Long‐term soil warming alters fine root dynamics and morphology, and their ectomycorrhizal fungal community in a temperate forest soil

Climate warming is predicted to affect temperate forests severely, but the response of fine roots, key to plant nutrition, water uptake, soil carbon, and nutrient cycling is unclear. Understanding how fine roots will respond to increasing temperature is a prerequisite for predicting the functioning of forests in a warmer climate. We studied the response of fine roots and their ectomycorrhizal (EcM) fungal and root‐associated bacterial communities to soil warming by 4°C in a mixed spruce‐beech forest in the Austrian Limestone Alps after 8 and 14 years of soil warming, respectively. Fine root biomass (FRB) and fine root production were 17% and 128% higher in the warmed plots, respectively, after 14 years. The increase in FRB (13%) was not significant after 8 years of treatment, whereas specific root length, specific root area, and root tip density were significantly higher in warmed plots at both sampling occasions. Soil warming did not affect EcM exploration types and diversity, but changed their community composition, with an increase in the relative abundance of Cenoccocum at 0–10 cm soil depth, a drought‐stress‐tolerant genus, and an increase in short‐ and long‐distance exploration types like Sebacina and Boletus at 10–20 cm soil depth. Warming increased the root‐associated bacterial diversity but did not affect their community composition. Soil warming did not affect nutrient concentrations of fine roots, though we found indications of limited soil phosphorus (P) and potassium (K) availability. Our findings suggest that, in the studied ecosystem, global warming could persistently increase soil carbon inputs due to accelerated fine root growth and turnover, and could simultaneously alter fine root morphology and EcM fungal community composition toward improved nutrient foraging.


| INTRODUC TI ON
Tree fine roots (<2 mm in diameter) represent less than 2% of the total biomass in forests but contribute up to 33% to the global terrestrial net primary productivity (Jackson et al., 1997;McCormack et al., 2015). Fine roots are considered a major source of plant carbon (C) input into temperate forest soils, and their biomass growth, traits, turnover, as well as their exudation of labile C are important components in soil C cycling and storage (Keller et al., 2021;Rasse et al., 2005). In temperate and boreal forests, fine roots are typically colonized by EcM mycelium (Tedersoo et al., 2012). The symbiosis of roots and EcM fungi serves as an efficient nutrient foraging strategy for trees, especially in nutrient-poor forest soils (Lõhmus et al., 2006). The mycelium of EcM fungi greatly improves access to nutrient and water resources in soils that are hardly directly accessible for fine roots (Lindahl & Tunlid, 2015;McCormack & Iversen, 2019).
Climate warming is predicted to severely affect temperate forests during the 21st century (IPCC, 2021), and thus likely also the uptake and transport of water and nutrients by fine roots. Changes in fine root functions and the diversity and functional traits of EcM fungi may alter whole ecosystem C and nutrient cycles. Developing an efficient root system through change of biomass allocation to fine roots, changes in fine root morphology and root-associated microbial communities is required to maintain plant water and nutrient uptake under changing environmental conditions (Ostonen et al., 2011).
Trees can adapt their belowground surface area by modifying root biomass and fine root morphological traits to improve soil resource uptake and plant performance (Weemstra et al., 2020). FRB is driven by the balance between root growth and mortality and depends on environmental conditions, especially soil temperature and soil moisture (Salazar et al., 2020;Wang et al., 2021). Soil warming has been suggested to increase FRB due to decreased soil nitrogen (N) availability (Leppälammi-Kujansuu et al., 2013) but also to decrease FRB in temperate and boreal forests due to increased soil N availability (Dawes et al., 2015;Melillo et al., 2011;Wan et al., 2004;Zhou et al., 2011). The inconsistent findings across forest biomes illustrate the complexity of the interaction between increased soil temperature, N availability, and FRB (Wang et al., 2021). However, with ongoing high atmospheric N deposition in many forested regions, such as central Europe (Borken & Matzner, 2004;Talkner et al., 2019), where the current study took place, the availability of essential nutrients such as phosphorus (P) and potassium (K) may likely play a similar, or even a more important role, with regard to tree fine root responses to warming. Decreasing foliar P and K indicate poor availability of these nutrients in many European forests in recent decades (Jonard et al., 2015;Penuelas et al., 2020;Talkner et al., 2015), though the impact of soil warming in combination with low P and K availability on fine root systems is unknown.
Morphological traits of fine roots are indicators of trees' nutrient uptake efficiency and ecosystems' responses to changes in environmental conditions (Freschet et al., 2021;Ostonen et al., 1999).
Among those, traits such as specific root length (SRL, the length of a root per unit dry mass), root tissue density (RTD, the dry mass of root per unit volume of fresh root), specific root area (SRA, the ratio of fine root surface to fine root dry mass), root area index (RAI, the surface area of roots per soil surface area), mean root diameter (D, the average of all root diameter observations of a root diameter distribution), and root tip density (RTID, the number of root tips per soil volume) are used to describe the functional characteristics of fine roots (Freschet et al., 2021;McCormack & Iversen, 2019). For example, an increase in SRL indicates morphological adaptation toward thinner and longer fine roots, which are less resistant to stress and have a shorter lifetime but are more active metabolically (McCormack et al., 2015). This is seen as a strategy to increase nutrient acquisition from the soil at low biomass production (Weemstra et al., 2020). An increase in RAI and SRA also indicates adaptation toward increased ability to explore and take up soil resources. However, only few studies have yet considered the effect of soil warming on fine root morphology in forest ecosystems (Wang et al., 2021). In their meta-analysis, Wang et al. (2021) found that SRL and fine root diameter were irresponsive to experimental warming across a wide range of terrestrial ecosystems. Of the few studies conducted in forests, Parts et al. (2019) showed that trees increased their SRL and SRA with soil warming but decreased RTD. Björk et al. (2007) reported that warming increases SRL and SRA of roots <0.5 mm, while Leppälammi-Kujansuu et al. (2013) observed no effect on RTD and diameter of first-and second-order fine roots. Hence, similar to FRB, fine root morphology can respond differently to soil warming and associated changes in nutrient availability or moisture.
Due to the interaction between fine root traits and soil biota, understanding the response of the fine root systems to soil warming also requires studying root-microbial interactions (Bennett & Classen, 2020;Weemstra et al., 2017). EcM and root-associated bacteria are particularly important because they produce extracellular enzymes that release nutrients from soil organic matter in the vicinity of fine roots (Averill & Hawkes, 2016;Averill et al., 2014).
The diversity of EcM and root-associated bacteria might further increase the effectiveness of nutrient acquisition from different depths and locations in the soil (Leake, 2001). However, changes in root-associated microbial communities have received little attention in the context of soil warming. Because of the temperature dependence of soil enzymatic activities, soil warming directly influences nutrient cycling processes (Staddon et al., 2002). This may lead to shifts in EcM community composition (Solly et al., 2017;Treseder et al., 2016) and EcM diversity. Mycorrhizal associations greatly enhance the roots' surface area and, therefore, their water and nutrient uptake capacity (Smith & Read, 2008;Weemstra, 2017). EcM fungal traits like exploration types are good predictors of ecosystem processes because they integrate species functions in functionally redundant communities (Koide et al., 2014). Exploration type is a trait that connects the morphology and differentiation of ectomycorrhizal hyphae to differences in nutrient acquisition strategies (Defrenne et al., 2019). Agerer (2001) classified EcM into exploration types and assigned specific functions to those. For example, EcM with contact, short, and medium distance exploration types (low EcM biomass) might be preferred at sites with high N availability (Hobbie & Agerer, 2010), while long-distance exploration types (high EcM biomass) might be necessary in nutrient-poor sites (Tedersoo, 2017). Thus, EcM exploration types likely are going to be affected by warming if elevated temperatures alter the availability of soil nutrients or the nutrient requirement of trees.
The effect of soil warming on fine root dynamics and rootassociated bacterial-fungal communities in forests has largely remained unresolved, as most studies were short in terms of duration and differed in experimental approaches (Wang et al., 2021;Yuan et al., 2018). The long-term forest soil warming experiment at Achenkirch in the Austrian alps, where regional climate models predict an above-average increase in temperature (up to 3.3°C at the end of the 21st century) as compared to global average warming (Gobiet et al., 2014;Smiatek et al., 2009), offers an excellent opportunity to increase our understanding of how soil warming affects FRB, fine root production, fine root morphology, and rootassociated bacterial and fungal communities in temperate forests.
To achieve this, we studied fine roots in control and warmed (+4°C) plots in 2012 and 2019, 8 and 14 years after starting the experimental soil warming treatment, respectively. We hypothesized that (1) soil warming increases FRB after 8 or 14 years because of the globally positive response of FRB to warming (Wang et al., 2021) and the potential decline in soil P and K availability in warmed soil. We also expected (2) increases in SRL, SRA, and root tip density in warmed soil, which are linked to an increase in the absorptive capacity of fine roots. We did not expect significant changes in root-associated microbial community composition because previous results showed no soil warming effects on soil and root microbial community composition (Kuffner et al., 2012;Schindlbacher et al., 2011). However, because exploration types are directly linked to fine root morphology, we expected (3) a shift in the relative abundance of EcM exploration types toward long-distance exploration types after 14 years of soil warming.

| Study location and experimental design
This study was performed in the Achenkirch soil warming experiment located in a mountainous forest in the Austrian Alps (11°38′21″ E; 47°34′50″ N) at 910 m a.s.l. (Schindlbacher et al., 2007). The 140-year-old forest is composed mainly of Norway spruce (Picea abies L. H. Karst.) and few other less abundant trees species, including European beech (Fagus sylvatica L.) and silver fir (Albies alba Mill.).
The mean annual air temperature and mean annual precipitation from 1988 to 2017 were 7°C and 1493 mm, respectively. The soil type was classified as a shallow Rendzic Leptosol. It consists of thin organic L and F horizons, a mineral A-horizon with about 15-20 cm thickness, with a bulk density of ~0.5 g cm −3 , and an underlying C-horizon deriving from dolomite. The carbonic A-horizon had a pH of ~7, a C:N ratio between 15 and 18, and stored ~120 Mg ha −1 of organic C. The L/F-horizons stored about 10 Mg C ha −1 . Based on the "Austrian bioindicator grid" analyses (https://bfw.ac.at/rz/bfwcm s2.web?dok=3687), average element concentrations of live Norway spruce needles for the period 2000-2005 were 12.1 mg N g −1 , 0.94 mg P g −1 , and 3.5 mg K g −1 . Further site details have been published elsewhere (Herman et al., 2002;Schindlbacher et al., 2011).
The soil warming experiment comprised six blocks of paired 2 × 2 m plots (each pair consisting of one control and one warmed plot), established in 2004 (n = 3) and 2007 (n = 3). Six plots were warmed (hereafter termed warming treatment) using heating cables (Etherma, Salzburg, Austria) installed at 3 cm soil depth and at a distance of 7.5 cm. In the other six plots (hereafter termed control treatment), heating cables were installed, but not heated to account for the disturbance created by their installation. The heating system was controlled by a service unit that automatically kept a 4°C difference between the control and warming treatment throughout the snow-free period (April-December). On a half-hourly interval, soil temperature was recorded at 5 and 15 cm mineral soil depth. A detailed description of the experimental setup is given in Schindlbacher et al. (2007Schindlbacher et al. ( , 2009).

| Fine root sampling and processing
The sampling of fine roots (<2 mm in diameter) took place within one day at the end of the growing season in October 2012 and October 2019 using soil corers of 5 cm diameter and 20 cm in length. Three control and three warmed plots (warmed since 2004) were sampled in October 2012 (n = 3), whereas all 12 plots were sampled in October 2019 (n = 6). Ten soil cores were randomly taken from 0 to 10 and 10 to 20 cm soil depth in each plot. The sampling depth was less than 20 cm at a few sampling points because of the shallow soil and the dolomitic bedrock. In total, 120 samples were taken in 2012 and 240 samples in 2019. Samples were immediately transferred into zip lock plastic bags, stored in cooling boxes filled with ice packs, and transported to Bayreuth, Germany, for further processing.
In the laboratory, the soil cores were stored in a +2°C climate chamber and processed within a maximum of 3 weeks. The complete processing protocol included fine root washing, sorting into live and dead roots under a microscope, and finally scanning with a flatbed scanner. Roots were separated from the soil by wet sieving. We used a 0.63 mm sieve (Retsch Technology GmbH, 42781 Haan, Germany) to process the soil cores by hand using tap water until all soil and other impurities were removed from the fine root fraction. After washing, stones were picked using tweezers, dried in small aluminum dishes, and weighed to correct soil mass. Roots were then transferred into 500 ml glass beakers filled with water and ice and washed in an ultrasonic bath to remove residual soil particles attached to the fine roots. Right after washing, fine roots were sorted into a live and dead fraction under a binocular microscope. We used the morphological characteristics described in Wu (2000) and Burke and Raynal (1994). Herbaceous roots (white, light, and succulent) were excluded during sorting. Live roots were resilient and flexible, reddish, with several lateral root tips, while dead roots were soft, dark, and easily breakable. After separation, the dead root fraction was dried in an oven at 60°C for 3 days, and the dry weight on a soil area basis (g m −2 ) was determined.
Fine root production was measured using ingrowth cores. In October 2019, after the soil coring described above, five ingrowth polypropylene mesh tubes per plot (5 cm diameter, 20 cm long, 6 mm × 4 mm mesh) were inserted into the cored soil holes. They were filled with root-free mineral sieved soil deriving from the same study site. Soil bulk density was adjusted to ~0.5 g cm −3 (see above) by alternating filling and compaction of the soil using a funnel and a piston. After 12 months, the ingrowth cores were sampled, and roots grown within a year were processed as described above. Fine root production was calculated on a soil area basis (g m −2 year −1 ). We also estimated absorptive fine root biomass (aFRB) by multiplying mean root tip weight (see next paragraph for detailed information) by root tip number per m 2 (Ostonen et al., 2017).

| Fine root morphology
The live fine root fraction was scanned in 20 × 25 cm transparent trays filled with cold water using a flatbed scanner (EPSON Perfection V700; SEIKO EPSON CORP, Japan) with the following setting: scanning resolution of 400 dpi; pixel classification method based on grey level and dark root on white background. Care was taken to avoid root overlap within the tray, and images were analyzed using the software WinRhizo TM Reg 2008 (Regent Instruments Inc., Canada). Due to the many replicates, we scanned on average 80% of fine roots collected at 0-10 cm depth, while the complete sample collected at 10-20 cm depth was scanned. Images were analyzed for fine root length (cm), average diameter of the fine roots (mm), fine root volume (cm 3 ), the number of root tips, and surface area (cm 2 ). Based on those basic parameters, the following morphological fine root traits were calculated: specific root length, expressed as the ratio of root length to dry mass (SRL; m g −1 ); specific root area, expressed as the ratio of root surface to dry mass (SRA; cm 2 g −1 d.w); root area index, expressed as the ratio of root surface to soil surface (RAI; m 2 m −2 ); root branching intensity, expressed as the number of root tips per root length (RBI; tips cm −1 root length) and root tip density, the number of root tips per soil volume (RTID; tips m −3 soil). We recalculated root volume as the sum of all diameter class' averages (Freschet et al., 2021;Rose, 2017) and estimated RTD as the ratio of root dry mass to root volume (RTD; g cm −3 ). Right after scanning, root tips were excised and dried at 60°C for 2 days to determine root tip weight (RTW; mg). Mycorrhizal root colonization was determined under the microscope by random examination of intact root fragments. The roots to be examined were placed in a petri dish, and using different magnification levels allowed careful examination of the presence of emanating hyphae and rhizomorphs. Roots colonized by EcM were sampled using scalpels (ca. 20-30 root tips per samples), transferred into 5 ml Eppendorf ® tubes, and stored at −24°C until DNA extraction. The live fine root fraction was finally dried in an oven at 60°C for 3 days to determine fine root mass. Root biomass was expressed on a soil area basis (g m −2 ).

| Fine root chemistry
Dried fine root samples were ground using a ball mill (MM 400; Retsch Technology GmbH). Fine root chemistry was assessed on fine roots sampled in 2019. Element concentrations of P, Na, K, Ca, Mg, Fe, and Mn were determined after 65% HNO 3 digestion in a microwave device (Mars 5, CEM, Germany). For the digestion, 8 ml 65% HNO 3 was added to 100 mg dried and ground fine roots. Digests were then transferred to 50 ml volumetric flasks, diluted with deionized water, and filtered through nylon filters (0.45 µm). Element concentrations in the extracts were determined using ICP-OES (Optima 3200 xl; Perkin Elmer, Germany) and AAS (SpectraA 220 Z; Varian, USA). The organic C and total nitrogen (TN) concentrations were analyzed with an elemental analyzer (EA1110; CE Instrument, Italy).

| DNA extraction and amplicon barcoding
We extracted DNA from 48 samples collected during the second fine root sampling in 2019 (two technical replicates per plot and soil depth, 10 root tips per sample) using the ChargeSwitch ® gDNA plant kit (Invitrogen™; Carlsbad, USA) following the manufacturer's instructions. Amplicon libraries were prepared in two consecutive PCR runs using a liquid handling workstation (epMotion ® 5075; Eppendorf, Hamburg) and a thermocycler with a high-pressure lid (peqstar 384X HPL). Amplicon libraries were prepared of the V3 and V4 regions of the 16S rRNA gene as recommended by Illumina (2013) and otherwise processed as detailed in the following for fungi. The fungal ITS2 region was first (PCR1) amplified from the DNA extracts using the primers gITS7 (Ihrmark et al., 2012) and ITS4g ("CGCTTATTGATATGCTTAAGT", as modified after Schlegel et al. (2018)). The PCR1 primers ("nGS grade," Metabion, Planegg, Germany) included the fungus-specific sequences prepended by "Tags" (barcode sequences for sample identification of 4-7 bp in length, Table S3) and the recommended overhang adapter sequences (https://suppo rt-docs.illum ina.com/SHARE/ Adapt erSeq/ illum inaadapt er-seque nces.pdf). The reactions were conducted in a total volume of 4 µl, including the primers (370 nM, each), 2 µl OneTaq ® 2X Master Mix (NEB, Frankfurt am Main), and 0.2 µl DNA extract.
The thermocycler kept the prevailing 95°C for another 2 min after the reaction mixes were inserted. Amplification was achieved in 25 cycles of 94°C for 20 s, 45°C for 40 s, and 68°C for 55 s, followed by a final extension at 68°C for 7 min. Purification of the PCR1 products was achieved by adding 3.1 µl of a mixture of Exonuclease (3.2 U) and Shrimp Alkaline Phosphatase (0.32 U, both NEB, Frankfurt am Main) to each PCR1 product and incubating the total volume of 7.1 µl at 37°C for 25 min, followed by 80°C for 15 min. Index (barcode sequences of 8 bp in length) and adapter sequences (P5 and P7, respectively) were appended to the PCR1 products in the second PCR (PCR2), utilizing the overhang adapter sequences (Table S4).
The reactions were conducted in a total volume of 8 µl, including the primers (270 nM, each), 4 µl OneTaq ® 2X Master Mix, and 2.5 µl of purified PCR1 product. Equal volumes (1 µl) of root-derived fungal and bacterial PCR2 products were pooled separately. The two pools were purified using CleanPCR ® Nucleic acid Clean up (CleanNA, GC biotech B.V.) as recommended by the manufacturer. To discriminate against short fragments, such as primer dimers, CleanPCR ® beads were applied in a volume corresponding to 70% of the volume of the pooled PCR2 products. The sequencing service of the Faculty

| Sequence data processing
Sequence reads were pre-sorted by the sequencer according to the inserted dual index sequences, that is, each combination including a unique forward and a unique reverse index. Within index combinations, reads were assigned to samples according to dual index combinations, using the open-source software QIIME version 1.9.0 (Caporaso et al., 2010). During demultiplexing, only reads with at most one ambiguity base were retained, and a quality filter (threshold phred 19) was applied: reads were truncated after nine consecutive low-quality base calls and only retained if at least 35% of the entire sequence consisted of consecutive high-quality base calls.
Only the R1 reads were further processed. Reads were grouped according to the lengths of the tag-sequences and length adjusted using the FastX toolkit (http://hanno nlab.cshl.edu/fastx_toolk it/). CD-HIT-OUT (Li & Chang, 2017) was used for de novo clustering sequence reads to operational taxonomic units (OTUs) based on 97% sequence similarity. Sequences representing an OTU were assigned to taxa using QIIME and the UNITE database v8 (Kõljalg et al., 2013) as a reference. Initially, unassignable OTUs were assigned as detailed by Peršoh et al. (2010). A table coding the read counts per sample and OTU was generated and rarefied to 1646 reads per sample using the rrarefy function of the R package vegan (Oksanen et al., 2019).
For improved comprehensibility, the OTU identification number is prepended by the assigned genus name in the following.
Taxonomic and functional assignments were verified by species-level identification or sequence comparison for abundant and discriminative OTUs, in particular for OTUs assigned to the genus Sebacina.
Assignment and sequences of the OTUs from identified mycorrhiza taxa are provided in Table S7. Subsequent analyses were exclusively based on this (re-standardized) subset of mycorrhizal fungi. From the bacterial dataset, sequences of plant plastids were removed before normalization.

| Statistical analysis
All statistical analyses and graphs were performed using R version 4.0.3 (R Core Team, 2020). The R packages ggplot2 (Wickham, 2016) and gridExtra (Auguie, 2017) were used for data visualization. Potential outliers in the data set were first identified visually with boxplots and then tested with Rosner's test using the package EnvStats (Millard, 2013). Root traits were checked for normality using the Shapiro-Wilk test and, when needed, were square-root transformed before analysis to meet the assumption of normality.
Because of the layout of our control and warmed plots in pairs, we tested the effect of warming on FRB, necromass, morphological traits, and nutrient contents with paired t tests (α = 0.05 was used as the significance level).
We used Fisher's alpha, Shannon-Wiener, and Pielou's evenness to characterize EcM fungal and root-associated bacterial communities at the two soil depths. Fisher's alpha is a parametric diversity index that assumes that the species abundance distribution follows a log series distribution. It was calculated as S = a × ln (1 + n∕a); with S, the number of taxa, n is the number of individuals, and a is the We used linear mixed-effects models to test the effect of treatment and soil depth on fungal and bacterial diversity using the package lme4 (Bates et al., 2015). Treatment and soil depth were fixed factors and plot, inserted as random factor. Post-hoc pairwise comparisons among groups were conducted using the package emmeans version 1.5.4 (Lenth, 2021). To test the effect of treatment and soil depth on fungal and bacterial communities, we used permutational  Agerer (2001Agerer ( , 2006: contact; contact or medium-distance; short distance; contact, short or medium distance; medium distance and long distance. Because contact or medium and contact, short or medium distance, were rare, we reclassified exploration types into short distance (contact; short distance) and long distance (medium distance and long distance) types.
Paired t tests were performed to test for differences in exploration types between control and warming treatments.
We used the classification method program (CLAM) implemented in the package vegan to evaluate the preference of OTUs for either control or warming treatments. The CLAM statistical approach uses a multinomial model to classify OTUs in either generalist (i.e., well distributed between the two treatments) or specialist species (preference in one treatment) based on their relative abundance in the different treatments (Chazdon et al., 2011).
We carried out principal component analysis (PCA) to assess the interrelation between fine root traits, soil nutrients, EcM exploration types, and bacterial and fungal diversity indices using the R package FactoMinerR (Lê et al., 2008) and factoextra (Kassambara, 2017).
The variation of the EcM fungal community with treatment and soil depth was visualized with non-metric dimensional scaling (NMDS) using the function metaMDS of the package Vegan.

| Fine root biomass, necromass, and fine root production
Total FRB increased by 17% (from 355 to 414 g m −2 ) by soil warming in 2019 (Figure 1). The effect of soil warming on FRB was stronger at 10-20 cm (+34%) than at 0-10 cm soil depth (+12%); however, the absolute increase was greater at 0-10 cm soil depth. Soil warming decreased fine root necromass by 14% in 2019 (from 85 to 74 g m −2 for control and warmed plots, respectively). Absorptive fine root biomass measured in 2019 increased by 22% in warmed plots (76.6 and 93.4 g m −2 in control and warmed plots, respectively) and represented approximately 23% of the total FRB at 0-20 cm soil depth, irrespective of treatment (Table 1).
During the first sampling campaign in 2012, mean FRB amounted to 513 and 582 g m −2 in control and warmed plots, respectively. Soil warming had no significant effect, although there was a trend to higher root biomass (+13%) in the warmed plots (Figure 1). We also found that FRB decreased with soil depth on both occasions. In comparison to 2019, soil warming did not affect fine root necromass in 2012 (Table 1).
In the warmed plots, fine root biomass production in ingrowth cores increased strongly, by 128% from 99 to 225 g m −2 year −1 between October 2019 and October 2020 (Figure 2). While newly produced roots showed no changes in average diameter, SRL and SRA, their root tissue density and root tip density increased by 11% and 168%, respectively, in the warmed plots ( Figure 2).

| Fine root morphology
In 2019, morphological analyses of fine root samples revealed that soil warming increased SRL, SRA, RAI, RTID, and RTW. The average diameter of fine roots was also affected, especially at 0-10 cm (Table 1) Lowercase letters indicate significant differences between treatments (t test; p < .05) plots, respectively), but no effect was observed at 10-20 cm soil depth. RBI and RTID also increased in warmed plots (Table 1). The most significant increase of the latter two was observed at 0-10 cm soil depth. RTD was not affected at 0-10 cm depth (0.42 and 0.43 g cm −3 , in the control and warming treatment, respectively). Similar fine root morphological patterns were also observed in 2012. SRL, SRA, and RTID significantly increased by 11%, 43%, and 23%, respectively. On the other hand, fine root diameter and RAI were not affected by soil warming in 2012 (Table 1).
We observed an increase in the proportion of fine root length in the 0-0.2 mm diameter class in the warming treatment ( Figure   S1). We also noted a decrease in the proportion of fine root length with increasing diameter of fine roots, where warming became non-significant. Root length of the first diameter class (0-0.2 mm) accounted for up to 52% of the total root length, irrespective of treatment and soil depth. The first three diameter classes (0-0.2, 0.2-0.4, and 0.4-0.6 mm) contributed about 86% to the total fine root length in both treatments ( Figure S1).

| Nutrient concentrations in fine roots
The concentration of the macronutrients N, P, K, and Mg did not differ between control and warmed plots, although they showed a tendency to decrease with soil warming (Table 2). However, C and Ca concentrations significantly decreased by 5% and 28% with soil warming at 0-10 and 10-20 cm soil depth, respectively. The concentration of the micronutrients Fe and Mn was not affected by soil warming, except Na, which decreased by 40% at 10-20 cm soil depth.  (Figure 4).

TA B L E 1
Means ± SE of absorptive fine root biomass, fine root necromass, and morphological traits of live roots at 0-10 and 10-20 cm soil depth in control and warmed plots in 2012 (n = 3) and 2019 (n = 6). Different letters indicate significant differences between control and warming treatments, tested separately for each soil depth (t test; p < .05) Abbreviations: aFRB, absorptive fine root biomass; FRN, fine root necromass; RAI, root area index; RBI, root branching intensity; RTD, root tissue density; RTID, root tip density; RTW, root tip weight; SRA, specific root area; SRL, specific root length.

Fine root traits
The linear mixed-effects models showed that warming, but not soil depth, significantly affected bacterial diversity. Root-associated bacterial communities were altered in warmed plots, as indicated by the Shannon-Wiener diversity index and Pielou's evenness (Table S1). For EcM fungi, not treatment but soil depth significantly affected all diversity measures (Table S2). PERMANOVA revealed a significant effect of soil depth (p = .003) and warming treatment (p = .011) on EcM fungal community composition (Table 3), as also indicated by the NMDS ordination ( Figure S7). Soil depth and treatment explained approximately 34% of the total variation observed in the EcM community at the site.
Bacterial community composition was significantly affected by depth (p = .021), but not by treatment (Table 3).

F I G U R E 2
Mean (± SE) fine root biomass production (a) and morphological traits of fine roots from ingrowth cores in control and warmed plots between October 2019 and October 2020 (n = 6): specific root length (b), specific root area (c), average diameter (d), root tissue density (e), and root tip density (f). Different letters indicate significant differences between control and warming treatments (t test; p < .05) According to the classification method program (CLAM), soil warming affected EcM fungal OTUs differently (Figures 3 and 4).  TA B L E 2 Mean ± SE element concentrations; C:N, N:P, N:K ratios of live fine roots in control and warmed plots at 0-10 and 10-20 cm soil depth in 2019 (n = 6). Different letters within each row indicate significant differences between control and warming treatments tested separately for each soil depth (t test; p < .05)

F I G U R E 3
Abundance of ectomycorrhizal fungal genera with >1% relative abundance across all samples in control (a) and warmed (b) plots at 0-10 cm soil depth and their contribution to dissimilarity (c) determined by SIMPER analysis. Numbers behind genera of fungus are OTU numbers. Asterisks indicate a significantly higher abundance in either control or warmed plots

| Relationship between fine root traits, soil nutrients, EcM exploration types, and fungal and bacterial diversity
At 0-10 cm soil depth, the first and second principal components (PC1 and PC2) explained 40% and 22% of the variance, respectively  Note: Significance level given as **p < .01; *p < .05.
Abbreviations: df, degrees of freedom; MS, the mean sum of squares; Pseudo-F, F value by permutation; p (perm), p values based on more than 9000 permutations; SS, the sum of squares; % Var, the percentage of variation explained.

| DISCUSS ION
We explored how long-term soil warming impacts the fine root system and associated ectomycorrhizal fungi and root-associated bacterial communities in a temperate mountain forest. We found that FRB, aFRB, and fine root production were consistently higher in warmed plots, indicating greater root litter input. Soil warming also changed the morphology of fine roots, including higher RAI, SRL, SRA, and RTID, which increased the absorptive surface of the tree root systems for nutrient and water uptake. The community composition of ectomycorrhizal fungi was also affected by soil warming. Overall, our results suggest that the fine root system responded to warmer soil temperatures, although soil warming did not affect N availability. The low availabilities of P and K in the soil and low contents in roots and needles indicate a strong limitation of these nutrients in this temperate forest. Decreases in soil P and K availability have likely contributed to the changes of the root system in the warmed plots.

| Effects of soil warming on fine root biomass
In agreement with our first hypothesis, our results showed that soil warming increased FRB, and this increase was significant after 14 years of soil warming. Absorptive fine root biomass also increased with warming by 22%. Since there are no indications that soil warming changed soil N availability (Heinzle et al., 2021), we postulate that the increase in FRB and aFRB with soil warming in our study could be linked to the low P and K availability in the warmed plots, as explained by the optimal partitioning theory (Bloom et al., 1985).
We found a tendency of increasing FRB and aFRB in warmed plots with decreasing soil P and K ( Figure 6). Mean P concentration in fine roots (0.4 mg g −1 in warmed plots) was below the global mean fine root P concentration (0.9 mg g −1 ) (Gordon & Jackson, 2000). This P deficiency is further supported by the high N:P ratios of fine roots, indicating an imbalance between N and P in root tissues. Needle P concentration at the field site (0.9 mg g −1 ) also indicated P deficiency, being below the critical P concentration of 1.2 mg g −1 for Norway spruce needles (Ilg et al., 2009). Average fine root K concentration (1.3 mg g −1 ) was also below the global mean K concentration (2.8 mg g −1 ) in fine roots (Gordon & Jackson, 2000). The few existing studies on nutrient levels in fine roots of Norway spruce (0.7-1.7 mg P g −1 and 2.2-4.4 mg K g −1 ; Borken et al., 2007;Brunner et al., 2002;Genenger et al., 2003) illustrate the strong P and K deficiency at the study site.
At Achenkirch, the K concentration in soil solutions was persistently below the detection limit of 0.25 mg L −1 in all plots during the growing season (unpublished data), suggesting that K availability is strongly limiting, and K nutrition relies on K input by atmospheric deposition and ecosystem recycling. By contrast, Ca and Mg concentrations were very high in fine roots and soil solutions, resulting from the weathering of the underlying dolomite bedrock. Phosphorus is mainly supplied through weathering, desorption, and organic matter mineralization and is immobilized by sorption, precipitation, and microbial uptake processes (Bünemann, 2015). Phosphorus limitation causes strong plant-microbe competition for labile P resources, and the low P availability suggests that P might be a key player in driving the increase in FRB and fine root biomass production, especially at 0-10 cm soil depth. Recent findings at our site showed that longterm soil warming decreased total P in both organic and inorganic forms (Ye Tian et al., unpublished data). One would expect that soil warming increases the availability of P and K due to the increasing mineralization of litter and soil organic matter. However, those nutrients might become depleted in the long run due to high turnover rates, increased leaching losses, or intense competition for uptake between trees and microbes . The availability of F I G U R E 5 Box plots showing mean relative abundances (%) of ectomycorrhizal fungi grouped into short-distance (a) and long-distance (b) exploration types in control and warmed plots at 0-10 and 10-20 cm soil depth these elements was low in the warmed plots, and the plant-microbe competition for these limiting nutrients therefore was likely strong.
Increasing the FRB and the absorptive surface (see below) strengthens the competitiveness for nutrient uptake against other plants and non-root-associated soil microorganisms. Therefore, increased FRB and fine root production might be a key plant strategy to efficiently take up P and K, which seems to have become more limiting for trees in warmed plots.
Our result has been confirmed globally in a meta-analysis, demonstrating that soil warming increases fine root biomass production and FRB by 30% and 9%, respectively (Wang et al., 2021).
However, they attributed the increase in FRB to the stimulation of growth due to high photosynthetic rates and an increase in soil N mineralization. The response might, however, also be due to decreased soil moisture in the warmed plots, which is a common phenomenon in soil warming experiments (Xu et al., 2013), but which has not consistently been shown at the Achenkirch site, as the high site-level precipitation frequently resets any decrease in soil moisture in warmed plots, back to the levels in control soils. Therefore, while soil moisture certainly is a driver of tree fine root biomass and production, at the Achenkirch site, increased plant demands for soil P and K seem to be the major trigger of the fine root responses as observed in the warmed plots.

| Effects of soil warming on fine root morphology
In agreement with our second hypothesis, we found significant increases in SRL, SRA, and RTID in warmed plots at both sampling occasions. This suggests a tree strategy to form long roots with a large surface area and short lifespan (Weemstra et al., 2020), which improve nutrient and water uptake as well as soil exploration in warmed plots. This aligns with an acquisitive resource plant strategy (McCormack & Iversen, 2019;Weemstra et al., 2017). There was also a tendency toward a decrease in RTD in warmed plots, indicating lower costs for production of fine roots, which then, however, are less stress-resistant, but have faster metabolic and growth rates (Birouste et al., 2014). SRL and RTID tended to be negatively correlated with soil nutrients at both soil depths. RTD and soil K were positively correlated, indicating a resource conservation strategy ; blue arrows represent root-associated microbial diversity (Fischer diversity of bacteria and EcM fungi); red arrows represent fine root traits measured (aFRB, absorptive fine root biomass; D, fine root diameter; FRB, fine root biomass; RTD, root tissue density; RTID, root tip density; SRL, specific root length) and black arrows are soil properties (Soil P, total soil phosphorus; soil N, total soil N; soil K, soil K + ). Due to their strong correlation with SRL, SRA, and RAI were removed from the plots. At 0-10 cm, bacterial diversity was not important for both PCs, and therefore removed for graph visualization. 95% Ellipses are shown for soil K ( Figure 6). As mentioned earlier, the low availability of soil  Jackson et al. (1997). Björk et al. (2007) reported an increase in SRL and SRA, but no effect on RTD of fine roots <0.5 mm in a soil warming experiment in a boreal forest in Northern Sweden. These changes in fine root morphology were also linked to increasing plant nutrient uptake efficiency. However, changes in fine root morphological traits observed in this study differ from the meta-analysis of Wang et al. (2021), who found no response to warming. They explained this non-response by the limited number of studies and the high variability of morphological traits.
At 0-10 cm soil depth, the mean diameter of fine roots increased in warmed plots in 2019, while no change was observed in 2012. This contradicts the assumption that SRL and fine root diameter are negatively correlated (Bergmann et al., 2020). However, the proportion of root length in the 0-0.2 mm diameter class tended to increase (+38% in both soil depths) in the warmed plots ( Figure S1), supporting the optimization theory. Low diameter fine roots represent the most active part of the root system, which is highly relevant for plant nutrient and water uptake from soils. It is also this absorptive root size fraction that shows little secondary development, high metabolic activity, and high mycorrhizal colonization, which make them most responsive to changes in soil environmental conditions (McCormack et al., 2015). On the other hand, fine roots in the higher diameter classes were less affected, likely due to their high content of non-structural carbohydrates, making them more resistant to warming (Eissenstat et al., 2000;McCormack et al., 2015). These changes in fine root morphology imply that warming may profoundly alter tree nutrient and water uptake.

| Effects of soil warming on root-associated fungal and bacterial communities
Contrary to our third hypothesis, soil warming did not affect the relative abundance of EcM exploration types. With the observed changes in FRB and their morphology, one could expect a shift in the differentiation of extraradical hyphae, as fine roots are the primary source of C for EcM fungi (Koide et al., 2014). EcM fungi are usually patchily distributed in soils due to the heterogeneity in the distribution of soil nutrients (Luis et al., 2005). This micro-heterogeneity and patchiness might make it hard to detect significant responses of the EcM community at the species level. Fungal traits might contribute more clearly to shifts in ecosystem processes in the context of soil warming. For example, at 0-10 cm soil depth, we observed a 15% increase in the relative abundance of the EcM genus Cenococcum in warmed plots (Figure 3). This wide host and habitat range (Trappe, 1962) short-distance exploration type fungus associates well with all tree species present at our study site. Its high melanin content makes it more drought stress tolerant (Koide et al., 2014;LoBuglio, 1999) and might reduce eventual drought stress for the host tree species. Cenococcum mainly acquires NH 4 + as an N-source but also has well-developed proteolytic abilities (LoBuglio, 1999). However, at 10-20 cm soil depth, Sebacina-28 and Boletus-2 were most abundant in warmed plots (Figure 4). An increase in the relative abundance of Sebacina in warmed plots, a hydrophilic EcM genus with a short-distance exploration type, which requires low plant energy investment, might be beneficial for the host to satisfy its increased demand for water and nutrient uptake at low C investment. This fits well to changes in the fine root morphology as discussed above (increases in SRL, SRA, RAI, and RTID), because alterations in hostplant nutrition might induce a direct shift in host-tree C allocation to EcM fungi (Lilleskov & Bruns, 2001;Treseder, 2004). The increase in the relative abundance of the genus Boletus, a long-distance exploration EcM species, appears conflicting, but it is very long and highly differentiated hydrophobic mantles and rhizomorphs, which avoid hyphal water and nutrient leakage when transported over long distances (Agerer, 2006), may indicate increasing plant demand for nutrients such as P and K. The increase in the relative abundance of long-distance EcM, like Boletus at 10-20 cm soil depth in warmed plots, was also observed in other soil warming studies , although it was mainly related to increasing water uptake from deeper soil layers. a boreal forest (Allison & Treseder, 2008). In our study, EcM communities were dominated by host-generalist taxa, which are known to be less sensitive to changes in environmental conditions (Mucha et al., 2018). The dominance of these host-generalist taxa might be the reason why we did not find an effect of soil warming on EcM fungal diversity. In addition, relatively high atmospheric N deposition at the site, about 12 kg ha −1 year −1 (Herman et al., 2002), may have potentially shifted the competitive capabilities of the EcM fungi or decreased the tree dependence on mycorrhizal N acquisition (Clemmensen et al., 2006;Lilleskov & Bruns, 2001;Treseder, 2004).
Root-associated bacterial community diversity (Pielou's evenness and Shannon-Wiener index) increased with soil warming.
Greater bacterial diversity is beneficial for the ecosystem as a whole, because it promotes metabolic activities and efficient nutrient mineralization (Nautiyal & Dion, 2008). A more diverse and evenly distributed bacterial community might have greater resilience and functional stability in relation to warming (Cleland, 2011). This indicates that, in this forest ecosystem, root-associated bacterial communities may have maintained their ability to perform ecosystem multifunctionality with soil warming and fits well with the observed sustained increase in soil respiration in warmed plots since the beginning of the experimental warming manipulation at the site (Schindlbacher et al., 2011(Schindlbacher et al., , 2015. Soil depth, related to a strong change in physicochemical properties, which greatly influences soil microorganisms, affected bacterial community composition in our study. A wide range of edaphic factors such as soil nutrients, the quality and quantity of litter inputs, and root-derived C could affect the composition of soil and root-associated bacterial communities (Baldrian, 2017). Because those factors change with soil depth, a corresponding shift in the root-associated bacterial community is expected. In our study, one of the most evident changes through the soil profile was the decrease in FRB with soil depth, which may affect root exudation, a crucial C source for rootassociated bacteria.
It has to be noted that the experimental warming setup had some limitations. Only a limited area of soil was warmed, whereas the above-ground parts of the tree vegetation remained unaffected and experienced ambient temperatures. Thus, we cannot exclude varying fine root responses if the whole rooting area of individual trees or the whole forest would have warmed. For instance, if soil warming increases nutrient availability, one can anticipate root (in)growth from the surrounding soil into the warmed plots. This would result in an overestimation of the warming effect on FRB growth as well as stocks. Under the preconditions in our study, such an artefact can rather be excluded since it is unlikely that trees invest into root in-growth in warmed plots and progressively nutrient-depleted soils at the expense of ingrowth in unwarmed soil with higher nutrient availability. A general limitation of soil warming studies is that we could not predict how climate warming and the associated above-ground physiological responses will affect below-ground C allocation, fine roots, and EcM dynamics. Despite these constraints, we provide important insights into long-term warming effects on tree fine root dynamics and the connected soil C and nutrient dynamics.
In conclusion, our findings suggest that soil warming profoundly changed FRB, fine root production, root morphology, and the community composition of EcM fungi, which may have strong implications on fine root functions in temperate forests. The response of fine root systems to soil warming is linked to the availability and acquisition of soil nutrients which can differ among forests. The limited soil P and K availability align well with the observed responses in fine root biomass and morphology, though other more general physiological responses to warming, like faster growth (Pregitzer et al., 2000), may have contributed to the observed changes. Compared to the strong warming response of fine roots, the effects on rootassociated microbial communities seem limited, at least for the parameters measured in this study. This is surprising because the root system is seen as a continuum of roots, symbiotic fungi, and bacteria (Freschet et al., 2021;Ostonen et al., 2017). More plots and seasonal replicates seem necessary to increase the statistical power of microbial community analyses and to assess their potential effects on the fine root systems. The long-term warming response of tree fine roots may have strong implications on ecosystem C dynamics. Assuming steady-state conditions between production and mortality, fine root C input to soil is around 106 g C m −2 year −1 in warmed plots versus 48 g C m −2 year −1 in control plots, based on fine root production in ingrowth cores and fine root C concentration ( Table 2). Because of the disturbances associated with ingrowth cores, the estimated C input only represents an approximation to the surrounding soil. Other long-term soil warming studies showed that the stimulatory effect of temperature on soil CO 2 efflux decreased over time (Melillo et al., , 2017. However, this was not yet the case at the Achenkirch site (Schindlbacher et al., 2009), indicating that the increase in FRB and fine root production increased the root system C input into the warmed soils by root exudation and fine root turnover. How mechanistically the continued increase in soil CO 2 efflux is linked to FRB, turnover, production, exudation, and morphological changes is currently under investigation, using field root exudation experiment and fine root radiocarbon dating methods. The consistent patterns of root responses during more than a decade of intensive soil warming indicate that changes in the fine root system are not of transient nature but likely persistently affect tree and soil C dynamics. Ploompuu for assistance with fine root processing; and Gerasimos Makis Gkoutselis, Theresa Janssen for their help with DNA extractions. Open Access funding enabled and organized by Projekt DEAL.

CO N FLI C T O F I NTE R E S T
The authors have no conflict of interest to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are openly available in Dryad at https://doi.org/10.5061/dryad.jwstq jqbj.