Bacterial and fungal communities in sub-Arctic tundra heaths are shaped by contrasting snow accumulation and nutrient availability

Abstract Climate change is affecting winter snow conditions significantly in northern ecosystems but the effects of the changing conditions for soil microbial communities are not well-understood. We utilized naturally occurring differences in snow accumulation to understand how the wintertime subnivean conditions shape bacterial and fungal communities in dwarf shrub-dominated sub-Arctic Fennoscandian tundra sampled in mid-winter, early, and late growing season. Phospholipid fatty acid (PLFA) and quantitative PCR analyses indicated that fungal abundance was higher in windswept tundra heaths with low snow accumulation and lower nutrient availability. This was associated with clear differences in the microbial community structure throughout the season. Members of Clavaria spp. and Sebacinales were especially dominant in the windswept heaths. Bacterial biomass proxies were higher in the snow-accumulating tundra heaths in the late growing season but there were only minor differences in the biomass or community structure in winter. Bacterial communities were dominated by members of Alphaproteobacteria, Actinomycetota, and Acidobacteriota and were less affected by the snow conditions than the fungal communities. The results suggest that small-scale spatial patterns in snow accumulation leading to a mosaic of differing tundra heath vegetation shapes bacterial and fungal communities as well as soil carbon and nutrient availability.


Introduction
High-latitude soils store approximately half of the global soil organic carbon, and hence it is of major concern how global climate change will affect decomposition rates and C flux to the atmosphere in these regions (Tarnocai et al. 2009, Schuur et al. 2015 ).Arctic r egions ar e warming up to four times faster than the global av er a ge (Rantanen et al. 2022 ), with the cold season months warming at a m uc h faster rate than the growing season (Mikkonen et al. 2015, Rantanen et al. 2022 ).In northern Fennoscandia, in addition to increasing winter month temper atur es, the number of frost days has declined, exceptionally cold winter days have decr eased, while exceptionall y warm days and the number of freeze-thaw cycles have increased (Mikkonen et al. 2015, Kivinen et al. 2017, Lepy and Pasanen 2017 ).Precipitation is also increasing in high-latitude ecosystems (Groisman et al. 2005, McCrystall et al. 2021 ), which influences the amount of snow that accumulates during winter.The depth and insulating properties of the snowpack, ho w ever, depend on the form of precipitation.As the extr emel y warm winter days are increasing in Arctic ecosystems it is predicted that the precipitation will be more and more in the form of rain instead of snow, and the thickness and timing of snowpack may be reduced (Bintanja andAndry 2017 , McCrystall et al. 2021 ).
Furthermor e, incr eased winter thaw and rain on snow events diminish the insulating properties of snow (Serreze et al. 2021 ) and may lead to colder soil temper atur es during winter (Groffman et al. 2001 ).In addition to direct changes in precipitation and soil temper atur e, climate warming is altering wintertime soil temperatur e r egimes also indir ectl y thr ough c hanges in the dominant v egetation.Tundr a ecosystems are undergoing strong expansion and increase of shrubs (Sturm et al. 2001b ), which may control the wintertime soil temper atur es by tr a pping mor e snow that leads to increased insulation (Sturm et al. 2001a ).Due to the complex w ays b y whic h climate c hange is altering snow cov er, ther e is a high need to better understand the role of snow cover for soil micr obial comm unities.
The wintertime microbial activity contributes to a significant part of annual carbon cycling in tundra ecosystems (Oechel et al. 1997, Fahnestock et al. 1999, Euskirchen 2012et al. 2012, Natali et al. 2019 ), and in boreal and Arctic soils overwinter CO 2 effluxes may e v en exceed plant carbon uptake during the growing season (Natali et al. 2019 ).Microbial activity and CO 2 flux continue in tundra soils through the winter down to −20 • C (Natali et al. 2019 ).In Fennoscandian tundra soil, bacterial activity down to −16 • C was identified with stable isotope probing with distinct communities growing at different subzero temperatures (Gadkari et al. 2019 ).Around and below 0 • C, soil microbial activities become, ho w e v er, incr easingl y temper atur e sensitive (Mikan et al. 2002, Sullivan et al. 2008, Tilston et al. 2010 ), and microbial substrate utilization may shift to decomposition of more labile organic matter (OM) such as fresh litter and root exudates or microbial biomass and by-products further affecting the C and nutrient cycling (Mikan et al. 2002, Schimel et al. 2004, Grogan and Jonasson 2005, Sturm et al. 2005 ).Under a thick snow cover, soil temperatures are decoupled from air temperatures and may remain close to 0 • C throughout the coldest months (Schimel et al. 2004, Männistö et al. 2013, Pattison and Welker 2014, Convey et al. 2018, Way and Le wk owicz 2018, Rixen et al. 2022 ).Snow cover conditions thus control cold season microbial activities and deeper snow may enhance winter microbial respir ation e v en to the degree that it switches the ecosystem annual net carbon exchange from a sink to source (Nobrega and Grogan 2007, Natali et al. 2019 ).Ho w ever, increased cold season micr obial r espir ation ma y ha ve legacy effects to the following gro wing seasons.Sno w manipulation experiments have indicated that incr eased micr obial activity during winter may lead to depletion of labile C and over time to reduced growing season CO 2 emissions (Semenchuk et al. 2016, Sullivan et al. 2020 ).Moreover, incr eased nitr ogen (N) miner alization under deeper snow cover and associated higher soil temper atur es may have significant consequences for plant growth and consequently carbon cycling especially in the nutrient limited tundra ecosystems (Schimel et al. 2004 ) The current understanding of the effects of snow depth on soil micr obial comm unities mainl y deriv es fr om experimental manipulation of the snowpack.In subalpine grassland and temperate deciduous forest soils, strong changes were reported in the bacterial and fungal communities under reduced snow cover during winter, but these differences leveled out during the growing season (Aanderud et al. 2013, Ga vazo v et al. 2017 ).In boreal forest soil, no effect of snow depth or snow properties was detected in bacterial or fungal communities either before or after spring thaw or in the late growing season (Männistö et al. 2018 ).On the other hand, strong shifts were reported in microbial communities of alpine grasslands during spring thaw and these were linked to shifts in microbial functions and biogeochemical fluxes suggesting that changes in the timing of spring thaw may have important consequences for the ecosystem functioning (Broadbent et al. 2021 ).In acidic tundra soils, bacterial community structure was significantly affected by snow depth, which was associated with changes in edaphic factors (Ricketts et al. 2016). Semenova et al. ( 2016 ) reported changes in the abundance and community structure of saprotrophic, ectomycorrhizal, plant pathogenic, and lichen-and bryophyte-associated fungal guilds with deeper snow.These c hanges wer e not entir el y associated with shifts in vegetation but there were indications that fungal communities in the Arctic may exhibit faster turnover which is influenced e.g. by soil nutrient availability and dynamics of other microbial groups.Experimental manipulations of snow cover conditions have shown both differing vegetation (Olofsson et al. 2009 ) and higher soil microbial N and bacterial counts (Buckeridge and Grogan 2010 ) with deepened snow .T ogether, these studies indicate that there may be lar ge differ ences in the r esilience and sensitivity of bacterial and fungal communities to changes in winter soil temper atur e and moisture in different soil ecosystems.In addition, experimental manipulations of snow depth are often of short duration, and may thus not necessarily depict long-term changes that would be mediated by the combination of differing v egetation, nutrient av ail-ability, and temper atur e r egimes, whic h all c hange in r esponse to differing snow cover.
In tundr a landsca pes, the small-scale v ariation in topogr a phy, which in turn affects wintertime snow accumulation, leads to a mosaic of habitats with differing snow conditions and dominant v egetation ov er a gr adient fr om lo w and absent sno w cover to snow-accumulating (SA) sites (Oksanen andVirtanen 1995 , Niittynen et al. 2020 ).Exposed ridges that are windswept (WS) remain nearl y snow-fr ee thr oughout the winter, while shelter ed slopes and depr essions accum ulate snow alr eady earl y in winter that remains until early summer.The WS and SA habitats exhibit widely div er gent wintertime soil temper atur es and the duration of the snow-fr ee period, whic h giv es rise to div er gent plant cov er a ge, amount of litter, and nutrient availability (Niittynen et al. 2020 ).In Fennoscandian tundra, Empetrum nigrum ssp.hermaphroditum heaths dominate the WS heaths while communities rich in Vaccinium myrtillus L. occur as strands between Empetrum heaths and grass-dominating snow-beds (Oksanen and Virtanen 1995 ).Under the greater snow accumulation, the shrubs are higher and species such as Betula nana L. and Salix spp.are intermixed with V. myrtillus .Both tundra heath types are characterized by a continuous bryophyte and lichen cover.To date, little is known about how WS and SA habitats differ in microbial community composition and seasonal trends, and the consequences of these changes in community composition on carbon cycling.Soil microbial communities under a thick sno w lay er with soil temperatures around 0 • C have stable conditions that likely enable active soil organic matter (SOM) degradation throughout the winter (Schimel et al. 2004 ), whereas WS habitats can experience very low temperatures during winter (Niittynen et al. 2020 ), leading to a need for the community to adjust to such cold conditions.Microbial communities could, thus be expected to be highl y div er gent between WS and SA tundra habitats.
In this study, we utilized tundra heaths with long-term natural differences in wintertime snow accumulation.The effect of contr asting snow accum ulation and subsequent differ ences in winter soil temper atur es on micr obial biomass as well as bacterial and fungal comm unity structur e wer e assessed during differ ent seasons .T her e ar e indications that the bacterial communities during the growing season are relatively similar in soils of both habitat types and are dominated by str ess-toler ant, oligotr ophic bacterial taxa such as the Acidobacteriota (Männistö et al. 2013 ).Ho w e v er, earlier studies of the same tundra sites indicated strong shifts in the r elativ e abundance of dominant bacterial phylotypes in WS heaths especially during spring thaw suggesting that lo w er winter temper atur es and mor e fr equent fr eeze-tha w cycles ma y ha ve str ong contr ol ov er the micr obial biomass and comm unity structure with a stronger effect in the WS than in deep snow habitats (Männistö et al. 2013 ).In this study, we e v aluated bacterial and fungal biomass using phospholipid fatty acid (PLFA) and quantitativ e PCR (qPCR) anal yses and c har acterized the comm unity structures b y rRN A gene and ITS amplicon sequencing of soil sampled in mid-winter, early, and late growing seasons .T he aim was to identify k e y taxa of the bacterial and fungal communities linked to differences in winter snow accumulation.We hypothesized that (1) differences in the snow accumulation with strong differences in winter soil temper atur es lead to div er gent patterns in bacterial and fungal biomass between WS and SA tundra heaths during the mid-winter and early growing seasons.We predicted that the mor e se v er e fr eezing and fr equent fr eeze-thaw cycles in WS tundra heaths are associated with lo w er microbial abundance due to lo w er microbial activity, and higher turnover, whereas SA tundra heath harbor more stable microbial communities across seasons.
We further hypothesized that (2) differences in the winter conditions are associated with differences in the bacterial and fungal comm unity structur e, sho wing a distinct cry otoler ant micr obial comm unity structur e under WS tundr a heaths during winter, with only weak differences in the bacterial and fungal communities between the summer and the winter under SA heaths.

Field site
The study site was located in the north side of Mt Pikku-Malla fjeld in Malla Nature reserve, Kilpisjärvi, north-western Finland (69 • 03 50 N, 20 • 44 40 E).The mean annual precipitation is 420 mm and the mean annual temper atur e is −1.9 • C (Aalto et al. 2018 ).The bedr oc k at the sampling sites of this study was formed from siliceous r oc k materials r esulting in acidic barren soil where heaths dominated by the dwarf shrub E. nigrum spp.hermaphroditum (Hagerup) Böcher prevail (Eskelinen et al. 2009 ).The tundra heaths are exposed to heavy winds, which together with the differences in topogr a phy dr amaticall y influence snow accum ulation.Areas with high snow accumulation (up to ≥ 1 m) are located in depressions and areas sheltered from the winds, while WS areas r emain essentiall y snow-fr ee thr oughout the winter ( Figur e S1 , Supporting Information ).Soil temper atur e under thic k snow cov er is more stable, remaining close to 0 • C throughout the winter, while in the WS heaths, the temper atur e follows air temper atur e and may drop down to −15 • C (Männistö et al. 2013 ;Fig 1 ).In addition to soil temper atur e and snow accum ulation, v ariations in topogr a phy r esult in alternating patterns in v egetation.Empetrum nigrum dominated especially in the WS heaths while Vaccinium myrtillus was more abundant in the SA heaths.Vaccinium vitis-idea and Vaccinium uliginosum were common in both habitat types.Under the higher snow accumulation, the shrub height was higher and species such B. nana L. and Salix spp.were abundant.Both tundra heath types were characterized by a continuous bryophyte and lic hen cov er.

Soil sampling
Four plots (2 m × 2 m) r epr esenting WS slopes (i.e.dominated by E. hermaphroditum ) and four plots corresponding to SA biotopes (i.e.dominated by V. myrtillus ) were selected and marked based on earlier snow cover estimates (Männistö et al. 2013 ) and the type of vegetation.All plots were within 300 m from each other and at least 25 m apart.Soil temperature was recorded once every hour using Hobo U10 temper atur e loggers (Onset Computer, Bourne, Massachusetts) that were buried 3-5 cm below the soil surface .T he soil was sampled from WS and SA tundra heaths in F ebruary, J une , and September 2013 from the top 5 cm (humus layer) using a soil corer (diameter ca. 2 cm).Composite soil samples of five soil cores were taken from each plot.In June and September, the samples were sieved in the field using a 2-mm mesh and immediately frozen in a liquid nitrogen dry shipper.February samples were transported frozen to the lab and sie v ed after brief tha wing.T hree subsamples (0.3 g) were taken from each composite sample and stored at −80 • C until thawed for DN A/RN A extraction.

Soil physico-chemical analyses
The dry matter content of the soil was determined by drying the samples (105 • C, 12 h) and OM content was analyzed by loss on ignition (475 • C, 4 h).Soil pH was measured in 1:5 (v:v) soil:water suspensions (Denver Instrument Model 220).A subsample of ∼3 g fresh soil was extracted for 2 h with 50 ml of 0.5 M K 2 SO 4 .Dissolv ed or ganic carbon (DOC) concentr ations in these extr acts wer e anal yzed with a TOC-VCPH/N Total Organic Carbon Analyzer (Shimadzu Corporation, K yoto , Ja pan).Total nitr ogen, NH 4 -N and NO 3 -N concentr ations wer e anal yzed via flow injection analysis (Quic kc hem 8000 FIA Anal yzer, A83200, Zell weger Anal ytics, USA).Extr actable or ganic nitr ogen (N) w as calculated b y subtracting inorganic N concentrations from the total.Microbial C and N wer e extr acted fr om the samples using 0.5 M K 2 SO 4 after c hlor oform fumigation for 18 h (Brookes et al. 1985 ), and then analyzed as total extractable N and DOC.Phosphorus was analyzed colorimetricall y (Mur phy and Riley 1962 ).

Nucleic acid extraction and cDNA synthesis
Total genomic DNA and RNA were extracted from ∼0.25 g of soil with slight modifications as described earlier (Männistö et al. 2016 ) using a CTAB-based method by Griffiths et al. ( 2000 ).Three r eplicate extr actions wer e pr ocessed fr om eac h of the eight plots.Hexadecyltrimethylammoniumbromide (CTAB; 650 μl) extraction buffer and phenol-c hlor oform-isoamyl alcohol (25:24:1; pH 8.0; 650 μl) were added together with a mixture of beads to the sample tubes follo w ed b y bead beating on a Pr ecell ys 24 Dual homogenizer (Bertin Tec hnologies, Montign y-le-Br etonneux, Fr ance) for 30 s at 5500 r m −1 .The bead mixture contained 0.1 mm glass beads (0.3 g), 1.0 mm ceramic beads (0.7 g), and two large (3.5 mm) glass beads (Bio Spec Products Inc., Bartlesville, OK, USA).Samples were further processed as described in Männistö et al. ( 2016 ).DNA samples wer e tr eated with RNAse A (T hermo Scientific , Waltham, MA, USA) and RNA samples with DNAse I (Thermo Scientific) and converted to cDNA using the Re v ert Aid H Minus First Strand cDNA Synthesis kit (Thermo Scientific).All solutions used for RNA extr action wer e tr eated with 0.1% diethylpyr ocarbonate.RNA and DN A concentrations w ere measured using a Qubit fluorometer and Quant-iT RNA and dsDNA HS assa y kits (T hermo Scientific), r espectiv el y.
qPCR was performed using the Bio-Rad CFX96 Real-time thermal cycler (Bio-Rad) and SsoAdvanced Univ ersal SYBR Gr een Supermix (Bio-Rad).16S rRNA gene cop y numbers w ere quantified using the primer pair Eub341F (CCTA CGGGA GGCA GCA G) and Eub534R (A TT ACCGCGGCTGCTGG) (Muyzer et al. 1993 ), and fungal ITS2 region copies with the primer pair fITS7 (GTGART-CA TCGAA TCTTTG) (Ihrmark et al. 2012 ) and ITS4 (TCCTCCGCT-T A TTGA T A TGC) (White et al. 1990 ).All qPCRs were run in technical triplicates of 20 μl and contained 10 μl Supermix, 0.5 μl of each primer (10 mM), 7 μl ddH2O, and 2 μl template in a 100-fold dilution.PCR conditions for bacterial anal ysis wer e 98 • C for 2 min follo w ed b y 40 c ycles of 98 • C (5 s), 56 • C (20 s), and for fungal analysis 98 • C for 3 min follo w ed b y 40 c ycles of 98 • C (15 s), 61 • C (60 s), (following a plate read).Genomic DNA from Granulicella mallensis MP5ACTX8 isolate was used as a bacterial and Laccaria laccata isolate as a fungal standard.Amplicons were then combined in equimolar concentrations for sequencing.The pooled 16S rRNA gene amplicon libraries were sequenced using Ion Torrent Personal Genome Machine (Thermo Scientific) at the University of Jyväskylä, Finland.One set of samples (DNA samples of February) was sequenced in 2014 using a 316 v2 chip and all other samples in 2015 using a 314 v2 chip.The sequencing chemistry and Ion Torr ent softwar e was updated multiple times between these two runs, potentially affecting the r esults, and ther efor e the bacterial DNA sequences fr om February were not compared with the other sampling seasons (see the section "Statistical Analyses").Amplification of the ITS2 region of fungal rRNA operons was performed as a 2-step pr ocedur e r ecommended by Berry et al. ( 2011 ).The first amplification step was done in triplicate for each sample (1 μl of 1:50 dilution) in a 10μl reaction and the second step in a single 50 μl reaction for each sample (1 μl of PCR product from step 1), both amplification steps using Phusion High-Fidelity DNA Pol ymer ase (Thermo Sci-

Bioinformatics
Bacterial sequence reads from the two sequencing efforts (total 3 005 260 and 2 666 670 r eads) wer e dem ultiplexed, quality filter ed and merged with the following adjustments using QIIME 1.9.1 (Capor aso et al. 2010 ): minim um length of 300 bp, maxim um of one mismatch in primer sequences, and minimum mean quality score of 25 within 50 bp window size.We employed a r efer ence-based operational taxonomic unit (OTU) picking against SILVA release 128 99% identity r efer ence database (Quast et al. 2013 ), removing c himer as, and clustering all sequences using USEARCH 6.1 (Edgar 2010, Edgar et al. 2011 ) with a sequence similarity value of 97%.Taxonomy was assigned for r epr esentativ e sequences using a naïve Bayesian RDP classifier (Wang et al. 2007 ) against SILVA 99% identity majority taxonomy strings with a confidence threshold of 50%.Re presentati ve sequences from each OTU were aligned to the SILVA 99% identity r efer ence alignment using PyNAST (Caporaso et al. 2009 ) and a phylogenetic tree was built using FastTree (Price et al. 2009 ).After removing singletons (OTUs r epr esented by a single sequence) and alignment failures from the data, 1 249 249 r eads wer e obtained fr om all samples with an av er a ge of 10 498 reads per sample (min 3314 and max 32 820).For downstream anal yses, all samples wer e r ar efied by r andom sampling (without replacement) to an equal sequence number of 3300 to minimize bias due to different sequencing efforts across samples.
ITS sequence reads (total 701 430 reads) were demultiplexed and quality filtered with the following adjustments using QIIME 1.9.1 (Ca por aso et al. 2010 ): minim um length of 200 bp, the maximum length of 600 bp, maximum of one mismatch in primer sequences, and minimum mean quality score of 20 within 50 bp window size.We employed a c himer a c hec k using UCHIME (Edgar et al. 2011 ) and sequences were then clustered into OTUs using UCLUST (Edgar 2010 ) with a sequence similarity value of 97%.Taxonomy was assigned for r epr esentativ e sequences using BLAST (Altschul et al. 1990 ) against UNITE v.7.1 97% threshold reference database (Kõljalg et al. 2013 ) with 90% identity.After removing singletons (OTUs r epr esented by a single sequence) and nonfungal hits from the data, 404 561 reads were obtained from all samples with an av er a ge of 6321 r eads per sample (min 1486 and max 10 579).For downstream analyses, all samples were rarefied by random sampling (without replacement) to an equal sequence number of 1400 to minimize bias due to different sequencing effort across samples.
Raw sequence data and associated metadata were deposited in GenBank with the Bioproject accession no.PRJNA1080106.

Sta tistical anal yses
Differences in bacterial and fungal comm unity structur e wer e anal yzed using perm utational anal ysis of v ariance (PERMANOVA; Anderson 2001 ) and visualized with principal coordinate ordination (PCO).For the PERMANOVA and PCO analyses, bacterial and fungal OTU data were Hellinger-transformed and Bray-Curtis dissimilarity matrices used as the resemblance matrices.Habitat (WS , WS , or SA) and sampling season (winter, early, or late growing season) were used as fixed factors and plot as a random factor nested in habitat.When significant interactions were detected for habitat and season, pairwise PERMANOVA was performed on the r espectiv e terms.All PERMANOVA anal yses wer e performed with 999 random permutations PERMANOVA and ordination analyses were performed using the PERMANOVA + add-on (Anderson et al. 2008 ) for PRIMER v7 (Clarke and Gorley 2015 ).Data analysis indicated that bacterial communities of winter DNA samples deviated str ongl y fr om all other samples (deriv ed fr om DN A or RN A) as illustrated by a PCO ordination ( Figure S2 , Supporting Information ).T his deviation ma y be , at least partially, due to a sequencing bias as the winter DNA-derived bacterial communities were sequenced earlier than the other samples.Samples of the different sequencing runs were sequenced using a different chip and sequencing chemistry and the also the server software was updated between the runs .T her efor e , to a void potential sequencing bias the February DNA data were not compared with other sampling points or with RNA-derived community structure in the statistical analysis.
The effect of habitat, sampling season, and their interaction on microbial biomass proxies (16S rRNA gene and ITS copies, total PLF As, bacterial PLF As, fungal PLF A, and bacterial/fungal PLF A ra-tio) and soil physico-chemical properties as well as on the abundance of 10 most abundant bacterial and fungal genera were tested using linear mixed effect model (LME) with habitat and season as fixed factor and plot as random factor.Sampling season was assigned as repeated factor with site as a subject and AR1 as the cov ariance structur e.When significant inter actions were detected for habitat and season, habitats were further tested separ atel y with a least significant difference test as a post hoc test under the linear mixed model.Logarithmic transformations were used as necessary to meet the assumptions the linear mixed model.LME tests for microbial biomass and soil par ameters wer e conducted using IBM SPSS 29.0 software.Distance based linear modelling (DistLM) (Legendre andAnderson 1999 , Anderson et al. 2008 ) was used to determine the extent to which soil variables explain bacterial and fungal community structure in WS and SA tundra heaths.Multicollinearity between variables was first tested using the Draftsman Plot function and spearman correlations in PRIMER v7 (Clarke and Gorley 2015 ) and from variables that correlated by more than 0.9, only one was picked.Of the nitrogen forms, total and organic N correlated by 0.96 and N org was, ther efor e omitted fr om the anal ysis.Logaritmic tr ansformations were used for the same variables as for the LME tests (N tot , NH 4 , P tot , and P mic ).Marginal tests identified the influence of individual soil variables on bacterial (DN A and RN A) and fungal comm unity structur e without considering the effect of other v ariables.To identify the soil parameters that in combination explained bacterial and fungal community structure, DistLM model was utilized with stepwise selection pr ocedur e and corrected Akaike information criterion as the selection criteria with 999 permutations .T he resulting models were visualized using distance-based r edundancy anal ysis (dbRDA) plots.DistLM analyses and dbRDA plots were done performed using the PERMANOVA + add-on (Anderson et al. 2008 ) for PRIMER v7 (Clarke and Gorley 2015 ).

Soil physico-chemical properties and microbial biomass in WS and SA tundra heaths
Soil temper atur e differ ed str ongl y in WS vs. SA tundra (Fig. 1 ).The av er a ge soil temper atur e fr om Se ptember 2012 to Se ptember 2013 was 0.1 • C in the WS plots and 2.4 • C in SA plots.During the coldest months (December-March), the average soil temperature was −7.1 • C and −1.0 • C in the WS and SA plots, r espectiv el y.On the other hand, due to earlier spring thaw and less insulating vegetation cover, the WS plots were warmer in June with an av er a ge soil temper atur e of 8.4 • C compared to 7.0 • C in the SA plots.
LME model indicated that there were no significant differences in soil pH, moisture, or OM content between WS and SA heaths, but they differed in nutrient availability.Total nitrogen (N tot , F = 21.34,P = .004),organic nitrogen (N org , F = 17.053,P = .007),NH 4 -N ( F = 7.534, P = .034),and total phosphorus (P tot , F = 10.675,P = .017)av ailability wer e significantl y higher in the SA than WS heaths throughout the sampling year.On the other hand, N stored in microbial biomass (N mic , F = 5.353, P = .060)tended to be higher in the WS than under SA heaths.Sampling season had a significant impact on soil N availability, the concentrations of all N forms were at their lowest in samples collected in late growing season (T able 1 ; T able S1 , Supporting Information ).Ho w e v er, when the habitats were tested separately, sampling season had a significant effect on N tot only in the WS heaths ( P = .012between winter and early growing season and P < .001 between other sampling seasons), while NO 3 availability was significantly different between winter and earl y gr owing season ( P = .008)and early and late growing season ( P = .001) in WS heaths, and winter and early growing season ( P = .049)and winter and late growing season ( P = .003)in SA heaths.Microbial biomass was estimated using soil PLFA analysis and qPCR of bacterial 16S rRNA gene and fungal ITS region copy numbers (Figs 2 and 3 ).The LME model indicated that there were no significant differences in the abundance of total PLFAs between the different tundra heaths or sampling dates.Bacterial PLFAs were higher in the SA than WS plots and these decreased to w ar d the late growing season, but the differences were not statistically significant.The fungal PLFA marker was in higher abundance in the WS heaths ( F = 16.208,P = .007)and tended to be lo w er ( F = 2.732, P = .086)in the early and late growing season samples .T he fungal to bacterial PLFA ratio was higher in WS heaths ( F = 82.804,P < .001)and was lowest during June ( F = 4.728, P = .019)(Fig. 2 ; Table S1 , Supporting Information ).
Bacterial 16S rRNA gene cop y numbers w er e significantl y different ( F = 31.165,P = .001;)between WS and SA heaths as well as at different sampling seasons ( F = 38.792,P < .001)and there was a significant habitat × season interaction ( F = 54.388,P < .001;Table S1 , Supporting Information ).This was attributed to the significantly lo w er cop y numbers in the WS plots during late growing season ( P < .001).LME test indicated no significant main effect of habitat on fungal copy numbers, but when the habitats were tested separ atel y, pairwise tests indicated that in WS heaths, ITS copy numbers wer e significantl y lower P < .001) in late growing season than in winter or earl y gr owing season (Fig. 3 ; Table S1 , Supporting Information ).

Bacterial community structure in WS and SA tundr a hea ths
Bacterial community composition was c har acterized by 16S rRNA gene (r epr esenting total comm unity) and r e v erse tr anscribed 16S rRNA (r epr esenting activ e comm unity) amplicons.Due to a potential bias associated with the differences in sequencing (see the section "Materials and methods"), winter bacterial DNA samples were excluded from the multivariate statistics (PCO and PER-MANOVA) and not compared to data of the other seasons.
PCO ordination indicated the greatest differences between the DN A (total) and RN A (acti ve) deri v ed bacterial comm unity structur es ( Figur e S3 , Supporting Information ), whic h separ ated the samples into two main groups along the first axis that explained most of the variance.To delineate the effects of habitat and sampling season on the total vs. activ e comm unities, the DNA and RNA-derived datasets wer e anal yzed separ atel y.PCO ordination grouped both total and active communities from WS and SA tun-  S1 , Supporting Information ).Significance le v els: * * * , P < .001;* * , P < .01;* , P < .05;and NS, not significant.F igure 3. Bacterial 16S rRN A and ITS copy numbers in WS and SA tundra heaths sampled in winter, early, and late growing season.Values are means ± SE, N = 4. Significant differences of between the habitats (H) or sampling season (S) were analyzed using LME model ( Table S1 , Supporting Information ).Significance le v els: * * * , P < .001;* * , P < .01;* , P < .05;and NS, not significant.dr a heaths separ atel y (Fig. 4 ).This was supported by PERMANOVA analysis in which habitat explained slightly more of the comm unity v ariation than sampling season (Table 2 ).Pairwise PER-MANOVA analysis for the WS and SA habitats separ atel y indicated that the active bacterial communities were significantly different in all sampling seasons ( P = .001-.004) except in SA heaths between early and late growing season.Total bacterial communities wer e differ ent in WS heaths ( P = .001)but not in SA heaths ( P = .265)between early and late growing season.
Both tundra heath types were dominated by Actinomycetota, Pseudomonadota (mainly class Alphaproteobacteria), and Acidobacteriota (Fig. 5 ).At the genus le v el, the most dominant Actinomycetota were Acidothermus spp., Mycobacterium spp., and unknown taxa of the class Acidimicrobiales .T he most abundant Acidobacteriota were Granulicella spp., Bryobacte r spp., Edaphobacter spp., unknown Acidobacteriia (subdivision (SD) 1), and unknown SD 2 Acidobacteriota.The most abundant Pseudomonadota were unknown members of the Acetobacteraceae, Roseiarcus spp., Bradyrhizobium spp., Variibacter spp., Rhodoplanes spp., unknown gener a of Caulobacter aceae and Xanthobacter aceae, and the DA111 group of Alphaproteobacteria.The same genera dominated both in the DNA and RNA amplicons but with differences in the r elativ e abundances (Fig. 5 ; Figur e S3 , Supporting Information ).Of the Alpha pr oteobacteria, unknown gener a within the Acetobacter aceae and Xanthobacter aceae wer e mor e abundant in the F igure 4. PCO or dination of active (RN A derived) and total (DN A derived) bacterial communities and total fungal communities in WS and SA tundra heaths.Winter DNA-derived bacterial communities are excluded from the ordination due to possible technological bias in the data (see the section "Materials and methods").
RNA-deriv ed comm unity while Brad yrhizobium spp.wer e mor e abundant in the DNA.Similarly, Acidothermus spp.(Actinomycetota) wer e significantl y mor e abundant in RNA while Mycobacterium spp.were more abundant in the DNA-derived community.Of the Acidobacteriota, unknown members of SD2 and SD1 were more abundant in the DNA-than RNA-derived community.
Comparison of the active bacterial communities between WS and SA tundra heaths revealed relatively small differences in the dominant bacterial groups (Fig. 5 ; Figure S4 , Supporting Information ).Of the 10 most dominating active bacterial genera, unknown genera within the family Acetobacteraceae and members of Acidobacteriota were significantly affected by the habitat (LME, P < .05),but this interacted with the season.Acetobacteraceae wer e mor e abundant in WS heaths onl y in earl y and late growing season.Of the Acidobacteriota, unknown Acidobacteriaceae wer e mor e abundant in the SA tundr a heaths in the earl y and late growing season, Granulicella spp.were more abundant in WS tundra heaths but only in the early growing season, while members of SD2 Acidobacteriota were more abundant in SA tundra in the late growing season.The abundance of differ ent gener a

Fungal communities in WS and SA tundra heaths
PCO and PERMANOVA anal yses indicated that fungal comm unities wer e structur ed mor e by the habitat than by sampling season (Fig. 4 and Table 2 ).Mor eov er, the comm unities wer e mor e dispersed in the SA than WS heaths.Pairwise PERMANOVA analysis indicated that there were no differences between winter and early gr owing season comm unity structur e in the WS tundr a heaths while all other sampling seasons differed in WS and SA heaths ( P = .001-.049).Fungal communities under both tundra heaths w ere dominated b y Basidiom ycota, Ascom ycota, and Mucorom ycota.Basidiomycota were the most abundant fungi under both WS and SA heaths and wer e especiall y dominant in the winter samples.Of the Basidiomycota, the order Agaricales dominated in both habitats with Clavaria spp. as the most dominant genus in WS heaths and Cortinarius spp.more abundant in the SA heaths.Group B Sebacinales were more associated with the WS heaths.Members of the order Helotiales were the dominant Ascomycota both in WS and SA heaths while members of Chaetothyriales were more abundant in the WS heaths (Fig. 6 ).
Of the 10 most abundant gener a, Clav aria spp., unknown Sebacinales , unknown Chaetothyriales , and unknown Ascomycota wer e significantl y (LME; P < .05)mor e abundant in the WS tundra heaths while Mortierella spp.were more abundant in the SA heaths.Members of Cortinarius wer e mor e abundant in the SA heaths, but due to the high variation between samples, the difference was not significant.Rhizoscyphus and other Helotiales were abundant in both heath types and tended to be more abundant in the WS heath, but the difference was not statistically significant ( Figure S5 , Supporting Information ).

Effect of soil physico-chemical properties on bacterial and fungal community structure
Distance based linear modeling was used to determine the contribution of soil factors on bacterial and fungal community structur es.Mar ginal tests indicated that except for NO 3 , all nine tested soil factors had a significant impact on both bacterial and fungal community structure ( Table S2 and Figure S6 , Supporting Information ).Total nitrogen explained most of the variation of bacterial and fungal comm unity structur e and was also among the variables in the best DistLM model that predicted the community structur e. Activ e, RNA-deriv ed bacterial comm unity structur e was pr edicted best by N tot , P mic , OM%, and N mic .while total, DNA derived bacterial community (with only early and late growing season samples) was predicted by pH, N tot , and N mic and fungal community by Ntot, Nmic, pH, and OM% ( Figure S6 , Supporting Information ).The models explained 20.6%, 20.0%, and 26.3% of the variation in RN A-derived, DN A-derived bacterial and fungal community structure, respectively.

Discussion
WS and SA tundr a heaths ar e c har acterized by lar ge differ ences in soil temper atur e.In addition to strong differences in the winter soil temper atur es, ther e is a differ ence in the amplitude of annual temper atur e v ariation.Due to earlier sno wmelt and lo w er insulation fr om v egetation, the WS heaths ar e associated with high-temper atur e fluctuation and, depending on the air temperatur es, also pr one to fr equent fr eeze-thaw cycles during spring (Männistö et al. 2013 ; this study).These differences in topogr a phy lead to a spatially heterogeneous snow cover depth and soil temper atur es that in turn form predictable patterns in the dominant vegetation (Oksanen andVirtanen 1995 , Niittynen et al. 2020 ), but so far, how these differences drive microbial community composition has remained uninvestigated.Supporting our hypothesis that the topogr a phic differ ences and associated consequences on v egetation, soil temper atur es , nutrient a vailability as well as the quantity and quality of OM modify soil microbial community composition, PLFA, qPCR, and sequencing of 16S rRNA genes and the ITS region indicated distinct differences in the bacterial and fungal communities in WS vs. SA tundra heaths.Contrasting our hypothesis, ho w e v er, the seasonal trends were either uniform between the WS and SA tundra heaths or amplified to w ar d the end of the growing season rather than being greatest in winter when the difference in soil temperatures was the greatest.These results imply that differences in the bacterial and fungal communities in WS and SA tundra heaths across different seasons were likely driv en by v egetation, soil nutrients, and carbon substr ates r ather than by soil temper atur e per se .
PLFA and qPCR analyses indicated generally higher fungal abundance in the WS tundra heaths while bacterial PLFAs tended to be higher in the SA tundra heaths with higher nutrient availability.This is in line with other studies from tundra showing positiv e corr elations of bacterial abundance with nutrient availability (Eskelinen et al. 2009, Stark et al. 2012 ).Contrary to our hypothesis, ther e wer e onl y minor differ ences in the bacterial and fungal gene copy numbers between WS and SA heaths in winter, but differ ences wer e ma gnified to w ar d the end of the gro wing season when both bacterial and fungal gene copy numbers dropped in the WS heaths, coinciding with dr asticall y lo w er soil N and P concentrations .T he parallel trends in bacterial and fungal gene copy numbers and nutrient availability indicate that limitations in soil nutrient a vailability ma y ha v e contributed to incr eased micr obial turnover during the growing season in WS heaths, which supports earlier findings of strong competition for nutrients between plants and microbes in Arctic nutrient-poor soils (Jonasson et al. 1996, Stark and Kytöviita 2006, Stark et al. 2023 ).Owing to a strong nutrient limitation, tundra soil carbon and nitrogen cycles are strongly coupled, and microbial N immobilization and even soil microbial biomass may be regulated by plant nitrogen uptake (Jonasson et al. 1999, 2001, Schimel and Bennett 2004 ).Following the patterns of plant nitrogen uptake, soil nitrogen availability often decreases during the growing season, leading the soil microbial communities to become incr easingl y N-limited to w ar d the end of the growing season (Weintraub and Schimel 2005a,b , Stark and Kytöviita 2006, Wallenstein et al. 2009 ).This induces a strong seasonality of microbial biomass and activities that may be at their highest in early spring and late autumn when the plant activity is at its lo w est (Stark andKytöviita 2006 , Stark andVäisänen 2014 ).The decrease of bacterial and fungal gene copy numbers to w ar d the end of the growing season in the nutrient-poor WS tundra heaths is, thus in line with microbial communities being driven more by nutrient limitations than dir ectl y by the winter temper atur es.
Another factor that may have contributed to the decline in microbial biomass in WS tundra heaths to w ar d the end of the growing season could be decreased availability of labile C that would incr ease turnov er of micr obial biomass.In contr ast, some studies suggest that increased wintertime decomposition under increased snow depth may decrease microbial respiration in the follo wing gro wing season due to reduced availability of labile C substr ates (Semenc huk et al. 2016, Sulliv an et al. 2020 ).Differences in vegetation between WS and SA sites, ho w e v er, likel y cause differences in soil C and N by multiple mechanisms.Empetrum nigrum that was especially dominant in the WS sites produce allelophatic compounds and slowly decomposable litter that gener all y deceler ate soil nutrient and carbon cycles (Bråthen et al. 2010 , Vowles andBjörk 2019 ).Mor eov er, the known and putative ErM fungi, whic h wer e mor e dominant in the WS sites, pr oduce r ecalcitr ant necr omass that further incr eases the stability of the OM and reduces the availability of labile C and N forms (Clemmensen et al. 2015(Clemmensen et al. , 2021 ) ). Snow-cov er r elated to topogr a phy may, thus be an important microclimatic driver of both microbial community composition and SOM dynamics.
The higher fungal PLFA abundance and fungal-to-bacterial ratio in the WS tundra heaths ( Table S1 , Supporting Information ) was associated with clear differences between the fungal communities in WS and SA tundra heaths.While known ericoid mycorrhizal fungi, such as Rhizoscyphus spp.and other Helotiales (Clemmensen et al. 2015, Leopold et al. 2016 ) were abundant in both, WS and SA tundra heaths (Fig. 6 ), ectomycorrhizal fungi, especially the genus Cortinarius , were more abundant in some sites of the SA tundra heaths corresponding to higher abundance of ectom ycorrhizal d warf shrubs ( B. nana and Salix spp.) under SA heaths.The low abundance of Cortinarius spp. in WS heaths may, ho w e v er, also be explained by the sensitivity of these fungi to freezing (Ma et al. 2011 ) as some Cortinarius species have been shown to benefit fr om incr eased snow depth in dry tundra sites of Alaska (Morgado et al. 2016 ).Members of the genus Clavaria were the most abundant fungal taxa at the genus le v el and significantl y mor e abundant in the WS tundra heaths .T he most abundant O TU of the whole fungal data set ( > 11% of all sequence reads) was related to Clavaria sequences from Arctic soils (Deslippe et al. 2012, Dahl et al. 2017 ) and Clavaria argillacea sampled from Greenland, indicating an association to Arctic ecosystems.Clavaria has been reported as the dominant taxa in tundra soils of Alaska (Semenov a et al. 2016 ), Gr eenland (Vo říšk ová et al. 2019 ), and in Raisduoddar fell in northern Norway located close to our study site (Ahonen et al. 2021 ).Clav aria spp.ar e consider ed sa pr otr ophic, but they have been found abundantly in hair roots of ericoid shrubs such as V. uliginosum (Yang et al. 2018 ), Vaccinium corymbosum (Li et al. 2020 ) and bidirectional nutrient transport between a Clavaria and ericaceous plant species have been reported (Englander and Hull. 1980 ), indicating that they may have symbiotic associations with the ericoid shrub vegetation in nutrient-poor tundra heaths.Clavaria was identified as one of the most dominant taxa associated with roots of the Ericaceae shrub Cassiope tetragona growing in Sv albard (Lorber au et al. 2017 ), further suggesting its importance in Ericaceae shrub-dominated tundra heaths .T he higher abundance in the more nutrient-poor WS tundra heaths in this study suggests a role of Clavaria spp. in nutrient accessibility and transport between/for the ericoid v egetation.Furthermor e, members of Clavaria and Clavariaceae have been associated with freezethaw tolerant fungal communities in soil from Northern Sweden (Perez-Mon et al. 2020 ), suggesting their to more extreme winter conditions, which may contribute to their abundance in the WS heaths.
Other fungal taxa that were consistently more abundant in the WS tundra heaths were members of the order Chaetothyriales and the family Serendipitaceae of the order Sebacinales (group B).Sebacinales comprises a div erse gr oup of cryptic organisms that ar e consider ed to form symbiotic r elationships with man y types of vegetation (Leopold et al. 2016, Weiß et al. 2016 ), including ericoid mycorrhizal associations with Ericaceae shrubs (Selosse et al. 2007, Vohník et al. 2016 ).Similar to Clavaria , Sebacinales was reported as the dominant order of the roots of C. tetragona and Bistorta vivipara in Svalbard where they were suggested to play an important role in the Arctic tundra ecotone either as mycorrhizae or as endophytes (Blaalid et al. 2014, Lorberau et al. 2017 ).Ho w e v er, the ecological role of the Sebacinales in these ecosystems require further r esearc h as OTUs both in this study and those fr om Sv albard had low sequence similarities to known Sebacinales, a taxonomicall y and functionall y a div erse gr oup with man y unknown r epr esentativ es (Oberwinkler et al. 2013 ).
PERMANOVA analysis indicated significant differences in the bacterial communities between WS and SA tundra heaths during all sampling seasons but contrary to our hypothesis there wer e r elativ el y minor differ ences in the dominant bacterial taxa during winter.Bacterial communities were dominated by members of Actinomycetota, Pseudomonadota (Alpha pr oteobacteria), and Acidobacteriota.These groups have been shown to dominate the soil and mycosphere of ericoid shrubs (Timonen et al. 2017 ), indicating their link to the shrub vegetation and/or associated fungi.Actinomycetota, Pseudomonadota, and Acidobacteriota wer e r eported as the dominant taxa also in metagenomes and metatranscriptomes of tundra soil sampled from the same area (Pessi et al. 2022, Viitamäki et al. 2022 ).Our earlier studies of the same tundra habitats indicated a high dominance of Acidobacteriota in clone libraries of both WS and SA tundra heaths (Männistö et al. 2013 ).In this study, the dominance of Acidobacteriota was lesser with a greater abundance of Actinomycetotaassociated reads .T his incr ease in the shar e of Actinomycetota is likely due to differences in the DNA and RNA extraction protocol with stronger beat beating conditions used in our current protocol (described in Männistö et al. 2016 ).The most abundant bacterial OTUs were associated with the Actinomycetota genus Acidothermus which comprised ca.20% of all reads and was especially dominant in the RNA-derived bacterial community both in WS and SA tundra heaths.Acidothermus spp. was r ecentl y r eported as one of the most dominant genus-le v el taxa also in metagenomes and transcriptomes of soils sampled from the same area (Viitamäki et al. 2022 ).They wer e especiall y abundant in acidic shrubdominated heaths, indicating that this taxon may have an important role in the organic-rich, nutrient-poor tundra soils.Members of Acidothermaceae were abundant also in alpine soils, where they increased with expansion of ericaceous shrubs (Broadbent et al. 2022 ) further indicating their association with ericaceous vegetation.The only described species of this genus is a thermophilic cellulose degrader (Mohagheghi et al. 1986 ) indicating that the role of tundra soil Acidothermus spp.may be associated with the decomposition of the lar ge plant-deriv ed or ganic stoc ks.As Acidothermus spp.belong to the order Fr ankiales, whic h contain wellknown nitrogen-fixing species (Gtari et al. 2012 ) that may rise the question of whether the high abundance of these Actinomycetota is connected also to the nitrogen limitations of the habitat.
The most abundant class-le v el taxa in both WS and SA tundra soils were members of Alphaproteobacteria.The abundance and role of many of these alphaproteobacterial taxa may be associated with nitrogen acquisition as they are associated with known nitr ogen-fixing gener a (Tsoy et al. 2016 ).Bradyrhizobium and other Alpha pr oteobacteria wer e found to be dominant members of the denitrifier populations in the tundra soils of Kilpisjärvi where they wer e r e ported to encode terminal o xidases that are acti ve both under highl y aer obic conditions and those with high oxygen affinity (Pessi et al. 2022 ).This type of adaptation would give a competiti ve ad vantage for growth under waterlogged (spring) and dry conditions.Bradyrhizobium and many other alphaproteobacterial taxa have been also reported as dominant lignin-decomposing taxa in Alaskan tundra soils (Tao et al. 2020 ), indicating that in addition to a putative role in nitrogen cycling their abundance may be explained by their role in the decomposition of recalcitrant OM in or ganic-ric h soils .T he r elativ el y small differ ences between bacterial communities in WS and SA tundra heaths may be due to the versatility of the dominant bacterial groups, with putative roles both in decomposition and nutrient acquisition.Ho w e v er, additional studies are needed to understand the mechanisms associated with these communities.
Although differences in vegetation and its mycorrhizal associations wer e likel y important determinants for the soil microbial comm unity composition, soil physico-c hemical pr operties contributed significantly to both the bacterial and fungal community structure.Based on the DistLM model, the analyzed soil variables (pH, OM, and different N and P forms) explained r oughl y 20% and 26% of the variation in bacterial and fungal community structur es, r espectiv el y.Apart fr om nutrient av ailability, pH and SOM were k e y determinants of acti v e bacterial and fungal comm unity structur e.Of the anal yzed soil pr operties, particularl y soil pH has been shown to control the bacterial community structure in the same area (Männistö et al. 2007 ) as well as bacterial and fungal comm unity structur e globall y (Lauber et al. 2009, Tedersoo et al. 2014 ).Although there were no significant differences is soil pH between the WS and SA sites, and the variation among different sample replicates was small, soil pH contributed significantly to the variation of bacterial and fungal community structures also in this study.Ho w e v er, as the soil pH w as belo w 5 in all sites, the bacterial and fungal comm unities wer e dominated by acid-tolerant and oligotr ophic taxa, whic h may constitute an important factor behind the similarities of the communities between WS and SA tundra heaths.

Conclusions
As hypothesized, we found clear differences in the soil microbial communities between WS and SA tundra heaths .Snow-co ver related to topography may, thus be an important microclimatic driver of both microbial community composition and soil C and N dynamics.WS heaths experienced very low wintertime soil temper atur es and strong seasonal fluctuations, which drastically differ ed fr om the mor e stable soil temper atur e r egimes of the SA heaths, but contrary to prediction, the soil microbial communities differed the most during the late growing season rather than during winter.Further contrasting predictions, we did not detect distinct cryotolerant communities in WS heaths during winter.A higher abundance of fungal PLFAs and a lo w er nutrient availability in the WS heaths together with the differing fungal community composition between the WS and SA heaths suggested that these communities were represented by stress-tolerant organisms adapted to nutrient-poor and cold soils.Pr e vious studies have indicated that these types of microbial communities may be r ather insensitiv e to extr eme temper atur es (Männistö et al. 2018 ).
Instead, the seasonal patterns in microbial biomass and comm unity composition closel y follo w ed the seasonal patterns in soil nutrient av ailability, especiall y in the mor e nutrient-limited WS tundra heaths .T hese r esults suggest that, r ather than dir ectl y thr ough low wintertime soil temper atur es, topogr a phic differ ences sha pe soil micr obial comm unities thr ough modifying the dominant vegetation and soil nutrient a vailability.T his could partially explain why different snow-manipulation experiments have found inconsistent effects of snowpack on microbial community structure and function, ranging from no effect to transient or legacy effects on bacterial and fungal communities in tundra, forest, and alpine soils (Aanderud et al. 2013, Morgado et al. 2016, Mundr a et al. 2016, Ric ketts et al. 2016, Gav azov et al. 2017, Männistö et al. 2018, Vo říšková et al. 2019 ).Future experimental studies on Arctic soils should, thus include se v er al habitat types with div er gent dominant vegetation and nutrient le v els to fully separate the direct and indirect effects of temperature.
In the future, the ongoing climate warming will lead to more fr equent fr eeze-thaw c ycles and a lo w er duration and insulation of the snow cover (Bintanja and Andry 2017, McCrystall et al. 2021, Serreze et al. 2021 ).According to our findings from current topogr a phic gr adients in snow accum ulation, owing to the high str esstolerance of soil microbial communities in acidic soils with low temper atur es (Männistö et al. 2018 ), these climatic changes may potentiall y hav e a minor r ole for futur e soil micr obial comm unities.Instead, shifts in microbial community composition in response to climate warming will likely be lar gel y mediated by shifts in the dominant vegetation and corresponding effects on mycorrhizal associations as well as substrate and nutrient availability for soil micr oor ganisms.Assuming that the ongoing expansion of deciduous shrubs continues across the cir cumpolar Ar ctic (e.g.Myers-Smith et al. 2020 ), counterintuitiv el y, soil micr obial communities similar to what we detected for SA tundra heaths might, ther efor e incr ease in cov er a ge.

Figure 1 .
Figure 1.Soil temper atur e in the WS and SA tundr a heaths.Values ar e means obtained by data loggers in eac h plot ( N = 4).Stars denote the sampling points for bacterial and fungal community analyses.

First
amplification of the V1-V3 region of the 16S rRNA gene was done in duplicates for each DN A/cDN A sample in 20 μl reactions containing 1 μl of 1:50 diluted DNA template, Dream-Taq DNA Pol ymer ase (Thermo Scientific), 0.3 μM of eac h primer (27F 5 -A GA GA GTTTGATCMTGGCTCA G-3 ; Lane 1991 , and 518R 5 -A TT ACCGCGGCTGCTGG-3 , Muyzer et al. 1993 ), 3.2 μg of bovine serum albumin and 0.2 mM of dNTP mix.In the second amplification step, primer IonA_bc_27F included adapter IonA (5 -CC ATCTC ATCCCTGCGTGTCTCCGAC-3 ) and unique 10-12 bp long barcode sequences before the primer 27F and primer P1_518r included P1 5 -CCTCTCT A TGGGCAGTCGGTGA T-3 before the primer 518r to allow Ion Torrent sequencing and assignment to specific samples .T he cycling regime for the first PCR included denaturation of 95 • C for 5 min, follo w ed b y 25 c ycles of 94 • C 30 s, 55 • C 30 s, and 72 • C 1 min, and a final elongation step of 72 • C for 10 min.The second PCR reaction contained 1 μl of the reaction product of first amplification, and only 15 cycles were run.PCR pr oducts wer e cleaned using the Agencourt AMPure XP magnetic beads purification system (Bec kman Coulter, Br ea, CA, USA) and quantified with Qubit dsDNA HS Assay Kit (Thermo Scientific).
entific), Phusion HF Buffer, and 0.2 mM of dNTPs and each primer (ITS7 5 -GTGARTCA TCGAA TCTTTG-3 , Ihrmark et al. 2012 and ITS4 5 -TCCTCCGCTT A TTGA T A TGC-3 , White et al. 1990 ).In the second amplification step, ITS7 primer included the adapter P1 5 -CCTCTCT A TGGGCAGTCGGTGA T-3 and ITS4 primer included the adapter IonA 5 -CC ATCTC ATCCCTGCGTGTCTCCGACTC AG-3 and unique 10-bp long barcode sequences at the beginning of the primer to allow Ion Torrent sequencing and assignment to specific samples .T he cycling regime for the first step was: initial denaturation of 98 • C for 1 min, follo w ed b y 25 c ycles of 98 • C 10 s, 54 • C 20 s, and 72 • C 30 s, and a final elongation step of 72 • C for 7 min.For the second step cycling regime was the same as in the first except for changing the cycle number to 10 and the annealing temper atur e to 57 • C. Fungal ITS amplicons were sequenced at Biocenter Oulu Sequencing Center (Univ.Oulu, Oulu, Finland).PCR pr oducts wer e first cleaned using the Agencourt AMPure XP magnetic beads purification system and BioMek4000 Laboratory Automation Workstation (Beckman Coulter) follo w ed b y purity c hec king and quantification using MultiN A (MultiN A Micr oc hip Electr ophor esis System and DNA-1000 kit; Shimadzu, K yoto , Ja pan) and PicoGr een (Thermo Scientific) according to manufacturers' instructions.Amplicons were then combined in equimolar concentrations for Ion Torrent PGM sequencing with 314 v2 chip.

Figure 2 .
Figure 2. Bacterial and fungal PLFAs and their ratio in WS and SA tundra heaths sampled in winter, early, and late growing season.Values are means ± SE, N = 4. Significant differences of between the habitats (H) or sampling season (S) were analyzed using LME model ( TableS1, Supporting Information ).Significance le v els: * * * , P < .001;* * , P < .01;* , P < .05;and NS, not significant.

Figure 5 .
Figure 5. Abundance of bacterial classes (A) and dominant genera (B) in WS and SA tundra heaths sampled in winter, early, and late growing season.

Figure 6 .
Figure 6.Abundance of fungal orders (A) and genera (B) in WS and SA tundra heaths sampled in winter, early, and late growing season.UK = unknown.

Table 1 .
Soil physico-chemical properties in WS and SA tundra heaths.Values are means with SE in brackets.

Table 2 .
PERMANOVA main test for bacterial (derived from RNA and DNA) and fungal community structure in WS and SA habitats (Ha) acr oss thr ee differ ent seasons (Se) in four plots per habitat.Significance le v els: P < .05ar e underlined, P < .01ar e indicated by bold.
a For the bacterial DNA PERMANOVA anal yses, onl y samples from early and late growing season were included.