Fungal and bacterial diversity of Svalbard subglacial ice

The composition of fungal and bacterial communities in three polythermal glaciers and associated aquatic environments in Kongsfjorden, Svalbard was analysed using a combination of cultivation and amplicon sequencing. 109 fungal strains belonging to 30 mostly basidiomycetous species were isolated from glacial samples with counts up to 103 CFU/100 ml. Glaciozyma-related taxon and Phenoliferia psychrophenolica were the dominant species. Unexpectedly, amplicon sequencing uncovered sequences of Chytridiomycota in all samples and Rozellomycota in sea water, lake water, and tap water. Sequences of Malassezia restricta and of the extremely halotolerant Hortaea werneckii were also found in subglacial habitats for the first time. Overall, the fungal communities within a glacier and among glaciers were diverse and spatially heterogenous. Contrary to this, there was a large overlap between the bacterial communities of different glaciers, with Flavobacterium sp. being the most frequently isolated. In amplicon sequencing Actinobacteria and Proteobacteria sequences were the most abundant.

Core microbiome analysis showed that 28 OTUs were shared between all subglacial ice samples, of which 13 belonged to Basidiomycota, 7 to Chytridiomycota, 5 to Ascomycota, and 3 to unassigned Fungi. Hortaea werneckii and Malassezia restricta were among the species present in all the subglacial samples.
The principal component analyses (PCoA) of ordination patterns revealed four separate sample clusters based on phylogenetic distance metrics (weighted uniFrac distance) (Fig. 4A). The three axes of PCoA accounted for  www.nature.com/scientificreports www.nature.com/scientificreports/ Diversity of bacteria (amplicon sequencing) revealed a prevalence of Actinobacteria and proteobacteria. After subtraction of chloroplast sequences, the dataset was composed of 1,793,737 assembled sequences (for 12 samples in total), corresponding to 712 different features OTUs. Number of reads per sample ranged from 280,884 (C-ice-2) to 62,627 (Sea-H). Shannon indices for bacteria varied over a broad range from H′ = 2.5 (C-ice-1) to H′ = 5.5 (Sea-F) indicating large differences in the diversity of individual samples. Samples with lower bacterial diversity indices were both samples of clear ice (C-ice-1, C-ice-2), as well as four of the six subglacial ice samples (S-ice-A1 and A2, S-ice-B2 and S-ice-C1). Nevertheless, significant differences were found between all three types of samples (i.e., subglacial ice, clear ice, sea water, pseudo-F = 4.54; p = 0.012; the number of permutations = 999). However, no statistically significant differences were found in the pairwise PERMANOVA comparisons of sample types. Rarefaction curves suggested that Illumina sequencing depth was sufficient to capture the dominant phylotypes in all samples (Supplementary Material).
The principal component analyses (PCoA) of ordination patterns revealed five separate sample clusters based on phylogenetic distance metrics (weighted uniFrac distance) (Fig. 4B). The three axes of the PCoA accounted for 84% of the variation. Samples of subglacial ice (S-ice-A1, S-ice-B1 and S-ice-C1), clear ice and lake water clustered together. The second cluster was composed of the two subglacial ice samples S-ice-A2 and B2 that shared a very similar bacterial composition. Sea water samples formed the third cluster, while glacial water and tap water were clearly separated from all the other samples.

Diversity of cultivable bacteria reflected differences observed with NGS. 52 bacterial isolates
were obtained from subglacial ice and glacial meltwater samples. In Midtre Lovénbreen all isolates from S-ice A1 belong to Flavobacterium sp., while S-ice A2 showed the highest bacterial subglacial diversity. Although Flavobacterium sp. again prevailed, isolates belonging to Cryobacterium sp., Hymenobacter sp., Massilia sp., Polaromonas sp., Pseudomonas graminis, Spirosoma sp., and an uncultured species belonging to Oxalobacteraceae family, were also retrieved. In Pedersenbreen glacier a similar pattern was observed. Sample S-ice B2 was monopolized by Flavobacterium sp., while S-ice B1 additionally harboured Cryobacterium sp., Massilia sp., Paenibacillus antarcticus and an uncultured species belonging to Oxalobacteraceae family, closely related to Actimicrobium antarcticum.
Glacial meltwater had the most diverse cultivable bacterial community. The most frequently isolated species were Pseudomonas sp. (including P. frederiksbergensis and P. graminis), followed by Flavobacterium sp., and Massilia sp. Among sporadically isolated species were Polaromonas sp., Sphingomonas sp., unidentified Burkholderiaceae, and Sphingorhabdus sp.

Discussion
Glacier ice is one of the most challenging natural environments for life. Nevertheless, glacial habitats harbour a wide diversity of microorganisms, both prokaryotes and eukaryotes. Some microbes deposited on the glacial surface gradually travel to deeper ice layers. They can survive in ice as viable, frozen "living fossils" 26,27 , while others remain metabolically active and possibly multiply in the veins of brine between ice crystals 28 . Microbial communities also develop in deep ice or subglacial ice at the base of polythermal glaciers, where melting occurs due to gravitational pressure resulting in thin layers of liquid water, often with high concentrations of salts and enriched with the inorganic material of the glacier bed. These glacial populations likely represent an important share of global biological activity 13,[16][17][18][19] that has so far been poorly characterized. A deeper understanding of the microbial biodiversity of this unique and rapidly disappearing ecosystem would contribute to our understanding of the composition and dynamics of subglacial microbial communities and their roles in the Arctic ecosystems. Therefore, fungal and bacterial diversity of oligotrophic subglacial and glacial ice and meltwater and nearby marine water, moraine lake water and tap water were investigated both by cultivation and with molecular methods, the latter being applied on samples of subglacial ice for the first time.
Fungi were abundant in all samples as shown by both cultivation and amplicon sequencing. Frequencies in the three glaciers differed considerably. Midtre Lovénbreen harboured the lowest number of cultivable fungi (up to 100 CFU/100 ml), followed by Vestre Brøggerbreen (up to 990 CFU/100 ml) and Pedersenbreen (1000 CFU/100 ml) including yeasts at 500 CFU/100 ml) (Supplementary Material). The large proportion of yeast in Pedersenbreen was confirmed by amplicon sequencing, where the highest number of sequences belonged to Microbotryomycetes (75%). Sixty-eight % of the cultured yeast species belonged to the Microbotryomycetes and the remaining 32% to the class Tremellomycetes (although the latter were absent in amplicon sequencing results). A previous cultivation study 4 reported even higher yeast counts in Svalbard glacial ice (up to 4000 CFU/ml), while Turchetti et al. 29 reported similar counts in Alpine subglacial sediments (100 to 1000 CFU/g). Both studies reported several orders of magnitude lower numbers of yeast cells in supraglacial ice and sediments indicating an enrichment of yeast in subglacial environments. Such enrichment could be supported by the presence of organic carbon in subglacial sediments 17 , deriving from permafrost soils that are overridden by the advancing glaciers and finely ground by subglacial abrasion processes. This organic material consists of cyanobacterial mats, plant material, and roots, which are readily biodegradable by microbial activity 17 .
Our results show a significant spatial variability of fungal communities within individual glaciers as well as between glaciers, suggesting/highlighting that the microbial diversity in glacial ice is much more than just a collection of deposited propagules. Many species were recovered from only one of the two samples of the same glacier (Fig. 6A) indicating a heterogeneous nature of the glacial habitats, a phenomenon already described by Luo et al. 30 . In contrast, meltwater flowing from under the glacier was very similar in fungal composition to the nearby subglacial ice sample (cf. the hierarchical clustering dendogram of Midtre Lovénbreen glacier (Fig. 6B)). www.nature.com/scientificreports www.nature.com/scientificreports/ Differences in the microbial diversity might be explained by the fact that basal ice is composed of individual layers of ice with physico-chemical heterogeneity favouring the establishment of microhabitats 14 . Basal ice microbiota is not only influenced by the Aeolian inoculum deposition at the surface, but also by its anisotropic structure at the bottom. Spatial and/or temporary differences in deposition together with environmental filters acting on the resident microbiota can be envisaged to lead to a patchy distribution of species within the glacier and at the same time point to a low amount of microbial migration within the glacier, which would instead result in a homogenous distribution of species.
Analysis of the ITS2 amplicon discovered sequences assigned to the phylum Chytridiomycota in all samples, with the highest abundance in tap water, followed by harbour sea water, lake water, and glacial meltwater. In the past isolations of chytrids from extreme habitats were rare. However, recent metagenomic studies revealed their widespread occurrence in a broad range of extreme habitats such as snow in Europe and America 31 , Antarctic soils 32,33 , Patagonian glacier ice, meltwater and sediments 34,35 . In Arctic marine environments they are even the dominant fungi [36][37][38][39][40][41] . This is to the best of our knowledge the first report of Chytridiomycota in Arctic glacial ice and tap water. Chytridiomycota in household tap water probably originate from the water source itself, which in case of Ny-Ålesund water supply system is a surface lake filled by the snow meltwater 42 , and are possibly enriched during transportation to the final user. Tap water and to a lesser extent moraine lake water (from a different lake than the one providing the tap water) were also populated by unidentified Rozellomycota, an early diverging hyper-diverse lineage of fungi also known as Cryptomycota 43 that have been documented almost exclusively by NGS analyses of environmental samples 44 . These uncharacterized fungi dominate fungal communities in different aquatic ecosystems, such as temperate freshwater lakes [45][46][47][48] , coastal and marine environments [49][50][51] , snow 31 and polar aquatic systems 36,39,52,53 . Their abundant presence in glacial and related samples is in agreement with these findings, since the nature of samples analysed in this study is also aquatic, and/or is possibly linked to the ability of Rozellomycota to mycoparasitise chytrids 54,55 .
The amplicon sequencing showed strong dissimilarities between samples obtained from the same glacier and between glaciers, an observation also confirmed by cultivation. For instance, Libkindia masarykiana was abundantly present in the sample S-ice-B2, but was present in very low numbers in the other samples. C-ice-1 was characterized by Dothideomycetes, while C-ice-2 was dominated by Schizopora flavipora otherwise known as a tea plant pathogen. Also, fungi in the sea water sampled in the harbour were different from the seawater fungi sampled in the middle of fjord (which was more closely related to subglacial ice samples (Fig. 4A)).
Among the most unexpected results of our study was the presence (as shown by NGS) of the black yeast Hortaea werneckii (38%) in Lovenbreen glacier (S-ice-A1), never reported before in any glacial environment. The fungus was present also in all the other subglacial samples, but in lower frequencies. This extremely halotolerant black yeast is the dominant fungus in hypersaline waters of salterns worldwide 56 , and is additionally found in other marine related environments (e.g. surface sea water, sea sponges, corals, fish, halophites) 57-59 and deep sea [60][61][62] . Furthermore, H. werneckii is known as the etiological agent of human tinea nigra, a superficial infection of salty soles and palms in humans, occurring in tropical and subtropical areas 63,64 .
The second sample of Lovenbreen glacier (S-ice-A2) was characterized by the presence of Malassezia restricta (39%), the most common human skin-related fungus after M. globosa [65][66][67] . Initially Malassezia spp. were thought to be specifically associated with mammalian hosts, but culture-independent studies revealed their presence in very diverse marine and terrestrial ecosystems, such as cone snails 68 , Antarctic soils 69,70 , deep-sea sediments 71 , deep-sea hydrothermal vents 61,72 , nematodes 73,74 , corals 75 , and sponges 76 . Amend 72 found that ribosomal DNA sequences of Malassezia in the above-cited studies are nearly identical to sequences of human-associated isolates, suggesting that they either diverged in their habitats very recently or that they are capable of large adaptability and high stress tolerance enabling them to colonize environments as different as human skin and subglacial ice. The fact that this species was not detected by cultivation may be due to its slow and fastidious growth and lipophilic nature, usually not taken into consideration in standard growth media 77 .
Cultivation of various samples confirmed some of the similarities between sample sites as observed by NGS: the most frequently isolated yeast species belonged to the basidiomycetous Phenoliferia psychrophenolica (former Rhodotorula psychrophenolica) and were similar to Glaciozyma antarctica, both frequently recovered from persistently cold habitats 78,79 . Glaciozyma-related taxon isolates differed from Glaciozyma antarctica CBS 5842 by 27 nucleotides in the LSU marker and additionally contained a deletion of 5 nucleotides towards the end of the sequence, suggesting the possibility of a new taxon. Glaciozyma species are obligate psychrophiles, with a maximum growth temperature below 20 °C 78 and found in cold habitats such as Alpine and Apennine glaciers 28,79 , Antarctic sea water and soil 81 , and Greenland glacial ice 81 . Glaciozyma antarctica, formerly Leucosporidium antarcticum 78 , has been widely studied due to its biotechnologically important production of antifreeze proteins and cold-active enzymes 83,84 . The genome of Glaciozyma antarctica was sequenced and analysed only recently 85 , revealing systems of psychrophilic response that are present only in this species and might be associated with temperature variations. The higher expression of such unique genes associated with cold adaptation unique to G. antarctica, together with other physiological strategies such as modelling its membrane lipid composition 86 , could explain the high abundance of the genus in cold environments and its superior ability, compared to other species, to overcome stressful conditions associated with extreme habitats. Genus Phenoliferia has been proposed by Wang et al. 79 to accommodate a clade of species segregated from the genus Rhodotorula, characterised by the ability to assimilate phenol as the sole carbon source at 10 °C 87 . In particular P. psychrophenolica and P. glacialis can degrade up to 12.5 and 5 mM concentrations of phenol, respectively 86 . P. psychrophenolica has so far only been isolated from Alpine and Apennine glaciers 28,80,86 and from the Gulkana Glacier in Alaska 25 and the glacier Austre Broggerbreen on Svalbard 88 .
Additionally, many different yeasts species were isolated from our samples. These belonged mainly to genera Rhodotorula, Mrakia, Filobasidium and Naganishia. Rhodotorula is a ubiquitous genus commonly isolated from cold habitats 29  www.nature.com/scientificreports www.nature.com/scientificreports/ known to be part of the core yeast communities in cold environments 89 , with some species recovered exclusively from cold habitats and for this reason considered endemic and other species ubiquitous 89,90 . The ability to grow over a wide temperature range, its membrane fluidity stability, marked adaptability, nutritional versatility and polyextremotolerance 91,92 , together with the abundant production of photo-protective pigments 93 , make Rhodotorula spp. adept in withstanding extreme climates. R. svalbardensis is a novel psychrophilic species recovered from Svalbard glacier cryoconite holes 7 and Greenland glacial ice 82 . We refer to the species as R. svalbardensis pro. tem. (pro tempore, temporary taxonomic placement) as suggested by Turchetti et al. 94 due to its uncertain phylogenetic placement. Psychrophilic genus Mrakia has so far been restricted to the coldest areas of the world, such as Greenland, Antarctica, Siberia, Patagonia and Alaska 95,96 . The Principal Component Analysis based on cultivation data (Fig. 6A) revealed a strong correlation between the presence of the genera Mrakia and Glaciozyma-related taxon. Both yeasts were formerly classified as belonging to the Leucosporidium genus, characterised by an absence of visible carotenoid pigments production 97,98 . Glaciozyma-related taxon and Mrakia sp. (gelida or frigida) also share ecological traits: they are psychrophilic yeasts unable to grow at temperatures higher than 17 °C, with distribution consequently restricted to habitats with persistently low temperatures 97,98 . Further explanation of the co-occurrence of Mrakia and Glaciozyma-related taxon will be possible once the metabolic and other characteristics of the latter have been characterised further. Filobasidium species are widely spread, from very arid environments 99 to Italian glacial meltwaters 100 and basal ice of the High Arctic 4 . Species of the genus Naganishia are among the most resistant organisms to UV radiation 101 : they are found in many extreme environments such as high elevation soil of the Atacama Desert 101,102 , of the Antarctic Dry Valleys 81 , and of subglacial ice in Svalbard 4 .
Among cultivated filamentous fungi Penicillium crustosum, P. bialowiezense, and Pseudogymnoascus sp. prevailed. The frequent presence of Penicillium spp. in cold environments has been documented in various habitats, from soil to basal ice in the Arctic and Antarctic 6,103,104 . Sampling of subglacial ice of Pedersenbreen in 2001 6 showed a high abundance of P. crustosum and this was confirmed in our study by cultivation and NGS from sample S-ice B1 originating from the same glacier. Also, amplicon sequencing showed a higher presence of Penicillium (1.3% of sequences) in this subsample compared to all other samples and glaciers (<0.2%). On the other hand, Pseudogymnoascus sp. was isolated exclusively from one glacier (Broggerbreen) (Fig. 6A), while no sequences belonging to this genus were recovered by culture-independent methods. Representatives of this cold-adapted genus were previously found in an alpine glacier of the Qinghai-Tibet Plateau 79 .
Other sporadically isolated filamentous fungi belonged to the genera Lecanicillium, Thelebolus, Tetracladium, Leptosphaeria, and Mortierella, all of which have previously been isolated from Antarctic soil samples 105 . Mortierella sp. was described as a decomposer of granite after glacier retreat 88 . Isolates in our study from S-ice B1 and G-wtr A closely related to Venturia (GenBank accession no. AB916509) were also isolated from feathers of a Barnacle goose on Svalbard 106 . Isolates of Thelebolus sp. (S-ice C1) closely related to Thelebolus strain with the GenBank accession no. AB916508 were also isolated from bird feathers 106 . Genus Thelebolus has been described initially as an aquatic taxon from lake microbial mats in Eastern Antarctica 107 . Now it is recognized as one of the most common genera in polar soil crusts adjacent to glaciers 88,108 .
The overlap of the cultivation and amplicon sequencing results was modest, with only 8 out of 30 species found by cultivation also present in the amplicon sequencing results, again showing that each of the two approaches has its strengths and that cultivation and non-cultivation methods complement each other, as noted before by several authors investigating other environments 82 .
In line with previous studies 13,114,115 culture independent results identified Proteobacteria and Actinobacteria as the most abundant phyla in almost all samples. On the one hand the composition of the bacterial populations varied between samples from the same glacier, but were on the other hand similar between some sample pairs collected from different glaciers. For example, subglacial ice samples S-ice-A1 (Lovenbreen) and S-ice-B2 (Pedersenbreen) were both dominated by Sulfuricurvum sp., a novel genus isolated from an underground crude-oil storage cavit 116 . Sulfuricurvum is a facultatively anaerobic, chemolithoautotrophic bacterium previously isolated from subglacial lakes in Iceland 117,118 . It uses elemental sulphur, sulphide and thiosulfate as electron donors during microaerobic (1% O 2 ) and anaerobic respiration 118 . Sulphur is common in subglacial bedrocks, favoring microbial sulphide oxidation, and sulphate reduction 19 . Thus Sulfuricurvum spp. might play a role in the biogeochemical cycling of sulphur in the subglacial ice, which is in direct contact with the underlying bedrock.
While the fungal diversity of different glacial samples differed considerably even between samples from the same glacier, the differences in bacterial diversity was much smaller, with Micrococcus (Actinobacteria) being the most abundant species in all samples. Micrococcus has been previously found in ice, for example in 20.000 years old Bolivian core ice samples 119 and Malan Glacier ice cores 120 .
Cultivation dependent results for bacteria overlapped to some extent with the results of the amplicon sequencing: 17 species of bacteria were isolated, 12 of them belonging to the phylum Proteobacteria. However, the most commonly isolated species was Flavobacterium sp. (Bacteroidetes), which represented less than 4% of sequenced amplicons. Members of this genus are commonly isolated from aquatic ecosystems 121 . Flavobacterium species are considered psychrotolerant, since they have been isolated from several polar habitats including Antarctic soil, lakes, marine sediment and sea ice [122][123][124][125] , high Artic snow cover 1 , and frozen soil in China 126 . The genus contains the causative agent of bacterial cold water disease, F. psychrophilum 127 . Among other isolated bacterial species were Massilia sp. and Polaromonas sp. (Oxalobacteraceae and Comamonadaceae families, respectively). Massilia members were originally isolated from clinical samples 128 and characterized as aerobic and facultatively psychrophilic 129 . Later they were found in various habitats such as freshwater 130  www.nature.com/scientificreports www.nature.com/scientificreports/ soil 134,135 , and phyllosphere 136,137 and also from glacier ice cores in China from which two novel species, M. eurypsychrophila and M. glaciei, were recently described 138,139 . Cold-active bacteria of the genus Polaromonas (class Betaproteobacteria) have also been recognised as significant components of glacial microbiomes before 140 .

conclusions
We investigated the microbial diversity of subglacial ice and nearby aquatic environments of three polythermal Arctic glaciers located in the Kongsfjorden area on Svalbard. The main findings of the study are: • Subglacial ice of polythermal glaciers harbours a rich and spatially heterogenous fungal and bacterial diversity. • Several as yet undescribed species were isolated, including strains related to the genera Glaciozyma and Rhodotorula. • The first amplicon sequencing of the subglacial ice mycobiota uncovered species previously not reported from the Arctic ecosystems, such as the extremely halotolerant Hortaea werneckii and the skin pathogen Malassezia restricta. • Of additional interest was the high abundance of the elusive phylum Chytridiomycota and their parasites Rozellomycota in the snow-derived tap water, but of Chytridiomycota also (and even more surprisingly) in the glacial ice.
In summary, the first study combining the cultivation and amplicon sequencing to investigate the diversity of both fungi and bacteria in subglacial ice of polythermal glaciers has shown a rich diversity of species, an unexpected spatial heterogeneity of (particularly fungal) communities and has found several species that are either new to science or have never been found in glacial ice before.

Materials and Methods
Site and samples description. Fieldwork was conducted during the end of July and beginning of August 2017 in the Kongsfjorden area (Ny-Ålesund) of the Svalbard Archipelago, located at 79°N, 12°E, Norway. Subglacial ice, clear ice, sea water, lake water and glacial meltwater samples were collected from the locations shown in Table 1 and Fig. 1. Subglacial ice and clear ice floating on seawater were sampled by chopping the ice from the glacier edges or from the floating growlers (smaller fragments of ice) with surface-sterilized tools and transferred into sterile Whirl-Pak ® plastic bags. Glacial melt-water, moraine lake water, seawater and tap water were directly collected into sterile plastic bottles. All samples were transported to the lab for preparation within five hours of collection and processed at the NERC Arctic Research Station, Ny-Ålesund. Glaciers studied are small valley glaciers ending on land with similar surface areas (5-5.6 km 2 ): Midtre Lovénbreen, Pedersenbreen, and Vestre Brøggerbreen. They are all non-temperate valley glaciers with polythermal characteristics 141 , and subject to substantial glaciological and ecological studies 142,143 . The glacier thermal regime is important in defining the presence of water at its bed and consequently exerts a firm influence on the subglacial drainage system development 12 . cultivation-based fungal and bacterial diversity analyses. The surface layer of subglacial ice samples was melted in a sterile container at room temperature and discarded. The remaining ice was washed with sterile water, transferred to another sterile container and melted. Five, 20 and 50 ml aliquots of subglacial ice samples and glacial meltwater were filtered through Milli-pore membrane filters (0.45 μm pore size) in duplicate. Due to the high sediment content of the subglacial ice and glacial meltwater, the samples were vortexed and left to sediment for one minute before filtering. Filters were placed on two enumeration and four different selective agar media with either low water activity (a w ) or low nutrient content. Since fungal diversity is usually neglected, greater effort in fungal isolation was attempted. The media used were therefore DRBC -a general-purpose enumeration medium 144 ; MY10-12 -a medium for the isolation of xero-and halo-tolerant fungi with 10% glucose and 12% NaCl (a w = 0.880) 145 ; DG-18 -a medium for detection of moderate xerophiles (a w = 0.946) 146 ; SNA and  www.nature.com/scientificreports www.nature.com/scientificreports/ MM -two nutrient-poor media for the isolation of oligotrophic fungi and R2A -a low nutrient enumeration medium for heterotrophic microorganisms (both bacteria and fungi) 147 . To prevent the growth of fast-growing prokaryotes, chloramphenicol (50 mg/l) was added to all media, except for R2A agar. Plates were incubated at 10 °C for up to 8 weeks at the University of Ljubljana, where subsequent procedures were carried out. For every medium the average number of colony forming units (CFU/100 ml) was counted and calculated.

S-ice A1 S-ice A2 G-wtr A L-wtr S-ice B1 S-ice B2 S-ice C1 S-ice C2 C-ice 1 C-ice 2 Sea H Sea F Tap
Fungal and bacterial identification. DNA was extracted from pure fungal and bacterial cultures up to one week after incubation on a malt extract agar (MEA) and R2A media, respectively. DNA from filamentous fungi was extracted by mechanical lysis of approx. 1 cm 2 of mycelium according to Van den Ende and de Hoog (1999) 148 . For yeast-like and bacterial strains, DNA was extracted using PrepMan Ultra reagent (Applied Biosystems) according to the manufacturer instructions. For filamentous fungi a fragment of rDNA including ITS region 1, 5.8S rDNA and ITS region 2 (referred from here on as the ITS amplicon) was amplified using ITS5 and ITS4 primers 149 . Polymerase chain reactions (PCR) were performed using Thermo Scientific Taq DNA Polymerase according to the manufacturer's protocol. Reactions were run in a PCR Mastercycler Ep Gradient (Eppendorf) with an initial denaturation of 2 min at 95 °C, followed by 30 cycles of denaturation at 95 °C for 45 s, annealing at 54 °C for 30 s, and elongation at 72 °C for 2 min, with a final elongation at 72 °C for 4 min. For identification of Penicillium spp. strains, the partial β-tubulin gene (benA) was amplified and sequenced with Ben2f and Bt2b primers 150 . Initial denaturation at 95 °C for 1 min was followed by 35 cycles of denaturation at 95 °C for 30 s, annealing at 53 °C for 30 s, and elongation at 72 °C for 1 min. Final elongation was at 72 °C for 10 min. Cladosporium strains were identified using partial actin (act) sequences, amplified with ACT-512F and ACT-783R primers 151 . Initial denaturation at 94 °C for 5 min was followed by 45 155 and compared against the GenBank database using the BLAST software 156 . Maximum likelihood methods implemented in PhyML 3.0 157 were used to build phylogenetic trees after aligning the sequences with similar type and reference sequences from the GenBank database in MEGA5 155 (Supplementary Material). Bootstrap values were calculated from 500 replicate runs. A principal component analysis (PCA) based on presence/absence data of fungal species in the studied environments was performed to determine which species best explained the variance among samples. The PCA was performed in R with the prcomp function 158  Fungal and bacterial diversity analysis with amplicon sequencing. DNA was extracted from filtered biomass (0.45 μm pore size, Millipore) of subglacial ice (S-ice), clear ice (C-ice), sea water (Sea), glacial meltwater (G-wtr) and moraine water (L-wtr). All filters were placed in 1.5 ml microcentrifuge tubes containing RNAlater ® (Sigma-Aldrich Company Ltd., UK) and immediately frozen at −20 °C, until further analysis in the laboratory setting. DNA was extracted from filters using the PowerWater DNA Isolation Kit (MoBio Laboratories Inc., California) and then from the same filters using the PowerLyzer PowerSoil DNA Isolation Kit (MoBio), according to the manufacturer's instructions with a slight modification to increase the DNA yield and quality. To increase efficiency of fungal cells lysis an additional heating incubation at 65 °C for 10 min was used after adding PW1 solution. DNA from both isolation methods was pooled together and stored at −80 °C until PCRs were performed.
For the analysis of fungal diversity, Illumina Miseq V3 (300 bp paired-end, performed by Microsynth AG, Switzerland) sequencing was carried out on the ITS2 region of the ITS rDNA gene amplified using the primers ITS4-Fun (5′-AGCCTCCGCTTATTGATATGCTTAART -3′) and 5.8S-Fun (5′-AACTTTYRRCAAYGGATCWCT -3′) 160 . Amplification was carried out in a PCR Mastercycler Ep Gradient (Eppendorf) with initial denaturation of 2 min at 98 °C, followed by 20 cycles of 10 s at 98 °C, 25 s at 54 °C and of 25 s at 72 °C, with a final elongation of 7 min at 72 °C.
For both fungi and bacteria, the first step polymerase chain reactions (PCR) were performed in house using www.nature.com/scientificreports www.nature.com/scientificreports/ Paired-end reads were quality checked and trimmed (minimum quality score 20) and analyzed with QIIME2 2018.8 software package (Quantitative Insights Into Microbial Ecology) 162 . Since the assembly of the paired end reads was largely unsuccessful, only single-end forward reads were used in the subsequent analysis. The reads were denoised, the tree was constructed by FastTree on a mafft alignment and rooted at midpoint and the alpha and beta diversity indices were calculated. For assigning the taxonomy to sequences, the 99% cut-off GreenGene database 163 was used for training the feature classifier for bacteria, and the dynamically clustered UNITE ITS database 164 was used for fungi. Abundances in each sample were normalized to the number of sequences in the least abundant sample. Due to the high amount of amplified chloroplast DNA in lake water samples and in order to investigate the bacterial diversity in all samples in a comparable way, chloroplast sequences were excluded with a taxonomy-based filtering. Shannon index was calculated to study the alpha diversity. The distance and dissimilarity matrices were determined through Unifrac distances to visualize the ordination and clustering of the bacterial and fungal community composition for beta diversity analyses. The ordination patterns based on phylogenetic distance metrics were evaluated using principal coordinate analysis (PCoA). Differences in microbial community composition between samples types were assessed by non-parametric permutational analysis of variance (PERMANOVA). All analyses above were performed in QIIME2 2018.8 software package (Quantitative Insights Into Microbial Ecology) 162 . To evaluate community similarities between samples, a hierarchical clustering analysis of the taxa abundance in the communities was performed with the pheatmap and hclust packages in R 158,165 . Sequences. Bacteria GenBank MK670504-MK670553.

Data availability
The data generated during the current study are available in the GenBank repository, and are included in this published article (and its Supplementary Information files).