Contrasting sensitivity of soil bacterial and fungal community composition to one year of water limitation in Scots pine mesocosms

Abstract The soil microbiome is crucial for regulating biogeochemical processes and can, thus, strongly influence tree health, especially under stress conditions. However, little is known about the effect of prolonged water deficit on soil microbial communities during the development of saplings. We assessed the response of prokaryotic and fungal communities to different levels of experimental water limitation in mesocosms with Scots pine saplings. We combined analyses of physicochemical soil properties and tree growth with DNA metabarcoding of soil microbial communities throughout four seasons. Seasonal changes in soil temperature and soil water content and a decreasing soil pH strongly influenced the composition of microbial communities but not their total abundance. Contrasting levels of soil water contents gradually altered the soil microbial community structure over the four seasons. Results indicated that prokaryotic communities were less resistant to water limitation than fungal communities. Water limitation promoted the proliferation of desiccation tolerant, oligotrophic taxa. Moreover, water limitation and an associated increase in soil C/N ratio induced a shift in the potential lifestyle of taxa from symbiotic to saprotrophic. Overall, water limitation appeared to alter soil microbial communities involved in nutrient cycling, pointing to potential consequences for forest health affected by prolonged episodes of drought.


Introduction
Pr ecipitation pr ojections in futur e climate scenarios ar e v ariable and uncertain; ne v ertheless, gener al circulation models pr oject a consistent increase in air temperature and, thus, evapotranspiration in Eur ope, r esulting in decr eased soil moistur e in man y r egions (Vogel et al. 2018, van der Linden et al. 2019. In general, the reduction of soil moisture will force soil microbes to either avoid or tolerate water-limited conditions while striving for nutrient and energy sources that become spatially less available (Manzoni et al. 2014 ).
Soil water availability controls soil microbial community dynamics in three fundamental ways: as a resource, solvent, and transport medium (Tecon and Or 2017 ). All of these are important for regulating the function of the soil microbiome (Schimel 2018 ). Ther efor e, a declining soil water content can create envir onmental str ess for micr obes (Sc himel et al. 2007 ). Mor eov er, w ater scar city can dir ectl y affect the soil micr obiome by cr eating osmotic stress, leading to microbial death and cell lysis (Csonka 1989, Turner et al. 2003. Ho w e v er, it is expected that some taxa e volv e dr ought r esistance , e .g. by accumulating osmol ytes (Sc himel et al. 2007, Warr en 2014, pr oducing exopol ysaccharides (Roberson and Firestone 1992 ), forming thick cell walls (Potts 1994 ), or entering a dormant state (Lennon and Jones 2011 ).
Although bacteria endure desiccation by shifting to dormancy, producing cysts, or inhabiting small soil pores (Schimel et al. 2007 , Lennon andJones 2011 ), fungi are often considered to be more tolerant to water limitation than bacteria (Schimel et al. 2007, Strickland and Rousk 2010, Manzoni et al. 2012. Important assets of fungi under dry conditions, include osmol ytes, thic k cell walls, and melanin (Schimel et al. 2007 ). Moreover, the higher toler ance is mainl y attributed to the ability of fungi to cr eate lar ge hyphal networks that maintain nutrient and water translocation over long distances and allow the scavenging of small soil pores filled with water (Allen 2007 , Joergensen andWichern 2008 ). Soil ecosystems dominated by fungi ar e, as suc h, assumed to be more resistant to desiccation than soils dominated by bacteria (Yuste et al. 2011, De Vries et al. 2012. Furthermore, it has been proposed that fungi have lo w er nutrient requirements than bacteria (Güsewell andGessner 2009 , Strickland and, and that fungi can utilize a broader range of nutrient resources (Sterner and Elser 2003 ).
In addition to the mentioned direct physical effects, altered water availability may impact the soil microbiome indirectly through changes in vegetation or substrate supply (Nielsen and Ball 2015 ). Plants undergo a set of physiological reactions in response to water deficit. Adaptations to lower soil water contents result inter alia in changes in the amount and chemical quality of plant litter input to the soil (Cromer et al. 1984 ) and altered root growth and exudation profiles (Brunner et al. 2015, Hasibeder et al. 2015. For example, the quality of litter input decreases under drought through enhanced production of recalcitrant structural compounds by plants (Pugnaire et al. 2019 ). These changes consecutiv el y slow down miner alization r ates and nutrient release, further affecting the functional structure and activity of the micr obial comm unities (Bolton et al. 1992, Grayston et al. 1998 ). In dry soils, an increased abundance of microbial genes involved in the degradation of complex plant polysaccharides has been observed, while the quantity of microbial genes targeting less complex oligosaccharides of fresh organic matter inputs often decreases (Bouskill et al. 2013, Martiny et al. 2017. T herefore , it is pr oposed that dr ought-r elated decr easing plant vitality causes a community succession to w ar d taxa capable of degrading complex structures of dead plant material enhancing saprotrophic taxa (Vorisk ov a and Baldrian 2013, Kielak et al. 2016b, Baldrian 2017, Herzog et al. 2019. Concurr entl y, symbionts suc h as ectomycorrhizal fungi (EcM) and certain diazotrophic bacteria depend on their host plants (Nehls 2008, Carvalho et al. 2014, Mercado-Blanco et al. 2018. Ther efor e, the abundance and richness of symbiotic taxa are expected to decline with changing environmental conditions due to impair ed inter action with the host (Churchland and Grayston 2014, Knoth et al. 2014. Impair ed tr ee gr owth and physiology hav e been documented in se v er al ar eas of the Eur opean Centr al Alps, as in Switzerland, Italy, and Austria (Vertui and Tagliaferro 1998, Rebetez and Dobbertin 2004, Rigling et al. 2018, with Scots pine ( Pinus Sylvestris L.) being one of the dominant tree species of inner-Alpine forests (Leuschner and Ellenberg 2017 ). Although average annual precipitation has remained constant over the last decades, there is evidence that climate warming has increased evapotranspiration rates and that water has become a main factor affecting the vitality and stress resilience of trees (Rigling et al. 2013 ).
Pr e vious findings fr om a long-term irrigation study conducted in a xeric mature forest stand (Hartmann et al. 2017 ) indicated that water-limiting conditions favor oligotr ophic, metabolicall y v ersatile, and dr ought-toler ant taxa. Ther e is, ho w e v er , little information available on potential changes in microbial communities under different levels of water limitation over the whole growing season of trees. To assess whether observ ations fr om field studies can be confirmed under controlled experimental conditions during the growing season, we conducted a one-year mesocosm experiment with Scots pine saplings and natural soil from a xeric matur e for est stand (Herzog et al. 2014(Herzog et al. , 2019.
The main goal of this study was to assess whether and how soil prokaryotic and fungal communities respond to different levels of experimental water limitation in Scots pine mesocosms. Mor eov er, w e w er e inter ested in understanding if variations in soil micr obial comm unities ar e r elated to alter ations in tr ee gr owth and soil physicoc hemical pr operties. We hypothesized that sev er e water limitation and altered resource availabilities would cause community succession to w ar d an enhanced abundance of sa pr otr ophic taxa ca pable of degr ading complex or ganic compounds of dead plant material. At the same time, we expected that water limitation would decrease symbiotic taxa through c hanges in physicoc hemical soil pr operties. Mor eov er, we pr esumed that desiccation-tolerant taxa and oligotrophic bacteria would increase under water str ess. Finall y, we hypothesized that the composition of soil fungal communities would be more resistant to water limitation than prokaryotic communities.

Experimental set-up
For the experiment, natural forest soil was collected in a xeric forest in the upper Rhone Valley (Pfyn forest, Canton Valais, Switzerland, 46 • 18 16.1 N, 7 • 36 44.8 E, and 600 m a.s.l.) in November 2018. The soil (P ar ar endzina, de v eloped on an alluvial fan) was collected at the forest margins of the Pfyn Natur e P ark, below the for est canopy. For the soil collection, an excavator was used to r emov e the 3-6 cm deep organic horizon (Oe horizon), and ∼4 t of soil were taken from the upper ∼35 cm of the mineral soil (sand/silt/clay %: 49/43/8, skeletal material: 20%-50%, and bulk density: 1.3 g cm −3 ). The soil was transported in soil bulk bags to the greenhouse facility at the Research Station for Plant Sciences Lindau (ETH Zurich, Switzerland) and stored outside on wooden pallets , co v er ed for pr otection a gainst r ain. The soil was homogenized with a steel riffle splitter, and large stones were removed. In February 2019, part of the soil was used for the initial planting of 2-year-old Scots pine saplings ( Pinus Sylvestris L.) in pots (6 L volume). The Scots pine had been gr owing fr om seeds in a common potting substrate (seed origin: Leuk, Switzerland, 980-1250 m a.s.l.) at the tree nursery of the Swiss Federal Research Institute for Forest, Snow and Landscape Researc h (WSL, Birmensdorf, Switzerland). The pur pose of this initial planting was to acclimate the saplings to the forest soil for the duration of one growing season.
In September 2019, 18 of the Scot pine saplings (which had by then experienced three growing seasons) wer e individuall y tr ansplanted in pots with a size of 32 cm height × 69 cm diameter (100 L volume) at the greenhouse facility. During the transplantation of each Scots pine, the pots were filled with a 2-3 cm layer of stones (10-15 kg) and 20 cm of forest soil (100-110 kg). For three months, the plant-soil systems (subsequentl y r eferr ed to as "mesocosms") w ere w atered twice per week with 2 L of local rainwater reaching a volumetric water content (VWC) of ∼30% (close to field capacity, which was ∼35% for the soil, with a pF of 1.8). In January 2020, the mesocosms were assigned to three different irrigation treatments in a randomized design to minimize spatial effects (i.e. variability in shading). The thr ee le v els of irrigation were: sufficient water suppl y (contr ol; 30% VWC, n = 6), decr eased amount of water (intermediate; 40% reduction in VWC of control, n = 6), and water str ess (se v er e; 75% r eduction in VWC of contr ol, n = 6) (Fig. 1 ). The intermediate tr eatment r epr esents the maxim um for ecasted deviation of precipitation from the normal under a future climate in Southern Switzerland (period: 2081-2100) compared to 1981-2010 without climate change mitigation (CH2018-Climate Scenarios for Switzerland 2018 ). The se v er e tr eatment was chosen to maximize the effect of w ater stress. Ho w ever, the soil w ater content was k e pt at a le v el, at whic h the sa plings r eceiv ed a minim um of water not to suffer from permanent damage and maintain vitality. The greenhouse temperatures were regulated to simulate seasonal c hanges experimentall y (Supplementary Table 1) and constantl y contr olled, together with the humidity, whic h was k e pt ar ound 50%-70% thr oughout the seasons (Supplementary Table  1). The mesocosms were equipped with soil sensors continuously measuring VWC, soil water potential, and soil temper atur e (Ter os 11, Teros 21, Meter Group, Pullman, WA, USA).

Scots pine structure and photosynthetic capacity
The structure of the Scots pine was monitored by measuring height and diameter e v ery month as a proxy of aboveground plant productivity. The height was measured, including the buds, and the seasonal increment was calculated as tree growth. The diameter was measured at two angles, and the mean was taken. Here the seasonal increment was calculated as radial growth. The needle fall was collected each season on a PE net (mesh size 3 × 3 mm) placed above the pot. The collected needle litter was weighed, dried for one week at 40 • C, and redistributed on the soil surface according to the av er a ge needle fall for each of the three irrigation treatments . T he net photosynthetic assimilation was measur ed e v ery season to assess the maxim um r ate at whic h leav es ar e able to fix carbon during photosynthesis as a pr oxy of plant vitality (Johnstone et al. 2013 ). Ther efor e, 25 south exposed needles were enclosed in the 2 × 3 cm chamber of the LI-COR LI-6400 (LI-COR Biosciences, Lincoln, N.E., USA), and net photosynthetic assimilation (A net ) was measured under 400 μmole mol −1 CO 2 , 1000 PAR, local humidity and temper atur e, and a stomatal ratio of 1.

Soil sampling
Bulk soil sampling was performed at different stages of plant gr owth and temper atur e settings -Se ptember 2019, Jan uary 2020, Mai 2020, July 2020, October 2020, and January 2021-in all pots ( n = 18) with a slide-hammer corer (30 cm length, 5.5 cm diameter) to a depth of 25 cm. In the following text, these sampling time points will be termed "seasons" to account for differences in plant growth and gr eenhouse temper atur es and r eferr ed to as "A utumn-19, " "W inter-20, " "Spring-20, " "Summer-20, " "A utumn-20, " "W inter-21". After sampling, the soil samples were k e pt cooled, transported to the lab, and immediately stored at 4 • C.

Physicochemical soil properties
T he da y after eac h sampling, the fr esh soil samples wer e sie v ed to 4 mm, and lar ge or ganic r esidues wer e sorted out manuall y. Fresh soil samples were used to analyse gravimetric water con-tent (GWC, 10 g of soil) and inorganic nitrogen (NH 4 + , NO 3 − , 10 g of soil) and to perform DNA extraction (0.250 g). Soil samples for molecular anal ysis wer e fr ozen at −20 • C until further processing. The rest of the soil samples were dried at 40 • C to constant weight and sie v ed to 2 mm for measur ements of soil pH, total carbon (TC), inorganic carbon (IC), organic carbon concentration (C org ), and total nitrogen concentration (TN). Soil GWC was assessed by weighing a subsample of 10 g of soil before and after drying at 105 • C for at least 48 h. Ammonium (NH 4 + ) and nitrate (NO 3 − ) concentrations of the soils were determined by extraction of 10 g soil with 50 ml 2 M KCl solution in a 1:5 soil solution ratio and shaken for 1 h at 180 r/m. Afterw ar d, the soil-KCl solution was filtered with a 150 nm ashless filter paper (Whatman No. 42), and the clear solution was stored at −20 • C until further analyses. NH 4 + and NO 3 − concentrations were determined colorimetrical with a spectrophotometer v-1200 (VWR, Radnor , P A, USA) following Forster ( 1995 ) for NH 4 + and Doane and Horwáth ( 2003 ) for NO 3 − . Soil pH was measured in a 1:2.5 solution containing 10 g of dried soil and 25 ml of 0.01 M CaCl 2 solution. Samples were shaken horizontally at 180 r/m for 1 h and stored overnight to allow sedimentation before measurement with a pHmeter (VWR, Radnor , P A, USA). TC and nitr ogen concentr ations were determined with an elemental analyser LECO 628 (LECO, St. Joseph, MI, USA). IC was measured with the pressure-calcimeter method following Sherrod et al. ( 2002 ). The C org was calculated as follows: C org = TC-IC and used to determine C org /TN, which will be subsequently referred to as C:N ratio.

Microbial gene abundance
DNA extr action fr om 0.250 g fr esh soil was performed with the DNeasy Po w erSoil Pr o Kit (Qia gen, Hilden, German y) according to the manufacturer's instructions using the QIACube System (Qia gen, Hilden, German y). Quality and quantity of extracted DNA wer e measur ed via UV/VIS spectr ophotometry with the QIAxpert System (Qia gen, Hilden, German y). The abundance of taxonomic groups -bacteria and archaea, hereafter termed as prokaryotes, and fungi-was assessed using a SYBR ® Gr een-based qPCR a ppr oac h. Potential amplification inhibition induced by unintentional co-extraction of contaminants was tested across all samples by spiking pGEM-T plasmid (GenBank ® Accession No. X65308; Promega, Madison, WI, USA) into the soil DNA at equimolar concentrations in all samples and amplifying a region on the plasmid using specific primers SP6 and T7 (Micr osynth, Balgac h, Switzerland). The qPCR standar ds w er e pr oduced fr om purified PCR pr oducts obtained by pooling DNA from randomly selected samples.  (Vainio and Hantula 2000 ), were chosen to target the 18S rRNA gene (V7-V8 region). All qPCR reactions were performed in a final volume of 25 μL containing a final concentration of 0.75 μM of primer, 1x Sso Ad vanced™ Uni versal SYBR ® Green Supermix (Bio-Rad Laboratories , Hercules , C A, USA), and 10 ng of DNA template . T he conditions for the qPCR assays w ere as follo ws: 3 min for enzyme activation at 98 • C, follo w ed b y 15 s for denaturation at 95 • C, 30 s primer hybridization at 52 • C for 16S and 50 • C for 18S, r espectiv el y, and 30 s for elongation at 72 • C, for 35 cycles. To verify the amplification specificity, melting curv es wer e gener ated by incr easing the temper atur e fr om 75 • C to 95 • C by 0.5 • C e v ery 5 s at the end of the amplification cycles. All qPCRs were performed in technical triplicates in a thermocycler CFX96 Touch Real-Time System (Bio-Rad Laboratories, Hercules , C A, USA), and the results were documented and analysed using the CFX Maestro software (Bio-Rad Laboratories , Hercules , CA, USA). To c hec k for between run differ ences, eight standard samples were repeated in every plate and compared. The qPCR efficienc y w as betw een 90% and 100% with an R 2 > 0.99 for all runs.

Microbial community structure
Changes in prokaryotic and fungal community structure were investigated with DN A metabar coding of ribosomal marker genes. PCR amplification of the prokaryotic 16S rRNA gene (V3-V4 region) was performed with the primers 341F and 806R (Frey et al. 2016 ). Primers ITS3ngs and ITS4ngs were used to amplify the fungal ITS2 region of the rrn operon (Tedersoo and Lindahl 2016 ). PCR reactions contained 1X of GoTaq ® G2 Hot start Master Mix from Pr omega (Pr omega, Madison, WI, USA), 0.4 μM of the primers, and 40 ng of DNA template in a final volume of 25 μL. Cycling conditions for the PCR reactions consisted of a pol ymer ase activ ation step at 95 • C for 5 min, follo w ed b y denaturation at 95 • C for 40 s, primer hybridization at 58 • C, r espectiv el y, at 55 • C for fungi for 40 s, and elongation at 72 • C for 1 min, repeated in 30 cycles. PCR amplification was carried out in technical triplicates, and triplicates were pooled prior to sequencing. Pooled PCR products were sent to the Functional Genomics Center Zurich (FGCZ, Zurich, Switzerland) for indexing PCR. Indexed PCR products were purified, quantified, and pooled in equimolar r atios. Pr esequencing on the Illumina MiniSeq platform (Illumina, San Diego, CA, USA) was performed to inform library repooling in order to achieve optimal read count distributions across all samples. Final sequencing was executed on the Illumina MiSeq platform (Illumina, San Diego, CA, USA) using the v3 chemistry for PE300 reads.

Bioinformatics
Sequencing data were processed using a customized bioinformatics pipeline lar gel y based on VSEARCH v2.21.1 (Rognes et al. 2016 ). Primers used in PCR were trimmed with CUTADAPT v3.4 (Martin 2011 ), allowing for 1 mismatch. Bowtie2 v2.4.5 (Langmead and Salzberg 2012 ) was used to filter for PhiX contamination by aligning the r eads a gainst the PhiX genome (accession NC_001422.1). T he fastq_mer gepairs function in VSEARCH was used to merge trimmed paired-end reads, and the fastq_filter function was applied for quality filtering with a maximum expected error of 1 (Edgar and Fl yvbjer g 2015 ). Sequences were dereplicated using the derep_fulllength function in VSEARCH and delineated into amplicon sequence variants (ASVs) applying the UNOISE algorithm (Edgar 2016a ) of VSEARCH with an alpha of 2 and a minsize of 8. The UCHIME2 algorithm (Edgar 2016c ) implemented as the uchime3_denovo function in VSEARCH was used to identify and r emov e potentiall y c himeric ASV sequences.

Sta tistical anal yses
All statistical analyses were computed in R Version 4.1.3 (R Core Team 2022 ) using RStudio Version 2022.02.1 + 461 (RStudio Team 2022 ). For all tests, a P-value < .05 was considered significant unless mentioned otherwise. To test the effect of the treatment and the sampling time point (season) on the investigated soil parameters (GWC, pH, C:N ratio, C org , TN, NH 4 + , NO 3 − ), plant parameters (height, diameter, needle litter, and A net ) and estimated copies of the 16S rRNA gene and the 18S rRNA gene, the data were fitted to linear mixed effect models . Here , the lme function of the pac ka ge nlme v.3.1-157 (Pinheiro et al. 2022 ) was applied using the r estricted maxim um lik elihood method "REML " (Me yer 1989 ). Wher eas, irrigation tr eatment (contr ol, intermediate, se v er e, n = 6) and sampling time point (Winter -20, Spring-20, Summer -20, A utumn-20, W inter-21) were used as factor variables with inter action. The gr eenhouse (gr eenhouses 1 and 2) and the pot number (pots 1-18) were assumed to be a nested random effect. When visual inspection of diagnostic residual plots revealed that data deviated from the assumption of homoscedasticity or normality, the data were transformed using log (for TN), log 10 (for GWC, NO 3 − , 16S gene copies), or sqr t (C:N ratio, NH 4 + ). The data wer e bac k-tr ansformed for further tests using the ref_grid function of the pac ka ge emmeans v.1.7.3 (Lenth et al. 2022 ). To test the effect of the treatment and the sampling time point (season), pairwise comparisons were estimated by using marginal means with the emmeans function of the same pac ka ge while adjusting for multiple comparisons with the Sidak method. To identify and display significant differences between groups, the function cld of the pac ka ge multcomp v.1.4-18 (Hothorn et al. 2022 ) was a pplied. Sequencing depth was investigated using barplots and r ar efaction curves with the rarecurve function in the vegan pac ka ge v.2.6-2 (Oksanen et al. 2022 ) ( Supplementary Figure 2). To account for differences in sequencing depth, changes in α-diversity (observ ed ric hness, Pielou's e v enness, and Shannon div ersity) and βdiv ersity (Br ay-Curtis dissimilarity) wer e calculated fr om 100-fold iter ativ el y subsampled and squar e-r oot tr ansformed ASV count tables (Martiny et al. 2017, Hemk eme yer et al. 2019. Here, the functions rrarefy , specnumber , diversity , and vegdist in vegan were applied. The effect of treatment and sampling time point (season) on αand β-diversity were assessed using univariate or multivariate perm utational anal ysis of v ariance PERMANOVA (Anderson 2001 ) and PERMDISP (Anderson 2006 ) with 999 permutations, as implemented in the adonis2 and betadisper functions of vegan . Pairwise comparisons between factor le v els wer e performed using the pairwise.perm.manova function from the package RVAideMemoire v.0.9-81-2 (Hervé 2022 ). Differences in β-diversity were assessed b y unconstrained or dination using principal coor dinate analysis (PCO) (Go w er 2015 ) with the cmdscale function. Constrained ordination was performed using canonical analysis of principal coordinates (CAP) (Anderson and Willis 2003 ) implemented as the CAPdiscrim function of the B iodiversityR pac ka ge v.2.14-1 (Kindt 2022 ), with 999 permutations, setting the factors treatment and sampling time point (season) as constr aining factors. Her e, the CAP reclassification success rate provides a quantitative estimation of the degree of discrimination between treatment groups. The effects of measured soil physiochemical properties and plant par ameters on micr obial comm unities wer e obtained using the PERMANOVA test. Additionally, factors labeled as significant in the PERMANOVA test were further used as a constraining factor in building a parsimonious model executing the function ordistep in vegan , and the significant factors w ere display ed b y distancebased r edundancy anal ysis (db-RDA) (Legendr e and Andersson 1999 ) using the dbrda function in vegan . The response of individual taxonomic groups from phylum to genus level to w ar d w ater limitation was assessed using univariate PERMANOVA based on Euclidean distances via the adonis2 function with 9999 permutations on a ggr egated data at eac h taxonomic le v el, i.e. summing up the read counts of ASVs assigned to the same taxonomic group (Supplementary data). To adjust for multiple testing, q-values (Storey and Tibshirani 2003 ) were calculated using the qvalue function of the R pac ka ge qv alue v.2.24.0 (Stor ey et al. 2022 ) and q-v alues < .05 considered significant, and q-values < .1 as marginally significant. The data were z-transformed to visualize changes in relative abundance under each treatment, and the relative change in abundance compared to the control treatment was calculated.
A correlation-based indicator species analysis (De Cáceres and Legendr e 2009 , De Cácer es et al. 2010 ) was performed on significant ASVs from the PERMANOVA analysis to determine the asso-ciation strength of each ASV with a sampling time point (season) and treatment or a combination, therefore . T he function multipatt of the pac ka ge indicspecies v.1.7.12 (De Cáceres et al. 2022 ) with a max.order of 10 was used. P-value adjustments for multiple comparisons were performed using the false discovery rate correction according to Storey ( 2002 ), and associations were considered significant at q-value < .05. Bipartite association netw orks w er e cr eated using the softwar e Cytosca pe v.3.9.1 (Shannon et al. 2003 ). The bipartite networks were generated following Hartmann et al. ( 2015 ) by using the sampling time point (season) and treatment combination as source nodes and the ASVs as target nodes, with edges corresponding to positive associations of ASVs with sampling time point (season) and treatment combinations as obtained from the indicator species analysis Networks split by phyla were constructed using the Allegro Fruchtermann-Reingold algorithm (Fruchterman and Reingold 1991 ) with edges weighted according to the association strength.
For inferences about the potential lifestyle of the taxa, literatur e r esearc h was completed in conjunction with Fa pr otax v1.2.4 (Louca et al. 2016 ) and FUNGuild v.1.0 (Nguyen et al. 2016 ) for prokaryotes and fungi, respectively.

GWC in soils
During the first four months of the experiment, the GWC of the soils was k e pt at ca. 30% in all mesocosms . T her efor e, the GWC of the soils remained similar during the Winter-20 season (Fig. 1 ). Following the start of the three irrigation treatments (control, intermediate water limitation, and se v er e water limitation), the GWC of the soils consistently differed among the treatments throughout all samplings ( P < .0001, Table 1 ). Throughout the sampling time points (season), the soils of the control mesocosms had gravimetric water contents ranging between 17% and 36%, wher eas, the v ariability within the tr eatment gr oup was highest in Summer-20, ranging between 10% and 28%. Soils of the mesocosm treated with intermediate and severe water limitation had gravimetric water contents ranging from 10% to 25% and 4% to 9%, r espectiv el y (Fig. 1 ). The GWC of the mesocosms exposed to the thr ee tr eatments significantl y v aried acr oss sampling time points (season, P < .0001, Table 1 ). The highest water content values were measured in Spring-20 for all treatments, and the lo w est values in Summer-20 (Fig. 1 ).

Scots pine saplings response to treatments
The Scots pine saplings grew mainly from Spring-20 to Autumn-20 (subsequently referred to as growing season). The main increase in height and diameter occurred in the summer season, during which the saplings grew on average by 6.2-8.4 cm in height (Supplementary Table 2) and by 0.2-1.4 mm in diameter (Supplementary Table 2). Growth of the saplings, as measured by height and diameter, was influenced by the sampling time point (season), but we did not find consistent differences in growth among treatments (Table 1 ). Ho w ever, saplings under the control treatment sho w ed a lar ger r adial gr owth (Supplementary Table 2), and r adial gr o wth w as significantly influenced b y the interaction between treatment and sampling time point (season) ( P = .014).
As it is typical for e v er gr een pine tr ees, the amount of needle fall was highest in Autumn-20 (Supplementary Table 2). The photosynthetic assimilation was significantly affected by water limitation treatments ( P < .0001) and sampling time points (season, P < .0001). Across all sampling time points (seasons), A net was Effects of treatment (T) ( n = 3), season (S) ( n = 5), and their interaction ( n = 15) on soil physicochemical parameters, abundance of taxonomic markers (16S rRNA gene, 18S rRNA gene), and tr ee gr owth wer e tested with linear mixed effect models (lme) and displayed with the F-ratio (F) and le v el of significance (P). Significant r esults ar e indicated with bold numbers. GWC = gr avimetric water content; TN = total nitr ogen concentr ation; NH 4 + = ammonium concentr ation; NO 3 − = nitr ate concentration; C org = organic carbon concentration; and Photo. A net = photosynthetic assimilation.  Table 2). In the control mesocosms, the greatest photosynthetic assimilation was measured in Summer-20, whereas the other mesocosms sho w ed the largest values in Spring-20 (Supplementary Table 2).

Soil pH, carbon, and nitrogen concentr a tions
Soil pH was influenced by water limitation ( P = 0.0020) and sampling time point (season, P < .0001) ( Table 1 ). The uppermost pH was measured in Winter-20, with values around 7.1 ( Supplementary Figure 1A). During the growing season of the Scots pine sa plings, soil pH decr eased for all irrigation treatments to 6.8-6.9 ( Supplementary Figure 1A). After the start of the irrigation treatments, soil pH remained more basic in the control mesocosms compared to the soils treated with intermediate and se v er e water limitation ( Supplementary Figure 1A).
Soil C org and total nitr ogen concentr ations (TN) differ ed significantly among sampling time points (season) ( Table 1 ). C org tended to decr ease ov er the experimental period, except in the soils of the control mesocosms ( Supplementary Figure 2E).
TN was similar in Winter-20 and Spring-20 but declined throughout the following sampling time points (seasons) ( Supplementary Figure 1C). TN was lower in soils treated with waterlimiting conditions compared to the control; ho w ever , this difference was statistically not significant (Table 1 ).
Soil C:N ratios increased during the experiment ( P < .05) (Supplementary Figure 1D). At the beginning of the irrigation treatments soil C:N ratio ranged between 16 and 20. In the following, the soil C:N ratio increased, ranging between 17 and 24 in Summer-20 and 17-29 in Autumn-20 and Winter-21.
Ammonium (NH 4 + ) and nitrate (NO 3 − ) concentrations decreased during the growing season of the saplings, and for all treatments, the lo w est values w er e observ ed in Summer-20 (Supplementary Figure 1F and G).

Abundance of taxonomic groups
Water deficit did not affect prokaryotic and fungal abundance (Table 1 , Supplementary Figure 3A, B). Ho w ever, the sampling time point influenced the abundance of fungal ( P = .0007 , Table 1 ) and prokaryotic ( P = .0221, Table 1 ) gene copies . T his w as driven b y the significantl y lar ger estimated copy numbers detected in Summer-20 (Supplementary Figure 3B).
Water limitation had little effect on α-diversity (examined by observ ed ric hness, Pielou's e v enness, and Shannon div ersity), but α-diversity was strongly influenced by the sampling time point (season) ( Table 2 , Supplementary Figur e 3). Pr okaryotic e v enness was higher in the control treatment as compared to the water limiting treatments (Table 2 ). Prokaryotic α-diversity increased during Spring-20 and Summer-20 and subsequently declined during Autumn-20 and Winter-21 across all irrigation treatments . T he lar gest v ariability was observ ed in summer ( Supplementary Figure 3C-E).
The observed fungal richness decreased over the course of the experiment and differed between treatments already at the beginning (Supplementary Figure 3, Table 2 ). From Summer-20 onw ar d, observ ed ric hness r emained lar gest for the se v er e water deficit treatment ( Supplementary Figure 3F-H). A similar response to the treatments was observed for Pielou's evenness and Shannon diversity ( Supplementary Figure 3F-H). Ho w ever, statistical tests indicated that the sampling time point (season) only significantly affected Pielou's e v enness and Shannon diversity (Table 2 ).
Water limitation altered prokaryotic ( P < .001, Table 2 ) and fungal ( P = .002 , Table 2 ) β-diversity, explaining 3% and 4% of the variance, r espectiv el y (Table 2 ). In comparison, effects of the sampling time point (season) on β-diversity were stronger, explaining 13%-17% of the v ariance, r espectiv el y (Table 2 ). For prokaryotes, the treatment effects depended on the sampling time point (season, P = .003), wher eas, ther e was no tr eatment-by-time inter action for fungi.
T he C AP-r eclassification success r ates indicated that each combination of sampling time point (season) and tr eatment gr oup harbored distinct prokaryotic microbial communities (Fig. 2 A), wher eas r eclassification success incr eased to w ar d later sampling points, indicating that treatment groups become statisticall y mor e distinct. Ho w e v er, lo w er r eclassification r ates for Table 2. Effects of irrigation treatment and sampling time point (season) on soil microbial α-diversity and β-diversity.

Observ ed ric hness (S) Pielou's evenness (J) Shannon di v ersity (H) Bray-Curtis dissimilarity
Prokaryotes Effects of treatment ( n = 3), season ( n = 5) and their interaction ( n = 15) on prokaryotic and fungal α-diversity and β-diversity assessed by univariate ( α-diversity) and m ultiv ariate ( β-div ersity) perm utational anal ysis of v ariance (PERMANOVA). Values indicate the F-r atio (F M ), the le v el of significance (P), and the explained v ariance (R 2 ). Significant heter ogeneities of v ariance assessed by perm utational anal ysis of univ ariate dispersion (PERMDISP) ar e displayed with the F-r atio (F D ) and the le v el of significance (P). Significant r esults ar e indicated with bold numbers. fungal comm unities wer e attributed to smaller differ ences in the composition of these communities (Fig. 2 B).
The composition of prokaryotic communities was mainly shaped by GWC, pH, TN, C org , and soil temper atur e as obtained by statistical tests (Table 3 , 4 , Supplementary Figure 4C). The same parameters influenced fungal communities but in a differ ent ma gnitude (Table 3 , 4 , Supplementary Figur e 4D). Ov erall, soil temper atur e -whic h was similar to the gr eenhouse air temper atur e-was the main driver among the measured properties during Summer-20, Autumn-20, and Winter-21 (Supplementary Figure 4C and D).

Taxa sensitive to water deficit
After correction for multiple testing, 749 out of 19 840 (3.8%) prokary otic ASVs w ere observed to be significantly (PERMANOVA; q < .05) influenced b y w ater deficit. Out of these, 90 ASVs could be assigned at the genus le v el. 257 ASVs with a strong (F-ratio ≥ 10) and significant ( q < .05) change in relative abundance were selected for the indicator species analyses and subsequent construction of the bipartite association network. Since only 13 out of 2686 fungal ASVs r esponded significantl y to water deficit ( q < .05), a less strict threshold of q < .1 was a pplied, r esulting in 130 fungal ASVs (4.8% of total ASVs) that were used for indicator species analysis and bipartite network construction, with 21 ASVs assigned to genus le v el. Significant c hanges in r elativ e abundances were also statistically evaluated at all taxonomic le v els by a ggr egating the data based on summarized ASV counts from genus to phylum le v el.
A decr ease in r elativ e abundance under intermediate and sev er e water deficit compared to the control was observed for most of the r esponsiv e ( q < .05) prokaryotic phyla (Fig. 3 A). Treatmentsensitive ASVs were broadly distributed across the taxonomic hier arc hy, r e v ealing substantial r esponse heter ogeneity within individual phyla.
While , the o v er all abundance of the phylum Proteobacteria declined under water deficit, this was heterogenous at the genus le v el. The gener a Roseococcus , Caulobacter , Hyphomicrobium , Pedomi-crobium , Rhodobacter, Coxiella , and Massilia decreased in r elativ e abundance under water deficit (Fig. 3 B). On the contrary, gener a suc h as Erythrobacter , Novosphingobium , Sphingomonas , Arenimonas , and Lysobacter increased under water deficit (Fig. 3 B). The phyla Acidobacteriota and Methylomirabilota increased in relativ e abundance compar ed to the contr ol, but onl y under intermediate and not under se v er e water deficit (Fig. 3 A).
Onl y the r elativ e abundance of Actinobacteriota and the archaeal phylum Thermoplasmatota significantly ( q < .05) increased under se v er e water deficit compar ed to the contr ol tr eatment (Fig. 3 A), with a strong enrichment of the Actinobacteriota also clearly visible at the ASV level (Fig. 2 C). Salient examples of bacterial genera belonging to the Actinobacteriota with ASVs increasing under se v er e water deficit included Blasterococcus , Nakamurella , Nocardioides , Actinophytocola , Patulibacter, and Solirubrobacter (Fig. 3 B). No assignment at the genus le v el could be obtained for the archaeal phylum Thermoplasmatota.
Mucoromycota was the only fungal phylum that changed due to the irrigation treatment and declined in relative abundance under se v er e and intermediate water deficits compar ed to contr ol conditions (Fig. 3 A). Examples of fungal genera belonging to the Mucoromycota phylum with ASVs decreasing under intermediate and se v er e water deficit included Absidia and Umbelopsis (Fig. 3 B).
Fungal phyla that sho w ed a stable abundance under water limitation included Basidiomycota and Mortierellomycota (Fig. 2 D), although these phyla sho w ed different responses to water limitation at the genus le v el (Fig. 3 B). Salient examples of fungal genera belonging to these phyla included Amphinema , Cystobasidium , Par atritir achium , Udeniozyma, and Mortierella for the Mortierellomycota (Fig. 3 B). An ov er all tr end to w ar d an incr eased r elativ e abundance of ASVs under intermediate and se v er e water deficit was observed for the fungal phylum Ascomycota (Fig. 2 D). Since the phylum Ascomycota was dominant in all samples, this increase w as, ho w e v er, contr asting at the genus le v el, e.g. with the gener a Niesslia , Trichophaea , and Helicodendron increasing and Curvularia , Pulvinula , Staphylotrichum , Tuber , and Stachybotrys decreasing under water deficit (Fig. 3 B). The CAP reclassification success rates providing a quantitative estimation of the degree of discrimination between the groups are provided next to each treatment and time point (season). The CAP equivalent to Pillai's trace test (with P -value in brackets) indicating the overall effect size is provided at the top of the plot. (C and D) Prokaryotic and fungal bipartite association networks showing the ASV distribution (circles = prokaryotes, triangles = arc haea, squar es = fungi) across the different sampling time points (hexagon; seasons, WT = Winter, SP = Spring, SM = Summer, and AT = Autumn) and treatments. Node sizes are scaled by read counts (square root) and color-coded by phylum-level assignment. Edges correspond to significant associations between ASVs and samples based on indicator species analysis . T he edge-weighted (weighted by ASV association strength) "Allegro F ruchterman-Reingold" algorithm w as applied to the netw ork, whic h clusters samples with higher connectivity ( = similar comm unity structur e).

Effect of soil water content on soil microbial communities
One year of experimental soil water limitation did not alter the total abundance of prokaryotes and fungi (Supplementary Figure  3A and B). Ne v ertheless, water limitation sha ped the composition of the soil microbiome of the Scots pine mesocosms (Fig. 2 , 3 ), confirming that water deficit promotes distinct microbial communities . T his finding corr obor ates r esults fr om long-term field observations indicating that changes in forest irrigation patterns significantl y alter ed micr obial comm unities due to differ ent abilities to cope with water scarcity and modified substrate availabilities (Hartmann et al. 2017 ).  We observ ed negativ e r esponses to water limitation for most of the r esponsiv e pr okaryotic phyla but onl y for one fungal phylum. The decrease in relative abundance of sensitive phyla was mainl y gr adual and mor e substantial under se v er e water limitation (Fig. 3 A). While the fungal α-div ersity r emained higher under water limitation, the opposite pattern was observed for prokaryotic α-diversity (Supplementary Figure 3). Fungal communities a ppear ed to hav e a gr eater toler ance to w ar d w ater limitation than pr okaryotic comm unities (Figs. 2 B, 2 D, 3 A, 3 B). This observation aligns with a study by de Vries et al. ( 2018 ), who observed a greater resistance of fungal networks compared to bacterial networks under dr ought. Contr ary, other studies found a gr eater toler ance of bacterial communities under simulated global change (Martiny et al. 2017 ) or minor effects of rainfall exclusion on community compositions (Ren et al. 2018 ). Mor eov er, Sayer et al. ( 2017 ) reported shifts in soil fungal and bacterial community structure in response to long-term drought and a substantial loss of fungal taxa. Since the results presented in this study have been generated over one year, more significant changes in fungal community composition might only become evident at longer time scales. Ho w e v er, some first changes in the composition of fungal communities could already be detected (Fig. 3 A and B ).
The greater tolerance of fungal communities to ongoing water limitation (Figs. 2 B, 2 D, 3 A, 3 B) is likely explained by the ability of fungi to create large hyphal networks, which allows them to enter water-filled small soil pores and, thereby, maintain nutrient and water uptake over long distances (Allen 2007 , Joergensen andWichern 2008 ). Uncertainty remains whether the tolerance to water deficit indicates an adaptation of soil-inhabiting fungi or rather the presence of abundant dormant propagules, which can withstand adverse environmental conditions and regain metabolic activity once favorable conditions return (Lennon and Jones 2011, Barnard et al. 2013, Meisner et al. 2018 ). The soil used in the study was collected from a xeric forest and stored for some months before the start of the experiment. Hence, the observ ed toler ance of fungal comm unities might be r elated to the presence of abundant dormant fungal taxa, which can withstand unfa vorable en vironmental conditions and regain metabolic activity upon r e wetting (Lennon and Jones 2011, Barnard et al. 2013, Meisner et al. 2018 ).

Effect of seasonality on soil microbial communities
The seasonal sampling time points pr ofoundl y gov erned the variability in the composition of both fungal and prokaryotic communities (13%, r espectiv el y, 17%, Table 2 ). Seasonality is well known to affect the activity and composition of microbial communities thr ough alter ations in soil moistur e and temper atur es (Moor e-Kucer a and Dic k 2008 , Cruz-Martínez et al. 2009, Kuffner et al. 2012, Vo řišk ov a et al. 2014, Žif čák ová et al. 2017. Soil temperature is essential to microbial growth, metabolism, and physiology (Schimel et al. 2007, Madigan et al. 2018. Ho w e v er, high soil temper atur es also cr eate unfavor able conditions setting inhibitory boundaries for microbial growth (Sheik et al. 2011 ). In our experiment, although soil water content was car efull y controlled with volumetric water sensors in each mesocosm, some confounding effects of seasonal changes in soil moisture could not be avoided (Fig. 1 ). Higher soil e v a potr anspir ation r ates primaril y r educed soil water contents during summer due to temper atur e incr eases (Fig. 1 ). The concurr ent incr ement in soil temper atur es and lo w er soil moistur e le v els during the growing season likely stimulated the proliferation of communities tolerant to higher temperatures and dehydration and might have The observed effect of seasonality on soil microbial communities was also likel y r elated to the periodic differences in plant growth and litter production (Myers et al. 2001, Wardle et al. 2004, Högber g 2006. Tr ees typicall y pr ovide r eadil y av ailable C substrates during the growing season for microbes stimulating their activity and growth (Hobbie 2006 ). These seasonal differences in C and nutrient availabilities could thereby explain the strong influence of the seasonal time point on the microbial communities. Although water limitation affected the photosynthetic C assimilation of the saplings in our study, abov egr ound gr owth and needle litter production remained sustained during the growing season (Supplementary Table 2). This pattern likely led to the peak in pr okaryotic α-div ersity during the growing period of the Scots pine sa plings (Supplementary Figur e 3C, D, and E) and might hav e favored the higher abundance of microbes in summer as estimated by measuring 16S and 18S rRNA gene copy numbers (Supplementary Figure 3A and B). Moreover, the strong influence of the sampling time point could be related to successional changes in tree de v elopment after the establishment of the mesocosms.

Desicca tion-toler ant microbial taxa
Soil microbial communities were dominated by phyla commonl y observ ed in soil, including Proteobacteria, Actinobacteriota, and Acidobacteriota, as well as Basidiom ycota and Ascom ycota (Fig. 2 C, D ). The fungal phyla Ascomycota and Basidiomycota a ppear ed to toler ate water limitation, r esulting in a dominance of these phyla in all our samples (Fig. 2 D). This finding corr obor ates the results published by Hartmann et al. ( 2017 ), who observed a stable abundance of Ascomycota and an increase of Basidiomycota under long-term dry conditions in a Scots pine forest.
The bacterial phylum Actinobacteriota significantly increased in r elativ e abundance with increasing soil water limitation ( Fig. 3 A). Observations by Hartmann et al. ( 2017 ), coincide with our results as the abundance of Actinobacteriota increased in water-limited Scots pine forest soils. In our experiment, the abundance of aerobic genera commonly found in sandy and r oc ky soils (França et al. 2016, Sghaier et al. 2016, like Blasterococcus , Nakamurella , Actinoph ytocola , and Nocardioides , incr eased with incr easing water deficit (Fig. 3 B). Furthermore, the genera Patulibacter and Solirubrobacter, whic h ar e thermophilic and often associated with arid soils (Acosta-Martínez et al. 2014, Albuquerque et al. 2014, Bastida et al. 2017, Tóth et al. 2017, also increased under water deficit (Fig. 3 B). As a r esult, the abov e-mentioned taxa might be tolerant to water deficits and could serve as indicators of soils with low soil water content.
The archaeal phylum Thermoplasmatota increased in relative abundance under se v er e water deficit (Fig. 3 A). Thermoplasmatota seem tolerant to w ar d envir onmental str ess because they ar e found in hot and acidic environments (Tripathi et al. 2015 ). Thus, the increase of Thermoplasmatota in the mesocosms affected by se v er e water deficit potentially contributed to prokaryotic comm unities, whic h ar e mor e toler ant to w ar d envir onmental str ess.
Contrary to Hartmann et al. ( 2017 ), who observed an increase of Acidobacteriota under dry conditions, in our study, Acidobacteriota were not affected by intermediate water limitation (Fig. 3 A). This differing result might be explained by the fact that the se v erity of water limitation was more profound in our study than in the field site studied by Hartmann et al. ( 2017 ). Howe v er, the observed sustained abundance of Acidobacteriota under intermediate water-limiting conditions suggests a toler ance thr eshold of Acidobacteriota to w ar d moder atel y arid conditions. Acidobacteriota are ubiquitous, desiccation tolerant, and adapted to nutrientlimited environments (Kielak et al. 2016a ), as they are commonly found to be abundant in environments characterized by low resource av ailability (Fier er et al. 2007, Ko y ama et al. 2014. This adaptability could explain the observed resistance to w ar d intermediate water-limiting conditions . Moreo ver, it has been observed that Acidobacteriota produce extracellular polysaccharides that can create moist micro-niches (Kielak et al. 2016a ), which might benefit other bacteria during episodes of dry conditions. Barnard et al. ( 2013 ), also found a r elativ e incr ease of Actinobacteriota and a decrease of Acidobacteriota with summer drydown and concluded that these contrasting drought-associated changes in abundance might reflect different desiccation-related bacterial life strategies . T his reasoning could also serve as an explanation for the pattern observed in our study.

Changes in microbial communities related to altered soil properties
We observed a decrease in soil pH under water limitation (Supplementary Figure 1A), which could be explained by an accumulation of cations due to reduced leaching and plant uptake. A significant relationship between soil pH and community structure of both prokaryotic and fungal communities was observed in our mesocosms (Tables 3 , 4 ). Ho w e v er, the r elationship with pH w as lo w er for fungal as compared to prokaryotic communities, as also observed in previous studies (Rousk et al. 2010 , Siles andMargesin 2016 ). In our study, fungi wer e mor e sensitiv e to v ariations in other soil physioc hemical pr operties as compar ed to v ariations in soil acidity (Table 3 ).
Changes in soil pH can dir ectl y affect soil microbes by altering their competitive fitness or impairing the growth of individual taxa that cannot survive when soil pH falls outside a certain range (Lauber et al. 2009, Madigan et al. 2018. Ther efor e, the decr easing pH observed in our study might have imposed considerable stress on microbes that was only tolerated by certain taxa. For example, we found an increase of Actinobacteriota associated with a decrease in soil pH. Ho w e v er, the str ong influence of soil pH on soil micr oor ganisms can also be explained through indirect effects (Lammel et al. 2018 ). For example, soil pH shapes several soil characteristics important for micr obes, suc h as e.g. solubility of ions and nutrients and salinity , Brady and Weil 2013, Zhalnina et al. 2015, Madigan et al. 2018, Solly et al. 2020. In turn, these indirect effects of soil pH may also drive the observed c hanges in comm unity composition and thus could explain the strong influence of soil pH on microbial communities in our study.
C org also significantly influenced the composition of microbial communities (Table 4 , Supplementary Figure 4C, 4D). Organic C is known to be the primary energy substrate for microbes under oxic conditions; thereby, a reduction of C org can rapidly alter micr obial comm unities to w ar d oligotr ophic micr obial gr oups, whic h can survive under limited C resource availability (Aldén et al. 2001, Drenovsky et al. 2004, Soong et al. 2020. Ther efor e, in our study, C org potentially contributed to a proliferation of competitive taxa able to survive under a low availability of organic C and specialized in the degradation of more recalcitrant C compounds. During the experiment, we observed an increase in the C:N ratio related to a decline in TN (Supplementary Figure 1C, D). Increases in soil C:N ratios are often used as a proxy indicating an accumulation of complex organic compounds -such as ligninwhic h favors micr obes with the ability to decompose r ecalcitr ant structures (Baldrian et al. 2012, Van der Wal et al. 2013, Lindahl and Tunlid 2015, Žif čáková et al. 2017 ). We detected a lo w er abundance of Acidobacteriota in samples with a greater C:N ratio, which could be related to their adaptation to oligotrophic conditions (Naether et al. 2012 ). Hartmann et al. ( 2017 ) hypothesized that taxa, which can metabolize more recalcitrant compounds have an advantage when the input of easily degradable plant resources is limited. In our experiment, complex compounds likely accumulated in the mesocosms under the reduced and severe water limitation treatments , fa voring the abundance of saprotrophic taxa. With higher C:N ratios, Basidiom ycota, Ascom ycota, and Actinobacteriota increased in abundance (Fig. 3 ). Man y gener a of the Actinobacteriota are saprophytic and were shown to be highly resistant to desiccation and C starvation (Ventura et al. 2007, Rosenberg et al. 2014, Mohammadipanah and Wink 2016. Also, the dominant fungal phylum Basidiomycota comprises -next to other functional groups as symbionts -many saprotrophic genera (Tedersoo et al. 2010 ), which degrade complex compounds, including cellulose and lignin (Baldrian 2008 ). For instance, the sa pr otr ophic genus Par atritir achium (Tedersoo et al. 2014 ), significantl y incr eased in abundance with water limitation (Fig. 3 B). Man y gener a belonging to the highly abundant phylum Ascomycota are further described as sa pr otr ophs (Tedersoo et al. 2010 , Vorisk ov a andBaldrian 2013 ). For example, we found a significant increase of the saprotrophic genera Niesslia (Tedersoo et al. 2014 ) and Helicodendron (Tedersoo et al. 2014 ) (Fig. 3 B) . Next to being symbionts , most of the Ascomycota and Basidiomycota possess the ca pability to degr ade cellulose (De Boer et al. 2005 ). Consequentl y, some sa pr ophytic populations might suffer fr om r esource competition during decomposition (Sollins et al. 2007 ).

Rela tionship betw een microbial communities and Scots pine saplings
Mor e extr eme episodes of dr ought, whic h lead to soil moistur e le vels below the wilting point, are a major predisposing risk for tree mortality e v ents (McDo w ell et al. 2022 ). To ensure that the Scots pines remained vital, the saplings received a minimum of water in our experiment. The photosynthetic assimilation and abovegr ound gr owth of the Scots pine sa plings wer e alter ed by se v er e water limitation. Ne v ertheless, we observ ed no dir ect effect of the studied abov egr ound plant par ameters on micr obial comm unities . T his study did not focus on belowground plant traits because we could not destructiv el y sample the mesocosms but collected differ ent soil cor es fr om the same soil system while the experiment was still ongoing. Ne v ertheless, our findings r e v eal significant changes in the abundance of symbiotic taxa associated with c hanging physicoc hemical soil properties under drier conditions with potential consequences for de v eloping young trees.
Our results indicate that the abundance of the proteobacterial famil y Rhizobiales significantl y declined under intermediate and se v er e water deficits compared to the control (Fig. 3 B). Rhizobiales are well-studied associates of plants, and they commonly exert beneficial functions for their hosts , e .g. through N-fixation (Delmotte et al. 2009 ). T hus , their decline does corr obor ate our hypothesis of a decrease of symbionts under water stress . Furthermore , Pr oteobacteria wer e shown to be sensitive to short-term drought scenarios (Bouskill et al. 2013, Chodak et al. 2015, supporting our finding. Mor eov er, our observ ation is in line with the r esults fr om a forest field site where Hartmann et al. ( 2017 ) found an increase of the phylum Proteobacteria in plots under irrigation, presumably linked to their copiotrophic lifestyle, which is found in environments with rich plant-C inputs (Fierer et al. 2007 ). This observation could also explain the better plant performance (growth and photosynthesis) under the control treatment in which the relationship with symbionts might have sustained the saplings. Furthermore, we found a decrease of diazotrophic bacteria such as the endophytic genus Paenibacillus (Fig. 3 B) , which is known to impr ov e the nutrient status of plants through phosphate solubilization and N-fixation under dry and nutrient-poor conditions (Bal et al. 2012, Puri et al. 2018, additionally supporting our expectation of a decline of symbionts . T he observ ed decr ease of N-fixing bacteria ma y ha ve consequences for plant nitr ogen av ailability and might partially explain the reduced soil ammonium concentrations under drier conditions. Also, the fungal phyla Basidiomycota and Ascomycota encompass many symbiotic taxa (Tedersoo et al. 2010 , Voriskova andBaldrian 2013 ), of which we observed a decrease and thus supporting our h ypothesis. F or example , the famil y Tuber aceae (Ascomycota), known as truffles, forms symbiotic associations with plants (Tr a ppe et al. 2009 ), and was found to decr ease in mesocosm with water limitation . Further salient examples of putative EcM declining in abundance under water deficit included Amphinema (Basiodmycota ) and Pulvinula (Ascomycota) (Fig. 3 B) (Rinaldi et al. 2008, Tedersoo et al. 2010 . In our study, the phylum Mucoromycota was the only fungal phylum significantly affected by water limitation (Fig. 3 A). As opportunists, Mucorom ycota thri ve on readily available C-sources suc h as monosacc harides fr om r oot exudations (Dix and Webster 1995 ). These C-sources might be limited under water deficit as fresh and easily degradable C inputs of host plants might be impaired due to reduced photosynthetic activity under low soil moisture . T his lack of easily degradable C inputs could explain the observed significant decrease of this phylum in our study and matches with an increase of this phylum in irrigated soils of a Scots pine forest observed by Hartmann et al. ( 2017 ). Moreover, it is proposed that some families of the fungal phylum Mucoromycota ar e involv ed in symbiosis with plants (Field et al. 2015 ). This finding suggests that Mucoromycota might form mutualistic associations, indicating a functional ov erla p with EcM fungi (van der Heijden et al. 2015 ), further supporting our hypothesis of reduced symbionts under water stress.
Ov er all, our findings indicate that soil water limitation in Scots pine forest soils will likely lead to a decrease in the relative abundance of N-fixing bacteria and symbiotic microbial taxa of trees, with cascading consequences on plant nutrition and forest health.

Concluding Remarks
Our study r e v ealed that under water limitation, soil micr obial communities in Scots pine mesocosms are rather shaped by alterations in soil properties such as pH and C:N ratio than by the direct influence of measured sapling growth parameters. Our findings further demonstrate a strong effect of seasonal changes in soil temper atur e and soil water content and a gr adual r esponse of most sensitive phyla to contrasting levels of water limitation. These results contribute to our understanding of how soil microbial communities adapt to different thresholds of water limitation and support the current view that soil prokaryotes are generally mor e sensitiv e to water limitation than soil fungi.
The pr esented r esults indicate that water limitation promotes the pr olifer ation of micr obial gr oups toler ant of envir onmental stress. Shifts in microbial community compositions induced by water limitation led to an accumulation of desiccation-tolerant taxa, potentially altering critical functions provided by the forest soil micr obiome, suc h as nutrient cycling. Our data further suggest a shift related to the potential lifestyle of microbes as we recognized a decrease of symbiotic and an increase of saprotr ophic taxa, potentiall y affecting water and nutrient av ailability for young tr ees. Ov er all, our findings contribute to a better understanding of how ecosystem functions mediated by soil microorganisms may be impaired in forests affected by drier conditions.

Ac kno wledgments
We thank Christian Hug and Marcus Schaub of the Swiss Federal Research Institute for Forest, Snow and Landscape (WSL) for their support during the soil collection and for growing the Scots Pine seedlings . Furthermore , we are grateful for the intellectual input fr om Iv ano Brunner and Arthur Gessler for the project. We also thank members of the ETH r esearc h station for plant sciences in Lindau for the use of their equipment and facilities. We are particularl y gr ateful to Matti Barthel, Rafaela Conz, Brigitta Herzog and Britta Jahn-Humphrey for their technical support in the greenhouse and laboratory. We further acknowledge the help of Manon Longepierre, Adrian Fuhrmann, and Luisa Minich with analyses in the laboratory. We also thank Maria Domenica Moccia at the Functional Genomics Center Zurich (FGCZ) for providing the sequencing service on the Illumina MiSeq platform.

Supplementary data
Supplementary data are available at FEMSEC online.
Conflicts of inter est. The authors declar e that the r esearc h was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Funding
This work was supported by the Swiss National Science Foundation SNSF [grant number PZ00P2_180030] granted to [EFS].

Da ta av ailability
T he ra w sequencing reads of this study ha ve been deposited in the European Nucleotide Archive (ENA) at EMBL-EBI under accession number PRJEB53192 ( https:// www.ebi.ac.uk/ ena/ browser/ view/PRJEB53192 ).