Mycena species can be opportunist-generalist plant root invaders

Traditional strict separation of fungi into ecological niches as mutualist, parasite


INTRODUCTION
Among ecologists, a consensus is emerging that the classical assignment of fungal species into single trophic groups, including mycorrhizal, saprotrophic or pathogenic lifestyles, may be too simplistic (Baldrian & Kohout, 2017;Selosse et al., 2018).Some otherwise free-living fungi can invade plant roots and exist as either asymptomatic endophytes (where no harm or benefit is apparent) and then switch from endophytic into becoming pathogenic (e.g., Fusarium graminearum, Löfgren et al., 2018), forming orchid mycorrhizas (Piriformospora indica, Weiß et al., 2016), or forming ericoid mycorrhizal associations (Meliniomyces spp., Martino et al., 2018).Numerous genera of saprotrophic fungi have been screened for their ability to colonize Pinus sylvestris and Picea abies seedling roots in vitro (Smith et al., 2017).Several genera, including Mycena, Gymnopus, Phlebiopsis, Marasmius or Pleurotus, invaded roots apparently without decomposing dead tissue in the process.However, beyond the invasion, the nature of the interactions with the plant host (if any) remains unknown.
For our target genus Mycena, several lines of direct and indirect evidence point to their ability to invade and interact with living plant roots, at least in vitro.Mycena is one the largest genera in Agaricales (over 500 species), widespread across habitats and climate zones (Kühner, 1938;Maas Geesteranus, 1992;Rexer, 1994;Robich, 2003;Aronsen & Laessøe, 2016).Mycena species have been identified as potential orchid mycorrhizal symbionts (Guo et al., 1997;Martos et al., 2009;Ogura-Tsujita et al., 2009;Selosse et al., 2010;Zhang et al., 2012) or endophytes in photosynthetic moss tissue (Davey et al., 2013) and non-mycorrhizal brassicaceous plants (Glynou et al., 2018).They have also formed mycorrhiza-like structures in the roots of Vaccinium corymbosum in growth studies (Grelet et al., 2017).Recently, Thoen et al. (2020) showed that multiple species and individual strains of Mycena could colonize roots of Betula pendula seedlings in vitro, and formed a gradient of interactions from harmful to neutral to beneficial, with some species or strains able to transfer nutrients to the plant host.This is significant, as prior to this, Mycena sensu stricto (Moncalvo et al., 2002) (henceforth simply 'Mycena') was known primarily as quantitatively important litter and wood debris decomposers (Baldrian et al., 2012;Boberg et al., 2008;Kyaschenko et al., 2017).The spectrum of interactions seen by Thoen et al. (2020) is further noteworthy in light of the "waiting room hypothesis" (Selosse et al., 2009;van der Heijden et al., 2015) on mycorrhizal evolution, which suggests that the mycorrhizal habit evolves from saprotrophs gradually via neutral endophytic intermediate states.This hypothesis has remained controversial even though it is accepted that the mycorrhizal habit has evolved on numerous, independent occasions from saprotrophic ancestors (Kohler et al., 2015;Tedersoo & Smith, 2013).Thus, the genus Mycena may represent a promising research model for studying both ecological versatility in fungi and the possible ongoing evolution of fungi traditionally believed to be purely saprotrophic en route to developing mycorrhizal abilities.
Most evidence for trophic versatility in Mycena originates from in vitro studies, and it is uncertain to what extent this translates to the field.To investigate the trophic mode of fungi in natural environments, analysis of the isotope ratios of 13 C: 12 C and 15 N: 14 N (expressed as δ 13 C and δ 15 N values relative to known standards) can be applied directly to fungal carpophores and other field material.Ectomycorrhizal fungi are generally more enriched in 15 N and depleted in 13 C than saprotrophic fungi (Griffith, 2004;Hobbie et al., 1999;Hobbie et al., 2001;Kohzu et al., 1999;Mayor et al., 2009;Taylor et al., 1997).By comparing against fungi of known trophic status, the natural abundance of δ 15 N and δ 13 C can also give strong indications of the nutritional mode of fungi with unknown trophic status.Thus, Halbwachs et al. (2018) recently used this approach to strongly suggest that Hygrocybe, another genus traditionally believed to be saprotrophic, was most likely biotrophic with plants.
The occurrence and abundance of Mycena sequences retrieved from wild plant roots also suggest interactions with plant roots in situ.Such reports are particularly common from Arctic plants, including Bistorta vivipara, Cassiope tetragona, Dryas octopetala and Salix polaris (Blaalid et al., 2014;Botnen et al., 2014;Lorberau et al., 2017).In some cases, Mycena sequences comprised >30%-50% of the total reads, suggesting that they are not simply casual colonizers.The harsh and oligotrophic Arctic environments may stimulate otherwise free-living fungal genera (including Mycena) to explore new ecological niches (Botnen et al., 2014;Jumpponen & Trappe, 1998;Lorberau et al., 2017;Ryberg et al., 2009;Ryberg et al., 2011;Timling et al., 2012).Nevertheless, Mycena reads have also been recovered in high quantities from inside living Picea abies roots in temperate environments (Kohout et al., 2018).Overall, however, current information on the occurrence of Mycena and other saprotrophs in roots is unsystematic and too scattered to identify any clear patterns of their occurrence and abundance.
High throughput sequencing (HTS) studies of fungal communities in plant roots generally target mycorrhizal fungi (Bahram et al., 2011;Blaalid et al., 2014;Buee et al., 2009;Kaur et al., 2019;Tedersoo et al., 2010;Vasar et al., 2017).The workflow requires annotating hundreds or thousands of operational taxonomic units (OTUs)/clusters of fungi into ecological guilds.The genus level is traditionally considered the most relevant level for separating fungal taxa by nutritional mode (den Bakker et al., 2004;Fries & Mueller, 1984;Garnica et al., 2016;Molina & Trappe, 1994;Tedersoo & Smith, 2013).Although it is widely accepted that evolution is a continuous process, most studies still favour those conservative taxonomic ecological generalizations, often for practical reasons.Thus, modern ecological annotation software for amplicon analyses (Nguyen et al., 2016, Pölme et al., 2020) remains largely based on this view.In studies of mycorrhizal fungi in roots, fungal genera identified as being saprotrophic are therefore often at best briefly mentioned, reported as one lumped ecological group (Menkis et al., 2012;Tedersoo & Smith, 2013), or simply dismissed as accidental contamination (Liao et al., 2014).Thus, large quantities of potentially informative data are often ignored, leading to a potentially oversimplified view of root ecology and interactions.
Here, we analyse the occurrence of Mycena in the roots of wild plants in a range of ecosystems and investigate the potential trophic versatility of Mycena in the field by presenting: (1) a systematic analysis of data from 10 plant species from Arctic and temperate regions from previously published and newly generated ITS1/ITS2 HTS data sets from living plant roots, and (2) a comparison of the natural abundance of 13  We investigated four main research questions: (1) Are Mycena species (and other supposedly saprotrophic genera) systematically found associated with living plant roots in significant quantities?
(2) Do the data from 13C and 15N abundances support the view that Mycena species may form mutualist associations with plants?
(3) Are there indications of host preferences or specificity among invasive Mycena species?
(4) Are root invasions by Mycena or saprotrophic fungi more prevalent in arctic and alpine environments?

Analysis of ITS metabarcoding datasets from ectomycorrhizal plant roots
We data-mined our set of ectomycorrhizal plant host species systematically for Mycena presence in their roots.Our plant host sample consisted of 10 species in total, 3 amplified with ITS1 primers, and 7 with ITS2 primers from both previously published studies and new datasets (Table 1).We separated data into both OTUs (at the 3% threshold, retaining only those with ≥10 sequences) and amplicon sequence variants (ASVs) and investigated further OTUs/ASVs identified as Mycena (Table 2).As there were no major differences between the finer ASV-('haplotype') scale or at the coarser 3% OTU scale (Tables S8, S9), we focused on the OTU datasets in further analyses.
After quality sorting, a final dataset of 889,290 ITS1 sequences was clustered into 1193 OTUs for the ITS1 dataset and 992,890 sequences into 1032 OTUs in the ITS2 dataset (Table 1).For a detailed list of the sequence quality sorting steps on the ITS1 and ITS2 data and the respective counts for each host species, see Supplementary data and Tables S1-S6.Applying the 'coverage/completeness' method for assessing saturation (Chao & Jost, 2012), 111 samples failed to meet the 97% coverage cutoff value and were discarded.Although the iNEXT (Hsieh et al., 2016) extrapolations of observed species richness suggested that some slight undersampling remained in some samples (Figure S5), none of the 10 species showed correlations between Mycena infection levels (all p values >0.05,Table S7) with sampling depth.
For ITS1, 606 of 1193 OTUs (78.3% of the sequences) could be identified to genus level by SIN-TAX at the threshold of BPP >0.6; for ITS2, this number was 513 of 1032 ITS2 OTUs (84.5% of sequences).
The same SINTAX classification identified 13 Mycena OTUs (1.5% of all ITS1 sequences) in the ITS1 dataset and 14 Mycena OTUs in ITS2 (12.6% of all ITS2 sequences).In a second identification step, especially targeting Mycena, representative sequences of all OTUs were clustered together with our constructed Mycena ITS database of 576 sequences.This analysis revealed additional 7 ITS1 OTUs and 7 ITS2 OTUs not identified as Mycena by SINTAX at BPP >0.6.Thus, in total 20 ITS1 OTUs (2.1% of all sequences) and 21 ITS2 OTUs (15.8% of all sequences) could be identified as Mycena (s.str) with these two combined methods.
Of other (non-Mycena) taxa traditionally considered to be saprotrophic or endophytic, we found Sebacinales in the four Arctic host plants, the zygomycete Mortierella in B. pubescens, A. uva-ursi and A. alpinus (Scottish hosts), and Phialocephala in B. vivipara and Clavulinopsis/Clavaria in C. tetragona (Figure 2A-C).No other saprotrophic or endophytic genera made up more than 0.5% of the sequences in any of the host plants.The high and very variable infection levels and frequency patterns of Mycena were not found in other saprotrophic or endophytic taxa.

Mycena infection levels
Mycena infection levels reached up to 25%-80% of all reads in individual samples of 9 out of 10 examined plant host species (Figure 2A-C), with considerable intraspecific infection variation (Figure 1, Figure S3).Pinus sylvestris was the only plant host with consistently little or no infection by Mycena.For the ITS1 data, Mycena average read content for all 519 B. vivipara samples was significantly lower than for S. polaris (n = 20) and D. octopetala (n = 22; Figure 1A, B, Table S10).However, the S. polaris and D. octopetala data sets came from only one locality (Botnen et al., 2014), and when comparing them only with the B. vivipara dataset (n = 19) from the same locality, no significant differences were observed (one-way ANOVA, df = 2, F = 1.36, p = 0.263).When it comes to the ITS2 sequences, the host species could be roughly divided into three groups based on average Mycena infection level: (1) P. sylvestris with virtually no Mycena, (2) an intermediate group (median values about 5%-10%) with S. herbacea, A. alpine, B. nana and A. uva-ursi, and (3) B. pubescens and C. tetragona with median Mycena infection levels above 20% of all sequence reads.While all species (except P. sylvestris) harboured individual samples with few, if any, Mycena and some with >30%, there were still significant differences between these three rough categories.(Figure 1D, E, Table S11).

Environmental influences on Mycena
Disparities in sample size and study metadata only permitted limited testing of environmental influences on Mycena infection in three host species.In C. tetragona, Our plant samples consisted of the herbaceous ectomycorrhizal Arctic Bistorta vivipara, subshrub Dryas octopetala and the dwarf shrub Salix polaris (Blaalid et al., 2012;Blaalid et al., 2014;Botnen et al., 2014;Davey et al., 2015;Mundra et al., 2015;Yao et al., 2013); the Arctic ericaceous Cassiope tetragona (Lorberau et al., 2017) and the ectomycorrhizal conifer Pinus sylvestris (Jarvis et al., 2015) from Scotland; and new data also from Scottish (temperate) plants: the arbutoid mycorrhizal Arctostaphylos alpine and Arctostaphylos uva-ursi, ectomycorrhizal dwarf shrubs Betula nana and Salix herbacea, and the ectomycorrhizal trees Betula pubescens and additional Pinus sylvestris.The Betula pubescens samples all came from saplings of <1 m kept low by deer/sheep grazing; the other host plants collected were mature.
a Some samples with low sequence counts were omitted from our analyses.The numbers listed represent samples which were included in our final analyses.
there were no differences in Mycena infection levels between samples derived from drought and control plots applied by Lorberau et al. (2017) (two-tailed t-test, unequal variances, p = 0.57).In A. uva-ursi, the level of Mycena infection decreased significantly with increasing altitude (65-805 m above sea level; R 2 = 0.2, p < 0.0001, data not shown), which does not agree with the assumption (question 4) that increasingly stressful environments should facilitate infection with Mycena or saprotrophs.In B. vivipara, no correlations between Mycena infection level and annual mean temperature, latitude, or mean temperature of the wettest quartal were found (all adjusted R 2 < 0.01, all p > 0.25, see Table S12).A very weak correlation between decreasing Mycena species richness in roots and increasing the mean temperature of the wettest quartal (R 2 = 0.01, p = 0.04) disappeared when the Austrian outlier samples (which contained no Mycena) were excluded.
A chi-square test on the observed versus expected prevalence of Mycena in 222 B. vivipara host plants from 44 patches (a patch constituted multiple plants collected in close proximity) showed a significant, nonrandom association (c 2 = 65.22,df = 43, p = 0.01) of Mycena infections in host plants, suggesting that Mycena-infected B. vivipara were distributed in clumps.

Mycena phylogenetics of OTUs and species diversity
Among the 20 ITS1 and the 21 ITS2 OTUs that were identified as Mycena, we found no phylogenetic signal suggesting that root invasion might be linked to certain clades (Figure 3).Many of the same Mycena species were in both the ITS1 and ITS2 datasets; 12 ITS1-ITS2 pairs of OTUs clustered with >90% probability to the same branches.There were no indications of host specialization by Mycena species on particular host species, with large individual variations in all host plants (again except P. sylvestris) between which Mycena species that were found (Figure S3).Several species, such as Mycena epipterygia or Mycena leptocephala, occurred at infection levels of >10% in at least one individual of 6 of 10 host species or more (Figure S3).Indeed, the two OTUs were the only OTUs shared between C. tetragona from Svalbard and all Scottish host species (except P. sylvestris).

Stable isotope data
On average, carpophore values of δ 15 N and δ 13 C placed Mycena among the saprotrophs.They were higher in ™ 13 C and lower in ™ 15 N than the average of the remaining non-Mycena saprotrophs (Figure 4A-E).The ™ 13 C values of all saprotrophic species for all T A B L E 2 Total sequence counts for operational taxonomic units (OTUs) and amplicon sequence variants (ASVs) for each host species.regions were between À26% and À22%, except for a Phloeomana speirea at Finse (Figure 4A) at À27.1%, and one Mycena metata collection from Svalbard (Figure 4C) at À26.9%.However, there were striking anomalies (and intraspecific variations) in the ™ 15 N values for certain individual collections of Mycena, particularly Mycena pura, which varied between 1.7% for Mycena pura in Gribskov and 12.6% for Mycena pura1 at Finse.A t-test showed the Mycena pura1 and Mycena pura2 collections at Finse to be strongly and significantly higher in δ 15 N than the average for the other Mycena at Finse ( p < 0.0001), and slightly but still significantly higher in δ 13 C (p = 0.02).
In the stepwise regression of δ 15 N, genera were separated by up to 12% into 5 groups (Table S13).Relative to the mean, Mycena grouped at À3% with the litter decay fungi Calvatia, Lycoperdon and Rhodocollybia.The overall adjusted r 2 of the regression model was 0.56 (n = 253), with site accounting for 8.1% of the variance and the remaining 48.0% accounted for by genus.
In the stepwise regression of δ 13 C, genera were separated by up to 5% into 8 groups (Table S14).
Relative to the mean, Mycena grouped at +1% with both the ectomycorrhizal Rhizopogon, Ramaria, which appear to contain both saprotrophic and ectomycorrhizal members (Hobbie et al., 2002), and the litter decay fungi Calvatia, Lepista and Rhodocollybia.The overall adjusted r 2 of the regression model was 0.66 (n = 252), with site accounting for 11.7% of the variance, nitrogen concentration (%N) for 2.8% and the remaining 51.4% accounted for by genus.
The δ 15 N values for host plants in all 5 regions were all below 0%, which is significantly lower than both carpophores and the soil (Figure 4).Soil δ 15 N values were below 2%, with lower overall N content but higher δ 15 N values for deeper soil depths.The δ 15 N and δ 13 C values differed significantly between the five regions, but the amount of variance for both measures explained by region in the mixed linear model was <20% for both isotopes and well below that explained by sample type and genus/sample type in combination (>75% for both).Overall, the other fungal genera had isotopic profiles that matched their expected nutritional mode.
Mycena infection levels (fraction of read shares) and species richness at 97% coverage corrected for sampling depths (Chao & Jost, 2012) for the ITS1 (A-D) and ITS2 (E-H) datasets.Very little difference between the OTU and the ASV approaches were found.Letters xy-z denote host species «significance groups» as found by ANOVA + Scheffes multiple comparisons test for a significance at the p < 0.05 level.Species sharing one identical letter (x, y or z) do not significantly differ from each other in mean Mycena infection level/overall species richness (at 97% coverage).

Mycena found in roots are not surface contaminants
This study constitutes a rare systematic analysis of the occurrence of a supposedly saprotrophic genus, Mycena, in wild plant roots.The results indicate that Mycena species are frequent root colonizers of a taxonomic range of mycorrhizal host plants, although infection levels are very variable.However, here in our analyses presented here, Mycena infection was widespread in Arctic and alpine hosts as well as in temperate hosts.The general lack of host-specificity in Mycena was also universal.The ability to colonize living roots appears to be a widely shared trait across the Mycena phylogeny, consistent with the findings of Thoen et al. (2020).We argue that the systematic findings of Mycena in multiple plant roots should settle concerns such as Vohník (2020), who suggested that the high numbers of Mycena/Clavaria sequences in C. tetragona (Lorberau et al., 2017) might be explained by a lack of root cleaning/washing.All samples (including C. tetragona (Lorberau et al., 2017)) were indeed collected following standard practice for studies aimed at mycorrhizal fungi, with healthy-looking roots selected and either serially washed or surface sterilized, and the work performed in different laboratories.It is thus unlikely that the systematic recovery of large numbers of Mycena reads from multiple plants is mainly due to mycelia living commensally on the root surface.
In 9 out of 10 host plants, Mycena infections were present in many samples and varying from little or no infection up to >40%-50% of the recovered reads.No other supposedly saprotrophic fungal genus was widely found across the different plant species.This is consistent with the findings of Kohout et al. (2018)

What characterizes root invasion patterns by Mycena species?
More different Mycena species possess the ability to invade living plant roots (Smith et al., 2017;Thoen et al., 2020) than those we found to regularly do so in the field in this study (Figures 2, 3).The most frequent root-invading species: M. leptocephala, M. epipterygia.M. sanguinolenta, M. galopus and M. metata are not phylogenetically related but are all broad generalists as litter decayers, and thus, their ecological versatility appears to translate into root invasion, too.
There are known challenges in identifying Mycena species molecularly with ITS (Harder et al., 2013), and some morphospecies could not be separated by our analysis here; however, the most abundant root-invading Mycena were fairly easily identified at the 3% OTU threshold.This is a relatively high value for species separation, but consistent with the average intraspecific ITS sequence variation of 3.6% (ITS1) and 2.7% (ITS2) in our 576 collection database (Figure S3) and with the range for true phylogenetically concordant Mycena species in known species complexes as Mycena pura (Harder et al., 2013).Lower thresholds of, for example, 1.5% (Blaalid et al., 2012) do not give a better species identification, and the ASV and the 3% OTU datasets give very similar conclusions.Thus, it is a robust conclusion that root invasion by Mycena is a frequent but opportunistic phenomenon found mostly in broad litter generalists.
The question then is what constitutes an opportunity for Mycena invasion of a plant root.One answer could be that Mycena preferentially invade roots late in the season as many roots have become senescing and/or otherwise damaged.In twigs on trees, an effect termed 'late pruning' has been described, where wood-decay fungi invade still living but senescing or damaged branches towards the end of the growth season (Butin & Kowalski, 1986;Butin & Kowalski, 1990;Hendry et al., 1998).This is beneficial to the tree, as it causes dead branches to fall off more quickly, closes a possible entry point for pathogens, and allows the tree to spend the resources on new fresh branches.McKay (1968) and Keay and Brown (1990) showed a comparable effect in grass roots, where saprophytes such as Psilocybe semilanceata or Melanoleuca and Conocybe invaded dead cells and parts of senescing but otherwise healthy roots in growth experiments.A mechanism of 'late root pruning' of senescing roots by opportunistically early saprotrophic invaders could thus be invoked in an analogy to that in twigs.The small but significantly positive trend with increasing Mycena content and advancing season (June-October) in Arctostaphylos uva-ursi (Figure S6) suggests that this seasonal senescence effect does exist; however, the r 2 of only 0.08 does not suggest that seasonality is the main reason for root invasion.A disturbance may be another explanation.Defoliation by animal grazing is also known to affect plant root  health with local dieoffs of fine roots, the abscissions increasing the vulnerability of otherwise healthy roots (Sanderson et al., 1997).The B. pubescens in this study were tree saplings of >1 m which were kept low by very high grazing pressure from sheep and deer.All other species with high Mycena infection levels were smaller dwarf shrubs of variable sizes or herbaceous plants, also known to be frequent subjects to grazing by deer or reindeer (Kolari et al., 2019).The clumped (local) distribution of high Mycena infection levels in B. vivipara could be interpreted as offering some support for this theory.Finally, human management in forestry with the planting of dense monocultures of multiple even-aged seedlings in large cleared areas may favour saprotrophs present in the soil and create opportunities for them to invade multiple young and yet uninfected root systems, where they would otherwise tend to lose out in direct competition with ectomycorrhizal symbionts with dominant networks in more natural forest systems with uneven-aged trees and old trees left standing.
The complete absence of Mycena in our P. sylvestris (Figure 1) samples supports this hypothesis.As seen in the growth experiments by Smith et al. (2017), Mycena and several other saprotrophs are indeed able to invade roots in P. sylvestris; however, in vitro resynthesis experiments are necessarily done with tree seedlings, and such diaxenic in vitro resynthesis experiments are not realistic models of natural forests.Our P. sylvestris samples were mature trees from largely undisturbed, uneven-aged, and unmanaged stands within a national park, with strongly heterorhizic root systems where only the most distal feeder roots are neither suberised nor metacutinised, severely limiting colonization by fungi.For such mature trees, ectomycorrhizal colonization levels of the feeder roots can be close to 100% under field conditions, so the available surface for possible colonization by nonmycorrhizal fungi can be very limited.Indeed, the P. sylvestris in our study showed a near-complete dominance by one genus (Suillus, Figure 2D) and had an associated very low general root diversity (Figure 1G, H), consistent also with pre-HTP sequencing-era studies of undisturbed P. sylvestris roots (Jonsson et al., 1999).
In contrast, the roots of the fellow conifer (Picea abies) in Kohout et al. (2018) were heavily Mycenainfected (and had a high overall root fungal diversity), even though they were also sampled from fully mature trees ($80 year old).However, those Picea abies stands in that study were from a planted, largely evenaged and managed (and ultimately clear-cut) forest.Did the Mycena of Kohout et al. (2018) invade their Picea abies hosts already as seedlings, with the trees then 'carrying the seeds of their own destruction', as Parfitt et al. expressed it (2010), for the next eight decades?It has long been known that healthy sapwood tissue of trunks and twigs in trees and woody plants harbour propagules of wood-decaying poroid/corticioid basidiomycetes and xylarioid ascomycetes (Boddy & Griffith, 1989;Gilmartin et al., 2022;Parfitt et al., 2010).This has generally been interpreted as wood-decayers biding their time as seemingly commensal endophytes, waiting for the trunk and twigs, or parts thereof, to die off and turn into a suitable substrate (Parfitt et al., 2010), then having the advantage relative to later post-mortem-invaders.The findings of Kohout et al. (2018) appear to provide direct support for this interpretation of Mycena, which became dominant in the roots of the stumps left after the stand was clear-cut.If indeed the Mycena are latently present for decades, the fact that they invade the roots rather than the trunk or the twigs has further implications for their nutrition, as discuss further in the subsection below.
Summing up, it is conceivable that Mycena root invasion should be seen as an opportunistic fungal response to plants that are young and susceptible, affected by animal or human impact, or otherwise vulnerable to new infections.This is also consistent with what is generally known to facilitate attacks from known fungal parasites (Walters, 2011).In forest management, there is increasing attention to how traditional planting and harvesting practices shape the associated fungal communities and the vulnerability of the planted forests to known pathogens (Hoeksema et al., 2020).Our results here suggest that besides classical pathogens, intensive forest management practices could also favour opportunistic novel root invasions by Mycena and other normally saprotrophic fungi.

What effect does Mycena have on the host upon invasion?
If we accept that Mycena regularly invades seemingly healthy plant roots, the important issue (question 2) then is what impact Mycena invasion in a root has on the host.As described above, the advantage of being latently present on a substrate before its death is obvious.However, the very long life of wood plant hosts means that the time from the first invasion to the significant weakening and actual death of the host could well last for decades or more.As seemingly asymptomatic endophytic fungi always have the potential to opportunistically turn symptomatic, fungi may easily start interacting more measurably with their host during this time in their 'waiting room' (Selosse et al., 2009).It is not hard to envision how the lifestyle of a supposedly patient commensal endophyte could quickly evolve into parasitism or more active necrotrophy like in, for example, Armillaria, and, for Mycena or other saprotrophs that preferably invade roots rather than trunks and twigs, to armistice and mutualism.
Such a spectrum of responses reflects what was seen in the laboratory growth experiments by Thoen et al. (2020), where different Mycena species and conspecific strains displayed behaviours ranging from either little or no invasion into the Betula host plant root, over a primarily harmful or commensally endophytic effect, and ultimately with two strains of M. pura and M. galopus being able to transfer P to their host, with a similar plant growth effect on the host as the known ectomycorrhizal mushroom Paxillus involutus.In our field analyses here, while we cannot directly determine whether a Mycena carpophore sampled for analysis of stable isotopes is free-living or associated with a plant host, the variable isotopic patterns of particularly 15 N in Mycena here are broadly consistent with the patterns of variation observed in these growth experiments.
The average value of the isotopic signatures of Mycena do suggest a saprotrophic lifestyle as expected: The Mycena pura from Svalbard and Vettakollen (in mainland Norway) and a Mycena rosea1 from Gribskov had such saprotrophic isotopic signatures close to the expected.All of our collections of Mycena galopus also displayed saprotrophic profiles with limited similar intraspecific variations and no signs of mycorrhizal-like isotopic profiles, as one might have expected based on Grelet et al. (2017) or Thoen et al. (2020).
However, some individual collections did have profiles that suggested alternative modes of nutrition.Most clearly, two collections of Mycena pura in Finse have isotopic profiles that resemble those of mycorrhizal fungi.In addition, the isotopic profiles of Mycena pura and its close relatives Mycena rosea2 and Mycena diosma from Gribskov (in Denmark) are not far from those of the ectomycorrhizal Russula or Cortinarius (Figure 4).These species all belong to the Mycena section Calodontes which are notably hard to grow in culture, which has previously led to speculations about their nutrition (Boisselier-Dubayle et al., 1996;Harder et al., 2010;Harder et al., 2012;Perreau et al., 1992).On the other hand, Mycena pura was not commonly found in the root samples and only constituted a significant (>10%) fraction of the root community in one single B. vivipara individual (Figure S5).It is thus clear that Mycena-plant nutrient transfer is not a common phenomenon.
The δ 13 C values in those mycorrhiza-like Mycena (Calodontes) collections in Figure 4 are on the high side for average mycorrhizal fungi if their main carbon source was derived from recent photosynthesis.However, some supposedly ectomycorrhizal fungi such as Chalciporus piperatus (Lupt akov a & Mih al, 2020), also have high δ 13 C in carpophores.This has been used to argue that Chalciporus piperatus could be both saprotrophic (Högberg et al., 1999) or biotrophic on ectomycorrhizal fungi (Tedersoo et al., 2010).The 13 C enrichment of Chalciporus relative to autotrophic plants is comparable to our Mycena here.
The 15 N isotopic profiles of the same aforementioned mycorrhiza-like Mycena are more in line with that of mycorrhizal fungi, with the notable exception of the Mycena pura1 sample from Finse.This is surprisingly high in δ 15 N, more than even known mycorrhizal fungi, almost appearing to suggest a naïve pro bono net N donation to the plant.There could be other explanations for a high δ 15 N profile, mycoheterotrophic plants associated with ectomycorrhizal fungi are commonly also quite high in δ 15 N (Trudell et al., 2003), which is likely due to them preferentially accessing the 15 Nenriched protein component of these fungi rather than the 15 N-depleted chitin.However, Hobbie et al. (2020) found that wood-decay Mycena were no more than 1% enriched in 15 N relative to their source nitrogen.It thus seems unlikely that our Mycena here accessed 15 Nenriched nitrogen from co-occurring ectomycorrhizal fungi.Thus, the high δ 15 N values in Mycena pura from Finse do seem to reflect an unusually high source of δ 15 N or high sequestration of 15 N-depleted N in mycelial chitin prior to carpophore formation.Furthermore, the Mycena pura culture in the study by Thoen et al. (2020), which seemingly disproportionally transferred 32 P to its plant host compared to the 14 C it received in return, was originally derived from the same M. pura1 carpophore in Figure 4.This could be an unusual individual aberration or an indication of a recently evolved ability to transfer nutrients that is not yet well-adapted to a usual standard mutualistic relationship.Those results from the field and the laboratory are fairly consistent.
Overall, the bulk of the evidence suggests that the main ecological role of root-invasive Mycena is living commensally endophytically or as slight parasites within the root for some time.A similar role as largely asymptomatic (commensal) endophytes has been suggested for some Sebacinales (Blaalid et al., 2014;Botnen et al., 2014;Lorberau et al., 2017;Selosse et al., 2009) or as well as for dark septate endophytes (Newsham, 2011) in many Arctic plants.As this waiting process can be prolonged for years or decades in trees, and as some Mycena species have indeed evolved the ability to transfer nutrients in the meantime, this genus does provide an empirical example of saprotrophic fungi having progressed to the early intermediate stages from pure saprotrophy.This is in accordance with the predictions of the 'waiting room hypothesis' (Selosse et al., 2009;van der Heijden et al., 2015) of mycorrhizal evolution, which may lead to true reciprocal mycorrhizal Mycena-plant interactions one day.

Need for more directly targeted studies on saprotrophic root invasion
Until now, most data on the occurrence of saprotrophic fungi inside plant roots have arisen as a by-product of other research, and the sampling for these metabarcoding datasets here was not originally designed to investigate this question or to be analysed together.The differences (Figure 1) in average Mycena infection levels between 9 of the 10 host plants in our sample should be interpreted with the caution warranted by differences in sample sizes and site variations and in comparing two ITS regions (Harder et al., 2013).The reverse primer (ITS2_r) of the ITS1 primer set has a terminal mismatch with 99% of all Mycena species (Tedersoo & Lindahl, 2016), which suggests that Mycena content in the ITS1 dataset could be underestimated.If true, this would merely strengthen our overall conclusions about Mycena as an overlooked but significant root-invading genus; however, more studies directly targeting supposedly saprotrophic or endophytic (non-mycorrhizal) fungi in roots are certainly desirable.
To test our hypothesis that Mycena (or other saprotrophic) root infections result from opportunistic invasion under disturbance-related circumstances, future targeted metabarcoding root studies should directly analyse roots of multiple host species of different ages and disturbance levels in the field to identify particular factors that may affect root invasion.Besides disturbance, differences in nitrogen availability have also been shown to influence fungal nutrition facultatively (Peng et al., 2022), and nutrient variability in soil and its role in fungal nutritional mode shifts should certainly also be investigated further.Annotation databases should be continuously updated to reflect our best taxonomic knowledge, and further efforts should be undertaken to identify fungal OTUs or ASVs beyond the genus level, which may require more attention to detail than relying on SINTAX/RDP classification based on even the best possible databases.
Most importantly, more studies on direct ecological interactions between particular hosts and known fungal species are needed, specifically: (1) resynthesis hostfungus experiments with studies of nutrient and C transfer between the symbionts, and (2) stable isotope studies in the field that specifically target saprotrophic taxa.Direct visualization of root-invasive Mycena or other saprotrophic fungi with genus-or species-specific FISH hybridization probes (Schneider-Maunoury et al., 2020) would be particularly useful for conclusive demonstrations of the specific nature and degree of the root invasions in each case.
If fungal ecology can be versatile not only below the genus or even the species level, then this may lead to reconsideration of the high importance ascribed to nutrition as a decisive taxon-delimiting trait (as in Serpulaceae Skrede et al. (2011) or Clavariaceae Birkebak et al. ( 2013)).Redhead et al. (2016) proposed to split the monophyletic Amanita sensu lato into ectomycorrhizal Amanita sensu stricto and a new saprotrophic Saproamanita, precisely to make an ecological annotation in molecular studies easier, but this would be unnecessary with a greater appreciation of ecological versatility.This has important implications for the widely applied approach in HTP-sequencing/metabarcoding plant root studies where annotating ecology to a sequence with genus-level-based taxonomy could be misguided.

CONCLUSIONS
The investigation of the trophic status of the genus Mycena using sequence data from wild plant roots and δ 15 N and δ 13 C signatures yielded the following: (1) In 9 of 10 analysed herbaceous and ericaceous plants and tree mycorrhizal host plants from temperate, alpine and Arctic environments, Mycena was consistently present in living plant roots across species and in different environments, while other saprotrophic taxa were only occasionally present.( 2) The stable isotopic data on carpophores suggested that, although the genus Mycena is indeed mostly saprotrophic, strains of certain Mycena species can display an ecological versatility in the field and exchange nutrients with plants, consistent with previous results from in vitro resynthesis experiments.(3) Mycena infections were not generally more prevalent in Arctic environments or at higher altitudes, but we hypothesize that infection may be more prevalent under conditions of disturbance.(4) The ability to invade living plant roots is a feature of multiple Mycena species that do not discriminate between plant hosts.The evidence that fungal trophic modes may be variable on the species level and that within a large genus such as Mycena, there may be several potential trophic options in addition to pure free-living saprotrophy, raises intriguing questions about the general understanding and study of fungal ecology and the evolution of plant-fungus interactions.More research directly targeting root-associated fungi with unclear or unknown ecologies is required to resolve these questions.This study highlights the importance of continued detailed studies on interactions among organisms at the species level.Such studies would enhance data usage from broad, environmental metabarcoding approaches to community characterization.

Sample site and sample descriptions
Betula pubescens roots were collected at the RSPB Nature Reserve at Corrimony in north-west Scotland in August 2008.The trees were regenerating saplings at a maximum of 1 m in height, growing on moorland within heather-dominated vegetation on a site previously browsed by sheep and deer.Root samples (supporting 100-200 ECM tips) were taken from the trees by direct tracing fine roots from the main laterals.Roots from five trees from within a block were pooled to give one single sample.Roots from Salix herbacea, Betula nana, Arctostaphylos alpina and eight A. uva-ursi from an original biogeography study were collected from mountains across Scotland (Figure S1; Hesling & Taylor, 2013).
The remaining 68 samples of A. uva-ursi roots were from an altitudinal gradient study (and 9 additional P. sylvestris samples in addition to those from Jarvis et al. (2015) in this study), collected June-July 2011 in the Invereshie-Inshriach National Nature Reserve in the north-west of the Cairngorms National Park in Scotland (Figures S1, S2).Samples came from 9 elevation transects from 450 to 850 masl on a Calluna-Arctostaphylos subalpine heath with scattered Scots pine trees up until the tree limit at $650 masl.This was in close proximity to the P. sylvestris forest studied by Jarvis et al. (2015).
The previously published datasets of B. vivipara, S. polaris, D. octopetala and C. tetragona were all collected in Arctic and alpine tundra above the treeline in Arctic Norway, Iceland and Austria, and from grassland below the treeline in Scotland.For more details on the previously published data, we refer to the original publications.A more detailed description of the plant species targeted and the sample sites for the new data can be found in the supplementary data.

Preparation of roots and old amplicon libraries for previously published ITS2 datasets
For the three new ITS2 datasets/454 runs representing 5 of 7 host species (see bioinformatics below), roots were sampled and cleaned under a dissection microscope to remove visible soil debris, woody and non-target species' roots, then lyophilized in 2 mL tubes and milled using a steel bead in a mixer mill (RETSCH, Düsseldorf, Germany).The dry weight for DNA extraction was adjusted for each sample so that extracted mass was proportional to the total sample dry weight: extract weight (mg) = (42.50Â total sample dry weight (mg)) + 47.98.DNA was extracted using 96 well, DNeasy Plant Minikits (QIAGEN, Hilden, Germany).
PCR amplification of the ITS2 region was conducted on a 2720 Thermal Cycler (Life Technologies, Carlsbad, CA, USA) in 10 μL reactions: 5 μL diluted template; 40 μM of each nucleotide; 0.55 mM MgCl 2 ; 40 nM ITS7A primer (Ihrmark et al., 2012); 40 nM ITS 4 primer with a 3 0 8 bp tag (unique by ≥2 bp between samples); and 0.005 U/μL polymerase (DreamTaq Green, Thermo Scientific, Waltham, MA, USA) in buffer.Cycling parameters were: 94 C for 5 min then 25, 30, or 35 cycles at 94 C for 30 s; 57 C for 30 s; 72 C for 30 s; with a final extension of 72 C for 10 min.PCR products were checked using gel electrophoresis (dilutions/cycles adjusted if products were out with the range 1-10 ng μL À1 ), then purified using AMPure 96 (Beckman Coulter, Brea, USA).DNA concentrations were established using a Qubit 2.0 fluorometer (Invitrogen, Paisley, UK), samples combined in equal molar proportion, further purified using GeneJET PCR Purification (Thermo Scientific, Waltham, USA), and lyophilized.Adaptor ligation, 454-sequencing, and sequence adapter trimming were performed by the NERC genomics facility (Liverpool, UK) on one picotitre plate using the GL FLX Titanium system (Roche, Basel, Switzerland).

Bioinformatics
From the previously published studies of B. vivipara, a high-quality dataset of 119,054 sequences was compiled from Blaalid et al. (2012), 191, For the ITS2 dataset, we first analysed the two previously published studies of Jarvis et al. (2015) and Lorberau et al. (2017), and obtained 175,829 and 1,952,314 sequences, For the three unpublished ITS2 454 runs, 327,480 raw reads were obtained on a run with 104 Betula pubescens samples; 49,187 raw reads on a run combining altitude and biogeography samples, and 232,125 for a run with 16 first year biogeography samples.After denoising, chimera check, length, primer/base pair match, and quality controls, 121,587, 326,380 and 154,121 high-quality reads remained, respectively.In total, this amounted to 2,730,231 highquality ITS2 sequences of all fungal ecological groups.These were then used for clustering into OTUs/ASVs.The OTUs/ASVs were classified taxonomically with the non-Bayesian SINTAX classifier (Edgar, 2016) using the 8.2 utax eukaryote reference database (Abarenkov et al., 2020).
QIIME (Caporaso et al., 2010) 1.9.1 pipeline was employed for the 3 unpublished 454 runs through the same steps as in Jarvis et al. (2015) until the OTU clustering step.We retained those with a sequence length of 200-550 bp, only 100% match to in primer/ tag sequences, passed chimera check in UCHIME (Edgar et al., 2011), a sliding window quality check of 50 bp applied to identify low-quality regions (average Phred score <25).The resulting fasta files from the individual ITS1 and ITS2 runs were combined, and clustered first into OTUs at 97% similarity using vsearch (Rognes et al., 2016) and its usearch_global command function, and then into ASVs using the standard settings in UNOISE (Edgar, 2016).The R decontam package (Davis et al., 2018) with the default settings to remove likely contaminants based on the negative controls on a per-sample basis for each of the six different datasets in the ITS1 part, and on the single negative control samples in the Cassiope tetragona ITS2 dataset (no negative controls were sequenced in Jarvis et al. (2015) nor in any of the new ITS2 datasets).Non-fungal sequences ( p < 0.95), and OTUs and ASVs with respectively <10/<8 counts were removed.Finally, sampling saturation was assessed with the iNEXT package (Hsieh et al., 2016; see also Figure S5), and all samples not meeting coveragebased completeness (Chao & Jost, 2012) of 97% were discarded.

Mycena database
For the identification of Mycena sequence data to species level, we first generated 151 new sequences from herbarium specimens and personal collections, using the ITS1F/ITS4 primers and the PCR protocol of Gardes and Bruns (1993).All Mycena ITS full-length sequences from GenBank and from the UNITE database (1099 sq) were extracted.Sequences not identified to species level, which did not cover the regions amplified by the ITS1F-2/ITS3-4 primer target regions, and which were not inside the Mycena sensu stricto clade (Figure 3), or duplicates between both databases were discarded.Additionally, 14 complete Mycena sequences in GenBank were also discarded, as these were deemed to be misidentified (see Table S15), most of those from (Hofstetter et al., 2019).The final database comprised 576 high-quality sequences with 136 named Mycena species, 89 of which with ≥2 sequences.Sequences were aligned with the FFT-NS-i algorithm in MAFFT v 5 (Katoh & Standley, 2013).The complete (628 bp) and annotated Mycena pura EU517504 sequence was used to identify the ITS1 and ITS2 regions in the alignment.They clustered into 156 ITS1 and 139 ITS2 3% OTUs, respectively.Most species were correctly identified by a 3% separation threshold; however, for both regions, some OTUs contained two or more species (such as Mycena galericulata + Mycena megaspora and Mycena olivaceomarginata, Mycena citrinomarginata and Mycena albidolilacea), while other species were split into multiple OTUs (e.g., Mycena pura and Mycena epipterygia).Average intraspecific variation was 3.6% (ITS1) and 2.7% (ITS2; Figure S3).

Phylogenetics
The ITS phylogeny (Figure 3) was constructed by first aligning the selected high-quality subset of 89 ITS fulllength sequences (those with ≥2 sequences for each species) with the Q-ins-i algorithm in MAFFT (Katoh & Standley, 2013) for a final alignment of 1502 positions (gaps included), and then running a maximum likelihood with 1000 bootstrap replications in RaxML (Stamatakis, 2014), saving branch lengths.Then, the 20 ITS1 and the 21 ITS2 OTUs from the ITS1 and ITS2 datasets that were identified as Mycena s.s. were added to the alignment, the Q-ins-i alignment redone, and the OTUs mapped to the branches using the EPA algorithm (Barbera et al., 2019).The tree was visualized in FIGTREE v. 1.4.3 (http://tree.bio.ed.ac.uk/ software/figtree).

Collection of samples for isotope analysis
Fungal carpophores, host plants, and soil were collected from five different regions: Svalbard (Arctic Norway), Finse/Hardangervidda (alpine central Norway), Vettakollåsen (boreal forest) in southeastern Norway, Solhomfjell National Park (boreal forest) in South Norway, and Gribsø /Gribskov (North Zealand, Denmark), in a beech-dominated broadleaf forest patch (Figure S1) in 2015 and 2016.For more information and geographic coordinates of the field locations, see Figure S1 and legend.In Svalbard, the collection sites spanned several similar valleys on the southern banks of Isfjorden, with the sites separated by up to $60 km (Figure S1); for the other four remaining collection sites, samples were taken from an area that extended over no more than 1 km 2 .
Fungal carpophores, soil samples, and plants were dried with continuous airflow for 12-36 h at 70 C until dry.Plants and fungi were identified morphologically, and Mycena furthermore by ITS sequences.For Svalbard, some additional fungal samples were taken from dried mushroom specimens kept at the herbarium at Tøyen at the Natural History Museum in Oslo.It was assumed that individual carpophores collected within a distance of <50 cm between them originated from the same mycelium.Conspecific Mycena carpophores sampled from larger distances were treated as separate samples.Whenever possible, collections from a given site were triplicated or at least duplicated, using separate carpophores from the same collection.Fungi were divided into the three categories 'ectomycorrhizal', 'saprotrophic' or 'Mycena'.For every 36 samples analysed, internal replicates of material from two samples from the same carpophore were used to verify consistent machine functioning.Soil samples were taken from top-soil (A horizon, 0 cm) and mineral soils in $50 cm depths.We sampled soil from three different locations on the different sites.Plant samples were all taken from leaves.

Stable isotope analysis
Dried samples (plants, fungi, soil) were ground by hand, weighed (see Table S4) into 5 x 9 mm tin capsules (Sercon), closed and compressed.Samples consisted of 5 mg of fungi/plant, 10 mg of topsoil, or 20 mg of 50 cm depth soil.Samples were analysed for δ 13 C, δ 15 N, % C and % N by continuous flow with a Costech ECS4010 elemental analyser (Costech Analytical Technologies Inc, Valencia, CA) coupled with a DEL-TAplus XP isotope ratio mass spectrometer (Thermo Scientific, Bremen, Germany) at the University of New Hampshire Stable Isotope Laboratory.All carbon and nitrogen isotope data are reported in delta notation according to this equation: δX = [(R sample /R standard ) À 1] Â 1000 where X is 13 C or 15 N and R is the ratio 13 C/ 12 C or 15 N/ 14 N.All δ 13 C and δ 15 N values were normalized on VPDB (δ 13 C) and AIR (δ 15 N) reference scales with the following internationally calibrated standards and values: IAEA CH6 (210.45%),CH7 (232.15%),N1 (0.4%) and N2 (20.3%).Laboratory working standards included NIST 1515 (apple leaves), NIST 1575a (pine needles), and tuna muscle, as well as a Boletus quality control.

Statistics/graphics
Stepwise multiple regression models of fungal δ 15 N and δ 13 C were analysed with genus, site and %N as the independent variables.Because of the declining δ 13 C of atmospheric carbon dioxide, the year was also included as a continuous factor in the δ 13 C regression.Genus and site were categorical variables and year and %N were continuous variables.These statistical analyses were carried out in JMP 13 Pro (SAS Institute, Middleton, Massachusetts, USA).Models that minimized the Bayesian Information Criterion (BIC) were selected.The variance inflation factor of each model factor was also calculated, which measures multicollinearity.This approach allowed a test of whether Mycena generally grouped with saprotrophic or ectomycorrhizal genera without a priori setting up specific contrasts among Mycena, saprotrophic genera and ectomycorrhizal genera.
All other statistics were done in R using the 'phyloseq' 1.19.1 R package (McMurdie & Holmes, 2013) for combining and rearranging OTU tables and taxonomy information, and the heatmap.2function from the 'gplots' package (Warnes et al., 2016) for visualizing heatmaps.We applied a sequential ANOVA for Figure 1A-D C and 15 N in carpophores of Mycena with other fungi, soils and mycorrhizal host plants from 253 fungal collections, host plants and soils from 5 field locations (see Figures S1, S2 for a map of sampling sites).
low sequence counts were omitted from our analyses.The numbers listed represent samples which were included in our final analyses.
before and after clearcutting.There, Mycena differed from several other saprotrophic fungi by being present in the tree roots already before the clearcutting took place (and going on to dominate the root communities after clearcutting), whereas the other saprotrophs invaded afterward.It is thus a robust conclusion that Mycena can systematically invade the living roots of multiple plant roots in the lab as well as in the field.
at the 0.05 significance threshold with the Scheffe post-hoc test correction for multiple comparisons, using the 'agricolae' package (de Mendiburu, 2020).the specimens deposited in H, TUR, and OULU were produced as part of the Finnish Barcode of Life Project (FinBOL) funded by the Ministry of Environment, Finland (YM23/5512/2013), Otto A Malm's Donationsfond, and the Kone Foundation.We thank the European Commission (grant no.658849) and the Carlsberg Foundation (grant no.CF18-0809) for grants to C.B. Harder that made this research possible.C.B. Harder was financed by a grant from the Danish Independent Research Fund DFF/FNU 2032-00064B (SapMyc) at the time of writing.
Overview of all species and samples included in this study.
T A B L E 1 in their study on root communities in a Picea abies stand F I G U R E 2 Heatmap of Mycena species occurrence (A-C) and overall genus occurrence (B-D).Note the slightly different colour bars in each plot.Only Mycena species that made up >1% of at least one sample were included as separate rows in A-C.In B-D, only genera that comprised >5% of reads as an average in at least one host species had its own separate row.Besides these criteria, we also included Helotiales that could not be identified to generic level, but might still conceivably harbour ericoid mycorrhizal fungi or dark septate endophytes (as Phialocephala).ERM, ericoid mycorrhiza; ECM, ectomycorrhiza; Sap, saprotroph; end, endophyte.Note the near-complete dominance of ectomycorrhiza (particularly Suillus) in P. sylvestris (Ps).