This research was conducted in Mount Revelstoke and Glacier National parks of Canada (research permit#: GLA-2019-32862) (Figure 1). This region is characterized as an interior rainforest; lower elevations are in the Interior Cedar-Hemlock biogeoclimatic zone, which transitions to the Engelmann Spruce-Subalpine Fir zone at higher elevations. WBP occurs at subalpine elevations in mixed forests, often as a sub-dominant species to subalpine fir (Abies lasiocarpa), Engelmann spruce (Picea engelmanni) and mountain hemlock (Tsuga mertensiana).

Sampling occurred over four sites—three sites where mature trees and naturally regenerated seedlings were sampled and one site where planted seedlings were sampled (Figure 1, coordinates Supplementary Table 2). All of our sites were shared with other monitoring projects that provided data on site characteristics, tree health and transects to sample from. The mature tree and naturally regenerated seedling sites were shared with a project that is monitoring blister rust in the Canadian Rocky and Columbia Mountain regions (Smith et al., 2008, 2013; Shepherd et al., 2018). In this project, 10 m wide belt transects were established at these sites and health assessments conducted every 5 years. The planted seedling site (Saint Cyr) is also a monitoring site that tracks seedlings from putatively blister rust resistant WBP lineages. Seedlings in that project were planted in line transects and separated by > 1 m.

Climate, soil and vegetation characteristics for the sampling sites are summarized in Supplementary Table 2. The sites are centered around 2,000 m elevation and on southerly aspects. Mean annual temperature is approximately 0°C and mean annual precipitation averages 2,000 mm. The soils are Brunisols and Podzols (detailed classifications in Supplementary Table 2).

The Saint Cyr planting site was burned by a mixed severity wildfire in 2003; approximately 50% of the area burned was classified as moderate severity with the remainder equally split between low and high severity (K. Macauley, Parks Canada, pers. comms., 18 March 2020). Two groups of WBP were planted 14- (fall, 2017) and 15-years (fall, 2018) post fire, respectively. Seedlings were 2-years old when received from the nursery and were 4 and 3 years-old, respectively, at the time of sampling. Naturally occurring WBP had not been recorded at the site (pre- or post-fire). In the surrounding area (within 1 km), WBP occurs as rare sporadic individuals. The site is considered an ideal planting location because it meets the physical habitat requirements of WBP and the burn provides abundant open space for planting. However, the site’s recent history of fire and lack of naturally occurring WBP distinguish it from the naturally occurring WBP sampling sites and have significant implications for our results.

Thirty mature trees, 13 naturally regenerated seedlings and 10 planted seedlings were sampled in total; the distribution of samples between sites is recorded in Supplementary Table 2. Mature trees and planted seedlings were randomly selected from transects associated with other monitoring projects (see above). Equal numbers of planted seedlings were sampled from the 2017 and 2018 planting groups present at the Saint Cyr site. Naturally regenerated seedlings were randomly sampled from just outside the associated mature tree transects to avoid disturbing other monitoring projects. Mature trees sampled were a minimum of ∼5 m apart.

WBP commonly grows in multi-stemmed clumps and the outplanting at the Saint Cyr site emulated that form. In mature tree sampling, multi-stemmed trees or multiple adjacent stems whose roots could not be separated, were treated as single individuals. Occasionally, naturally regenerated seedlings germinated as a clump and could not be sampled as individual stems in the field but could be separated in processing; this added three additional individuals to our original target sample size of 10 (final sample size = 13). All naturally regenerated seedling samples were processed as individual samples but only one seedling from each of the clumps sampled was retained in comparative statistical analyses. Outplanted seedling clumps could be sampled as single stems without damaging neighboring seedlings.

Mature tree sampling followed the methods in Birch et al. (2021). Subsamples of fine roots and soil were collected from three random locations around each tree and then pooled into a single fine root sample and a single soil sample per tree. Excavated roots were followed from the bole of the tree outward until fine roots with a visible connection to the parent root could be found and the terminal fine root mass was then collected. This process was repeated at two other random locations around each tree. A soil subsample was collected adjacent to each fine root mass subsample. The L layer was removed and soil collected with a trowel to a 15 cm depth (this sample included the FH layers and surface mineral horizons). Equal volumes of the three soil subsamples were then pooled into a single sample. A tree core was also taken from the bole of each mature tree (Haglöf Increment Borer). The tree core was collected at the root collar, or at the earliest point that the stem became vertical, on the uphill side. In multi-stemmed clumps, the stem with the largest diameter was cored. Seedlings were excavated and collected in entirety. For each seedling, a single soil sample was collected adjacent to the seedling roots. To reduce possible disease transmission between trees or sample cross contamination, sampling tools were cleaned and sprayed with 10% bleach or alcohol between each tree sampled.

Fine root samples were stored at 4°C and processed within 48 h. Fine roots were rinsed free of soil and debris in water and then inspected for ECMF root tips. In the seedling samples, the entire intact root masses were scanned using WinRHIZO prior to collecting root tip samples. Root tips were examined using a 4X magnification dissecting microscope to identify ECMF colonization. Cross sections of each distinct ECMF morphotype were further examined using a 10X–40X compound microscope to confirm the absence of root hairs, presence of a fungal mantle, and presence of a Hartig net. Fifty ECMF root tips were randomly excised per sample and placed in a 1.5 ml microtube; occasionally, seedling samples contained less than fifty ECMF root tips, and all ECMF root tips that were present were thus collected. Samples were frozen and stored at −30°C. To reduce DNA cross-contamination between samples, trays and tools used for rinsing root masses were cleaned and submerged in a 10% bleach solution for 5 min between samples.

Soil samples were immediately frozen at −30°C post-field sampling and processed within 45 days. Soil samples were thawed and sieved to a final sieve pore size of 2 mm. A volume of 7.5 ml of each sieved soil sample was transferred to a Falcon tube and then re-frozen at −30°C for storage until DNA extraction and sequencing. The remainder of each sample was oven dried at 50°C and then stored in paper bags for subsequent soil chemistry analyses.

Processed fine root tip samples and 7.5 ml soil samples were shipped together on dry ice to the Integrated Microbiome Resource (Dalhousie University, Halifax, NS) for DNA extraction and subsequent next-generation sequencing.

DNA extraction and next-generation sequencing were completed for each pooled fine root tip sample and soil sample to identify the ECMF tree and soil communities, respectively. DNA was extracted using the QIAGEN PowerSoil DNA Kit, following the manufacturer’s protocol. The Internal Transcribed Spacer 2 (ITS2) region of fungi was PCR amplified and then sequenced following a 1-step amplicon library preparation protocol (Comeau et al., 2017), using modified primers ITS86(F) (Turenne et al., 1999) and ITS4(R) (White et al., 1990) that contained both the Illumina MiSeq adaptor and barcodes. Amplicon libraries were returned in the format of 1.8 Cassava demultiplexed paired-end fastq sequences.

Bioinformatics were primarily carried out in QIIME 2 (Bolyen et al., 2019). Low quality fragments were trimmed using Trimmomatic (threshold: sliding window = 5 bp, quality score = 20) (Bolger et al., 2014) and then primers were removed using CutAdapt (Martin, 2011). The ITS2 region was extracted with ITSxpress (Rivers et al., 2018). DADA2 (Callahan et al., 2016) was used for denoising, dereplication, chimera removal and production of amplicon sequence variants (ASVs). Taxonomy was assigned to ASVs using a trained Naïve Bayes classifier sourced from the Integrated Microbiome Resource (Comeau et al., 2017); the classifier was developed from the UNITE database (version 8.0, current to 2 February 2019) (Nilsson et al., 2019) with a 99% similarity threshold applied to species hypotheses and covered the entire ITS region. ASVs were filtered with three successive criteria to create the final dataset: (i) ASV frequency was > 0.1% of the dataset’s mean sample depth (corresponding to contaminant abundance between Illumina MiSeq runs) (Comeau et al., 2017), (ii) ASV was classified to the level of genus or species, and (iii) ASV was assigned to a taxon that is a known ectomycorrhizal associate. Ectomycorrhizal status was assigned to taxa with the FUNGuild database (Nguyen et al., 2016a). Taxa not described or belonging to multiple trophic guilds in FUNGuild were assigned ectomycorrhizal status manually; we relied on Tedersoo et al. (2010) as the authoritative source for ectomycorrhizal status. The main group requiring manual assignment was the Helotiales; only taxa from this group that were definitively described as ectomycorrhizal in Tedersoo et al. (2010) were retained in the dataset.

R (version 3.6.2) (R Core Team, 2019) was used for all data analyses. We relied primarily on the packages phyloseq (for microbiome data analyses) (McMurdie and Holmes, 2013) and ggplot (for graphics) (Wickham, 2016). Tree and soil communities were separated in analyses; within these communities, samples were typically aggregated by tree type (e.g., alpha diversity of ECMF from root tip samples (tree community) of mature trees). Prior to analyses, the raw dataset created in QIIME2 was rarefied to a library size of 1,250 reads (determined from the saturation point in number of reads vs. observed ASV curves); we did not discard samples below this library size because library size was interpreted to be an indicator of the abundance of ECMF tissue in a sample and not an artifact of inherent variability in amplification (see section “Amplification Levels”).

To describe the abundance of lineages and the composition of communities, we used the abundance metrics of number of occurrences and relative abundance; relative abundance was defined as the proportion of reads of a given ASV, species or genus in a sample. To investigate differences between tree types, we compared per-sample sequence counts, alpha-diversity and community composition. Alpha-diversity was estimated at the ASV level with Shannon’s diversity metric (H′=-∑i=1Rpiln⁡pi; p_i is the proportion of reads belonging to the ith ASV in a given sample). To test for differences in sequence count and alpha-diversity between tree types, we used ANOVA and pairwise t-tests or the non-parametric alternative Kruskal-Wallis and Dunn tests; all of these tests were housed in the rstatix package (Kassambara, 2020).

To quantify compositional differences, we used the Bray-Curtis dissimilarity metric (BCij=1-2CijSi+Sj; C_ij is the sum of the shared ASVs between two samples and S_i and S_j are the total number of ASVs in the samples i and j, respectively) (Bray and Curtis, 1957) and Principal Coordinate Analysis (PCoA). To better resolve compositional differences, we agglomerated ASVs to the level of genus before running these analyses. The first two axes of PCoA were used to construct ordination plots. Genera were added to ordination plots with weighted taxa scores to identify marker genera (vegan function wascores); to reduce plot clutter, only the genera making up 90% of the relative abundance in each of the tree types were plotted. Adonis and ANOSIM tests were used to test the significance of compositional differences between tree types; estimates of beta dispersion (vegan function betadipser) or built-in diagnostic plots were used to check for location vs. dispersion effects in these tests.

To test the effects of soil (pH, %C, %N, C:N) and tree (age, canopy kill) variables on ECMF composition, we used vector fitting to the first two axes of PCoA ordination (vegan function envfit). Samples associated with planted seedlings were removed from the dataset to isolate naturally occurring patterns. Tree variables were only tested as predictors in the tree community. Ordinations were created following the method described above. For all ordination-based methods, distance matrices and PCoA analyses were done in phyloseq (McMurdie and Holmes, 2013) and statistical testing was done in vegan (Oksanen et al., 2019).

Raw sequencing data contained 107 samples with a mean sample depth (sequences/sample) of 58,605. This data is available on NCBI’s Sequence Read Archive (PRJNA722685). Standard processing (isolating ITS2, denoising, dereplication, removal of chimeras) rendered 5,409 ASVs across 107 samples with a mean sample depth of 45,760. Taxonomic filtering to ECMF taxa identified at the genus or species level left 353 ASVs across 106 samples with mean sample depth of 21,242. Full summary statistics from the bioinformatic workflow broken down by sample type are in Supplementary Table 4. Representative sequences for each ASV are available on GenBank; the accession numbers for these sequences are in Supplementary Table 5.

The set of confirmed WBP ectomycorrhizal associates reported here as the tree community is the core contribution of this research. These ECMF were identified using next-generation sequencing and barcoding from true ectomycorrhizae samples. An important caveat is that identification of sequences in “meta-barcoding” (Somervuo et al., 2017) approaches, like that applied here, can often be wrong. This is especially true at the level of species and for some groups (e.g., Cortinarius is not well resolved by ITS2). Sequence errors, intraspecific variation and errors/limited coverage in the reference database can pose problems to correct identification (Somervuo et al., 2017; Lücking et al., 2020). The taxa reported here should be understood as “first diagnoses” (Lücking et al., 2020). Still, this dataset combined with previous research on the WBP ECMF community forms a profile of the reported associates of WBP with reasonable coverage across the WBP range; this list is compiled as Supplementary Table 1.

Sixty-seven species and twenty-six genera were identified in the tree community; an additional, non-overlapping, twenty species and nine genera were identified from the paired soil sampling. We carried out a qualitative confidence assessment of the taxa in the tree community based on frequency of occurrence and whether the taxa had been reported previously as a WBP associate (Supplementary Table 12). This left thirty-three species and twenty-one genera that we confirm as WBP associates with high or medium confidence; of these, twenty-three species and three genera had not been reported as WBP associates before.

Generalist ascomycetes, Cenococcum and Meliniomyces (=Hyaloscypha), were dominant and widespread. The Atheliales (generalists: Piloderma, Amphinema, and Tylospora) were also a dominant clade. A diverse spread of taxa occurred at medium abundances and frequencies—generalists: Wilcoxina, Hydnotrya, Pseudotomentella, Tomentella, Sistotrema and Endogone (=Jimgerdemannia) and high-elevation conifer associates: species of Cortinarius, Lactarius and Russula. Suilloids (Rhizopogon and Suillus) were, as reported elsewhere, a regular and important part of the community.

Helotian taxa were abundantly co-amplified in our samples. These taxa are known as secondary root associates and they are commonly co-amplified from ectomycorrhizae samples; the nutritional modes of these taxa are, however, debated and only a few have been proven truly ectomycorrhizal (Tedersoo et al., 2009). In reviewing the ectomycorrhizal status of these taxa following Tedersoo et al. (2010), we removed all except for the lineages Acephala macrosclerotiorum and Meliniomyces bicolor (=Hyaloscypha bicolor). Even if non-ectomycorrhizal, the presence of these taxa is worth reporting because they form a significant part of the WBP rhizosphere and have been linked elsewhere with poor stand health (discussed below).

The community documented here diverges from previous descriptions of the WBP ECMF community. There is a functional shift toward generalist and high-elevation conifer associates and away from Suilloids. The dominance of the wood inhabiting Atheliales clade (Piloderma, Amphinema and Tylospora) is particularly distinctive. The shift we observed is likely the result of sampling mixed forests here vs. relatively pure WPB forests in other studies. Host plant identity is a primary determinant of ECMF composition (van der Linde et al., 2018). In our study region, WBP is subdominant to subalpine fir, Engelmann spruce and mountain hemlock, where the Atheliales have previously been shown to be a dominant ectomycorrhizal guild (Hagerman et al., 1999; Walker et al., 2014). This contrasts with the other regions where the WBP ECMF community has been sampled, where pine was the dominant host. In addition, Piloderma can access organic nitrogen pools (Heinonsalo et al., 2015) and this may be advantageous in soils, like those at our study sites, that have accumulations of recalcitrant litter.

This shift in dominant guilds could also reflect our sampling approach. Sampling was biased toward roots that were in organic soil horizons and close to the bole of the tree, potentially selecting for wood-inhabiting fungi. Talbot et al. (2014), in contrast, sampled from both organic and underlying mineral horizons in WBP forests in the Yosemite region; they reported no analogous shift, but structuring between these layers has been documented in other high-elevation pine systems (Koizumi et al., 2018). Different survey methods (fruiting body surveys vs. next-generation sequencing and metabarcoding) could also amplify this functional shift because it is driven by resupinate and hypogenous taxa, but the shift is present in comparisons with studies that employ similar methods (Talbot et al., 2014; Glassman et al., 2015, 2017a).

The second mode by which our community diverges from communities described in other parts of the species’ range is compositional. At the species level, the communities diverge, whereas they align at the genus level and higher. Of the thirty-three species and twenty-one genera identified with medium or high confidence, 70% (23/33) of the species are newly described associates of WBP but only 14% (3/21) of the genera are new. This indicates there is regional endemism, as suggested by others (Talbot et al., 2014; Rosinger et al., 2018; Defrenne et al., 2019b). Talbot et al. (2014) documented the soil ECMF community across continental North America in the Pinaceae and emphasize endemism resulting from dispersal limitation as a primary determinant of ECMF composition. Rosinger et al. (2018) and Defrenne et al. (2019b) both provide evidence for endemic structuring of ECMF within the range of single host species but also join a growing body of evidence that identifies environmental variables (e.g., climate, edaphic properties and nutrient regimes) as determinants of ECMF composition (reviewed in van der Linde et al., 2018). Our study region is climatically distinct from other regions where sampling has occurred and environmental variables, in particular mean annual precipitation (Defrenne et al., 2019b), may contribute to this compositional divergence.

Our data also raises unresolved questions in the pine-specific Suilloid group. Suillus americanus (syn. S. sibiricus), S. discolor and S. subalpinus are five-needled pine or whitebark pine specialists with broad distributions (Cripps and Antibus, 2011; Nguyen et al., 2016b). None of these species were recorded in our data. The lineage we identified as S. punctatipes (=Boletinus punctatipes) sits in the Suillus placidus/subalpinus/anomalus species complex but is molecularly (with ITS2) and/or morphologically distinct from the related species in the complex (Nguyen et al., 2016b). S. americanus, S. discolor and S. subalpinus being absent from our data is surprising because these taxa do not appear dispersal limited, and greater congruence is expected among host specialists. S. americanus has been recorded with western white pine (Pinus monticola) in our study area and with WBP in Banff National Park (C. Cripps, pers. comms., 6 May 2021), which is a good indication that processes other than dispersal limitation are restricting it from our site. S. brevipes is a second Suillus taxa recorded in our data; finding it here is surprising because it is part of the well-established/luteus clade that is near completely restricted to hard-pine hosts, whereas WBP is a soft-pine (Nguyen et al., 2016b). S. brevipes has previously been reported from soils of WBP stands in the Yosemite region (Glassman et al., 2017a) and this would represent a second rare case of “host jumping” in that clade (Nguyen et al., 2016b).

The common errors in assigning taxonomy with meta-barcoding techniques (noted above) (Somervuo et al., 2017; Lücking et al., 2020) may contribute to the questions raised in Suillus and more generally to the compositional divergence with previous studies. Re-evaluation of our data with alternative approaches to classification [e.g., phylogenetic placement of reads in reference trees or probabilistic methods (PROTAX)] may increase confidence in species and genus names assigned to sequences (Somervuo et al., 2016; Lücking et al., 2020) and aid comparison with other descriptions of the WBP ECMF community.

The ECMF community on planted seedlings was clearly distinct from those on natural tree types (mature trees and naturally regenerated seedlings). Planted seedlings had fewer colonized root tips, significantly lower per-sample sequence counts, lower (though not significantly) diversity and were compositionally distinct; the relative abundance of the generalist Meliniomyces (=Hyaloscypha) was increased and numerous taxa prominent on natural tree types (e.g., Piloderma and Cortinarius) were reduced or absent. The most apparent difference, however, was observed visually when processing fine root samples. Planted seedlings had a completely distinct root structure, with far more fine-root mass than naturally regenerated seedlings. Moreover, with the exception of a single dominant morphotype that we expect is Meliniomyces (=Hyaloscypha), planted seedlings were visually non-mycorrhizal (though Russula, Suillus, and Cenococcum were identified at medium abundances through sequencing). The prolific fine root mass of planted seedlings may be an artifact of root development and fertilizer regimes in a nursery upbringing or it may be a foraging response to the lack of ectomycorrhizal partners that are the dominant source of nutrients (Defrenne et al., 2019a).

Differences in the ECMF community between the soils of the planting site and soils where naturally occurring trees were sampled were even more apparent than the differences observed from the mycorrhizal communities on tree roots. Largely, our results indicate that soils in proximity to planted seedlings did not contain ECMF tissue with DNA that could be amplified at this sequencing depth. Per-sample sequence counts and alpha diversity were significantly lower in soil from the planting site compared to soil associated with natural tree types, and Meliniomyces (=Hyaloscypha) was the single dominant genus.

Together these results reveal that the ECMF community occurring on planted seedlings did not match naturally occurring communities, nor did the soils of the planting site contain the appropriate inoculum to provide that community. This has not yet effected seedling survival (as of 2019, seedling survival was 82.4 and 94.2% for the 2017 and 2018 plantings, respectively) (Skinner et al., 2019) but we expect it will in time. High survival rates immediately after planting are common (Jenkins, 2017) and even obligately mycorrhizal tree seedlings can survive as non-mycorrhizal for a lag period (Collier and Bidartondo, 2009). Monitoring over 5–10 years will help isolate the effect of mycorrhizal status on WBP planted seedling survival (Jenkins, 2017). Note that our sample size for seedlings (natural and planted) was relatively small because of restrictions on the destructive sampling of a species protected by the Canadian Species at Risk Act and comparisons against mature trees were unbalanced. But the consistent and visibly apparent nature of the results observed at the planting site support the conclusions we make.

These differences in ECMF are largely attributable to the planting site. The planting site was located within the footprint of a 2003 mixed severity wildfire, largely lacks an ectomycorrhizal plant community and is not a naturally occurring WBP forest (presently or historically). The effect of fire on ECMF communities is an active field of research that has variable consensus. There is good evidence that fire shifts ECMF composition for long (∼50 year) timeframes and moderate evidence that fire reduces diversity and inoculum potential (measured with colonization) over decadal timeframes (∼22 and 11 years, respectively) (Karst et al., 2014; Dove and Hart, 2017; Taudière et al., 2017). We also know there is a suite of fungi, termed the “ectomycorrhizal fungal sporebank,” that is separate and compositionally distinct from the active ECMF community and exists in the soil as resistant propagules that colonize regenerating plant hosts post-disturbance (Glassman et al., 2015, 2016); in WBP forests, this community is dominated by Rhizopogon, Cenococcum, Wilcoxina and Thelephora (Glassman et al., 2015). ECMF sporebanks have been shown to persist through wildfires (Glassman et al., 2016) and it is likely this sporebank that WBP seedlings outplanted on burns are reliant on. The longevity of ECMF sporebanks is not well assessed but existing research is focused on pine associated taxa (predominantly Rhizopogon) and predicts that these sporebanks will remain viable for decades (Bruns et al., 2009; Nguyen et al., 2012); these two studies are first reports from a century long experiment that will contribute valuable knowledge to this discussion.

Paired with the influence of fire, is the importance of established trees as sources of inoculum for regenerating seedlings (Haskins and Gehring, 2005; Hubert and Gehring, 2008; Teste et al., 2009; Mueller et al., 2019; Simard et al., 2021). Established trees can provide inoculum directly [from their root system at immediate spatial scales (∼5 m)] (Teste et al., 2009) or indirectly [nearby stands that host ECMF that disperse to regenerating seedlings at local scales (∼1 km)] (Trusty and Cripps, 2011; Policelli et al., 2019). The species identity of the source tree can alter the community composition of the inoculum (Hubert and Gehring, 2008); conspecific or congeneric source trees are likely especially important for WBP because these trees will host Suilloid taxa. Even dead trees can provide a “legacy” of ECMF inoculum; prior to tree death, ECMF spore banks may form that persist to colonize successive generations (Mueller et al., 2019; Shemesh et al., 2019).

The combination of fire and lack of established ectomycorrhizal trees help explain the paucity of ECMF at our planting site. Fire would have reduced ECMF inoculum potential directly by burning live ECMF and reducing the density of the soil spore bank, and indirectly by killing ectomycorrhizal host plants. The absence of inoculation from a persistent, fire resistant soil spore bank is more puzzling. The significant lag time (14 or 15 years) between fire and planting may have led to the degeneration of the spore bank (contrary to viability estimates in Bruns et al., 2009; Nguyen et al., 2012); our knowledge of ECMF soil spore bank longevity is small and this process may well be modified in harsh high-elevation habitats. The lack of significant WBP stands at or near the planting site is another important variable; the nearest trees to the site are sporadic, isolated WBP ∼1 km away. These stands would have hosted Rhizopogon species that contribute to soil spore banks and facilitated colonization of seedlings by spore dispersal from outside the fire perimeter.

This interpretation aligns with (and builds upon) the only other study that has sampled ECMF from outplanted seedlings (Trusty and Cripps, 2011). Their results contrast with ours and document successful integration of planted seedlings with naturally occurring ECMF. They identify two key factors in this success: (1) seedlings were planted 1 year post burn and could access a viable soil spore bank, and (2) unburned WBP stands were present outside the fire perimeter that could provide Suilloid propagules (Trusty and Cripps, 2011). These same two factors likely explain the lack of integration of planted seedlings and native ECMF in our results.

No successional patterns were apparent in our data; we detected no compositional differences between mature trees and naturally regenerated seedlings, and tree age was not a significant predictor of ECMF composition. Similarly, none of the edaphic variables (pH, %C, %N, C:N) were significant predictors of ECMF composition. These results are surprising because these variables have been shown to be determinants of ECMF composition elsewhere (Twieg et al., 2009; Cripps and Antibus, 2011; Tedersoo et al., 2012; Glassman et al., 2017b; Koizumi et al., 2018; van der Linde et al., 2018; Defrenne et al., 2019b). In particular, Glassman et al. (2017b) and Koizumi et al. (2018) are studies of subalpine pine systems (WBP-Pinus contorta and Pinus pumila communities, respectively) at similar spatial scales that have good comparative potential. Both studies reported soil variables (dominantly pH) and tree age or life stage as strong drivers of ECMF community composition.

The continuous, mixed age nature of the forests we sampled may well explain why tree age was not significant. A shared feature of Glassman et al. (2017b) and Koizumi et al. (2018) is that individual trees were spatially separated (see Glassman et al., 2017a for an application of island biogeography to their system). Spatial separation requires seedlings to rely on soil spore bank ECMF and allows succession to occur on small spatial scales. The trees we sampled were in a continuous forest matrix where root systems of neighboring trees of different age classes overlapped and a common mycorrhizal network is likely present (Simard et al., 2012); this might promote homogeneity in ECMF communities and dampen or obscure successional processes.

Limitations in methodology may also contribute to not detecting significance in soil variables and tree age. We did not use multivariate models explicitly designed to account for the influence of multiple predictor variables simultaneously (e.g., Glassman et al., 2017b) and instead relied on simple vector fitting to ordinations. In soil sampling we did not separate organic and mineral horizons and Koizumi et al. (2018) stress that the significance of soil variables in their study was driven by stratification of ECMF in these layers. In addition, twelve soil samples were removed from analyses because of an error in sample processing, which reduced statistical power, and a second processing error allowed samples to be stored wet before air drying, which could have altered their properties.

Canopy kill was a significant predictor of ECMF composition in the tree community; diseased trees hosted simpler communities with compositional differences driven by increased abundances of Cenococcum and Amphinema. Shifts in ECMF composition in response to tree stress (drought, herbivory, parasitism) are well documented in pine systems and tend to move toward ascomycete and low-carbon demanding fungi (Karst et al., 2014). This generally aligns with our observations and might suggest Amphinema has low carbon requirements.

Recent evidence linking mountain pine beetle (MPB) outbreaks to ECMF (Treu et al., 2014; Karst et al., 2015) has shaped our interpretation of these results. Treu et al. (2014) showed that ECMF composition was altered and abundance reduced in MPB-affected stands. Karst et al. (2015) showed that these changes in ECMF mediated transgenerational effects that reduced seedling survival and the capacity to generate defense compounds. Helotian taxa were identified as a dominant feature of MPB affected stands (Karst et al., 2015). These taxa were abundant in our samples, which may signal ECMF are already responding to stressors and mortality of WBP in our region.

The link between tree health and ECMF is important because it recognizes that ECMF suffer from the decline of their host (Karst et al., 2014) and that the resulting changes in ECMF can influence the success of regeneration and the capacity of the host to resist stressors and pathogens (Karst et al., 2015). There is no reason to assume that this link does not apply to WBP as well and our results are a first indication that it does. These changes in ECMF are expected to disproportionately affect pine-specific Suilloid taxa and some are at risk of being lost from the landscape altogether (Cripps and Antibus, 2011; Treu et al., 2014; Osmundson, 2016). In addition, this link highlights the importance ECMF in WBP ecology. Some ecological or organismal processes that are critical to species recovery (e.g., production of insect defense compounds) may be mediated by ECMF.