Long-term increase in snow depth leads to compositional changes in arctic ectomycorrhizal fungal communities

Many arctic ecological processes are regulated by soil temperature that is tightly interconnected with snow cover distribution and persistence. Recently, various climate-induced changes have been observed in arctic tundra ecosystems, e.g. shrub expansion, resulting in reduction in albedo and greater C ﬁxation in aboveground vegetation as well as increased rates of soil C mobilization by microbes. Importantly, the net effects of these shifts are unknown, in part because our understanding of belowground processes is limited. Here, we focus on the effects of increased snow depth, and as a consequence, increased winter soil temperature on ectomycorrhizal (ECM) fungal communities in dry and moist tundra. We analyzed deep DNA sequence data from soil samples taken at a long-term snow fence experiment in Northern Alaska. Our results indicate that, in contrast with previously observed responses of plants to increased snow depth at the same experimental site, the ECM fungal community of the dry tundra was more affected by deeper snow than the moist tundra community. In the dry tundra, both community richness and composition were signiﬁcantly altered while in the moist tundra, only community composition changed signiﬁcantly while richness did not. We observed a decrease in richness of Tomentella , Inocybe and other taxa adapted to scavenge the soil for labile N forms. On the other hand, richness of Cortinarius , and species with the ability to scavenge the soil for recalci-trant N forms, did not change. We further link ECM fungal traits with C soil pools. If future warmer atmospheric conditions lead to greater winter snow fall, changes in the ECM fungal community will likely inﬂuence C emissions and C ﬁxation through altering N plant availability, fungal biomass and soil-plant C-N dynamics, ultimately determining important future interactions between the tundra biosphere and atmosphere.


Introduction
Artic ecosystems are beginning to exhibit significant shifts in ecosystem structure and function induced by changes in climatic conditions (Elmendorf et al., 2012;Tape et al., 2012). Despite interannual and regional variability, global mean surface temperature have consistently increased since the late 19th century (Collins et al., 2013). In the Arctic, temperatures have risen between 0.06 to 0.1°C per yr, while the global average increase has been ca. 0.017°C per yr during the past 30 years (Comiso & Hall, 2014). These temperature increases have already had major consequences, including accelerated summer ice loss, extended periods of open water in the Arctic Ocean and delayed autumn freeze up (Stroeve et al., 2014). At the same time, precipitation in the Arctic has increased (greatly exceeding the global average increase), especially during the cold season, where most precipitation falls as snow (Kattsov & Walsh, 2000;Screen & Simmonds, 2012). Additionally, state-of-the-art models predict further increases in the twenty-first-century, possibly by more than 50% of the current precipitation, leading to thicker snow cover (Collins et al., 2013;Bintanja & Selten, 2014). Deeper snow would have a suite of consequences for tundra ecosystems. These include protection from the abrasive wind (Liston et al., 2002;Sturm et al., 2005;Blok et al., 2015), warmer winter soil temperatures and increased soil moisture with subsequent effects on thaw depth and C storage (Natali et al., 2012(Natali et al., , 2014, N turnover DeMarco et al., 2011), plant phenology and mineral nutrition (Borner et al., 2008;Leffler & Welker, 2013;Pattison & Welker, 2014), vegetation composition (Wahren et al., 2005;Welker et al., 2005) and soil microbial respiration (Aanderud et al., 2013;Natali et al., 2014). However, with the exception of Buckeridge & Grogan (2008) that compared bacterial and fungal biomass growth responses, how arctic soil fungal communities may respond to changes in winter snow depth conditions is still largely unknown.
Microbial activity in the Arctic has been shown to increase due to higher winter soil temperatures inducing changes in the nitrogen (N) cycle dynamics, particularly in moist tussock tundra and less so in dry heath tundra in Arctic Alaska DeMarco et al., 2011;Natali et al., 2014;Pattison & Welker, 2014). In the Arctic, fungi are considered to constitute the bulk of soil microorganisms biomass (Callaghan et al., 2005) and have a key ecological role. Hobbie & Hobbie (2006) estimated that up to 86% of the N obtained by tundra plants is via mycorrhizal fungi, in exchange, plants can allocate between 10 to 20% of their photosynthatederived C to their fungal partners (Harley, 1971;Hobbie, 2006), constituting an important pool of soil C. Additionally, these exchanges might be positively correlated, i.e., increased allocation of plant C to the mycorrhizal partner might lead to increased uptake of N from the soil pool and subsequent delivery to the plant host (Talbot & Treseder, 2010). The limiting step in soil N cycling is the breakdown of macromolecular organic compounds, particularly the depolymerization of proteins (Schimel & Bennet, 2004;Jones & Kielland, 2012) that in high-latitude ecosystems has been correlated with fungal biomass (Wild et al., 2013), and is particularly attributed to ectomycorrhizal (ECM) and ericoid mycorrhizal fungi (Read & Perez-Moreno, 2003). Recently, several studies reported major changes in the arctic fungal mycorrhizal communities in response to summer warming (Deslippe et al., 2011;Geml et al., 2015;Morgado et al., 2015;Semenova et al., 2015), with the fungal community of moist tussock tundra typically showing more pronounced response than the dry heath tundra, including potential shifts in functional traits and the subsequent ecosystem processes. However, possible effects of increased winter soil temperatures on the richness and compositional structure of soil fungal communities have not yet been investigated.
Tundra plant community responses to increased winter snow depth include a combination of shifts in community composition as well as increases in net plant productivity (Borner et al., 2008;Natali et al., 2014) and plant N tissue concentrations (Leffler & Welker, 2013). At the community level, the general trends are increases in shrub coverage and litter layer and, decrease in lichens, bryophytes, and leaf C:N ratio (Welker et al., 2005;Wipf & Rixen, 2010;Pattison & Welker, 2014). Wahren et al. (2005) and Mercado-D ıaz (2011) reported (from the same experimental plots that we used in our study) an increased coverage of several species of deciduous and evergreen shrubs, and a sedge species. Although most of these plants are highly dependent on root-associated fungi, especially ECM fungi, to acquire soil nutrients, how soil fungal community changes in response to deeper snow remains uncertain. Here, we focus on ECM fungal community responses to long-term increased snow depth and the associated warming soil temperatures across the dry heath and moist acidic tussock tundra.
Ectomycorrhizal fungi represent the most taxonomically diverse fungal guild in the Arctic tundra (Gardes & Dahlberg, 1996;Geml et al., 2012;, and provide crucial roles in soil-root interaction, particularly in plant N uptake (Read et al., 2004;Ekblad et al., 2013) and in soil C dynamics (Clemmensen et al., 2013;Averill et al., 2014). Recently, an increasing amount of studies on fungal functional traits are amassing valuable insights into the role of the community structure in potential ecosystem functions (e.g. reviewed in Treseder & Lennon, 2015). For example, Hobbie & Agerer (2010), gathered evidences from d 15 N patterns and argued that ECM fungi have two main strategies for growth and nitrogen acquisition that match the extramatrical mycelium (EMM) characteristics of the ectomycorrhizae. ECM fungi with low abundance of EMM and hydrophilic mycorrhizae with contact, short-distance and medium-distance smooth hyphal exploration types (ETs) were argued to focus on uptake of labile nitrogen (N) forms, such as amino acids, ammonium, and nitrate. Supporting this hypothesis, many taxa composing this group showed limited growth in media of protein culture conditions (Lilleskov et al., 2011). On the other hand, the ECM fungi with high abundance of EMM with medium-distance fringe, medium-distance mat and long-distance ETs with hydrophobic rhizomorphs (or mycelial cords), likely focus on widely dispersed and spatially concentrated soil resources requiring efficient long-distance translocation. Such investment in exploratory hyphae is unlikely to rely on labile substrates under low nutrient availability; therefore, this group of taxa likely requires enzymes with hydrolytic and oxidative capabilities in order to access nonlabile N forms, such as proteins (Hobbie & Agerer, 2010). Supporting this hypothesis some studies pointed to increased exoenzyme activity in ECM fungi with abundant EMM (Tedersoo et al., 2012;Talbot et al., 2013). Another example of a fungal trait and a potential ecosystem function is the presence of melanin in hyphal cell walls, which was thoroughly discussed by Koide et al. (2014). Therefore, it seems reasonable to com-pare community composition of fungi with certain functional traits in light of contrasting environmental conditions which in turn might feedback to altered ecosystem functions.
This research focuses on the effects of long-term increased snow depth on ECM basidiomycete communities. Based on the evidence previously stated, we hypothesize that: 1) ECM fungal community composition is strongly affected by increased snow depth, and that the response will be more pronounced on the moist tundra community; and 2) changes in ECM fungal community composition will reflect altered patterns in vegetation, soil nutrient pools and moisture, induced by the increased snow depth. Therefore, we expect altered patterns in the ECM fungal community functional traits.

Study site and experimental design
The International Tundra Experiment (ITEX) (Henry & Molau, 1997;Welker et al., 1997) study site that we sampled is located on the northern foothills of the Brooks Range, at the Toolik Lake Field Station. The area lies in the Arctic tundra biome within the bioclimatic subzone E, which covers approximately 36% of the Arctic dry land surface . The mean air annual temperature is À7°C and annual precipitation ranges between 200 mm and 400 mm with approximately 50% falling as snow. The average snow depth is 50 cm (DeMarco et al., 2011). The snow fence experiment was established in the summer of 1994 in moist tussock and dry heath tundra (Jones et al., 1998;Walker et al., 1999;Welker et al., 2000). The snow fences are 2.8 m high and 60 m long, inducing leeward drifts of ca. 60 m long (Walker et al., 1999;Pattison & Welker, 2014). Our sampling was focused on the intermediate zone near the center of the experimental setup, corresponding to ca. 1-1.5 m winter snow depth. Although the deeper snow slightly shortens the growing season by ca. 5-8 days, this does not affect plant phenology significantly (Borner et al., 2008). The average winter soil temperatures at 2 cm depth were À2.9°C (AE0.2) and À4.7°C (AE0.2) in the increased snow depth plots and in the control plots, respectively (Pattison & Welker, 2014).
We sampled soil at the end of July 2012 from two tundra types, the dry heath and moist tussock tundra experimental sites. In each tundra type, we sampled five plots/replicates with increased snow depth and five plots with ambient snow depth ('control plots'). The control plots were located adjacent to the snow fences (the experimental setup was described in detail by Walker et al., 1999). Each replicate consisted of five soil cores of 2 cm diameter and 20 cm depth. Both organic and mineral layers were included in the soil cores, while undecomposed litter, moss, and coarse roots were removed. For each replicate the soil cores were thoroughly mixed and kept frozen until lyophilization. In total we sampled 100 soil cores across 20 plots of ca. 1 m 2 each.
The vegetation of the dry heath tundra is characterized by Dryas octopetala, Salix polaris, Vaccinium spp. and fruticoselichens, while the moist tussock tundra is dominated by Betula nana, Salix pulchra, and the sedge Eriophorum vaginatum. Detailed descriptions of the plant communities can be found in Walker et al. (1999) and Kade et al. (2005), and their detailed response to the altered snow depths in Walker et al. (1999), Wahren et al. (2005), Welker et al. (2005), Mercado-D ıaz (2011), Pattison & Welker (2014).

Molecular work and sequence quality control
The DNA extraction, PCR protocol, Ion Torrent sequencing and data clean-up procedures were described in detail elsewhere (Geml et al., 2014a). Briefly, for each sample we carried out two independent DNA extractions, using ca. 1 ml of lyophilized soil and pooled them to optimize extraction homogenization. In the PCR we targeted the ITS2 region of the nuclear ribosomal internal transcribed spacer that is currently accepted as the universal barcode marker for fungi (Schoch et al., 2012). We used primers fITS7 (Ihrmark et al., 2012) and ITS4 (White et al., 1990). The ITS4 primer was labeled with sample-specific Multiplex Identification DNA-tags (MIDs). The amplicon library was sequenced using an Ion 318 TM Chip by an Ion Torrent Personal Genome Machine (PGM; Life Technologies, Guilford, CT, U.S.A.) at Naturalis Biodiversity Center. For the initial clean-up of the raw sequence data we used the online platform Galaxy (https://main.g2.bx.psu.edu/root), in which the sequences were sorted according to samples. Primers and adapters were removed. We used a parallel version of MOTHUR v. 1.32.1 (Schloss et al., 2009) for subsequent sequence analyses. Sequences shorter than 150 bp and longer than 400 bp were removed following Blaalid et al.  Morgado et al. (2015) and Semenova et al. (2015), because these tend to be low-quality reads. The quality-filtered sequences were normalized following Gihring et al. (2012) by random subsampling so that each sample contained equal number of sequences. We then clustered the sequences into operational taxonomic units (OTUs) using OTUpipe (Edgar, 2010) with the simultaneous removal of putatively chimeric sequences using de novo and reference-based filtering using the curated dataset of fungal ITS sequences of Nilsson et al. (2011), with the default settings. We used a 97% sequence similarity clustering threshold following many other fungal ecology studies (e.g. O'Brien et al., 2005;Higgins et al., 2007;Geml et al., 2008;Geml et al., 2009;Amend et al., 2010;Tedersoo et al., 2010;Geml et al., 2012;Kauserud et al., 2012;Brown et al., 2013;Blaalid et al., 2013;Geml et al., 2014b;Davey et al., 2015). Global singletons were discarded from further analysis. The reference database published by Kõljalg et al. (2013) was used to determine the taxonomic affinity of the OTUs using USEARCH v7 (Edgar, 2010). OTUs with less than 80% similarity to any identified fungal sequence were excluded from the final analysis due to unreliable classification, and/ or uncertainty regarding their ecological role. A representative sequence of each OTU was deposited in GenBank under the accession numbers KP827673 -KP828017. Because GenBank only accepts sequences with more than 199 bp, and we included sequences ranging 150 to 400 bp in our dataset, we include a representative sequence of each OTU as Supporting Information (Data S1).

ECM fungal database and EMM determination
We followed Tedersoo & Smith (2013) to determine ECM basidiomycete genera. For most OTUs we used a ≥ 90% sequence similarity to determine genera. Because Sebacinales have a diverse ecology we selected ECM OTUs based on their supported phylogenetic placement (with ≥70% bootstrap and/ or ≥0.95 posterior probability) among sequences of taxa that were morphologically confirmed as ECM published by Glen et al. (2002), Urban et al. (2003), Ryberg et al. (2009) and.
To determine the EMM characteristics, we followed the work of Agerer (2006), Tedersoo & Smith (2013) and consulted the DEEMY database (http://deemy.de, accessed in November, 2014an information system for the characterization and determination of ectomycorrhizae). In the genus Russula, if no EMM information was available for the species of interest, we assumed the EMM characteristics based on the closest species with known characteristics. To determine the closest species, we followed the phylogenetic study by Miller & Buyck (2002). Similarly, for OTUs of the genus Hebeloma, we followed the phylogenetic study by Boyle et al. (2006).

Statistical analysis
For each replicate, we calculated rarefied OTU accumulation curves using the R package Vegan (Oksanen et al., 2012) and determined the Good's coverage (complement of the ratio between the number of local singletons and the total sequence counts) for the fungal OTUs to estimate the exhaustiveness of our deep sequencing effort. Beta diversity, for all ECM basidiomycetes OTUs and for the ECM genera with the highest OTU richness, in each tundra type and treatment combination was calculated following Baselga (2010) with the R package betapart (Baselga & Orme, 2012), with the function beta.multi using a Sørenson dissimilarity matrix. This function computes community dissimilarity accounting for the spatial turnover and nestedness. OTU presence was defined as more than four sequences on a per sample basis following the suggestion of Lindahl et al. (2013) to minimize false positives (e.g. OTUs that are common in one sample, but may be low-abundant contaminants in others). Due to uncertainty of sequence abundance as indicator of species abundance in the samples (Amend et al., 2010), we carried out analyses with two types of data transformations. First, we transformed the data into presenceabsence matrix. However, given that the most abundant and the rarest OTUs have equal weight in presence-absence datasets, we also wanted to see if taking read counts in consideration influences the results. Therefore, as a second dataset, we used square-root transformed sequence abundance to moderate the influence of OTUs with extremely high sequence counts, while maintaining some approximation of template abundance that may reflect ecological significance. We used PC-ORD v. 5.32 (McCune & Grace, 2002) to run nonmetric multidimensional scaling (NMDS) on a primary matrix of experimental plots by OTUs and a secondary matrix of plots by OTU richness per taxon, EMM characteristics and sequence counts. The dataset was subjected to 500 iterations per run using the Sørensen similarity (Bray-Curtis index) and a random starting number. We also calculated the Pearson's correlation coefficient (r) values between relative OTU richness per taxon and axes 1 and 2. We tested whether fungal communities were statistically different across the treatments using a multiresponse permutation procedure (MRPP) and determined any preferences of individual OTUs for either control or increased snow depth plots in dry and moist tundra using Indicator Species Analyses (Dufrêne & Legendre, 1997) as implemented in PC-ORD v. 5.32. We also tested for significant differences in OTU richness across the dry and moist tundra control and deeper snow plots, per taxa (genera) and EMM characteristics using Student's t-test. Sequence abundance was also compared between each tundra type and treatment combination using Student's t-test. Correlation coefficients were calculated as implemented in Microsoft Excel v. 2010 between the most OTU-rich genera and the hyphal exploration types that had been combined in two functional groups: (I) contact, short-distance, medium-distance smooth, and (II) mediumdistance fringe and long-distance ETs. The Venn diagram for the whole community and genera with higher OTU richness was also calculated, using the online version of the publication by Oliveros (2007).

Results
Through the pipeline: from raw data to taxonomic diversity We obtained 3 960 925 sequences with a median length of 268 bp. After quality control and random subsampling we retained 1 161 160 sequences with a mean length and standard deviation of 255.1 AE 52.7 bp. Clustering the sequences at 97% similarity generated 7015 OTUs, excluding global singletons and putative chimeric OTUs, of which 459 ECM basidiomycete OTUs were retained for further analyses (Data S2). Across all treatments, ECM fungi were represented by 23 genera classified in seven orders (Table 1, Fig. 1). Overall, Cortinarius and Tomentella were the most OTU-rich genera, with 125 (ca. 27%) and 124 OTUs (ca. 27%), respectively, followed by Inocybe (79 OTUs, 17%) and Russula (40 OTUs, 9%), with the remaining genera having less than 5% of the OTUs per genus. The order Agaricales had by far the highest OTU richness (224 OTUs, ca. 49%), followed by Thelephorales (128 OTUs, ca. 28%), Russulales (57 OTUs, ca. 12%), Cantharellales (33 OTUs, ca. 7%), Sebacinales (11 OTUs, ca. 2%), Boletales (4 OTUs, ca. 1%), and Atheliales (2 OTUs, ca. 1%). The analysis with sequence abundance resulted in similar patterns as the OTU richness analysis (Data S3). The recovered OTU richness was higher than in previous publications that used similar methods to investigate arctic ECM fungal communities, but genera diversity and patterns of genera richness were in general agreement (Bjorbaekmo et al., 2010;Geml et al., 2012;Morgado et al., 2015). The asymptotic rarefaction curves (Fig. 2a) and estimated Good's coverage (Fig. 2b) indicate that the deep sequencing allowed a very high OTU coverage and that most fungi in the samples were sequenced. It is important to note that OTU delimitation based on sequence similarity cut-off is a proxy for species delimitation with some inherent uncertainties. While the methodology we used is routine practice in fungal ecology and 2-3% ITS sequence divergence usually delimit different species in many basidiomycete lineages (Hughes et al., 2009), deep sequencing may slightly overestimate the number of species. Therefore, we consider the estimated absolute number of OTUs of secondary importance; rather, we think that the extent and direction of changes in richness among treatments (discussed below) are of primary importance, because these give us insights into trends that the fungal groups exhibit in response to experimental manipulations.

Overall results
The NMDS analyses of the presence-absence and square-root sequence abundance matrices gave similar results. However, the final stress values were somewhat lower in the square-root abundance ordinations, therefore, we continued with these for the main interpretations, while the presence-absence ordinations are shown in Supporting Information (Data S4). For the total dataset, the square-root abundance NMDS analysis resulted in a 2-dimensional solution with a final stress and instability of 0.1062 and <0.00001 respectively. The results of the Monte Carlo test indicated that the two dimensional solutions using the real data were significantly better than occurrences by chance (P < 0.01). The coefficients of determination for the correlations between ordination distances and distances in the original n-dimensional space were axis 1: r 2 = 0.599; axis 2: r 2 = 0.240; total r 2 = 0.839; orthogonality = 88.3%. The NMDS ordination plot was orthogonally rotated by the treatment to visualize correlations between snow depth effect and fungal community composition in general (Fig. 3a). The MRPP analysis indicated a clear distinction between dry and moist ECM community composition (P < 0.0000001, A = 0.14601). The NMDS and MRPP analysis with the presence-absence matrix results were similar (Data S4a). Beta diversity values in the dry tundra were similar among control and treatment plots and among plots per treat-ment; while in the moist tundra the values were generally higher among control and treatment plots than among plots per treatment. Species turnover was substantial in all comparisons (Table 2).
Across the ambient snow plots of both dry and moist tundra, Tomentella had the higher OTU richness with 103 OTU (ca. 29%), followed by Cortinarius with 78 OTUs (ca. 22%), Inocybe with 66 OTUs (ca. 19%) and Russula with 28 OTUs (ca. 8%). All the other genera had less than 5% of the OTUs per taxa and combined solely represented ca. 22% of the OTUs. On the other hand, across the deeper snow plots, Cortinarius had the higher OTU richness with 78 OTUs (ca. 36% of all OTUS), followed by Tomentella with 45 OTUs (ca. 20%), Inocybe with 28 OTUs (ca. 13%) and Russula with 24 OTUs (11%). All the other genera had less than 5% of the OTUs per taxa (Fig. 1). Differences between the ambient and deeper snow plots were also evident at the order level. Agaricales and Russulales had an increased OTU richness in deeper snow areas, while Thelephorales and Cantharellales had a decrease. Approximately 53% of the OTUs were only present in the ambient snow plots, ca. 24% were solely found in the deeper snow plots, and the remainder 23% present in both (data not shown). There was a significant decrease in OTU richness from the ambient to the deeper snow plots (P = 0.0377), with the control plots having on average 66.2 AE 24.5 OTUs, while the deep snow plots had 43.8 AE 28.4 OTUs (Table 1). Together the contact, short-distance and medium-distance smooth ET represented the group with highest OTU richness in the control plots with an average of 46.1 AE 17.9 OTUs per plot, while the medium-distance fringe and long-distance ETs group had an average of 16.6 AE 7.09 OTUs per plot. Comparing with the OTU richness values of the deep snow plots, the first group had a significant decrease (P = 0.0042, 24.6 AE 14.05 average OTUs per plot); whilst the later did not change significantly (P = 0.36, 15.2 AE 12.71 average OTUs per plot) (Data S5). The overall patterns of changes in functional traits were also depicted when comparing the OTUs in the ambient with the deeper snow plots (Fig. 1).

Dry heath tundra
The NMDS analysis of the square-root sequence abundance matrix resulted in a 2-dimensional solution with a final stress of 0.0763 and a final instability < 0.00001. The results of the Monte Carlo test indicated that the two dimensional solutions using the real data were significantly better than occurrences by chance (P < 0.01). The coefficients of determination for the correlations between ordination distances and distances in the original n-dimensional space were axis 1: r 2 = 0.667, axis 2: r 2 = 0.183, total r 2 = 0.849 and orthogonality = 77.8%. The NMDS ordination plot was orthogonally rotated by the treatment to visualize correlations between snow depth and fungal community composition in general, and the taxonomic groups and EMM characteristics in particular (Fig. 3c). The MRPP analysis indicated a clear distinction between control and deep snow ECM community composition (P = 0.0039, A = 0.04940). The NMDS and MRPP analysis of the presence-absence matrix results yielded similar conclusions (Data S4c). The groups with the strongest negative correlation (Pearson's correlations) with the increased snow depth were OTUs of the contact, short-distance and mediumdistance smooth hyphal ET (r = À0.950), Tomentella (r = À0.937), OTUs with hydrophilic hyphae (r = À0.935), Inocybe (r = À0.935), sequence counts (r = À0.912), total OTU richness (r = À0.896), Sebacina (r = À0.682), Tulasnella (r = À0.582), Sistotrema (r = À0.533). None of the groups showed a strong positive correlation (r > 0.5) with the increased snow depth. The control plots had on average 54 AE 23.29 OTUs per plot, while the treatment plots had 26.6 AE 8.02 OTUs per plot. This difference was statistically significant (P = 0.03). The comparison of sequence abundance also showed a significant decrease, in the same direction as the OTU richness patterns (Data S6).
Regarding taxonomic groups, Tomentella had the highest OTU richness in the control plots with 21 AE 13.55 OTUs per plot, followed by Inocybe with 10.4 AE 5.3. Interestingly, across the deep snow plots, both genera showed a significant decrease on the average OTU richness per plot (P = 0.022 and 0.047 respectively). On the other hand, OTU richness in Cortinarius was nearly unaffected and this genus had the highest mean richness in the increased snow depth plots (Table 1). At the order level, there was an abrupt decrease in the proportion of Thelephorales OTUs, from 36% in the control plots to 17% in the deep snow plots, while most other orders had an increase in proportion between the ambient and increased snow depth plots, with Agaricales having 55% of all OTUs (across the deep snow plots) (Fig. 1).
In the ambient snow conditions, the functional group with contact, short-distance and medium-distance smooth ETs had the highest OTU richness with an average of 42.2 AE 19.82 OTUs per plot; while the group with long-distance and medium-distance fringed ETs solely had on average 11.8 AE 4.82 OTUs per plot. Interestingly, in the deeper snow plots the first group had a significant decrease (P = 0.02) in OTU richness, while the latter group maintained similar OTU richness (P = 0.42). Regarding relative sequence abundance, the same patterns were depicted for the above-mentioned functional groups, although the difference between control and increased snow depth plots for the contact, short-distance and medium-distance smooth ETs was only marginally significant (P = 0.055).The vast majority of OTUs were only present in the control plots (60%), a smaller percentage was present in both the control and the increased snow depth plots (20%), and Fig. 1 Ectomycorrhizal basidiomycetes OTU richness, classified by taxonomic and functional traits in ambient snow and increased snow depth plots. The legend for each pair of graphics is organized by colors and in a clock-wise disposition. C, contact; S, short-distance; MDS, medium-distance smooth; MDF, medium-distance fringe; L, long-distance. only a minority (18%) was strictly present in the deeper snow plots (Fig. 4). The two genera with the highest OTU richness in the control plots, i.e. Tomentella and Inocybe, also follow this pattern. In comparison the percentage of OTUs that are solely present in the control plots in Cortinarius and Russula is considerably lower (Fig. 4). Indicator species analysis was significant (a = 0.05) for five (undetermined) Tomentella OTUs in the ambient snow plots, while no OTU had significant values in the deeper snow plots (Table 3).
The correlation coefficient between the taxonomic groups and ETs revealed significant positive and negative correlations among specific groups (Data S7). Across the ambient snow depth plots, Inocybe OTU richness showed a significantly negative correlation with Russula, and a significantly positive correlation with Tomentella. Interestingly, across the increased snow depth plots, the negative correlations of OTU richness between the previously mention genera decreased sharply to nonsignificant values. The correlations for subsampled sequence abundance were all nonsignificant.

Moist tussock tundra
The NMDS analysis of the square-root sequence abundance matrix resulted in a 2-dimensional solution with a final stress and instability of 0.0759 and <0.00001 respectively. The results of the Monte Carlo test indicated that all two dimensional solutions using the real data were significantly better than occurrences by chance (P < 0.01). The coefficients of determination for the correlations between ordination distances and distances in the original n-dimensional space were axis 1: r 2 = 0.607; axis 2: r 2 = 0.265; total r 2 = 0.872; orthogonality = 93.1%. The NMDS ordination plot was orthogonally rotated by the treatment to visualize correlations between snow depth effect and fungal community composition in general, and the taxonomic groups in particular (Fig. 3b). The MRPP analysis indicated a clear distinction between ambient and deeper snow ECM community composition (P = 0.0017, A = 0.0943). The NMDS and MRPP analysis of the presence-absence matrix yielded similar results (Data S4b). Inocybe and the group of OTUs with short-distance exploration type had a strong negative correlation with the increased snow depth plots, r = À0.757 and À0.775, respectively, as well as the OTUs with hydrophilic hyphae (r = À0.540) and Lactarius (r = À0.536). On the other hand, Alnicola and Laccaria OTU richness had the strongest positive correlation with the deeper snow plots with r = 0.696 and r = 0.510, respectively. The control plots had on average 72.4 AE 22.02 OTUs per plot, while the deep snow plots had 53 AE 31.01 OTUs per plot. Despite the considerable decrease, the difference was not statistically significant (P = 0.15). Overall sequence abundance showed a similar pattern with a nonsignificant decreasing trend from the control to the increased snow depth plots.
The genus with the highest OTU richness in the ambient snow areas was Tomentella with 20.6 AE 7.4 OTUs per plot, followed by Inocybe with 10.4 AE 5.3. On the deeper snow areas, while Tomentella had only a marginally significant (P = 0.07) decrease on the average OTUs per plot to 12.6 AE 8.33, Inocybe had a significant (P = 0.02) decrease to 3.2 AE 1.92 OTUs per plot. On the other hand, Cortinarius was the genus with higher OTU richness in the deeper snow plots with 9.4 AE 6.8, and showed no significant changes (P = 0.47) on OTU richness compared with the ambient snow plots (Table 1). Interestingly, in the deeper snow plots, on the order ranking, Thelephorales increased the percentage of OTUs while all the remaining orders decreased (Fig. 1), mainly due to the decrease in Inocybe, Russula, and Lactarius OTU richness. Regarding EMM characteristics, in the control plots the contact, short-distance and medium-distance smooth had on average 50 AE 17.13 OTUs per plot, while the long-distance and medium-distance fringed had 22 AE 5 OTUs  per plot. In the deeper snow plots the first group had a marginally significant decrease (P = 0.07) to 33.6 AE 15.08 OTUs per plot, while the latter group did not change significantly (P = 0.38, 19.4 AE 16.59 OTUs per plot). Most OTUs, 45% were only present across the ambient snow plots, 25% were present in both the control and deeper snow plots and 30% were only present in the deeper snow plots (Fig. 4). Tomentella and Russula followed the overall OTU distribution pattern. Conversely, Cortinarius had a contrary pattern, with a higher percentage of OTUs solely recovered from the deep snow plots -41%, 33% solely present on the control plots and 25% present in both the control and deeper snow areas. On the other hand, Inocybe had 63% of OTUs only recovered from the ambient snow plots, 23% recovered solely from the deeper snow areas and only 13% were found in both the ambient and deep snow plots (Fig. 4). OTUs from seven different genera had significant P-values (a = 0.05) in the indicator species analysis. Of these 13 OTUs were indicator of ambient snow plots and eight OTUs were indicator of deeper snow plots (Table 3). Nonmetric multidimensional scaling (NMDS) ordination plots of ECM basidiomycetes communities from the ambient and increased snow depth plots based on OTU sequence square-root abundance in (a) the whole community (dry and moist tundra), (b) the moist tundra, (c) the dry tundra. Vectors with |r| 0.5 are represented on the ordination plot. C, contact; S, short-distance; MDS, medium-distance smooth; MDF, medium-distance fringe; L, long-distance; Hi, hydrophilic; Seq counts, sequence reads.
The correlation coefficient of OTU richness between the taxonomic groups and ETs revealed a significant positive correlation among two groups in the ambient snow plots: Cortinarius -Russula, and between the group of OTUs with contact, short-distance and medium-distance smooth ET, and the group of OTUs with medium-distance fringe and long-distance ET (Data S7). In the increased snow depth plots, the significant positive correlation, among these groups, was also observed, and further extended to the pairs Tomentella -Cortinarius, and (marginally significantly) for Russula -Tomentella. None of the groups tested had a negative correlation either at the control or the deeper snow plots. The correlation patterns for subsampled sequence abundance were similar in the increased snow depth plots, and all nonsignificant in the control plots.

Discussion
The results presented here clearly show that long-term increase in snow depth alters ECM fungal community composition in moist tussock and dry heath tundra, with a considerable portion of OTUs not being resistant to the resulting changes in environmental conditions. Such local extinction of many taxa is inferred from the significant decrease in average OTU richness with increased snow depth and from the high compositional turnover among the control and deep snow treatment in both tundra types. We find the strong decrease in ECM fungal richness surprising, because vegetation studies revealed a strong increase in shrub cover and size in the moist and to a lesser extent in the dry tundra that is presumably coupled with greater root biomass, i.e. more colonizable habitat for ECM fungi. Although our results suggest that only a subset of ECM fungi can withstand the altered environmental conditions caused by the increased snow depth, there were some species present in the deep snow treatment that were not detected in the control plots. It is plausible that some of the taxa characteristic of the deep snow plots may be DC, dry heath tundra with ambient snow; DS, dry heath tundra with increased snow depth; MC, moist acidic tussock tundra with ambient snow; MS, moist acidic tundra with increased snow depth; D, Dry control and increased snow depth; M, moist acidic tundra control and increased snow depth. snow bank specialists and may, therefore, be absent from the zonal (moist) and ridge-top (dry) control plots.

Cortinarius
Recently it has been shown that ECM fungi are also sensitive to summer warming with the communities of the moist tundra showing a more evident response to summer warming in terms of altered community composition, trait patterns and OTU richness levels Morgado et al., 2015). The strongest response to increased snow depth was observed in the dry heath tundra, where besides community composition, there was also an overall signifi-cant decrease in abundance (sequence reads) and richness of the ECM fungal community in response to increased snow depth. The richness and read abundance of the moist tundra also showed a decreasing trend but did not change in a significant manner. This was surprising, as we anticipated that ECM fungal community in the moist tussock tundra would have a stronger response to deeper snow, because the plant community and N dynamics had been reported to be more strongly affected by the increased snow depth in the moist tundra Wahren et al.,  2005; Mercado-D ıaz, 2011). Together with our previous results , we argue that the dry tundra ECM fungal community likely is more sensitive to changes in snow depth than to summer warming. The increased snow depth not only elevates winter soil temperature, but also increases soil moisture content as well (Wipf & Rixen, 2010), the effect of which likely is greater in the dry tundra. Because ambient snow in dry tundra areas is very shallow and provides little protection against cold and desiccating winds, the above compositional differences of fungal communities between ambient and deep snow may be partly caused by both temperature and moisture level differences. Our correlation analysis of OTU richness and sequence abundance with the different snow depths showed different patterns among the tundra types. For example, in the moist tundra OTU richness and sequence abundance between Cortinarius and Tomentella showed a significant positive correlation in the increased snow depth plots, while no significant correlation was observed in the control plots. On the other hand, in the dry tundra no correlation was observed between these two groups in either snow depths. This suggests that in our study the interaction between OTU richness and potential species abundance likely are habitat-specific, and are in agreement with studies that addressed species-species interactions (Kennedy et al., 2007;Simard et al., 2012;Fransson et al., 2013). Collectively, our findings reflect the complexity of arctic tundra responses to predicted changes in summer and winter climate and the need to undertake comparative studies that include multiple ecosystem types even at reduced spatial scales (Welker et al., 2000;Walker et al., 2008;Sullivan et al., 2008a,b;Rogers et al., 2011;Christensen et al., 2013;Leffler & Welker, 2013;Sharp et al., 2013;Lupascu et al., 2014). Tomentella, the genus with the highest richness in the control plots of both tundra types, showed a sharp negative response to increased snow depth, with a significant sixfold decrease in average OTU richness and a majority of the OTUs disappearing in the dry tundra, as well as an overall decrease in proportional sequence counts. In the moist tundra, Tomentella richness also showed a decreasing trend, but in a less striking manner than in the dry tundra. The higher beta diversity in the increased snow plots, especially in the moist tundra, reflects an increase in species turnover and intrageneric community dissimilarity indicating potential species-specific responses to the altered conditions and patchiness distribution of this group of species. Nevertheless, the elevated number of indicator OTUs associated with the control plots (five in the dry and seven in the moist tundra) point to the sensitivity of this group to altered conditions. Moreover, two of these OTUs were very closely related with an OTU (KJ792685) that was negatively affect by increased summer temperatures in the dry tundra , further indicating that besides the general trends for the genus, at least one species of Tomentella, which is potentially widespread across the dry tundra, is very sensitive to summer and winter warming. Potential explanations to the observed patterns in our study may be linked with their functional traits and potential ecological roles. Tomentella and closely related genera (e.g. Pseudotomentella and Tomentellopsis) have melanized cell walls (Agerer, 1987(Agerer, -2002Agerer, 2006), which is not a common feature in ECM basidiomycetes (Kõljalg et al., 2000). Melanins can be produced by fungi, plants, and animals, and are dark macromolecules composed of phenolic and indolic monomers, often coupled with protein and carbohydrates (Butler & Day, 1998). They usually constitute a considerable portion of total fungal cell weight and likely require a considerable energetic investment (Rast & Hollenstein, 1977;Butler & Day, 1998). This feature has been extensively argued and was recently shown in physiological experiments (Fernandez & Koide, 2013) to increase tolerance to several environmental stressors, such as freezing (Robinson, 2001) and hydric stress (Singaravelan et al., 2008;Fernandez & Koide, 2013). The increased snow depth not only elevates winter soil temperature, but also increases soil moisture content (Wipf & Rixen, 2010), the effect of which likely is greater in the dry tundra. The deep snow conditions with elevated soil temperature and moisture might reduce the competitive advantages of melanin-producing Tomentella adapted to the abovementioned stress factors (e.g., drought, and very low temperatures). Additionally, Tomentella has either contact, short or medium-distance smooth hyphal ETs, which have been argued to be adapted to labile N soil pools (Hobbie & Agerer, 2010). The plant community responded to increased snow depth with a significant increase in the shrubs and litter layer (Wahren et al., 2005;Mercado-D ıaz, 2011), indicating a potential change in soil organic matter input and shifts in C:N ratio, that have been argued to be important regulators of arctic N dynamics (DeMarco et al., 2011). Therefore, it is possible that in the altered environmental conditions the combination of traits presented by Tomentella might not constitute an advantage in scavenging for soil nutrients, which might lead to a detrimental allocation of photosynthates by the ECM host and/or to competitive exclusion by other ECM fungi better suited to the altered conditions. The decomposition and turnover of ECM fungal biomass likely has a significant role in C and other nutrient dynamics (Wallander et al., 2001;Clemmensen et al., 2013;Ekblad et al., 2013). Melanized hyphae have been argued to be relatively long lived, slow growing (Robinson, 2001) and relatively resistant to decomposition (Coelho et al., 1997;Butler & Day, 1998;Butler et al., 2005), potentially representing a stable and recalcitrant component in the fungal biomass (Treseder & Lennon, 2015). In two laboratory experiments, Fernandez &  showed that the decomposition rate of ECM fungi was inversely correlated with the concentration of melanin and that the inhibition of melanin biosynthesis in an ECM fungi induced faster rates of decomposition. Moreover, Clemmensen et al. (2015), observed a correlation between the abundance of taxa with melanized hyphal content and higher carbon storage in the soil. If future climatic conditions lead to increased snow depth in the arctic tundra, the decreasing richness and relative abundance of Tomentella might contribute to soil C loss. Several other groups of rootassociated fungi also have melanized hyphae, such as ericoid mycorrhizal fungi and dark septated endophytes. We therefore highlight the need to address responses of those root-associated fungi to increased snow depth to better understand if the above-mentioned warming-induced trend of decreasing richness is common in melanized fungi or is specific to certain phylogenetic lineages. Despite the uncertainties, our evidence and grounded speculations are in line with the results by Natali et al. (2014) that addressed winter warming effects on C cycle dynamics and indicated a net soil C loss due to winter warming.
We observed an abrupt decrease in mean OTU richness of Inocybe from the control to the increased snow depth plots in both tundra types. However, a considerable proportion of Inocybe OTUs that were found in the increased snow depth plots were also found in the control plots, particularly in the moist tundra. These results suggest that although arctic Inocybe spp. seem to be very sensitive to climate changes, a resistant subset of the species may be able to withstand changes in the climatic conditions. Inocybe spp. were previously argued to be sensitive to altered environmental conditions, such as soil compaction (Hartmann et al., 2014) and summer warming . Additionally, there are evidences that in mature plant stands the rate of root-infection by Inocybe might decrease in sites with increased soil moisture (Fleming, 1984). It is possible that in the increased snow depth conditions, the lack of rizhomorphs and hydrophilic ectomycorrhizae of Inocybe (Agerer, 2006), a set of characteristics hypothesized to be adapted to labile N uptake (Hobbie & Agerer, 2010), might constitute detrimental traits in relation to other groups of ECM fungi. However, the increase in shrubs and litter layer might lead to potential patchiness of nutrient soil pools allowing for some species with hydrophilic hyphae to thrive in the increased snow depth.
The lack of significant changes in Cortinarius richness and the relative increase in overall sequence counts (a potential surrogate for relative abundance) between the control and increased snow depth plots indicate that this group might be more adapted to the altered conditions, and could become more dominant in the warming Arctic. A similar trend for this group was also observed in our previous work that reported ECM fungal responses to long-term summer warming . However, in contrast, in the present study most OTUs are not shared between the two treatments, indicating that, although average OTU richness does not change, there seems to be a considerable turnover in species composition. This suggests that only a subset of the OTUs present in the control plots are resistant to increased snow depth and that there is substantial functional variation within the genus that allows for the exploitation of new niches created in the altered environment by incoming species. Species of Cortinarius form a dense mycelium with medium-distance fringe ET and hydrophobic rhizomorphs (differentiated mycelial cords) (Agerer, 2001). This foraging strategy is adapted for efficient N absorption and nutrient translocation (Hobbie & Agerer, 2010). Additionally, at least some Cortinarius spp. were reported to have the capability to assimilate organically bounded N (Hobbie & Agerer, 2010), and to transcribe Mn-peroxidase genes (which are involved in the production of exoenzymes) in field conditions; these were further linked through co-localization of DNA abundance with exoenzyme activity that interacts in complex organic matter breakdown (B€ odeker et al., 2014). Another feature that may play a role is their physiological response to the altered conditions. Although, only a few studies focused on ECM fungal physiological responses to extreme cold, Ma et al. (2011) compared ECM fungi growth responses to very low temperatures (between À40 and +4°C) and freeze-thaw events of four ECM species from distinct lineages. Their results indicated that Cortinarius had the lowest tolerance to freeze-thaw events and the fastest growth when temperatures reach near 0°C. Because reduced temperatures, hydric stress, and freeze-thaw events inhibit the rate of chemical and microbial activity (Robinson, 2001), and given the characteristics and potential ecological role of Cortinarius spp., it seems feasible to argue that long-term increased winter soil temperatures, summer moisture, and reduced fluctuations in soil temperatures might convey advantages to this group over other ECM fungal groups.
The increasing dominance of fast growing ECM species with high EMM production and fast N mobilization might lead to increased C storage in the soil pool; however, this will be determined by biomass turnover. In an interesting work relating fungal traits, community structure and nutrient soil pools, Clemmensen et al. (2015) found a link between abundance of species with similar ETs to that of Cortinarius and low accumulation of root-derived soil C. Briefly, they hypothesized that the exploratory hyphae of already explored soil patches could be recycled in an autolytic process, leaving behind just the long-living rhizomorphs connecting the exploratory forefront of the EMM and the ectomycorrhizae. This strategy would enhance their nutrient acquisition and maintain or reduce their biomass, potentially reducing also energetic costs. Due to the potentially high turnover of this mycelium biomass, this theory also implies a reduction in stable soil pools, and therefore the long-term C and N sequestration.
Russula did not show any significant change in richness with increased snow depth, and a considerable portion of the OTUs were found both in the control and the treatment plots, indicating that many OTUs were resistant to altered conditions. While most Russula spp. lineages have an hydrophilic ectomycorrhizae and contact or short-distance ETs that lack rhizomorphs, other lineages in Russula have a medium-distance smooth ET and hydrophobic hyphae (Agerer, 2006). In our dataset, the Russula OTUs with short or contact ET did not show change in richness with increased snow depth. However, one OTU with contact or short-distance ET was indicator of the moist control plots while another was indicator of increased snow depth, suggesting speciesspecific responses to altered conditions. On the other hand, in the increased snow depth plots in the moist tundra, we observed a significant decrease in richness of OTUs that matched Russula species with hydrophobic and medium-distance smooth ET. These results indicate that even within a closely related group of species the functional diversity can differ. While some Russula species seem to have a considerable fast growth rate at low temperatures (Ma et al., 2011), others are considered to have a slow growth rate (Nygren et al., 2008). Moreover, there is evidence of intrageneric variability in N usage as well (Avis, 2012). It is possible that in some species the hydrophilic ectomycorrhizae allows for the rapid intake of labile N forms by the plants, without metabolizing, via diffusion through the mantle of the ectomycorrhizae directly to the plant-host root via the apoplast. This process would avoid energetic costs and the necessity of C allocation to the ECM fungi (Nygren et al., 2008). This may influence the competitive interactions between species with hydrophilic and hydrophobic mycelia.
In conclusion, our data provide first insights into the taxon-specific effects of increased snow depth on the ECM fungal community of Arctic tundra in Northern Alaska. We detected major shifts in ECM fungal community composition and its potential functional traits by coupling changes on fine scale taxonomic groups with their extramatrical mycelial characteristics. We postulate that ECM fungal community shifts induced by long-term increased snow depth likely stimulate C and N mobilization. However, the final balance induced by arctic ECM basidiomycete community in these nutrient pools will likely depend on the changes in the biomass of specific groups, particularly Tomentella and Cortinarius. Our results also highlight how the fundamental differences in tundra ecosystems control the nature of the existing fungal communities and their responses to deeper snow.