisher Scientific BioLite, 130189, Loughborough, United Kingdom) that were incubated at −2, 2, 4, 9, 15, 20, and 25°C for up to 5 weeks. Four flasks were prepared for incubation at each temperature. Those at 4 and 15–25°C were incubated in refrigerators or cabinets, the temperatures of which were monitored hourly using TinyTag Plus 2 TGP-4017 loggers (Gemini Data Loggers Ltd., Chichester, United Kingdom). Those at −2, 2, and 9°C were submerged vertically into coolant in recirculating water baths (Newsham and Garstecki, 2007), with all of the flask except the neck being submerged in the coolant. Temperatures were monitored regularly using a thermometer set into SEM in a dummy flask. Measurements of the diameter of each colony recorded at 7 d intervals were subsequently used to calculate radial extension rates (mm d^–1).

Sterile capillary matting (Gardman, Huntingdon, United Kingdom) was placed into 90 mm Petri dishes and was soaked with liquid SEM modified with PEG 8000 to water potentials of −0.66, −0.96, −2.22, −3.15, −4.90, −6.17, and −8.24 MPa, measured using a dew point potentiameter. Sterile black cotton cloth was placed on top of the matting, followed by sterile cellophane film (55 mm × 55 mm; Natureflex 28 NP, Innovia Films, Wigton, United Kingdom). Air bubbles were removed from under the cellophane with a sterile L-shaped spreader. Plugs (7 mm diameter), bored from the margins of colonies grown on SEM for 3 weeks at 15°C, were placed in the centre of the cellophane. The dishes, three of which were prepared for each water potential, were sealed with Parafilm and incubated in the dark at 15°C. Each colony was photographed at weekly intervals for up to 6 weeks. Using ImageJ (National Institute of Health, Bethesda, MD, United States), two measurements of each colony’s diameter were recorded at c. 90° to each other. Radial extension rates (mm d^–1) were subsequently calculated.

Measurements of extracellular phosphatase activities using p-nitrophenol assays were made using the methods described by Newsham et al. (2018), except that the substrate for the assays was 4-nitrophenyl phosphate disodium salt hexahydrate (Sigma-Aldrich, St Louis, United States), with the pH values of the acetate buffer being adjusted to 6.0 and 8.0 for acid and alkaline phosphatase assays, respectively. Semi-quantitative measurements of the activities of these two enzymes were also made at pH 5.4 and pH 8.5 with 2-naphthyl phosphate as the substrate using the methods described by Newsham et al. (2018) and API ZYM kits (bioMérieux Ltd., Basingstoke, United Kingdom). The same methods were also used to semi-quantitatively test for the activities of lipase, trypsin, α-chymotrypsin and α-galactosidase with the substrates 2-naphthyl myristate (pH 7.5), N-benzoyl-DL-arginine-2-naphthylamide (pH 8.5), N-glutaryl-phenylalanine-2-naphthylamide (pH 7.5), and 6-Br-2-naphthyl-αD-galactopyranoside (pH 5.4), respectively.

The ITS2 region of the fungus was entered into the GlobalFungi database (Větrovský et al., 2020) and the geolocations of exact hits were recorded. After excluding the data reported for Signy Island and Léonie Island reported by Cox et al. (2016), the details of substrates, soil pH, and mean annual temperature and precipitation were noted for the 285 exact hits in the database.

The effects of soil depth on the relative abundance of the fungus in DNA and RNA communities, and the influence of pH, temperature and water potential on the radial extension rate of the isolate, were tested with ANOVA and Tukey’s multiple range tests. Associations between the relative abundance of the fungus and edaphic factors were tested with linear and three part Gaussian regressions. Analyses were made in Sigmaplot 14 (Systat Software Inc., San Jose, CA, United States) and MINITAB 18 (State College, PA, United States).

The fungus was present in all 45 soil samples from Signy Island from which DNA libraries were made. Its mean relative abundance in DNA libraries from the island was 39% (range 19–66%; Supplementary Figure S1C). At Léonie Island, it was present in 42 of the 45 soil samples from which DNA libraries were made, with a mean relative abundance of 46% (range 0–92%; Supplementary Figure S1C). RNA of the fungus was detected in 43 out of 45 libraries constructed from Signy Island soil, but at a much lower relative abundance than in DNA libraries (mean relative abundance 2%, range 0–8%; Supplementary Figure S1F), and its RNA was detected in 36 of the 45 RNA libraries from Léonie Island, with a mean relative abundance of 17% (range 0–68%; Supplementary Figure S1F).

The relative abundance of the DNA of the fungus changed with depth at Signy island, with a doubling in its abundance from 25 to 50% of the DNA community as depth increased from 2 to 8 cm [F_{(2, 42)} = 37.35, P < 0.001; Figure 1A]. There were no apparent effects of soil depth on the relative abundance of the fungus in DNA and RNA communities at Léonie Island and Signy Island, respectively (Figures 1B,C), but its mean abundance in the RNA community at Léonie Island increased from 6 to 24% as depth increased from 2 to 8 cm [F_{(2, 42)} = 3.87, P = 0.029; Figure 1D].

Correlative analyses between the relative abundance of the fungus in the DNA and RNA communities at Signy Island and Léonie Island indicated that it was more frequent and active in soils with lower radiocarbon (¹⁴C) enrichment (Figures 2A,G), with significant negative linear associations between its abundance in the two communities and ¹⁴C enrichment (r = −0.670 and −0.743, both P < 0.001). There were also significant associations between the MRT of soil C and the relative abundance of the DNA and RNA of the fungus, with peak abundance of each nucleotide in soils aged approximately 1,000 and 1,200 years BP, respectively (r = 0.835 and 0.965, both P < 0.001; Figures 2B,H). The relative abundances of the DNA and RNA of the fungus were positively and linearly associated with soil δ¹³C value and C and N concentrations (r = 0.525–0.699, all P < 0.025; Figures 2C–E,I–K), whereas the abundance of its DNA was negatively associated with soil pH value (r = −0.646, P = 0.004; Figure 2F). Soil moisture concentration, which ranged from 58.4 to 76.2%, was not associated with the relative abundance of the fungus in the DNA or RNA communities (data not shown).

The LSU phylogeny placed the fungus in the order Helotiales. Type, CBS and voucher LSU sequences were able to refine the test sequence to matching to members of the families Pezizellaceae, Porodiplodiaceae, and Sclerotiniaceae, such as Chalara clidemiae and species of Calycina, Porodiplodia, Monilinia, Botryotinia, and Streptotinia (Figure 3), but with low support (posterior probability of 0.94) and without sufficient taxonomic coverage of genera available to gain any more precision. The ITS phylogeny similarly placed the fungus in the Helotiales, and indicated close matches to sequences assigned to the families Pezizellaceae, Hyaloscyphaceae, and Dermateaceae (Supplementary Figure S2). Comparisons of the ITS region sequence of the fungus with those in the UNITE database indicated that it bore close (> 99%) similarities to soil and root fungi from cold regions (Table 1). The best match was to the sequence of a fungus cloned from acidic (pH 4.4) soil under Eriophorum, Betula, Ledum, and Vaccinium spp. at Toolik Lake in the Low Arctic (Table 1). Other close matches were recorded with a Phialea species inhabiting the hair roots of the ericaceous plant species Phyllodoce aleutica growing in an alpine region of Japan, and members of the Helotiales in soil under the shrub Salix polaris growing at Abisko in sub-Arctic Sweden, in Poa flabellata roots from the Falkland Islands, and a member of the Helotiaceae from the roots of Tsuga canadensis growing in an alpine region of North America (Table 1).

On half strength PDA and OA medium, colonies of the fungus are thin and white with well-defined margins. Sclerotia and stroma are absent. Conidia are produced in masses at the apices of pigmented, flask-shaped phialides of length 25–37 μm and widths 5–7 μm (venter) and 2.0–2.5 μm (collarette; Figures 4A,B). The apices of phialides are not flared (Figure 4C). Conidia, which were not observed to form in basipetal chains, are hyaline, cylindrical with rounded ends, 0(–1) septate and measure 4.5–6(7) × 1.8–2.5 μm (Figure 4D).

The isolate is an acidophile, exhibiting maximum hyphal growth at pH 4.2–4.5, and slower growth at pH 6.1 compared with pH 3.9–5.6 (Figure 5A). It was also found to be a psychrotroph, with measurable hyphal growth at −2°C (0.04 mm d^–1) and optimum hyphal extension rate at 15°C (0.93 mm d^–1). Radial extension rate at 25°C was negligible (0.02 mm d^–1; Figure 5B). Measurements of hyphal growth on SEM modified with PEG 8000 to a range of water potentials indicated that radial extension rate was fastest at −0.66 MPa, with rapidly declining extension rates between −0.66 and −3.15 MPa, and no growth at below −4.9 MPa (Figure 5C).

Analyses using p-nitrophenyl assays indicated mean acid phosphatase activity of the isolate of 12 μM h^–1 p-nitrophenol liberated, with no apparent alkaline phosphatase activity (data not shown). API ZYM assays indicated strong acid phosphatase activity, and weak alkaline phosphatase activity. These assays also showed weak naphthol-AS-BI-phosphohydrolase activity, and no activities for lipase, trypsin, α-chymotrypsin and α-galactosidase (data not shown).

Analyses using the GlobalFungi Database indicated that fungi with ITS2 regions bearing 100% similarities to that of the fungus reported here are present on all continents (Figure 6). These analyses also indicated increased relative abundances (> 1%) of the fungus at high latitudes in the southern hemisphere and alpine regions in the northern hemisphere (Figure 6). DNA of the fungus has mainly been amplified from soils in forest, desert, tundra and grassland biomes (91.6% of records), with occasional records from litter, bryophyte tissues and roots. The locations in which the fungus has been recorded have a mean annual temperature of 5.0°C (range −10.4–17.7°C) and a mean annual precipitation of 1,077 mm per annum (range 206–4,251 mm). The mean pH value of the soils from which the DNA of the fungus has been amplified is 4.8 (range 2.6–6.1).