The Fungal Community of a High‐Arctic Semi‐Desert Ecosystem Is Robust to Two Decades of Doubled Summer Precipitation but Influenced by Plant Dominance

ABSTRACT

In high‐arctic semi‐desert ecosystems, climate projections suggest a doubling of the current ≤ 250 mm annual precipitation by 2075, increasingly falling as summer rain. This is expected to affect the balance among fungal ecological groups and stimulate decomposition over build‐up of soil organic matter. Here, we report the effects of 25 years of artificially doubled summer rainfall combined with nitrogen and phosphorus addition on soil fungal communities and organic matter stocks, in a semi‐desert near Zackenberg in NE Greenland entirely dominated by the ectomycorrhizal plant species Salix arctica , Dryas integrifolia and Kobresia myosuroides . Surprisingly, soil fungal biomass and community composition and soil carbon and nitrogen pools were mostly unaffected by the treatments, while the three plant species appeared to cause consistent small‐scale variation. The fungal communities were dominated by the ectomycorrhizal genera Cortinarius, Tomentella and Inocybe, but the evergreen Dryas supported a larger abundance of potentially nitrogen‐mining Cortinarius species, while the graminoid Kobresia was associated with a larger abundance of saprotrophic or root‐invasive Mycena species. Our results suggest that the soil fungal community and linked processes of this high‐arctic ecosystem were more robust to two decades of doubled summer rainfall than expected. Long‐term changes in belowground fungal communities will likely be mediated by changes in plant communities rather than driven directly by climate change.

1. Introduction

The ongoing climate change is particularly pronounced in the Arctic (IPCC ²⁰²³). For dry, high‐arctic ecosystems, the expected increase in rainfall in the otherwise particularly dry growing season (Bintanja and Andry 2017) could increase plant growth and affect organic matter decomposition. These changes have multiple potential cascading effects on plant cover, soil fungal and microbial communities, and nutrient cycling.

Arctic environments are generally characterised by low nitrogen (N) availability, and smaller plants associated with ecto‐ and ericoid mycorrhizal fungal symbionts dominate many vegetation types (Walker et al. 2018). In the Low Arctic, vast areas are covered by continuous tundra vegetation dominated by a mixture of ecto‐ and ericoid mycorrhizal shrubs, as well as non‐mycorrhizal sedges and mosses (Walker et al. 2018). However, the sparse vegetation in high‐arctic deserts and semi‐deserts can be dominated by a few plant species where ectomycorrhizal (ECM) host plants are far more abundant than non‐ or ericoid mycorrhizal host plants, as is the case with Salix, Dryas and Kobresia in Zackenberg (Christiansen et al. 2012). Mycorrhizal fungi provide their hosts with N and potentially other limiting nutrients and water through specialised mycorrhizal mycelia in the soil, and Hartig nets inside their hosts, thus alleviating nutrient‐ and potentially water limitation of their host. In return, the mycorrhizal fungi receive carbon (C) directly via plant photosynthate allocation to the roots, making them at least partly independent of C from decomposing soil organic matter. The mycorrhizal extramatrical mycelium extends from the plant roots into the surrounding soil, but both morphological organisation (Agerer 2001) and physiological capacities vary among fungal species. Ecto‐ and ericoid mycorrhizal fungi are generally the most abundant fungal ecological guilds in the well‐characterised Arctic funga (Blaalid et al. 2014; Geml et al. ²⁰¹⁵; Michelsen et al. ¹⁹⁹⁸; Morgado et al. ²⁰¹⁴), as a cold and often dry environment inhibits the lifestyle of organic matter degrading saprotrophs, leading to low decomposition rates. Root and mycorrhizal mycelial growth and turnover thus mediate C transport into the soil and may serve to build up a large fraction of the globally important arctic soil C stocks (van Huissteden and Dolman ²⁰¹²), in a similar way as observed in boreal forest (Clemmensen et al. 2013).

Warming and adequate soil moisture can lead to faster decomposition and mineralisation of N and P, and thus higher nutrient availability to plants (Jonasson et al. 1993; Nadelhoffer et al. ¹⁹⁹¹). Furthermore, increased precipitation may affect the accessibility of available nutrients (Schaeffer et al. 2013; Norby et al. ²⁰¹⁹). Several studies have investigated the short‐term impact of combined water and nutrient addition on plant responses (Wookey et al. 1995; Robinson et al. ¹⁹⁹⁸) and microbial activity (Christiansen et al. 2012), whereas the multi‐decadal responses of C pools and soil fungi are less resolved (but see Jespersen et al. ²⁰²²; Oldcorn et al. ²⁰²⁵).

Under such conditions, saprotrophic fungi may be more active and accelerate decomposition, while enhanced plant growth produces more plant litter, expanding the pool of organic C available to decomposers. The less nutrient‐starved plants conversely decrease the C allocation to roots and mycorrhizal fungi, and accordingly, restrict the competitive advantage which mycorrhizal fungi normally have under harsher environmental conditions. Hence, climate warming in the Arctic is expected to favour saprotrophic relative to ectomycorrhizal fungi (Leake et al. 2002; Geml et al. ²⁰²¹). Further, microbial degradation of organic matter shows a stronger response than plant primary production to temperature increases, which is the major causative mechanism behind the dreaded Arctic carbon feedback; this would have important consequences for multiple Arctic ecosystems (Bjorkman et al. 2018), and for the Arctic global warming feedback (Elmendorf et al. 2011).

However, the hitherto observed effects of the current climate change on plant communities and their fungi are not homogeneous across the Arctic. Taller ectomycorrhizal shrubs have generally increased in abundance as a response to climate change particularly in lower Arctic latitudes (Clemmensen et al. 2006; Henry et al. ²⁰²²), whereas the mostly non‐mycorrhizal graminoids have shown their strongest relative expansion in the colder and drier high‐arctic sites (Elmendorf et al. 2012).

Long‐term field experiments are essential for addressing the ecological responses and the carbon feedbacks associated with climate change. In a study on warming and fertilisation in High Arctic Canada, Lamb et al. (2011) found that while the plant community responded strongly to the warming, belowground soil bacterial functional groups and greenhouse gas fluxes failed to respond to similar degrees. However, several studies have found moisture rather than temperature per se to be the most important limiting factor for plant growth, particularly in the High Arctic (Arndt et al. 2019; Campbell et al. ²⁰²⁰; Wenzl et al. ²⁰²⁴), and there could be differences in belowground community responses between fungi and bacteria. Mediterranean fungi showed clear responses to increased summer drought stress associated with climate change (Castaño et al. 2018; Peñuelas et al. ²⁰¹⁸). However, there has been a lack of similar long‐term studies of precipitation effects on the fungal communities in High‐Arctic areas.

Here, we evaluated the effects of 25 years of increased summer precipitation and nutrient availability in a field manipulation experiment in a high‐arctic semi‐desert dominated by three ECM plant species, Salix arctica , Dryas octopetala and Kobresia myosuroides , in Northeast Greenland (Zackenberg). We assessed the fungal community composition, fungal and bacterial biomass and soil C and N stocks associated with each of the three ECM plant species in response to the field manipulations. Furthermore, we explored the C and N stable isotope composition in soils, plant tissues and fungal sporocarps to understand potential functional differences among the three plant species, their associated fungi and associated soil processes.

This long‐term irrigation experiment is almost unparalleled in terms of time span in the High Arctic. The dominance of a few ECM plant species makes the ecosystem a unique opportunity to understand soil processes driven by ECM fungi and plants and their co‐existing saprotrophic fungi. Vegetation responses have been documented after 25 years of treatment (Oldcorn et al. 2025) in this same field experiment, where water addition moderately increased the total vegetation cover, particularly of graminoids, mostly at the expense of bare soil rather than of other plants. Thus, our main hypotheses were (H1) that increased summer precipitation would increase the total soil fungal biomass; (H2) that both increased precipitation and N and P availability would stimulate fungi with pronounced decomposer activity (saprotrophic decomposing fungi) more than ECM fungi; (H3) that this would reduce soil C content and C/N ratio consistent with increased organic matter decomposition; and (H4) that magnitudes and patterns of these responses would depend on plant species related to differences in root and host specificity/preference traits of the three dominant plant species.

2. Materials and Methods

2.1. Site Description

The experimental area is high‐arctic dry heath/semi‐desert located south of the Zackenberg Research Station in Northeast Greenland (74°30′ N, 21°00′ W). The growing season lasts from mid/end of June to end of August. The semi‐desert has been classified floristically as part of the plant communities found on abrasion plateaus (Bay 1998; Bliss and Matveyeva ¹⁹⁹²), comprising ~50% bare soil. The vegetation is thus sparse and patchy and completely dominated by the three ECM plant species Kobresia (syn. Carex) myosuroides Vill., Dryas octopetala L. and Salix arctica Pall. The soil organic matter content (0–12 cm depth) is 5.4% ± 0.3%, and the soil has a pH of 6.86 ± 0.03. There are occasional snow beds making up 5%–10% that locally retain higher moisture levels, but the largest part of the ecosystem (up to ~60%) consists of dry heath as in the experimental area studied here (Greenland Ecosystem Monitoring [GEM] ^2024c). See Figure S1a–h for photos of the three plant species, the experimental site and important fungal fruit bodies.

The research station has a meteorological record from 1995, and mean annual precipitation ranged from 51 to 433 mm in the period 1995–2022 (GEM ^2024a), about 75% of which falls as snow during the cold season (September–May). The available data suggest that mean growing season precipitation has no clear trend for the past 30 years, at ~50 mm for 1996–2005 (Meltofte et al. 2008), with a similar average, but with large inter‐annual variation of 10–100 mm for the corresponding data for June–August 2015–2023 (Fouché et al. 2024; GEM ^2024a). The mean annual temperature for 1995–2022 was −8.8°C ± 0.2°C, with cold season temperatures of −13.3°C ± 0.2°C, and growing season temperatures of 4.8°C ± 0.2°C (GEM ^2024b). There has been a significant increase in mean annual temperatures and summer (June–August) temperatures of ~0.6°C/decade since 1995, and a comparable but non‐significant trend in December–February temperatures due to higher inter‐annual variability in the winter (GEM ^2024b).

The field experiment was established in 1996 on a dry gravel terrace adjacent to the river delta banks of the Zackenberg River, and it consists of 48 0.5 × 0.5 m plots placed in a grid with 120 cm distance between centres of neighbouring plots. The experiment is fully factorial with eight treatments randomised across six replicate blocks: control plots (C), plots with added nitrogen (N), phosphorus (P) or both (NP), and with all four either watered (W) or left at ambient precipitation (Illeris et al. 2003). The treatments have been applied as follows: in the summers of 1996, 1997, 2007 and 2012, plots were factorially fertilised either with N as NO₃NH₄ at a rate of 3.75 g N m⁻², or with P as Na₂HPO₄ at a rate of 0.81 g P m⁻². The fertiliser dose was intentionally kept low and infrequent to avoid excessive, unrealistic fertilisation. Assuming a growing season mineralisation rate of 0.6 g N m⁻² and 0.027 g P m⁻² for dry heath tundra, that is, the upper levels of biome estimates from Nadelhoffer et al. (1992) and Schmidt et al. (1999), respectively, the N and P additions approximately represent a doubling of summer N availability, and a fourfold increase of P availability since 1997. Thus, the additions are within a realistic range of changes in nutrient availability expected under the increasing temperatures, even if there is little information on long‐term impacts on soil nutrient availability (Oldcorn et al. 2025).

From 1997 and onwards, factorial water addition was achieved by watering each plot weekly with 2 L of water from the nearby river throughout the growing season, corresponding to 8 mm of extra precipitation per plot weekly, that is, 64–80 mm over a typically 8–10 weeks growing season, which more than doubled average growing season precipitation.

2.2. Sampling and On‐Site Treatment of Samples

Soil cores were sampled from the 48 experimental plots in late July, 2022. A cylindrical steel soil corer, 3 cm in diameter, was used to collect samples to a depth of 10 cm from the soil surface. Soil cores were taken in the immediate vicinity of each of the three dominant plant species ( Dryas octopetala , Kobresia myosuroides and Salix arctica ) whenever they were present in the plots. In eight of the 48 experimental plots, Salix arctica was absent, resulting in a total of 3 × 48 − 8 = 136 soil samples (Table S1). In the field, each soil core was immediately transferred into a tray and homogenised by hand and with scissors. The soil samples excluded any living aboveground plant parts but included the litter layer from the ground surface. Roots and stones larger than 5 mm in diameter were removed, while all smaller roots were kept in the sample.

The fresh soil was brought to the field lab in zip lock bags and weighed. A 20–25 g subsample was weighed into a paper bag, dried at 60°C for > 48 h and re‐weighed, for determination of water content. A dry soil subsample of ~10 g was used for all geochemical analyses as well as for DNA‐based work.

We also sampled reference material from monodominant patches of each of the three host plants across six patches immediately adjacent to the experimental site (3 species × 6 plots = 18 patches). The aim was to achieve plant species‐specific organic matter input types and root‐associated communities to aid interpretation of results from the manipulated experiment, where the three plant species were usually growing more intermingled between one another. From each monodominant vegetation patch, the focal plant was carefully excavated while sampling the following fractions: (1) recognisable litter from the ground (Salix, Dryas) or directly from the plant (Kobresia), (2) at least 20 green leaves from fresh shoots, (3) roots (< 2 mm diameter) directly attached to stem bases or rhizomes, (4) soil from the rooting volume, and (5) dead stem bases of Kobresia, which represent a long‐lasting category of organic input to the soil. In the field laboratory, root samples were washed three times and inspected under a stereo microscope. During this process, an additional sample of (6) dead roots (Kobresia, Salix) and (7) ECM root tips were sampled (38 root tip samples in total across the three plant species). Fungal fruit bodies were sampled in a 20 m zone around the experimental area (48 samples). All samples were dried at 60°C.

2.3. Sample Preparation and Geochemical Analyses

The dried soil, fungal and plant samples were crushed in a ball mill at 50 oscillations s⁻¹ for 1 min. Subsamples of the ground material of between 30 and 40 mg (adjusted to yield the same amount of C as in the standard material) were analysed for C and N concentrations and ¹⁵N and ¹³C natural abundances using a Eurovector elemental analyser (Eurovector, Pavia, Italy) coupled to an Isoprime isotope ratio mass spectrometer (Elementar, Cheadle Hulme, UK). Natural abundances of isotopes are expressed in the δ notation relative to international standards (Vienna Pee Dee Belemnite for C and atmospheric N₂ for N): δX _sample (‰) = 1000 × [(R _sample/R _standard) − 1], where R is the molar ratio of heavy X/light X. All samples were analysed with reference gas calibrated against international standards IAEA C5, CH6, CH7, N1, N2 and USGS 25, 26, 32, and drift corrected using peach leaves (NIST 1547) as internal standard. The SD of repeated samples for both N and C analyses was ≤ 0.2‰.

2.4. DNA Extraction and Fungal Amplicon Sequencing

DNA was extracted from 200 mg of the dried and milled soil with the NucleoSpin Soil kit, using SL1 lysis buffer and extraction enhancer (due to low DNA concentration) according to the manufacturer's instructions (Macherey‐Nagel, Düren, Germany). The dried and milled plant roots (50 mg) were extracted with lysis buffer SL2 without enhancer. To obtain DNA from fungal fruit bodies, we used lamellae of dried specimens with a standard CTAB‐chloroform‐isopropanol procedure (Gardes and Bruns 1993). DNA concentration was measured on a Qubit Fluorometer (ThermoFisher Scientific, USA), and extracts were diluted to 2 ng DNA μL⁻¹.

For fungal community sequencing we PCR‐amplified the internal transcribed spacer region 2 (ITS2) in technical duplicates using the forward primers gITS7 (soil samples) or fITS7 (root samples) (Ihrmark et al. 2012) and the reverse primer mix ITS4/ITS4arch at a 3/1 ratio (Sterkenburg et al. 2015). All primers were extended by a linker base (T), sample‐specific 8‐base identification tags (which differed in at least three positions) and a terminal base (C) (Clemmensen et al. 2023). Amplification was done with minimised numbers of PCR cycles (22–31) to reduce length‐based biases and obtain quantitative amplification (Castaño et al. 2020). We used between 0.05 and 3.2 ng of DNA per 50 μL PCR reaction with primer concentrations of 0.5 μM (ITS7), 0.3 μM (ITS4) and 0.1 μM (ITS4arch), and 0.2 mM dNTPs, 0.75 mM MgCl₂ and 0.025 U μL⁻¹ of DNA polymerase (Dream Taq, Fermentas, Sweden). Amplicons were cleaned using the AMPure kit (Beckman Coulter, Beverly, USA), concentrations measured fluorometrically with Qubit and amplicons mixed in equal amounts into two sequencing pools of 80 samples each. A mock community of equal abundances of 10 artificial ITS2 members with lengths varying between 142 and 599 bp (with a separate sample‐tag) was added after the pooling to monitor amplicon length‐based biases during sequencing (Castaño et al. 2020). Our mock communities were somewhat biased towards longer (500–600 bp) compared to shorter (200–400 bp) fragments (data not shown). Thus, shorter sequences, especially Ascomycetes, might have been discriminated against based on our use of PacBio Sequel II (Wang et al. 2014).

Five DNA extraction negative controls and two PCR negatives were also included. The 18 root samples (plus a negative control) were pooled with samples from a different experiment. The pools were further purified with the Cycle Pure kit (EZNA, Omega BioTek, Nocross, USA) and quality checked in a Bioanalyzer 2100 (Agilent). All pools were sequenced at SciLifeLab (NGI‐Uppsala, Sweden) following the addition of sequencing adapters via ligation. Libraries were prepared using the SMRTbell Template Prep Kit 1.0 and sequenced together on one PacBio Sequel II SMRT cell (Pacific Biosciences, Menlo Park, CA, USA).

For identification of fungal fruit bodies, we used a standard ITS1F‐ITS4 primer pair for PCR amplification in 20 μL reactions with both primers at a concentration of 0.5 μM (Gardes and Bruns 1993). Sanger sequencing was done by MACROGEN in the Netherlands, using the ITS4 primer for sequencing.

2.5. Bioinformatic Analysis of Fungal Communities

Sequences were quality filtered and clustered using the SCATA pipeline (https://scata.mykopat.slu.se/, accessed June 2023). Sequences were screened for primers (≥ 90% match required) and sample tags (at 100% match). Sequences with an average amplicon quality score below 20, a score below 7 at any position, and shorter than 140 bases were removed. Unique sequences occurring only once in the global dataset (all sequencing pools combined) were removed to reduce errors before clustering. Sequences were clustered into species‐level clusters (denoted ‘species’ in the following) using single linkage clustering with a minimum similarity of 98.5% to the closest neighbour required to enter clusters. We identified the sequences taxonomically by matching to the UNITE database (Abarenkov et al. 2024) with the PlutoF software (Abarenkov et al. 2010), using a 99% sequence similarity to reference sequences for identification to species. To further obtain statistical support for species identification, we used the Bayesian RDP classifier (Wang et al. 2007) in Mothur 1.48.2 (Schloss et al. 2009) with a posterior probability threshold of 80 BPP. After excluding the mock community sequences (8%) and non‐fungal sequences (20%), and samples with < 100 sequences (4 out of a total of 136 experimental samples), sequence counts of confirmed fungal species (193,237 reads) were converted to relative proportions out of the total sequences per sample (Table S2).

Based on the FunGuild database (Nguyen et al. 2016), we assigned species that could be identified to genus (or an otherwise ecologically informative taxonomic level) to the groups: ectomycorrhiza (ECM), saprotrophs (SAP), ericoid mycorrhiza (ERM), parasites, commensal endophytes, lichens and unknown/uncertain ecologies. The Sanger sequences from the fruit bodies (n = 48) were clustered into OTUs at the 98.5% level using the cluster_size command in Vsearch (Rognes et al. 2016) and taxonomically and ecologically classified based on the same databases.

2.6. Fungal and Bacterial Abundance

Total fungal and bacterial abundances were estimated by quantitative real‐time PCR (qPCR) using the iQ‐SYBR Green master mix (Bio‐Rad, Hercules, CA, USA) on a Bio‐Rad CFX Connect Real‐Time system. We targeted the fungal ITS2 region using the primers fITS7, ITS4 and ITS4arch (as above), and the bacterial V3 region of the 16S rRNA gene using the primers 341f and 534R (López‐Gutiérrez et al. ²⁰⁰⁴). The iQ‐SYBR Green reaction mix of 20 μL included 4 ng of template DNA, BSA (0.1%) and bacterial primers at 0.5 μM or fungal primers as for the amplicon preparation (above). Annealing temperatures were 60°C for bacteria and 56°C for fungi. Samples were first tested for inhibition by amplifying 1 × 10⁵ copies of pGEM‐16S plasmid (Promega, WI, USA) spiked into qPCR reactions with DNA template or non‐template controls, together with specific primers to amplify the plasmid (M13F and M13R). No inhibition was detected in any of the samples at the used template concentration. For each microbial group, we prepared a standard consisting of serial dilutions of known amounts of linearised plasmid containing each target fragment. Samples were run in duplicates on two consecutive plates, and results were repeated if the coefficient of variation in estimated copy numbers between duplicates was larger than 30%.

2.7. Statistical Analyses

The effects of the treatments on the geochemistry data and fungal abundances were evaluated using four‐factor ANOVA, initially testing the main factors water addition, nitrogen, phosphorus as well as plant species (Dryas, Salix or Kobresia) under which the soil was sampled, as fixed factors with all interactions, and block as a random factor. As analyses showed the effect of the plant species to always be strongly significant, we ran the same three‐factor ANOVAs on the dataset subdivided by each plant species. All significant (p < 0.05) or marginally significant (p < 0.10) effects are reported in the figures. All statistical analyses were conducted in R (4.3.2; R Core Team ²⁰²³), using the packages ‘car’ (Fox and Weisberg 2019) for the three‐way ANOVAs, ‘agricolae’ (de Mendiburu ²⁰²³) for the Tukey tests, ‘dplyr’ (Wickham et al. 2023) for dataset manipulations/transformations, and ‘effectsize’ for calculating the η ²‐values (Ben‐Shachar et al. ²⁰²¹). To evaluate whether the sequence in isotope relationship between δ¹⁵N and δ¹³C content in the leaves, litter, roots, root tips, soil and fungal ectomycorrhizal fruit bodies differed among the three plant groups (Dryas, Salix, Kobresia), we used an analysis of covariance (ANCOVA) framework. First, we combined all observations into a single dataset and fitted a linear model of the form:

$${\mathrm{\delta }}^{15}\mathrm{N}={\beta }_0+{\beta }_1{\mathrm{\delta }}^{13}\mathrm{C}+{\beta }_2\text{Group}+{\beta }_3\left({\mathrm{\delta }}^{13}\mathrm{C}\times \text{Group}\right),$$

with Group as a categorical factor with three levels representing our plant species. We separated all plant tissue and soil collected from its rhizosphere in three subsets for each plant species, but included all ectomycorrhizal fungi in each subset, as we had no way to identify their most likely plant host. The significance of the interaction term (δ¹³C × Group) was used to test whether slopes of δ¹⁵N and δ¹³C content differed between the three plant species. To obtain slope estimates, standard errors, degrees of freedom and p‐values for each species individually, we then fitted separate linear regressions of δ¹⁵N against δ¹³C within each plant group. These models were extracted using the broom package (Robinson et al. 2023). Pairwise comparisons of slopes between groups were conducted using the emmeans package (Lenth 2023), using the emtrends function for estimating group‐specific trends and the ‘contrast’ function with the pairwise method for comparing slopes and providing unbiased pairwise tests of differences.

The complete fungal community dataset (relative abundances) and the three subsets separated by plant species were analysed by nonmetric multidimensional scaling analysis (NMDS) on the Bray–Curtis dissimilarity index (Bray and Curtis 1957) using the metaMDS function of the R ‘vegan’ package (Oksanen et al. 2024). PERMANOVA tests on soil fungal community compositions and their possible associations with treatments, plant species and blocks were done using the adonis2 function, also of the R vegan package.

To calculate a distance‐decay relationship and test if the fungal communities in nearby plots were independent of each other, we performed Mantel tests (Douglas and Endler 1982) on the correlation between the matrices of Bray–Curtis dissimilarity and the geographic distances across the 48 experimental plots for the complete sample set (all three plant species), and for the three subsets separated by plant species. We created a distance matrix of the geographical distances for a Mantel test using the ‘vegan’ mantel function (Harder et al. 2019) between all pairs of the 48 experimental plots, with simple Pythagorean triangulation based on the 120 cm mean distance between all neighbouring plot centres, using the R package ‘sf’ (Pebesma and Bivand 2023). R packages ‘ggplot2’ (Wickham 2016) and ‘patchwork’ (Pedersen 2024) were used to create and arrange the graphics in the figures.

3. Results

3.1. Sequencing Output

After quality filtering and discarding sequence reads with unmatching tags, plant and mock sequences, contaminants (also found in the negative controls) and species with < 10 sequences across all samples, 193,237 high‐quality sequences remained from 132 experimental soils (151,437), 16 reference soils (28,671), 16 reference root samples (10,837) and 23 reference root tip samples (Tables S1 and S2). For root tips, we only retained samples that had at least one ECM sequence, while all other samples with ≥ 100 sequences were kept (Tables S1–S3). After this data trimming, 734 fungal species‐level clusters remained, of which 337 (76% of the sequences) could be identified to genus at the most conservative bootstrap level (80), and most (332) of these were found in the experimental soils, while the reference samples had subsets (Table S2). The sequences of 48 fungal fruit bodies belonged to 23 fungal species.

3.2. Fungal Communities in Soils and Roots

Basidiomycota (54%) and Ascomycota (41%) dominated the dataset, and 14 genera each comprised at least 1% of the communities in the experimental soil samples, and 9 of them at least 2% (Figure 1, Table S3). The most common fungal genera in both the experimental and reference soil samples and the reference plant roots were the ectomycorrhizal Inocybe, Cortinarius, Tomentella, Cenococcum, Pseudotomentella, Russula, the saprotrophic Cladosporium, Schizothecium, Phaeococcomyces and Mortierella, and the saprotrophic/endophytic/versatile Mycena, Serendipita and Hyaloscypha (Table S3).

FIGURE 1.

Bar plot of sequence fractions of the nine fungal genera that constituted ≥ 2% of the total soil fungal communities based on sequencing of ITS2 markers. Each bar represents the average fungal community composition in soils collected under each of three plants (Dryas, Salix, Kobresia) across eight field treatments in a high‐arctic semi‐desert at Zackenberg, NE Greelands. Numbers in small font denote the number of replicates for each bar (n = 2–6).

The R ² of 0.06 (p > 0.001) in the PERMANOVA analyses (Table 1) indicated a small but significant effect of plant species on the soil fungal communities; particularly Kobresia had a different community than Dryas and Salix (Figure 2a). Fungal community composition associated with Dryas and Salix, but not in Kobresia, was related to the experimental blocks (Table 1). The fungal communities in the experimental plots showed a weak clustering with the three combined treatments in the PERMANOVA tests with R ² of 0.06 (p = 0.03), but all clustering by treatments became insignificant once the community data were split by the respective plant species.

TABLE 1.

PERMANOVA tests on Bray–Curtis dissimilarity matrices of the fungal communities in the treatment plots, testing effects of treatments, plant species and blocks. Single or two‐way treatment interaction factors showed no significances or tendencies (all p‐values > 0.1).

Dataset	Model	df	R 2	Pr(F)
All data	Nitrogen:Phosphorus:Watering:Plant species	23	0.21	0.001***
All data	Plant species	2	0.06	0.001***
All data	Nitrogen:Phosphorus:Watering	7	0.06	0.03*
All data	Blocks	5	0.08	0.001***
Dryas	Nitrogen:Phosphorus:Watering	7	0.15	0.52
Dryas	Blocks	5	0.15	0.001***
Salix	Nitrogen:Phosphorus:Watering	7	0.17	0.99
Salix	Blocks	5	0.19	0.001***
Kobresia	Nitrogen:Phosphorus:Watering	7	0.15	0.34
Kobresia	Blocks	5	0.12	0.1.

Note: Significance codes: ‘***’ = 0.001, ‘*’ = 0.05 and ‘.’ = 0.1, and p‐values significant at the p < 0.01 level are highlighted in bold.

FIGURE 2.

NMDS plots of soil fungal communities across the 48 treatment plots. (a) Fungal communities under all three plant species analysed together; (b–d) fungal communities in soil below Dryas, Salix and Kobresia, respectively, analysed individually. Environmental parameters correlated with the ordination axes (p < 0.05) are displayed as arrow vectors with length reflecting the degree of correlation. All treatments containing water addition are depicted as full symbols (either circles [Dryas], squares [Salix] or triangles [Kobresia]); those with no water addition are shown as hollow symbols. Stress in the NMDS analysis were for all species = 0.26, Dryas = 0.26, Salix = 0.29 and Kobresia = 0.25.

When fungal communities were analysed across all three plant species, the first ordination axis was particularly correlated negatively with C% and C/N ratio and positively with δ¹⁵N, but there were fewer correlations between environmental factors and the ordination axes when the communities were analysed separately for the three plant species (Figure 2b–d). The relative abundances of ectomycorrhizal and saprotrophic fungi, as well as their ratio, however, were related to the ordination axes both in the full analysis and the analyses per plant species (Figure 2). There were differences in relative abundances of fungal genera and species associated with the three plant species, but little indication of plant species specialisation in the fungal communities, as seen by large overlaps in fungal communities among plants in both the experimental plots (Figure 2a) and the reference soil and root samples (Figure S2a,b).

Ectomycorrhizal fungi were the most abundant guild in all soils (Figure 1, Figure S2d), and even more so in the roots and root tips (Figure S2c,d, Tables S3 and S4). There was a striking similarity between the fungal communities in roots and soils in the reference plots (Figure S2c,d), the main difference being that Mycena was disproportionately more abundant in the roots when compared to soils, particularly of Kobresia, and of Dryas to a lesser degree. However, there was a large variation in community composition among individual samples, for roots as well as for soils (Figures S3–S6). In the root tips, almost only ECM fungi were found (Tables S3 and S4). Other identifiable ecological groups (parasitic, lichenised, endophytic fungi) were present but constituted < 5% of the total sequences in the dataset (Figure S7).

3.3. Treatment Effects on Individual Fungal Groups, Genera and Species

Across experimental soil samples the plant species affected the proportion of saprotrophic and ectomycorrhizal fungi, as well as soil C/N ratio (Table 2c). However, when analyses were separated by plant species, there were very few effects of the treatments on relative abundances of broad ecological groups (ECM/saprotroph), fungal genera (Figure 3, Figure S8, Tables S5–S7) or species (Figures S9–S11). Under Kobresia, N addition increased the abundance of Cortinarius in soil, while Pseudotomentella was slightly inhibited, but otherwise, there was no main factor response of any of the most abundant fungal genera to the treatments (Figure 3). The Mantel tests identified no associations between fungal community dissimilarity and physical distance of the experimental plots (i.e., no consistent distance‐decay relationship), neither across all three plant species nor separately by plant species (p > 0.16 for all, data not shown). Similarly, a Mantel correlogram with 6–10 distance classes over all possible distances among the experimental plots showed no autocorrelation above ~100 cm (data not shown). With a typical distance of 120 cm between each sampling point, this suggests that autocorrelation in our dataset is mainly occurring within each test plot rather than among plots.

TABLE 2.

(a–c) Three‐way ANOVA tables with degrees of freedom and p‐values for the effects of all three treatments and their interactions on ectomycorrhizal (ECM) (a) and saprotrophic (b) sequence fractions, and (c) C/N ratios in soil of all 48 experimental plots. Block was used as a random factor, and the three plant species as a separate fixed factor without interactions. Notice that the plant species under which the soil was collected is the only consistently and strongly significant factor in the analyses. For this reason, we split the subsequent analyses of potential treatment effects of N, P and watering on soil biotic and geochemical parameters into three separate parts for each respective plant species under which the soil samples were collected.

(a) ECM relative abundance
Model factors	X 2	df	η 2	p
Nitrogen (N)	1.068	1	0.003	0.3014
Phosphorus (P)	1.301	1	0.023	0.254
Watering (W)	0.067	1	0.002	0.7958
Plant species	18.638	2	0.138	1e‐04***
N:P	1.881	1	0.012	0.1702
N:W	0.966	1	0.004	0.3257
P:W	0.202	1	0.012	0.6528
N:P:W	0.378	1	0.003	0.5389
Model R 2 (marginal) = 0.151

(b) Saprotropic relative abundance
Model factors	X 2	df	η 2	p
Nitrogen (N)	3.036	1	0.002	0.0814
Phosphorus (P)	2.076	1	0.038	0.1497
Watering (W)	0.551	1	0.006	0.4581
Plant species	7.899	2	0.67	0.0193*
N:P	1.536	1	0.019	0.2152
N:W	1.4	1	0.02	0.2367
P:W	0.388	1	0.008	0.5331
N:P:W	0.013	1	0	0.9105
Model R 2 (marginal) = 0.126

(c) Soil C/N ratio
Model factors	X 2	df	η 2	p
Nitrogen (N)	3.553	1	0.003	0.0594
Phosphorus (P)	2.545	1	0.058	0.1107
Watering (W)	1.155	1	0	0.2826
Plant species	32.813	2	0.23	7.50E‐08***
N:P	0.778	1	0.005	0.3777
N:W	6.862	1	0.028	0.0088
P:W	0.056	1	0.015	0.8123
N:P:W	3.065	1	0.026	0.08
Model R 2 (marginal) = 0.281

Note: Significance codes: ‘***’ = 0.001, ‘*’ = 0.05 and ‘.’ = 0.1, and p‐values significant at the p < 0.01 level are highlighted in bold.

FIGURE 3.

Sequence fractions of soil fungi below the three plant species in each of the eight treatment types. Ectomycorrhizal fungi (ECM) (a–c), saprotrophic fungi (d–f) and the four most abundant fungal genera, the ectomycorrhizal Cortinarius (g–i), Inocybe (j–l) and Tomentella (m–o); and the saprotrophic/endophytic genus Mycena (p–r). p‐values below 0.1 of treatments are shown in each panel (three‐way ANOVA tests with N, P and W as treatment factors, with interactions and block as random factor). Notice the differences in y‐axis scales between the two ecological groups (a–f) and the four individual genera (g–r).

3.4. Treatment Effects on the Soil Biogeochemistry and Microbial Biomass

Overall, the experimental treatments imposed no strong effects on soil biogeochemistry, apart from the clear positive effect of water addition on soil moisture (Figure S4, Tables S8–S10). The soil moisture levels were elevated from 9%–10% to 11%–12.5% (Figure S4s–u), corresponding to changes in the absolute available water content of ~20%–40% measured a week after watering.

However, the addition of P reduced soil C/N under Dryas (Figure 4) and there was a water × P addition interaction on C stock due to a positive effect by P alone, which was eliminated in the combined water and P addition (Figure S4). The soil C concentration under Kobresia increased by water addition, and N concentration also tended to increase, but C stocks were not significantly affected (Figure S4). There were no other biogeochemical responses to the treatments.

FIGURE 4.

Soil C/N ratios and bacterial/fungal biomass of soils (qPCR copies mg⁻¹ soil C) below the three plant species in each of the eight treatment types. (a–c) C/N ratios; (d–f) bacterial biomass; and (g–i) fungal biomass. Small numbers below each plots displays number of successful replicates taken. p‐values below 0.1 of treatments are shown in each panel (three‐way ANOVA tests with N, P and W as treatment factors with interactions, and block as random factor). Treatment effects were analysed with and without log‐transformation.

We observed no general responses of fungal and bacterial biomass in the soil to the treatments across all three plant species (Figure 4d–i). The addition of P, however, increased both bacterial and fungal biomass in the soil under Salix (Figure 4e,h), although N addition appeared to decrease the positive effect of P on bacterial abundance. Under Kobresia, P also increased fungal abundance, but here, water interacted negatively with the P effect (Figure 4i).

3.5. Plant Species Effects on Soil Fungi and Biogeochemistry in the Experimental Plots

The identity of the dominant plant species affected the relative abundances of fungal groups and genera, biochemistry and microbial biomass in the soil immediately below in the experimental plots (Figures 5 and 6, Figures S12 and S13). Dryas was associated with a higher fraction of ectomycorrhizal fungi than Kobresia, while Salix was intermediate (Figure 5a), driven particularly by Cortinarius (Figure 5c) and Pseudotomentella (Figure 5f). Dryas and Salix also promoted higher soil δ¹⁵N signatures than Kobresia (Figure 6b). Conversely, soil under Kobresia had a higher C/N ratio (Figure 6), C% and C stock (Figure S13), while the soil C% and C stock was the lowest under Salix, and intermediate under Dryas. The soil under Kobresia had a higher relative abundance of presumed saprotrophic fungi, particularly Mycena (Figure 5b,d). Soil δ¹³C was lowest under Dryas and highest under Salix, and fungal abundance was lowest under Salix and Kobresia and about three times higher under Dryas (Figure 6d).

FIGURE 5.

Sequence fractions of the same soil fungal groups (a, b) and genera (c–f) as in Figure 3, separated by the plant species under which they were collected, irrespective of treatments. Averages with common letters to the right of the point are not significantly different from each other at the p ≤ 0.05 level (Tukey test). Notice the difference in the y‐axes between the panels a–b and c–f.

FIGURE 6.

Stable isotope ratios (a, b), soil C/N ratios (c) and fungal biomass (d) in soil in the treatment plots, separated by the plant species under which they were collected, irrespective of treatments. Averages with common letters to the right of the point are not significantly different from each other at the p ≤ 0.05 level (Tukey test). Notice the four different y‐axes between the four panels.

3.6. Stable Isotope Signatures of Plant, Fungal and Soil Fractions From the Reference Plots

There are shifts in δ¹⁵N and δ¹³C signatures (Figure 7, Figures S14 and S15, Table S11) among plant fractions and soils associated with the three plant species in our six reference plots, and of the fungal fruit bodies collected in the area. The trends are comparable overall, although with notable differences between Dryas and Salix on one side, and Kobresia on the other: (1) For Dryas and Salix, δ¹⁵N and δ¹³C signatures increase in the sequence: leaves < roots < root tips < soil < ectomycorrhizal fruit bodies. For Kobresia, the sequence is also first leaves and roots, but then soil is lower than root tips and ectomycorrhizal fruit bodies. (2) Additionally, the slope of the increasing δ¹⁵N/δ¹³C is less steep for Dryas and Salix than for Kobresia, mainly due to a smaller increase in δ¹³C between leaves and soil for Kobresia (about 2‰) than for Dryas and Salix (about 4‰). There were significant differences between the δ¹⁵N/δ¹³C slope for Kobresia and that of both Dryas (p = 0.016) and Salix (p = 0.003), whereas there were no differences in slopes between Dryas and Salix (p = 0.61, Figure S15a–c).

FIGURE 7.

Stable isotope ratios (‰) from fungal carpophores, soil and plant material. The ectomycorrhizal fungi are relatively more depleted in ¹³C and less depleted in ¹⁵N than the saprotrophic fungi. Abbreviation explanations for fungal carpophores grouped by ≥ 98.5% sequence similarities: Agar = Agaricus; Cort = Cortinarius; Gal = Galerina; Heb = Hebeloma; Inoc = Inocybe; Lep = Lepista; Lyc = Lycoperdon; Malc = Mallocybe; Russ = Russula. For plant material grouped by morphology: CoaRo = Coarse roots; Deadbs = dead base; Deadrt = dead roots; Litt = litter. For root tips grouped first by plant species, and then by their dominant OTUs: Ce11 = Cenococcum (OTU11); Co52 = Cortinarius (OTU52); Co6 = Cortinarius (OTU6); He34 = Hebeloma (OTU34); In181 = Inocybe (OTU181); Oi253 = Oidiodendron (OTU253); Ps3 = Pseudotomentella (OTU3); Ru9 = Russula (OTU9); Se13 = Sebacina (OTU13); Se484 = Sebacina (OTU484); To15 = Tomentella (OTU15); To5 = Tomentella (OTU5). For the three Salix root tip data points shown (NN), no sequence/OTU data on their fungi was obtained, and they were instead separated by rough ectomycorrhizal mantle morphology. See also Figures S14 and S15 which include statistical analysis.

As expected, the ectomycorrhizal fruit bodies were relatively lower in δ¹³C and higher in δ¹⁵N than the saprotrophs. Cortinarius, Hebeloma and Inocybe dominated among the fungal fruit bodies. Despite their abundance in the soil and root communities, no Mycena fruit bodies were located in the timeframe of the field work. The one Galerina had an isotopic profile closely matching the ectomycorrhizal fruit bodies. All three plants were abundant in relatively close vicinity of most fruit bodies, and it was not possible to discern a reliable dominant plant association for the fruit bodies from our field observations.

4. Discussion

4.1. High‐Arctic Semi‐Desert Surprisingly Resistant Belowground

Overall, this study demonstrates a surprising belowground stability of a high‐arctic semi‐desert ecosystem dominated by ECM plants despite more than 25 years of doubled summer rainfall. There were few detectable effects of water and nutrient additions on fungal communities, microbial biomass and soil biogeochemical factors. The lack of correlations between fungal community turnover and physical distance across our spatial sampling scheme suggests that there are no abundant mycelial networks expanding across sample plots that could have buffered responses to the treatments.

Earlier studies at the same experimental site found a significant initial and medium‐term effect of watering on the soil microbial biomass after 2 (Illeris et al. 2003) and 13 years of water addition (Christiansen et al. 2012). After 25 years, we found no unequivocal effects of watering on our qPCR‐based microbial biomass measurements. The two former studies were based on chloroform fumigation‐extraction which may give a slightly broader picture of microbial biomass compared to qPCR, which in this case was specific to fungi. However, our results do not support a long‐term effect of the watering on microbial biomass or fungal community composition. Thus, as the experiment has now run for close to the 30 years, which is the characteristic time frame of the global circulation climate models (IPCC ²⁰²³), the conclusion that the fungal community and biogeochemical properties of this high‐arctic semi‐desert belowground ecosystem is surprisingly resistant to the expected summer precipitation changes in the 21st century must now be considered fairly robust.

However, warming in itself could certainly also lead to changes in High‐Arctic dry heaths, with both increase in saprotrophs (Leake et al. 2002) and/or an eventual increase in the plant cover of ectomycorrhizal shrubs over time (Elmendorf et al. 2012).

Furthermore, increased winter precipitation (snowfall) could be another future source of increased moisture in Zackenberg. Studies in the low‐arctic (Alaska) have found both a belowground compositional change in ectomycorrhizal fungal communities and an aboveground increase in shrub cover as a consequence of naturally and artificially increased snow cover of 100–150 cm (Morgado et al. 2016; Semenova et al. ²⁰¹⁶; Geml et al. ²⁰²¹). However, in Zackenberg, winter snow covers do not reach more than 80 cm at most (often less), shrubs are low or prostrate only, and the snow gets unevenly dispersed by wind and it is predominantly found in snow beds that constitute only 5%–10% of the area, compared to the 60%–65% cover of dry heath with little snow accumulated (GEM ^2024c). Furthermore, the simultaneous effect of warming from a uniformly distributed snow cover over the winter cannot be readily separated from the increased moisture effect when it melts in the summer (Morgado et al. 2016).

Of course, if the global warming continues to progress according to the projections, fundamental changes to the High Arctic will be inevitable. Summer temperatures in Zackenberg have already increased significantly since the start of monitoring in 1995 (GEM ^2024a). We only have quality‐assured precipitation data since 2014 (GEM ^2024b), with no clear increase in this short window of time, but increased precipitation is strongly expected to follow along summer warming. However, more research will be needed to establish how quickly this will show and which specific climate change‐associated factors that will have the most decisive impacts on the Northeast Greenlandic plant and soil communities. Future field manipulation experiments that can test warming and summer/winter precipitation, both separately and combined, would be desirable.

However, just as the increased warming changed the plant community but not the belowground soil bacterial processes in Lamb et al. (2011), our results suggest that increased summer precipitation and moderately increased soil nutrient levels will affect the aboveground plant community, but not the belowground fungal communities and biochemical processes, of the dry high‐arctic semi‐desert for the foreseeable future.

4.2. Plant Effects on Fungi and Soil Processes

The major differences we found in the fungal community composition and soil characteristics were linked to the plant species identity under which the soil had developed. While we found few fungal species unique to one or either of our three host plants, there were nevertheless plant species preferences in both root and soil fungal communities, as seen by the overabundance of Mycena with Kobresia, and of Cortinarius with Dryas. The fungal communities, the biogeochemical soil factors and the stable isotope patterns all suggest that particularly Kobresia affected soil processes differently than both Dryas and Salix. Under Kobresia, the higher abundance of saprotrophs, particularly Mycena, was matched with a higher soil C concentration and C/N ratio, suggesting that larger inputs of dead plant parts, such as through annual root and leaf turnover, may support a larger community of fungal saprotrophs with Kobresia. This is also consistent with the δ¹⁵N and δ¹³C signatures of soil under Kobresia which were intermediate to those in Kobresia plant parts and ECM root tips, while under both Dryas and Salix, soil δ¹⁵N and δ¹³C signatures were clearly higher than those of corresponding root tips and instead resembled those of ECM fruit bodies. This may suggest that the accumulated soil organic matter under Kobresia had a higher proportion of undecomposed plant material, whereas organic matter under Dryas and Salix could be derived to a larger extent from ECM mycelium (Clemmensen et al. 2013).

For all three plant species the δ¹³C of leaf litter collected at the soil surface was lower than δ¹³C of living leaves, likely related to remobilization (before senescence) or initial decomposition of a ¹³C‐enriched C pool, such as starch (Göttlicher et al. 2006). Leaves and leaf litter in contrast did not differ much in δ¹⁵N signatures for any of the species, which would be consistent with saprotrophic decomposition and N retention in the early litter decomposition phase (Clemmensen et al. 2024). Stable isotope patterns for Dryas and Salix, which had the highest proportion of ectomycorrhizal fungi in their surrounding soil, displayed parallel increases in δ¹³C and δ¹⁵N in a sequence from leaves then roots, root tips and finally their associated soil, just below the values for the ECM fruitbodies (which could not be assigned as belonging to a specific host). This is a typical pattern found for ECM plants (Lilleskov et al. 2011), which arises as sucrose becomes more ¹³C‐enriched as it is allocated from source‐ to sink tissues in the plant and further to the ECM root tips and fungal mycelia, thus creating a strong gradual aboveground‐belowground δ¹³C‐gradient (Hobbie and Horton 2007). ECM fungi on the other hand fractionate strongly against the heavier ¹⁵N isotope when allocating N to the plant partner, which creates the above‐belowground gradient in δ¹⁵N, where particularly the ectomycorrhizal fungal tissues get enriched in ¹⁵N (Högberg et al. 1994). The somewhat greater increase in δ¹⁵N, together with the less drastic increase in δ¹³C, from Kobresia leaves to roots and soil compared to Dryas and Salix, suggests that Kobresia fractionates C isotopes less than the other two plant species. This could mean that Kobresia, while benefitting from ectomycorrhizal N supply (i.e., gaining ¹⁵N‐depleted N), deliver less of their C to the ectomycorrhizal fungi (Högberg et al. 1999). Taken together with higher C content, C/N ratio and more saprotrophic fungi in soils associated with Kobresia, this suggests that soil processes under Kobresia are dominated by a larger accumulation of plant remains undergoing continuous saprotrophic decomposition, and that Kobresia may contribute less to mutualistic ectomycorrhizal relationships than Dryas and Salix. Similarly, the higher Cortinarius abundance associated with Dryas relative to the two other hosts could be explained by Dryas being a better C provider to its symbiotic fungi than both Salix and Kobresia. Genera with greater mycelial biomass, such as Cortinarius, tend to have greater C demand, greater enzymatic capabilities to access soil organic nitrogen, rhizomorphs for long‐distance transport, and higher δ¹⁵N in their sporocarps (Hobbie et al. 2024). Evidence suggests that, if the soils are sufficiently undisturbed, Cortinarius can turn over biomass so slowly (Jörgensen et al. 2025, 2023) that it attains high standing mycelial biomass in the process despite slow growth.

More broadly, these results are consistent with the importance of changes in species or functional group composition of the vegetation and related fungal communities for potential longer‐term changes in soil processes (Hannula et al. 2019; Koranda and Michelsen ²⁰²⁴). Our targeted sampling of soils beneath each of the dominant ectomycorrhizal plant species in the experimental units made us able to correct for the plant effect on the fungal community shifts at the plot level, as it allowed us to characterise plant–soil–fungal relations for each plant species separately, and the associated fungal responses to treatments. So far, it is clear that the dominant plant species is more important for the soil fungal communities and the soil biogeochemistry than the direct impact of increased precipitation and moderately increased nutrient availability in Northeast Greenland. Grasses/graminoids have a high growth rate and root turnover (Elmendorf et al. 2012; Schläpfer and Ryser ¹⁹⁹⁶), and eventual changes in fungal communities brought about by climate change is likely to be mediated through changes in the plant communities rather than directly by physical climate effects: as Kobresia increases in abundance as identified by Oldcorn et al. (2025), more leaves and root litter and soil organic matter from this species will be available, and thus saprotrophic fungi are likely to increase along with it.

4.3. Fungi in the High Arctic

Ectomycorrhizal fungal host plants on Svalbard have previously been found to have a low degree of host specificity (Botnen et al. 2014), as our results also support (Figures 1 and 3, Figure S2a–c). This indicates that Arctic fungi are particularly adaptable to a wide range of potential host plants. Furthermore, it is now generally accepted that the traditional perception that fungal ecological guilds were supposedly conserved at the genus level is not true, which is particularly important for two of the most common genera identified here, Cortinarius and Mycena. Certain members of the Cortinarius genus are known to not only form ectomycorrhiza but also be able to exhibit decomposer activity to access N locked in organic matter (Bödeker et al. 2014; Lindahl and Tunlid ²⁰¹⁵). Conversely, several species from the mainly saprotrophic genus Mycena are additionally able to invade roots and interact with living plant hosts (Harder et al. 2023); and other fungal genera with similar ecologies have shown quite different species‐specific responses to climate change (Geml et al. 2015). Across the boreal forest, low N availability and soil organic stocks were linked to higher abundances of a subset of Cortinarius species which are hypothesised to have retained the Mn peroxidase genes (Lindahl et al. 2021), and the most abundant Cortinarius species (OTU4) in this study has a 100% match to C. diasemospermus which is part of this subgroup (Bödeker et al. 2014). The higher abundance of Cortinarius that we observed with Dryas suggests an increased ectomycorrhizal mining for organically bound N in the soils, which could be a factor driving the higher soil C/N ratio and δ¹⁵N, as well as the relatively low C stocks, below Dryas.

The most abundant genus traditionally considered mainly saprotrophic, Mycena, is also known to display several different degrees of interactions when invading a plant host (Thoen et al. 2020). The dominance of Mycena inside the plant roots (particularly Kobresia) in our data suggests that these fungi are not uniformly saprotrophic in this environment either. While we found no Mycena fruit bodies for stable isotope analysis, the fact that they are present in roots but not in the root tips suggests that their lifestyle is of opportunistic invaders, lying low waiting for living roots to die off and turn into suitable substrate, as described by Parfitt et al. (2010). The fact that they are even more abundant in the roots compared to the surrounding soils suggests that they actively preferred the roots as a substrate. Specifically, some arctic Mycena species have been shown to have enormously increased genome sizes (Harder et al. 2024) that enable them to fit into multiple different niches and lifestyles. The most abundant Mycena species here, M. leptocephala, is such a root‐invasive species with a huge genome. Thus, if we had found that water addition significantly increased Mycena abundance over that of Cortinarius, we could not necessarily interpret this as a certain shift towards saprotrophy. The ecological versatility not only in host choice, but more broadly between degrading and root‐invasive lifestyles in this fungal genus serves to make them adaptable and resilient to environmental change.

Arctic studies on arbuscular mycorrhizal (Rasmussen et al. 2022) and other root‐associated fungi (Burg et al. 2024) also suggest that not only is plant community composition essential for shaping the fungal communities, but conversely, the reciprocal benefit of the root‐associated fungal community may also improve the arctic‐adapted plant species' competitiveness under stress. Parisy et al. (2024) found that ‘rewiring’ (i.e., changes in interaction patterns) among fungi and hosts are greatly affected by changing environments across latitudes on a Panarctic scale, but also suggested that the high degree of flexibility in partner choices among arctic host plants and soil fungi might provide a stabilising force. Our findings support this suggestion.

5. Conclusions and Perspectives

Overall, our findings suggest that the fungal community of the high‐arctic semi‐desert in Zackenberg is resistant against environmental changes. In terms of our three main hypotheses regarding the treatments, we found that precipitation did not increase the total fungal biomass (H1); neither precipitation nor increased N and P availability significantly changed the relative abundances of fungal guilds (H2); and there was accordingly no effect on soil C or C/N ratio of doubled precipitation (H3). Long‐term enhanced summer precipitation (and associated nutrient release) simulating the predicted future high‐arctic climate change does not appear directly to lead to changes in the soil fungal communities away from dominance of carbon‐sequestering ectomycorrhizal fungi towards increased decomposer fungal dominance.

Our results from this high‐arctic dry tundra should not be extrapolated to the entire Arctic. The very different climatic and edaphic conditions in the low Arctic with increasing shrubification (Kemppinen et al. 2021; Myers‐Smith and Hik ²⁰¹³), larger C stocks and higher nutrient availability (Michelsen et al. 1998) certainly give rise to different predictions and responses to the ongoing warming there. Other regions of the High Arctic are characterised by different dominant vegetation cover and precipitation patterns than the study area here. Nonetheless, as this type of arid arctic tundra in Zackenberg is indeed representative for a major part of the high‐arctic region, the fact that it is more resistant to increased precipitation than hitherto believed should be noted as promising. However, even in the optimistic end of the carbon mitigation scenarios, we are still expecting impacts of inevitable climate changes throughout the rest of the 21st century (IPCC ²⁰²³), also in the high‐arctic tundra biome. The only clear effects on fungal guilds and soil biogeochemistry we found were associated with plant species‐specific effects (H4). Thus, if the anthropogenic climate change trend proceeds beyond the 21st century, our results suggest that the changes in the soil fungal communities and biogeochemical processes will be mediated through major changes in the plant community in the High Arctic.

Author Contributions

A.M. and K.E.C. designed and conceptualised the study, and collected the field samples. A.M. analysed the C/N concentrations and stable isotopes, and C.B.H. and C.C. did the qPCR analyses. C.B.H. did all other laboratory work, the statistical analyses and graphics, and wrote the draft of the manuscript. All authors helped rewrite and finalise the manuscript.

Funding

This work was supported by Danmarks Frie Forskningsfond (DFF/FNU 2032‐00064B) and Novo Nordisk Foundation (NNF24OC0089849).

Ethics Statement

All research complies with applicable laws on sampling from natural populations and animal experimentation (including the ARRIVE guidelines).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Pictures of the sampling site, the plants and some of the fungal fruit bodies found.

Figure S2: Fungi in the plant root samples.

Figure S3: Bar chart of fungi in the individual soil experimental samples by genera.

Figure S4: Bar chart of fungi in the individual soil experimental samples by ecology.

Figure S5: Bar chart of fungi in the individual coarse root samples.

Figure S6: Bar chart of fungi in the individual reference soil samples.

Figure S7: Supplementary soil physical/chemical data from the experimental plots, by treatments.

Figure S8: Treatment effects on the remaining abundant soil fungal genera (not shown in Figure 3).

Figure S9: Most abundant OTUs in the experimental soil plots under Dryas octopetala.

Figure S10: Most abundant OTUs in the experimental soil plots under Salix arctica.

Figure S11: Most abundant OTUs in the experimental soil plots under Kobresia myosuroides.

Figure S12: Abundant fungal genera (not in Figure 5) in the experimental soil plots by plant species.

Figure S13: Supplementary soil physical/chemical properties by plant species.

Figure S14: Comparisons of δ¹⁵N and δ¹³C fractions between plant and fungus tissue types.

Figure S15: Scatterplots of δ¹⁵N and δ¹³C fractions with linear trends for various plant and fungus tissue types.

Table S1: Number of total soil samples in the experimental site.

Table S2: Overall summary of the sequences used for analysis.

Table S3: The 14 most common fungal genera in the total dataset (soil + roots).

Table S4: Most abundant fungal OTUs in root tip samples.

Table S5: Full three‐way ANOVA tests of the treatment effects upon the relative abundances of fungal groups in the samples from soil experimental plots, collected under Dryas octopetala.

Table S6: Full three‐way ANOVA tests of the treatment effects upon the relative abundances of fungal groups in the samples from soil experimental plots, collected under Salix arctica.

Table S7: Full three‐way ANOVA tests of the treatment effects upon the relative abundancesof fungal groups in the samples from soil experimental plots, collected under Kobresia myosuroides.

Table S8: Full three‐way ANOVA tests of the treatment effects upon the soil biogeochemistry data in the samples from the experimental plots collected under Dryas octopetala.

Table S9: Full three‐way ANOVA tests of the treatment effects upon the soil biogeochemistry data in the samples from the experimental plots collected under Salix arctica.

Table S10: Full three‐way ANOVA tests of the treatment effects upon the soil biogeochemistry data in the samples from the experimental plots collected under Kobresia myosuroides.

Table S11: Stable isotope data from soil, plant and fungal material.

Data Availability Statement

All data and R code generated for this manuscript can be accessed at FigShare: https://figshare.com/s/33a2492f03ea790a0691.