The soil was sampled from browning Peninsula (Figure 1) in 2005 as a part of larger biodiversity project (Siciliano et al., 2014). Due to limitations of sampling site, such as erratic boulders, frost wedges, melt ponds and unstable slopes, the spatially explicit design uses three parallel transects which together provide a reasonable sample representation across the site (Siciliano et al., 2014). A total of 18 samples was selected for this investigation, from three transects, 2 m apart, with samples taken at variable lag distances of 0, 2, 100, 102, 200, and 202 m (Figure 1C). The soil samples (1 kg) were taken from top 10 cm of the soil profile, aseptically sieved and stored at −80°C until analyzed. Samples were taken from a number of different frost boils and all frost boils that were sampled from were given arbitrary number from 1 to 42.

Comprehensive physical and chemical properties of all 18 soil samples were obtained using standard approaches (van Dorst et al., 2014). Briefly, variables, such as location, elevation, slope and aspect were obtained using a combination of global positioning system (GPS), geographic information system (GIS) and site digital elevation models. Physical variables, such as grain size was measured by sieving soil particles larger than 2 mm and separating other particles based on grain size into mud (<63 μm) and sand (63–2000 μm). Chemical analysis, such as conductivity, pH, NO2−, NO3−, PO43− and SO4− was obtained using 1 in 5 dilution of soil on the dry mass basis (mg kg⁻¹). Total carbon (TC), total phosphorus (TP), and total nitrogen (TN) were measured by combustion and NDIR gas analysis (Rayment and Lyons, 2011). The measurement of other elements concentration, such as SO₃,P₂O₅, and Cl were determined using X-Ray Fluorescence (XFR) analysis.

The SSMS setup was performed following the protocol described by Ferrari et al. (2008) with slight modifications. Briefly, approximately 3–5 g of soil was placed into a sterile Tissue Culture Insert (TCI) (Merck Millipore; MA, USA) and the soil slurry was prepared by adding a few drops of water into the TCI and mixing well using a vortex. An inoculum was prepared by adding 1 g of the same soil to 10 ml of water and mixing well using vortex. The inoculum was then allowed to stand for 5 min to settle heavy particles and 100 μl of inoculum was filtered onto a 0.22 μm, 25 mm PolyCarbonate (PC) membrane (Merck Millipore; MA, USA) using filtration manifold. The PC membrane was then placed on top of pre-prepared TCI, within a six-well plate and incubated at 8°C for 21 days.

After 21 days of incubation, the presence of adequate growth of the micro-colonies on SSMS membranes was confirmed using epi-fluorescence microscopy following SYBR green staining (Invitrogen; Paisley UK). A quarter slice of a growth membrane was cut using a sterile scalpel blade and forceps and a thin layer of heated 1% agarose was then poured into a glass slide. The slice was inverted onto the agarose and allowed to dry at 30°C for 15 min. The growth membrane was taken from the slide and a drop of Vectashield mounting medium containing DAPI (Vector laboratories Inc, Burlingame) was added onto the remaining agarose. Subsequently, 10 μl of 1:100 dilutions of SYBR green was added to the agar. To stain fungi, 10 μl of 2X Calcofluor white stain (Sigma-Aldrich, Australia) was also added to the agarose and a coverslip applied. The microbial growth was then visualized using an Olympus BX61 microscope (Olympus, North Ryde Australia), with appropriate filters for green (Bacteria) and UV (fungi) fluorescence.

The prepGEM DNA extraction protocol was used as described in Ferrari et al. (2008) with some modifications. Firstly a quarter slice of each growth membrane was placed in a PCR tube with 99 μl of 1X Buffer (diluted from 10X) and 1 μl of prepGEM enzyme. Samples were mixed and placed in a thermocycler (Bioteke corporation), following the program of 37°C for 15 min, 75°C for 15 min, and 95°C for 15 min. The microfuge tube was centrifuged at 14,000 rpm for 3 min. The membrane slice was then removed with the help of sterile tweezers and 80 μl of DNA lysate was transferred to a new PCR tube. The genomic DNA was quantified using Quanti-iT™ Picogreen® dsDNA reagent (Invitrogen, Paisley UK) following manufacturer's instructions and stored at −20°C until used.

Genomic DNA was extracted in triplicate from the soil samples using the MP FastDNA™ SPIN Kit for soil (MP Biomedicals, NSW Australia), as described (van Dorst et al., 2014). The DNA extracts were quantified using Quanti-iT™ Picogreen® dsDNA reagent (Invitrogen, Paisley UK) on a black 96 well plate at an absorbance of 520 nm using a fluorescence plate reader (SpectraMax M3 Multi-Mode Microplate Reader; Molecular Devices, CA). The DNA lysate concentrations produced ranged from 0.21 to 0.37 ng μl⁻¹ and were stored at −80°C until used further.

Bacterial ARISA was performed to allow for the selection of replicates for tag sequencing. In comparison with the metabarcoding approach, ARISA is a cheap and rapid alternative, ideal for ensuring appropriate replication in large microbial ecology studies (van Dorst et al., 2014). All soil DNA was extracted in triplicate and then ARISA was used to evaluate technical replicates and ecological patterns. One of each of the technical replicates was then selected for sequencing. Briefly, ARISA PCR was carried out targeting the bacterial ITS region using the universal primers 1392f; 5′ GYACACACCGCCCGT 3′and 5′ MAX labeled 23Sr 5′GGGTTBCCCCATTCRG3′ (Fisher and Triplett, 1999; Hewson and Fuhrman, 2006). The PCR reaction mix included 0.5 μM of forward (1392) and reverse primer (5′ MAX labeled 23Sr), 1X buffer, 2.5 mM MgCl₂ (Promega), 0.25 mM of each dNTP (Promega), and 0.8 μg μl⁻¹ of BSA (Promega), 1 unit Taq polymerase with addition of water to final reaction volume of 25 μl. The cycling temperature was 94°C for 2 min denaturation temperature followed by 30 cycles of amplification at 94°C for 30 s, 55°C for 30 s, and 72°C for 30 s with a final extension 72°C for 5 min. The fragment separation was performed at Macquarie University on an Applied Bio-system 3730 fragment analyzer (Life technology), which was further interpreted via GENEMAPPER software (Life Technology). The sequences obtained were further filtered by removing the background fluorescence. Fragments were assigned to appropriate bins and final data were transformed into peak area verses sample matrix where the single peak represented single OTU (van Dorst et al., 2014). The samples were then compared for shared and unique bacterial OTUs using the Primer 6 version 6.1.13 and PERMANOVA version 1.0.3 software and a nonmetric multidimensional scaling (nMDS) plot and a cumulative rank abundance was generated using a Bray-Curtis similarity matrix (Clarke and Gorley, 2006; van Dorst et al., 2014).

After confirming a strong correlation between technical replicates using ARISA (van Dorst et al., 2014), a representative gDNA extract from each replicate was sent for barcode tag pyrosequencing to an external facility (MR DNA, Molecular Research Laboratory, TX USA) which was performed on the Roche 454 FLX titanium platform. The research laboratory followed a single step 30 cycle PCR using Hot StartTaq Plus Master Mix Kit (Qiagen, Valencia, CA) with the following conditions: 94°C for 3 min denaturation, followed by 28 cycle of 94°C for 30 s, an annealing temperature of 53°C for 40 s and elongation at 72°C for 1 min with a final extension step at 72°C for 5 min (Dowd et al., 2008). The bacterial universal primers, 28F and 519R were used to amplify the bacterial DNA (Dowd et al., 2008). For fungi, the ITS1 and ITS2 region was targeted with the primers ITS1 and ITS4 (Gardes and Burns, 1993). Following PCR, all amplicons from different samples were mixed in equal concentration and purified using Agencourt Ampure (Agencourt Bioscience Corporation; MA, USA).

Diversity indices were generated within MOTHUR (Schloss et al., 2009) and utilized as part of the multivariate analysis. The bacterial and fungal OTU sequence datasets were transformed to square root, standardized and (dis) similarity matrix was generated using Bray-Curtis coefficient (Clarke and Gorley, 2006; Anderson et al., 2008). Using a Primer 6 version 6.1.13 and PERMANOVA version 1.0.3 software PCO plots were configured for visualization and overlaid with normalized environmental data using Person correlation (0.5) (Clarke and Gorley, 2006).

Following analysis of total microbial diversity, six of the 18 soils, were selected for cultivation experiments. The selection was based on the nMDS plot generated and the clustering of bacterial OTU abundances was performed at community similarities of 40% (samples similarity threshold within frost boils). Each represented a major cluster of community and a different frost boils across the sample landscape (Supplementary Figure 2). The matrix generated for the nMDS plot was also used for analysis of similarities (one-way ANOSIM) and factor frost boil was tested for its significance against soil environment, bacterial communities, and fungal communities.

The environmental parameters were combined into a matrix and visualized with a draftmans plot. The right skewed variables, such as elevation, slope, TN, sand, mud, NO₂, NO₃, TKP, pH, TC, PO₄, SO₃, SO₄, P₂O₅ and conductivity were log transformed then all variables were normalized. The normalized environmental variables were used to create a resemblance matrix based on the Euclidean distance between samples (Supplementary Tables 2, 3). The environmental resemblance matrix was then used for a DistLM analysis in Primer and PERMANOVA (Clarke and Gorley, 2006; Ferrari et al., 2015). The selected criteria included adjusted R² and step-wise procedures.

Fungi and bacteria were isolated by suspending 1 g of soil with 10 ml of ultrapure milliQ water and further diluted to between 10² and 10⁴. Aliquots of 100 μl were transferred onto various media including BG11 (Sigma-Aldrich Castle Hill Australia), NA (Nutrient Agar, Sigma-Aldrich Castle Hill Australia), SCA (Starch Casein Agar- US Biological, Swampscott USA) and RAVAN (Watve et al., 2000) for bacterial isolation. For fungal isolation, PDA (Potato dextrose agar- Sigma-Aldrich Castle Hill Australia), CRBA (Crook Rose Bengal Agar) and MEA (Malt Extract Agar-Sigma-Aldrich Castle Hill Australia) were used. All media were prepared to 1X and 0.1X concentrations and incubated at 8°C for 1–2 months or RT (20 ± 1°C) for 7–15 days. Bacterial isolates were sub-cultured at least three times onto NA plates and fungi were sub-cultured on PDA plates. The use of 0.1X media and SSMS were to select for “K” strategists.

Following the isolation of pure cultures, genomic DNA was extracted from bacterial isolates using a simplified FastPrep bead beating method previously applied for fungal DNA extraction (Ferrari et al., 2011). For all fungal isolates, the Fast DNA® SPIN Kit (MP Biomedicals, NSW Australia) was used following the manufacturer's instructions.

All primers required for bacterial and fungal PCR amplification were obtained from Integrated DNA Technology (IDT; MCLeans Ridge, Australia). The reaction components used for bacterial 16S rRNA gene PCR were 10 μl of 5X Go Taq buffer (Promega; Madison, USA), 4 μl MgCl₂ (25 mM), 1 μl dNTP (10 mM), 40 mM 1 μl each Forward Primer (356F) and Reverse Primer (1064R) (Winsley et al., 2012), 0.25 μl (5 U μl⁻¹) Taq polymerase, 2 μl DNA, 5 ml BSA (0.1 μg μl⁻¹), 24.75 μl water (Nuclease-free water). The PCR program consisted of an initial denaturation of 95°C for 3 min, 35 cycles of 95°C for 35 s, annealing temperature of 60°C for 30 s and an extension step at 72°C for 5 min.

Fungal internal transcribed spacer (ITS) region genes were amplified using ITS1 forward primer and ITS4 reverse primer (Gardes and Burns, 1993) as described by Ferrari et al. (2011) with slight modification. The PCR reaction mixture consisted of 10 μl of 5X Go Taq buffer, 4 μl MgCl₂ (25 mM), 1 μl dNTP (10 mM), 40 mM 1 μl each Forward Primer (ITS1) and Reverse Primer (ITS4) (Gardes and Burns, 1993), 0.25 μl (5 U μl⁻¹) Taq polymerase, 2 μl DNA, 5 μL BSA (0.1 μg μl⁻¹) and 24.75 μl water (Nuclease-free water). The PCR program consisted of a 94°C initial denaturation step for 2 min, followed by 30 cycles of 94°C for 45 s, an annealing temperature of 55°C for 30 s and a final extension step at 72°C for 5 min. PCR amplicons were then visualized on 2% agarose gel stained with SybrSafe (Invitrogen, Paisley UK). The PCR product was treated with two restriction enzymes, Hinfl (Promega, Madison, USA) with the cleavage site 5′-G↓ANT C-3′, 3′-C TNA↑G-5′ and Rsal (Promega, Madison, USA) with cleavage site 5′↓GTAC-3′-3′ CATG↑ 5′. The reaction component included 0.17 μl restriction enzyme (Hinfl or Rsal), 2 μl 10X buffer B for Hinfl and Buffer C for Rsal, 0.2 μl 100X BSA, 2.63 μl water, and 15 μl of PCR product. The mixture was incubated overnight at 37°C and the digest was loaded onto 2% agarose gel and analyzed for band sizes. From selected isolates, PCR products (bacteria and fungi) were then purified by using the QIAquick PCR purification kit from Qiagen following the manufacturer's instructions. The purified PCR amplicons were sequenced at the Ramaciotti Center for Gene Function Analysis (UNSW, Australia) for Sanger sequencing using ABI 3730 Capillary sequencer. The data received were used for isolate identification by blasting (BLASTn) sequences against NCBI database (Johnson et al., 2008).

To compare the OTUs recovered using traditional cultivation and the SSMS from soil samples, the bacterial and fungal OTUs recovered using both cultivation methods were combined with the OTUs identified from total soil community sequencing. For cultured bacteria, the 16S rRNA gene sequences obtained from Sanger sequencing were realigned and trimmed to the same length as the tag sequences obtained from the soil and SSMS enrichments using the MOTHUR pipeline. The common bacterial and fungal OTUs were identified by clustering sequences at 97% identity using USEARCH (Edgar, 2010).

Bacterial and fungal sequences obtained from the SSMS have been deposited in NCBI under accession number PRJNA350563. For cultured bacterial isolates the sequences were deposited into NCBI under accession numbers KY432693-KY432726, and fungal isolates under KY432734-KY432749 and KY608091. Pyrosequencing data for the soil is available at https://data.aad.gov.au/metadata/records/soil_bacteria_fungi.

A total of 57,893 16S rRNA and 62,547 ITS quality checked bacterial and fungal reads were recovered after processing the raw datasets from total soil gDNA. Using a similarity threshold of 96% to cluster bacterial sequences and 97% to cluster fungal sequences (Siciliano et al., 2014), a total of 2063 bacterial and 279 fungal OTUs were obtained. Overall bacterial diversity was much greater than fungal diversity (Supplementary Table 1). Rarefaction curves revealed that fungal richness reached asymptote, whilst the bacterial OTU richness was still increasing (Supplementary Figure 1). Soil bacterial diversity detected by tag pyrosequencing spanned 36 phyla, 84 classes, 214 families and 407 genera, whilst fungal diversity was limited to four phyla, spanning 14 classes, 59 families, 104 genera and one unclassified fungi.

Actinobacteria dominated the bacterial communities present, comprising >36% of the total relative abundance across the site. Chloroflexi was the next most abundant phyla, followed by Acidobacteria, Cyanobacteria and Proteobacteria; together accounting for over 54% of total relative abundance within the entire 18 samples analyzed (Supplementary Figure 3). For fungal diversity, Ascomycota dominated all soil samples analyzed, accounting for 93% of the total relative abundance, followed by Basidiomycota (5%) and Zygomycota (1%) (Supplementary Figure 4).

A large proportion of bacterial and fungal OTUs were not shared between culture dependent and independent techniques (Figure 4). For example, 135 OTUs detected were common in soil and SSMS datasets; only four OTUs were common in SSMS and artificial media and one OTU was present in the artificial media and soil datasets (Figure 4). In fact, there were more unique bacterial genera retrieved than shared between techniques. For instance, 824 bacterial OTUs were not retrieved from culture-dependent techniques. Overall, only 13 unique OTUs were obtained from artificial-cultivation. Surprisingly, 306 bacterial OTUs recovered from SSMS were also not detected in the other two approaches suggesting minor taxa are enriched for using the SSMS (Figure 4).

The Shannon diversity index for fungi retrieved from the soil gDNA dataset was 3.51 markedly lower than for bacteria (5.5) (Supplementary Table 1). A similar recovery pattern of bacterial and fungal communities was observed from the SSMS enrichments with a Shannon index of 3.15 for bacteria and 2.57 for fungi. Combined, a total of 663 fungal OTUs were recovered using three approaches, yet no common OTU was recovered from all three methods used. Altogether, six of the fungal OTUs recovered were common in the soil and SSMS enrichments, four OTUs were common in enrichments and artificial media, while three OTUs were present in the soil and artificial media datasets. Similar to the bacterial data, 596 unique fungal OTUs were retrieved from the soil dataset only, 47 OTUs from SSMS enrichments and seven OTUs from cultivation (Figure 4).

Comparing recovery of individual phylum level OTUs from both soil and SSMS, SSMS recovered 39.5% of unique Proteobacteria, 23% of Actinobacteria, 27% of Bacteroidetes and 24% of Firmicutes (Supplementary Table 4). The shared OTUs were 10% for Proteobacteria, 11% for Actinobacteria, 12% for Bacteroidetes and 3.5% for Firmicutes. The recovery of fungal unique Ascomycota OTUs from soil was 90% and SSMS was 9%. For Basidiomycota unique OTUs obtained were 87% from soil and 7.23% from SSMS (Supplementary Table 5).

The soils of Browning Peninsula harbor a bacterial community dominated by Actinobacteria (up to a 36% relative abundance), followed by Chloroflexi, Cyanobacteria, Proteobacteria and Acidobacteria. The high abundance of Actinobacteria is in stark contrast to Arctic and other Antarctic sites commonly dominated by Proteobacteria, Cyanobacteria, and Acidobacteria (Steven et al., 2007, 2013; Koyama et al., 2014). For example, the McMurdo Dry Valleys, a well-studied site of Antarctica, has been found to be dominated by Proteobacteria (Clocksin et al., 2007; Cary et al., 2010); Miers Valley, is dominated by Bacteriodetes, Proteobacteria, and Actinobacteria (Stomeo et al., 2012), and Victoria Land is comprised of a heterogeneous bacterial community encompassing Proteobacteria, Acidobacteria, Actinobacteria, Chloroflexi, and Cyanobacteria (Kim et al., 2015). The variability in diversity structures in each of these sites may be due to variable organic matter content (Takebayashi et al., 2007; Bell et al., 2013; Ji et al., 2015). Previously, the dominance of Proteobacteria was linked with areas rich in higher organic matter (Bell et al., 2013) and moisture content (Horn et al., 2014; Niederberger et al., 2015), while dominance of Actinobacteria has been documented in arid soils (Niederberger et al., 2015), and linked to low moisture and nutrients (Takebayashi et al., 2007). Further biotic and abiotic factors, such as plants, penguins, birds, seal carcasses, and human activity are also reported to affect microbial community distributions across Antarctica (Cowan et al., 2014; Wang et al., 2015).

Although fungi are more tolerant of freezing and thawing compared to bacteria (Sharma et al., 2006), the fungal diversity of Browning soil was markedly lower than that of the bacterial community. Of those fungi present, Browning soil was dominated by Ascomycota (>97% of the total relative abundance) and was limited to four phyla only. In contrast, nearby Eastern Antarctic sites that do not contain frost boils harbor at least 60% more fungal richness than Browning Peninsula (Ferrari et al., 2015). A lack of tolerance of extreme variation in the salinity, combined with continuous mixing of soil due to cryoturbation, may have limited the fungal diversity at Browning Peninsula (Ferrari et al., 2015). Dispersal limitation across the site due to the presence of hard-edge barriers surrounding frost boils, may also contribute to the reduced fungal richness observed compared to highly connected Antarctic sites, as the larger size of fungi limits their dispersal compared to bacteria (Rao et al., 2011; Ferrari et al., 2015). Moreover, the lack of vegetation (Stewart et al., 2011), and fungal vectors including insects and animals could contribute to reduced fungal diversity in Antarctica compared with those of the Arctic. Fungi has been shown to have symbiotic association with plants (Geml et al., 2012; Timling et al., 2012) and no plant life has been documented at Browning Peninsula (Stewart et al., 2011).

The environmental factors chlorine, phosphate, aspect and elevation were observed to affect the bacterial community at Browning, whilst none of the analyzed environmental parameters were observed to be statistically significant to affect fungal community variation. Of the environmental parameters analyzed, chlorine was primarily correlated to the bacterial community composition, contributing 19% of the total biological variance. Chlorine was not significant to fungal community variation here, but has been previously observed to have strong effects on polar fungi (Siciliano et al., 2014). Elevation was also observed to affect the bacterial distribution and several in-depth studies of effect of elevation on microbial community have revealed dramatic shifts in the microbial taxa along elevation transects (Bryant et al., 2008; Cowan et al., 2010; Shen et al., 2013). It is also well established that the soil heterogeneity of Arctic and Antarctic soil depends on organic matter (total carbon) and nitrate (total nitrogen) availability (Michaelson et al., 2012; Shen et al., 2013; Siciliano et al., 2014). The reduced bacterial and fungal diversity of Browning Peninsula soil may, therefore, be a result of the relatively poor nutrient content of these soils (Chong et al., 2009). The phosphate content and aspect were also correlated to the bacterial community contributing to 19% of the total variation (Table 3). In contrast, fungal diversity distributions have been previously linked with pH (Connell et al., 2006; Arenz and Blanchette, 2011), conductivity (Arenz and Blanchette, 2011) and moisture content (Connell et al., 2006). However, none of the environmental factors were significantly correlated with the fungal community distribution here, which may be due to the unique landscape and limited diversity present.

The present investigation was the first study to culture fungi using the SSMS, with 30 non-lichenized microfungi including Toxicocladosporium, Exophiala, Hortaea, and Cadophora species selected for using the approach. From the two phyla recovered, Ascomycota members enriched for included Aureobasidium, Cladosporium, Nigrosphora and Cadophora, while Basidiomycota members included Malassezia, Hortaea, and Rhodotorula (Figure 3). Most of these fungal genera have been previously isolated from soils, plants and animals. However, the SSMS appeared to facilitate the cultivation of a wide range of microfungi, as demonstrated by the recovery of the endemic mesophilic psychrotolerant microfungi, such as Aureobasidium and Cladosporium (Zucconi et al., 2011) to host specific genera, such as Nigrosphora, Malassezia genera; a plant and animal pathogen (Hudson, 1963; Velegraki et al., 2015), Cadophora; a “soft rot” wood decomposer isolated from Antarctic hut wood (Blanchette et al., 2004; Ludley and Robinson, 2008), Rhodotorula; previously isolated from Windmill Island, Antarctica soils (Gesheva, 2010) and Hortaea; a halotolerant black yeast (Plemenitas et al., 2008). For bacteria, the SSMS enrichments led to an increase in Proteobacteria relative abundances up to 58%, highlighting the system's selective bias toward heterotrophs (Ferrari et al., 2008). Under the conditions used, the SSMS failed to enrich for Cyanobacteria, Chlorobi and Chloroflexi. Further modifications to include a light source may aid in the isolation of these phototrophic taxa.

The most commonly isolated bacteria Arthrobacter and Streptomyces were previously isolated from Antarctic cold deserts (Smith et al., 2006), post-glacial soils (Zdanowski et al., 2012) and Princess Elisabeth Station soil (Peeters et al., 2011), temperate soils and variety of other sources (Kaewkla and Franco, 2013). For fungal isolates, Geomyces, Thelobolus and Phoma have been previously recovered from Antarctic soils (Frate and Caretta, 1990; Azmi and Seppelt, 1998; Arenz et al., 2006; Connell et al., 2006; Arenz and Blanchette, 2011; Godinho et al., 2013). Interestingly, Chrysosporium sp., Mortierella gamsii, Mycelia sterilia, Phoma sp., and Thelobolus microspores have also been previously isolated from Browning Peninsula (Azmi and Seppelt, 1998). Similar to Devriesia the requirement for a heat pre-treatment to germinate cells before culturing into artificial media (Onofri et al., 2011) may have been a limiting factor in the diversity of cultured members. The recovery of cold-adapted bacterial strains, such as Dyella have been proposed to degrade single molecules, such as N-Acylhomoserine Lactones (Chen and Chan, 2012), and UV radiation-resistant fungi, such as Phoma (Hughes et al., 2003) may have significant biotechnological potential that warrants further investigation.

Of the 1284 bacterial OTUs and 663 fungal OTUs detected using multiple approaches here, tag pyrosequencing from soil gDNA detected 75% of the total bacterial and 91% of total fungal community diversity, which is in line with previous studies reporting ≤60% of the total bacterial diversity detected using sequencing alone (Vaz-Moreira et al., 2011; Stefani et al., 2015). It is well established that very low percentages of bacteria have been recovered into pure culture (Handelsman, 2004; Ferrari et al., 2008), and here 1.7% of the total bacteria diversity was recovered into artificial media. This recovery increased to 34% within the SSMS enrichments, highlighting the potential of this approach to recover novel species (van Dorst et al., 2016). This is comparable to the recovery of 50% of total bacterial diversity using the ichip method (a similar culturing technique) described by Nichols et al. (2010).

Only four of the total bacterial OTUs detected were present in all three techniques used. The absence of consistency between techniques, or a “culture clash” has been consistently observed between clone libraries and cultivation attempts (Maturrano et al., 2006). Previously, higher community diversity (2.4% of bacterial and 8.2% of fungal communities) has been shared between artificial media and tag pyrosequencing from hydrocarbon contaminated soils(Stefani et al., 2015; van Dorst et al., 2016). The rapid growth of microfungi, particularly yeasts on the SSMS may have occurred following the addition of water to the soils (Vishniac, 1996) with freely available water potentially aiding the germination of spores.

Microfungi are difficult to culture and enrichment of microfungi using the SSMS offers a new opportunity to harness the biotechnological potential of cold-adapted Antarctic fungi. Despite inherent bias associated with the techniques used, such as selection of phenotypic characteristics of colonies from artificial media (Tytgat et al., 2014), DNA extraction (Evans and Seviour, 2012; Plassart et al., 2012), PCR amplification (Schloss et al., 2011) and inadequate sequence-depth when carrying out 454 tag pyrosequencing (Kim et al., 2011; Kumar et al., 2011; Lagier et al., 2012), multiple approaches delivered a comprehensive view of the soil microbial community of Browning Peninsula, unveiling an ecosystem dominated by rare unclassified bacterial taxa combine with extremely low fungal diversity.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.