Species Associations Overwhelm Abiotic Conditions to Dictate the Structure and Function of Wood&#X2010;Decay Fungal Communit

Home , Exophiala, Herpotrichiellaceae, Lophodermium

Species associations overwhelm abiotic conditions to dictate the structure and function of wood-decay fungal communities

1,2,6 2 3 2 4 DANIEL S. MAYNARD, KRISTOFER R. COVEY, THOMAS W. C ROWTHER, NOAH W. S OKOL, ERIC W. M ORRISON, 4 5 2 SERITA D. FREY, LINDA T. A. VA N DIEPEN, AND MARK A. BRADFORD 1Department of Ecology and Evolution, University of Chicago, 1101 E 57th Street, Chicago, Illinois 60637 USA 2School of Forestry and Environmental Studies, Yale University, 370 Prospect Street, New Haven, Connecticut 06511 USA 3Institute of Integrative Biology, ETH Zurich, Univeritatstrasse€ 16, 8006 Zurich, Switzerland 4Department of Natural Resources and the Environment, University of New Hampshire, Durham, New Hampshire 03824 USA 5Department of Ecosystem Science and Management, University of Wyoming, 1000 East University Avenue, Laramie, Wyoming 82071 USA

Abstract. Environmental conditions exert strong controls on the activity of saprotrophic microbes, yet abiotic factors often fail to adequately predict wood decomposition rates across broad spatial scales. Given that species interactions can have significant positive and negative effects on wood-decay fungal activity, one possibility is that biotic processes serve as the primary controls on community function, with abiotic controls emerging only after species associations are accounted for. Here we explore this hypothesis in a factorial field warming- and nitrogen-addition experiment by examining relationships among wood decomposition rates, fungal activity, and fungal community structure. We show that functional outcomes and community structure are largely unrelated to abiotic conditions, with microsite and plot-level abiotic variables explaining at most 19% of the total variability in decomposition and fungal activity, and 2% of the variability in richness and evenness. In contrast, taxonomic richness, evenness, and species associations (i.e., co-occurrence patterns) exhibited strong relationships with community function, accounting for 52% of the variation in decomposition rates and 73% in fungal activity. A greater proportion of positive vs. negative species associations in a community was linked to strong declines in decomposition rates and richness. Evenness emerged as a key mediator between richness and function, with highly even communities exhibiting a positive richness–function relationship and uneven communities exhibiting a negative or null response. These results suggest that community-assembly processes and species interactions are important controls on the function of wood-decay fungal communities, ultimately overwhelming substantial differences in abiotic conditions. Key words: basidiomycete; biodiversity; community assembly; co-occurrence; diversity–function; interference competition.

INTRODUCTION traits, behaviors, and corresponding ecosystem process rates, with abiotic factors mediating the intensity and nature of A core goal in ecology is understanding how abiotic and these biotic effects (Maynard et al. 2017b). Despite our biotic factors interact to determine community structure understanding of how fungal interactions and abiotic condi- and function. Abiotic conditions—such as temperature, tions individually affect the functioning of wood-decay fun- nitrogen (N) availability, substrate quality, and moisture— gal communities, the relative importance of biotic vs. abiotic strongly influence microbial activity and metabolic rates, processes remains unclear. such that individual-level functioning (e.g., decomposition There are several ecological mechanisms that suggest that rate) is often correlated with abiotic factors (Boddy 1983, species interactions should be important drivers of structure Boddy et al. 1983, A’Bear et al. 2014). Yet environmental and function in wood-decay fungal communities. First, direct conditions alone are often poor proxies for microbial com- interference competition for space is the primary determinant munity function (Keiser et al. 2011, Bradford et al. 2014, of fungal survival over small temporal and spatial scales van der Wal and De Boer 2014). In wood-decay fungal com- (Boddy 2000). Individuals produce toxins, volatiles, and munities, isolating the factors driving this functional hetero- antimicrobial compounds to kill and inhibit growth of their geneity is particularly challenging (Crowther et al. 2015), in competitors (Hiscox et al. 2010, 2015, Maynard et al. 2017b). part because community-assembly dynamics in wood are The spatial arrangement of individuals and their unique colo- strongly influenced by stochastic processes (Hiscox et al. nization histories can thus have pronounced effects on com- 2016). These processes interact with microsite biotic and abi- munity structure and biodiversity (Coates and Rayner 1985, otic conditions to determine species colonization histories Fukami et al. 2010, Maynard et al. 2017a). In addition, these and resulting community structure (Fukami et al. 2010, direct combative interactions can have pronounced functional Crowther et al. 2015, Hiscox et al. 2016, Leopold et al. consequences: combat-induced stress leads to measurable 2017). Species interactions subsequently alter individuals’ increases in respiration, carbon (C) uptake, and enzyme production (Baldrian 2004, Hiscox et al. 2015, El Ariebi et al. 2016), ultimately reducing the efficiency by which Manuscript received 16 July 2017; revised 2 December 2017; accepted 8 January 2018. Corresponding Editor: Bitty A. Roy. wood-decay fungi use C for growth and biomass (Maynard 6 E-mail: [email protected] et al. 2017b).

801 802 DANIEL S. MAYNARD ET AL. Ecology, Vol. 99, No. 4

Despite the importance of combative interactions in wood- Ecological Research (LTER) site in Petersham, Mas- decay fungal communities, species co-occurrence patterns and sachusetts, USA from May 2013 to May 2015. The study site functional outcomes can also be dictated by facilitation and consists of 24 plots total, each 3 9 3 m, split evenly (n = 6) niche complementarity (Tiunov and Scheu 2005, Toljander among ambient controls, warming only (+W), nitrogen addi- et al. 2006, Ottosson et al. 2014). Fungal succession is often tion only (+N), and warming plus nitrogen addition (W+N) driven by changes in substrate quality in wood, with early treatments. The plots were established in 2006, with the +W fungal colonizers producing enzymes that can preferentially and W+N plots subjected to continuous +5°C above ambient degrade compounds such as cellulose and hemicellulose, and via underground cables, and with the +NandW+Nplots late-stage colonizers degrading more recalcitrant compounds amended with 5 g NÁmÀ2ÁyrÀ1 in the form of aqueous such as lignin (Frankland 1998, Rajala et al. 2012, Vorıskova NH4NO3 (~8 9 higher than ambient N deposition; see and Baldrian 2013, Ottosson et al. 2014). These different Contosta et al. 2011 for further experimental design). enzymatic capabilities of fungi can lead to facilitative relation- Wood blocks were cut from a single freshly felled red maple ships, whereby community-level activity is higher in complex (Acer rubrum L.) and dried to constant 7% moisture. We mixtures due to greater overall resource use (Blanchette 1978, selected red maple because it is locally abundant in the study Tiunov and Scheu 2005). In turn, these mutualistic interac- site (Contosta et al. 2011), because it is common throughout tions can translate into deterministic species co-occurrence eastern U.S. temperature forests (Prasad et al. 2012), and patterns at the landscape level (Niemela€ et al. 1995, Ottosson because it is of intermediate substrate quality, exhibiting mod- et al. 2014), which are intrinsically linked to decomposition erate decay rates (Harmon et al. 1986). Following Bradford stage, functioning, and community structure. et al. (2014), wood blocks were milled to dimensions of In contrast to biotic processes, the impacts of abiotic con- 12 9 14 9 2 cm, and a 1 cm deep 9 2 cm wide channel was ditions on individual fungal activity are somewhat more routed in the to allow for invertebrate colonization. A 4-mm straightforward. Wood-decay fungi (primarily species within thick clear plexiglass shield was subsequently screwed to the the Ascomycota and Basidiomycota) become metabolically top of the wood block to enclose the central chamber. more active at higher temperature and with adequate mois- Wood blocks were placed horizontally in the field in May ture, leading to corresponding increases in growth, enzyme of 2013, with six wood blocks in each treatment plot. One of production, and wood decomposition rates (Boddy et al. the +W plots had intermittent electrical outages since initial 1983, A’Bear et al. 2014, Crowther et al. 2015). Wood is a establishment in 2006, and so no wood blocks were placed low-N substrate (C:N > 100), such that N addition at low in this plot, resulting in n = 36 blocks each for control, +N, levels can stimulate fungal enzyme production and subse- and W+N treatments, and n = 32 for the +W treatment quent biomass production. However, at high levels, it can (n=138 total blocks). Leaf litter was brushed aside so that become toxic and ultimately inhibit activity, with this effect the wood blocks were in direct contact with the soil surface, being particularly pronounced in substrates with high lignin and each block was covered with a granite tile (12 9 14 cm) content (Knorr et al. 2005, van Diepen et al. 2015). Thus, in to prevent light access. Leaf litter was then redistributed the absence of biotic interactions, individual fungal activity, around the block, so that it sat within the litter layer. growth, and wood decomposition should be highest under warm, moist conditions, and at moderate N availability. Wood decomposition Nevertheless, the functional “costs of combat” (i.e., how much energy a wood-decay fungus diverts toward combat Wood blocks were left in the field for two years and removed rather than growth or decomposition) are likewise highest in May 2015. Loose debris and leaves were brushed from the under more optimal abiotic conditions (Maynard et al. surface and blocks were placed in separate bags and immedi- 2017b), thereby potentially disassociating monoculture ately stored on ice for transport. Once in the laboratory, the decomposition rates from community-level functioning. wood blocks were scraped of all residual debris, and weighed. Here we explore the relationships between abiotic conditions, Sawdust was collected from each wood block by drilling 36 fungal community structure, and community-level function. holes in a gridded pattern using a flame-sterilized 5-mm drill Using a long-term soil warming and N-addition experiment, bit. The sawdust was separated into two samples, with one set we explore how microsite variables (soil N, moisture, pH, of samples stored at À80°Cformolecularanalysisandasec- temperature) and plot-level treatments (+5°C, N addition) ondstoredat+4°C for quantification of fungal activity. The affect wood decomposition over the course of two years. Via remaining wood blocks were again weighed to calculate mass molecular identification of the resulting fungal communities, of the removed samples and dried at 65°C to constant mass, at we subsequently quantify how fungal community structure which point the final volume of the wood block was calculated relates to wood decomposition and fungal activity. Specifically, by water displacement. Decomposition was estimated by the we explore the extent to which community structure vs. abiotic percent change in wood density ([final dry mass or volume]/ conditions better explain the observed variation in (1) wood [initial dry mass or volume] 9 100), which is preferable to decay rates, (2) total fungal activity, and (3) species diversity. percentage of mass loss for separating fungal-mediated decomposition from invertebrate-mediated decomposition or environmental weathering (Crowther et al. 2015). METHODS

Study design Fungal activity (substrate induced respiration) The study was conducted at the Soil Warming and Nitrogen Fungal activity was quantified using a modified substrate- Addition (SWaN) plots at the Harvard Forest Long-term induced respiration (SIR) approach (Beare et al. 1990, Fierer April 2018 BIOTIC VS. ABIOTIC IN WOOD DECAY FUNGI 803 et al. 2003). Within 12 h of collection, 1 g dry mass equiva- 2012) and ITS4 (White et al. 1990) with High-fidelity Phu- lent of fresh, undried sawdust was placed into a 50-mL cen- sion polymerase (ThermoFisher Scientific, Waltham, Mas- trifuge tube, replicated twice for each block. To each tube, sachusetts, USA) with 30 PCR cycles and an annealing 2 mL of yeast extract was added (1 L deionized water plus temperature of 60°C (see Appendix S1 on primer design). 12 g autolyzed yeast). Autolyzed yeast serves as a labile C PCR aliquots were prepared in individual tubes rather than substrate that is easily metabolized by saprotrophic microbes plates to limit cross-contamination of sample identification and thus should not substantially favor one functional guild tags, which is commonly found at low levels in high-sequen- over another (Fierer et al. 2003). Tubes were continuously cing depth metabarcoding surveys (Nguyen et al. 2015). PCR agitated for 1 h at 20°C, flushed with CO2-free air, and incu- products for each sample were then combined and purified to bated for 3 h at 20°C, at which point total headspace CO2 remove products with fewer than 150 base pairs (bp) using an was determined using an infrared gas analyzer (IRGA; AxyPrep Mag PCR Clean-up Kit (Axygen Biosciences, Model LI-7000, Li-Cor Biosciences, Lincoln, Nebraska, Union City, California, USA). After purification, the concen- USA). Carbon dioxide concentrations were converted to rate tration of PCR products was measured using a Qubit dsDNA of C production, calculated as mass of C produced per hour High Specificity assay kit (ThermoFisher Scientific). PCR per dry mass wood. SIR is often used as a direct proxy for products were then combined into a single library, with each biomass in leaf-litter or soil microbial communities (Fierer sample at equal molar concentration. Three of the 46 samples et al. 2003, Frey et al. 2004) and it is highly correlated with did not yield adequate PCR products due to high levels of direct measures of fungal biomass (Beare et al. 1990). How- wood phenolics, resulting in 43 total communities. Sequenc- ever, because fungi can have widely different carbon-use effi- ing of the ITS2 region was performed using a dual-index ciencies, even when total biomass is equivalent (Maynard sequencing strategy (Kozich et al. 2013) on one Illumina et al. 2017b,c), we refer to SIR as a measure of “activity” so Miseq 2 9 250 bp paired-end sequencing run (Center for as to provide a more conservative interpretation of this met- Genomics and Bioinformatics, Indiana University, Bloom- ric. These fungal activity values were subsequently scaled ington, Indiana, USA). between 0 and 1 to facilitate interpretation. Reads were merged and dereplicated using USEARCH (Edgar and Flyvbjerg 2015), with a minimum overlap of 16 bp and a minimum sequence length of 150 bp. Extraction Soil analysis of the ITS2 region was conducted using ITSx (Bengtsson- Between May 2013 and May 2015, soil measurements were Palme et al. 2013) and clustered into operational taxonomic taken at the wood-block (hereafter microsite) and plot levels. units (OTUs) using USEARCH based on 97% sequence simi- Soil temperature and moisture measurements were taken at larity. Chimeras were detected and removed using de novo three times (May 2013, September 2014, and May 2015), with UCHIME implemented within the USEARCH pipeline. Tax- one temperature and three moisture measurements taken per onomy was subsequently assigned using QIIME (Caporaso microsite at each sampling period. In May 2015, three soil et al. 2010) by comparing representative sequences to the samples were collected from within each plot, each taken UNITE database (Koljalg et al. 2014) using BLAST from within 15 cm of a wood block. Leaf litter was cleared (Altschul et al. 1990). Following the recommended conserva- and the top 5 cm of the organic horizon was sampled using a tive filtering threshold for Illumina data (Bokulich et al. standard soil corer. Soils were sieved to 2 mm and sorted to 2013), OTUs with total reads <0.005% of all sequence reads remove residual roots, large organic debris, and rocks. Sam- were removed so as to minimize sequence errors. Samples ples were dried at 65°C, ground to a fine powder, and ana- were normalized using “cumulative sum scaling” (Paulson lyzed for total %N and %C using a Costech ESC 4010 et al. 2013) in the metagenomeSeq package in R to account Elemental Analyzer (Costech Analytical Technologies, Valen- for unequal sampling depth. OTUs were subsequently cia, California, USA). In September 2015, a second set of assigned to functional guilds using FUNGuild (Nguyen et al. soils samples was collected from the organic horizon in a sim- 2016). The effect of the abiotic treatment conditions (+W, ilar manner as described above and used to estimate soil pH +N) on individual OTU abundances was assessed using the at the plot level (a separate set of samples was required due to DESeq function in the DESeq2 package in R, using a nega- insufficient soil amounts from the initial collection). Soil sam- tive-binomial model. P values were adjusted for multiple ples were placed into soil slurries (20 mL DI per 10 g soil; 1:2 comparisons using the Benjamini-Hochberg method. v:v), which were stirred three times over the course of 30 min, and pH was measured using a standard probe. Community diversity metrics Basic diversity metrics were calculated for each wood DNA extraction, amplification, and bioinformatics block, including species richness, Shannon diversity, Simpson We randomly selected two out of every six wood blocks in diversity, and Pielou’s evenness. Despite potential issues with each plot for molecular community analyses, totaling 46 total unequal gene copy numbers across individual taxa, the com- wood blocks (23 plots 9 2 blocks). For each wood block, bination of presence/absence data and community-aggregate DNA was extracted from 50 mg of sawdust using a MoBio metrics (i.e., evenness) is suggested to better capture commu- PowerPlant DNA isolation kit (MoBio Laboratories, Carls- nity-level differences (Nguyen et al. 2015). Thus, evenness bad, California, USA). Following the protocol described by was calculated using the full OTU table to reflect differences Morrison et al. (2016), PCR amplification of the internal in community structure that are otherwise not reflected in transcribed spacer (ITS) was performed in triplicate using presence/absence data. Shannon diversity and Simpson diver- dual-index (Kozich et al. 2013) primers fITS7 (Ihrmark et al. sity were highly correlated with richness (q > 0.95), so 804 DANIEL S. MAYNARD ET AL. Ecology, Vol. 99, No. 4 richness is used throughout. Species were ranked from rarest relate to two functional outcomes (decomposition rate, fun- (rank = 1) to most common (rank = 257) based on the num- gal activity) and two diversity indices (taxonomic richness ber of communities in which they occurred. For each wood and evenness). We used an information theoretic approach to block, we then calculated the mean rank of all species present quantify the relative importance of biotic vs. abiotic variables in that community. This “average frequency rank” metric (Burnham and Anderson 2002). This method was selected serves as a complementary aggregate measure of community because our primary goals were to robustly estimate effect structure that potentially captures the relative importance of sizes and to quantify the overall explanatory power (i.e., R2) rare vs. common taxa. attributable to biotic vs. abiotic. Model selection and parame- ter estimation were performed using the Akaike information criterion corrected for sample size (AIC ) model averaging Quantifying species associations c (see Appendix S1), which penalizes overfitting and is recom- Co-occurrence patterns were estimated via latent-variable mended when the number of variables is relatively large Bayesian Markov chain Monte Carlo (MCMC) methods (Burnham and Anderson 2002, 2004). A trade-off of using using the boral package in R (Hui 2016). This approach par- AICc (or similar information-theoretic approaches) is that titions pairwise associations into an environmental-based variables can be retained regardless of their statistical signifi- component, which reflects shared responses to environmen- cance (i.e., even if P>0.05). Thus, although P values are tal gradients, and a residual component, which captures reported and were adjusted for multiple comparisons, these unmeasured species relationships after accounting for envi- should not be interpreted as strict statistical tests. However, ronmental conditions (Warton et al. 2015). The environ- in line with convention we refer to terms as “statistically sig- mental variables included in the latent variable model were nificant” when P < 0.05. All models included a plot variable the initial treatment conditions (+W, +N, W+N), soil % as a random effect and R2 values for each model were esti- moisture, soil N, soil temperature, wood moisture, wood % mated using the likelihood ratio test R2 in the MuMIn pack- N, and soil pH. Note that we included both the treatment age in R (R Core Team 2017). conditions (+W and + N) and the measured soil temperature and soil %N values because there was meaningful variation RESULTS in the measured values within treatment conditions (Appendix S1: Fig. S1) and our primary goal was to ensure Fungal community structure that we accounted for as much between- and within-treatment variation as possible (though see Appendix S1 for A total of 2,465,887 fungal sequences were identified across alternate analyses). The resulting residual correlation matrix all samples (an average of 57,346 reads per sample). After (given by the latent variables) was used to estimate separate quality filtering, this resulted in 257 unique OTUs across pairwise species associations for all pairs of taxa, and the all wood blocks with an average of 80 Æ 22 OTUs environmental correlation matrix (given by the OTU-speci- (mean Æ SD), and with a minimum of 43 OTUs and a maxi- fic regression coefficients) was used to estimate abiotic co- mum of 123 OTUs across all samples (Fig. 1C; Appendix S1: occurrence between taxa. Table S1). Only 17 OTUs were present in abundances >1% of In order to aggregate these pairwise species correlations global sequence reads (Appendix S1: Table S2), and only 3 into a single community-level metric, we subsequently calcu- OTUS were present in abundances >5% of all reads. However, lated the proportion of positive vs. negative pairwise correla- the dominant taxa in each wood block were highly variable, tions among species in each community. We predicted that with the five most abundant OTUs in each community this community-level metric would reflect different underly- encompassing 158 of the 257 total OTUs across all samples ing dynamics or ecological interactions (e.g., facilitative or (Appendix S1: Table S3). Each community contained, on combative), and thus potentially correlate with functioning average, 7.8 Æ 4.7 taxa present in abundances >1% of local or diversity. This community-level co-occurrence metric was sequence reads within that community, and 2.9 Æ 1.4 taxa first calculated using the residual correlation matrix, which present in abundances >5% of reads. is referred to as the proportion of “biotic” positive associa- On average, each OTU was present in 13 Æ 12 samples, tions. This same process was then repeated using the envi- such that a random sampling of 10 wood blocks would con- ronmental correlation matrix, yielding an estimate of the tain approximately 85% of the total OTUs across the site proportion of species with shared environmental preferences (Appendix S1: Fig. S2). Nine of the taxa were present in at (referred to as “abiotic” associations). Only those pairwise least 40 wood blocks, and 72 taxa were present in at least co-occurrences deemed statistically significant at the one-half of the wood blocks. Conversely, 101 OTUs were P = 0.05 level were retained so as to avoid spurious relation- found in less than 10 of the communities. Of the 240 OTUs ships (Ovaskainen et al. 2010, Veech 2014). This approach comprising <1% of all reads, five were found in at least 40 yielded a measure ranging from 0 to 1, with 1 being a com- wood blocks: the mycorrhizal fungus Cenococcum geophi- munity where all species exhibited positive correlations and lum; an unknown fungus in the Dermateaceae family; the 0 being a community where all species exhibited negative plant pathogen Lophodermium agathidis; an unknown fun- correlations. See Appendix S1 for calculations. gus in the Herpotrichiellaceae family, and the saprotroph Exophiala moniliae. Approximately 65% (166/257) of the OTUs were classified Statistical modeling into functional guilds. Of these, 28% were classified as unde- Linear mixed effects models (LMMs) were used to investi- fined saprotrophs, 13% as wood saprotrophs, 12% as mycor- gate how abiotic variables and community characteristics rhizal fungi, 11% as endophytes or plant pathogens, and 2% April 2018 BIOTIC VS. ABIOTIC IN WOOD DECAY FUNGI 805

FIG. 1. (A) Decomposition, (B) fungal activity, (C) taxonomic richness, and (D) evenness across treatments and soil moisture conditions. Shown are violin plots giving the distribution of values for each of the four treatment conditions (control, N addition [N], warming [W], and N+W), split by high vs. low moisture (defined by the 50th percentile across all wood blocks). Each tick mark indicates one observa- tion, and the thick black line denotes the mean response for each group. Fungal activity and evenness were scaled to between zero and one. Nitrogen addition correlated with a slight reduction in fungal activity (P=0.06), but otherwise there were no significant relationships. as litter saprotrophs. In terms of OTU abundances, the three respectively), consistent with what has been seen for long- most dominant OTUs were the general saprotroph Mega- term N addition at the nearby Chronic Nitrogen Amend- collybia rodmanii (16% of reads, present in 42 samples), the ment Experiment (Frey et al. 2014), and as has been wood-decay fungus Crepidotus cesatii (15% of reads, present documented in substrates with high lignin content (van in 19 samples), and an unknown wood-decay fungus in the Diepen et al. 2015). Despite a lack of statistical significance, genus Phanerochaete (10% of reads, present in 40 samples; +N addition exhibited a potentially large positive interaction Appendix S1: Table S2). with warming on fungal activity (15% Æ 10% for a 1 SD increase, P > 0.10; Fig. 1, Table 1). Abiotic factors and functioning Abiotic factors and community structure Decomposition rates and fungal activity (i.e., substrate- induced respiration) varied widely across abiotic conditions, There were no significant relationships between phyla, resulting in no significant differences between treatments order, or class and abiotic treatments (Appendix S1: (Fig. 1A, B; Appendix S1: Table S1). Across all wood blocks Table S4, Figs. S7–S18). There were likewise minimal differ- (n = 138), the mean (ÆSD) decomposition percentage was ences in the proportion of functional guilds across the treat- 30% Æ 4.5%, with a minimum of 5% and a maximum of 57%. ment conditions (Appendix S1: Table S5); however, OTUs Fungal activity exhibited a >50-fold difference across commu- classified as litter saprotrophs accounted for a significantly nities (Fig. 1B), ranging from 0.08 to 5.01 mg CÁminÀ1Ágdry higher proportion of reads in the W+N plots relative to the wood À1 with a mean of 1.17 Æ 0.96 (SD) mg CÁminÀ1Ágdry control or +N only plots (2.6% in W+N vs. 1.52% and 0.9% woodÀ1. Decomposition rate was moderately correlated with in control or +N plots, respectively; Appendix S1: Table S5), fungal activity (q = 0.62, P=0.001; Appendix S1: Fig. S3). and mycorrhizal fungi likewise accounted for a slightly Collectively, abiotic conditions explained 7% of the varia- higher proportion of sequence reads in the W+N plots relation in decomposition rate and 19% in fungal activity tive to control (13.25% vs. 11.81%). (Table 1). Average soil moisture, N addition, and soil pH Twenty-seven taxa showed significant variation in abun- were retained in both models, yet their effect sizes were small dance in response to either N addition or warming and not statistically distinguishable from zero. Nitrogen (Appendix S1: Table S6). Of these, six are known to be wood addition showed consistent negative effects on decomposi- saprotrophs, including two fungi that exhibited a positive tion and activity (À3% and À9% reduction per unit increase, response to warming (Scytalidium spp. and Phanerochaete 806 DANIEL S. MAYNARD ET AL. Ecology, Vol. 99, No. 4

TABLE 1. Fungal community function and structure in relation to abiotic conditions.

Function Structure Factor Decomposition Fungal activity Richness Evenness Warming (+5°C) 0.11 6.98 Nitrogen (treatment) À0.03 À0.09 À8.78 0.06 Nitrogen 9 Warming 0.15 Soil moisture 0.01 À0.02 À2.83 0.05 Soil pH 0.01 À0.03 3.13 Soil %N 0.06 Observations 138 138 43 43 R2 0.07 0.19 0.02 0.02 Notes: Shown are the model-averaged regression coefficients. Warming (W), N addition (N), and N+W retained their original (binary yes/ no) coding, whereas all other variables were standardized to have a mean of zero and a standard deviation of one. Abiotic variables were overall poor predictors of function and structure, explaining 7% and 19% of the variability in decomposition and fungal activity, respectively, and 2% in richness and evenness. No variables showed statistically significant relationships due to high variability among outcomes (Fig. 1). A blank cell indicates a variable that was not retained in any of the top models for that outcome. spp.), two that exhibited a significant negative response to fewer taxa per unit increase in soil moisture, with corre- warming (Xenasmatella vaga and Scytalidium lignicola), one sponding increases of 6% Æ 6% and 5% Æ 3% in evenness. that exhibited a significant negative response to N addition (Scytalidium spp.), and one that exhibited a significant posi- Biotic factors and function tive response to N addition (Fibroporia norrlandica). None of the remaining 230 taxa exhibited statistically significant In contrast to abiotic variables, the combination of fre- relationships with elevated temperature or nitrogen after quency rank, evenness, positive associations, richness, and adjusting for multiple comparisons. their interactions explained a large fraction (52%) of the total Overall, abiotic conditions were poor predictors of the variation in decomposition (Table 2, Fig. 2). Soil moisture community structure, with abiotic conditions explaining was the only abiotic factor that was retained, but its effect size ~2% of the variation in richness and evenness, and with no was small and not significant (2.0% Æ 1.2% increase per unit variables yielding statistically significant trends (Table 1). increase in soil moisture; Table 2). Communities with higher The top-ranked models for richness and evenness likewise proportions of positive biotic associations exhibited a signifi- retained +N and soil moisture, suggesting 8.78 Æ 6.8 fewer cant decrease in decomposition of 5% Æ 1.6% per unit taxa in the +N plots relative to control plots, and 2.83 Æ 3.3 increase in positive associations (P=0.013), corresponding

TABLE 2. Fungal community function and structure in relation to biotic and abiotic factors.

Function Structure Factor Decomposition Fungal activity Richness Evenness Warming (+5°C) 0.11 Nitrogen (treatment) Nitrogen 9 Warming Soil moisture 0.02 0.02 Soil pH À0.05 Soil %N 0.06 Proportion of positive associations (biotic) À0.05* À0.10** À11.08*** 0.05 Proportion of positive associations (abiotic) À0.02 À4.19 Richness 0.03 0.09** na na Evenness 0.03 0.01 na na Frequency rank À0.03 0.07** À10.49*** 0.14*** Frequency rank 9 Evenness 0.10*** 0.14*** na na Evenness 9 Richness 0.07** 0.06* na na Proportion of positive associations (biotic) 9 Warming 0.13** na na Observations 43 43 43 43 R2 0.52 0.73 0.69 0.51 Notes: Shown are the model-averaged regression coefficients for each model. Warming, N addition, and N+W retained their original (binary yes/no) coding, whereas all other variables were standardized to have a mean of zero and a standard deviation of one. Abiotic factors remained largely uninformative of both function and structure, whereas the proportion of biotic associations, frequency rank, evenness, and richness exhibited significant main effects and/or significant interactions, explaining >50% of the variation for each outcome. A blank cell indicates a variable that was not retained in any of the top models for that outcome and a value of na indicates variables that were not included as predictors for that outcome. *P < 0.05; **P < 0.01; ***P < 0.001. April 2018 BIOTIC VS. ABIOTIC IN WOOD DECAY FUNGI 807

FIG. 2. Fungal activity and decomposition showed similar relationships with community structure. Shown are the model-based regression trends identified in Table 2. The center lines denote the mean response, and the colored band in (A) denotes the 95% confidence interval for the overall mean. The two different colored bands (in B–F) denote “high” and “low” values for the interaction variables based on the 5th and 95th percentile values across the full data set. Note that the points show the raw data, and are not adjusted for the significant relationships identified in Table 2. Circles denote control plots; squares denote +W (warming) plots; diamonds denote +N plots; and triangles denote W+N. to a decrease in decomposition from ~40% to ~15% across the interaction). The effect of evenness on decomposition also full range of biotic co-occurrence values (Fig. 2A). Highly depended on whether the community was comprised of rare even communities exhibited a strong positive richness-decom- vs. common taxa (Fig. 2C; P<0.001 for the interaction). position relationship, whereas highly uneven communities Community properties explained 73% of the total vari- exhibited a negative relationship (Fig. 2B; P<0.001 for the ability in fungal activity, with the general patterns nearly 808 DANIEL S. MAYNARD ET AL. Ecology, Vol. 99, No. 4 identical to those for decomposition (Table 2, Fig. 2D–F). positive biotic associations and frequency rank correspond- A significant decrease in fungal activity in the control plots ing with an increase in evenness of 5% Æ 2.4% (P=0.20) was largely offset under warming (Fig. 2D; P = 0.001 for and 14% Æ 2.6% (P<0.001), respectively. Abiotic variables the interaction). In communities with high evenness, an remained largely uninformative of evenness or richness, even increase in richness corresponded with an increase in fungal after accounting for biotic factors, with only soil moisture activity; yet this relationship was essentially constant in exhibiting a slight positive effect on evenness (2.4% Æ 2.3% communities with low evenness (Fig. 2E; P = 0.015 for the increase in evenness per unit increase in moisture). interaction). Conversely, an increase in evenness corresponded to a strong decrease in fungal activity among rare DISCUSSION taxa and an increase among common taxa (Fig. 2F; P < 0.001 for the interaction). Positive abiotic correlations Abiotic environmental variables were largely uninformative exhibited a slight albeit nonsignificant negative relationship of function and structure even after accounting for biotic fac- with activity (Table 2). tors, despite the marked differences in abiotic gradients gener- ated by the warming and N-addition treatments (Fig. 1; Appendix S1: Table S1). In contrast, community structure Biotic factors and community structure exhibited strong, consistent relationships with functioning, Combined, species associations and frequency rank explaining 52% of the total variation in wood decomposition explained 69% of the variability in richness. Indeed, the pro- and 73% of the variation in fungal activity. Richness, evenness, portion of positive biotic correlations within each commu- and biotic co-occurrence patterns all exhibited significant nity was strongly correlated with species richness, both in effects, yielding largely consistent trends between decomposi- univariate analysis (Fig. 3A) and after accounting for addition and fungal activity (Fig. 2A–Cvs.2D–F). Similarly, com- tional community metrics (Table 2). Communities with a munity properties explained 69% and 51% of the variation in high proportion of biotic associations contained less than community richness and evenness, respectively, with abiotic one-half of the number of unique taxa (~40 vs. 100 OTUs) factors explaining negligible additional variation. than communities with equivalent proportions of positive Previous studies of wood decomposition have observed sub- and negative associations (Fig. 3A; R2 = 0.45, P<0.001). stantial variation in rates both within and across sites (Boddy Communities exhibited a decrease of 10.49 Æ 2.3 taxa per and Swift 1983, Yatskov et al. 2003, Dickie et al. 2012, standard deviation increase in frequency rank (P<0.001), Bradford et al. 2014). This variation can be attributed both to and a decrease of 11.08 Æ 3.1 taxa per unit increase in the deterministic factors such as substrate-mediated succession proportion of biotic associations (P<0.001). As with fun- (Chapela et al. 1988, Griffith and Boddy 1990, Frankland gal activity, abiotic associations exhibited a slight negative 1998), and to stochastic assembly processes such as priority relationship with richness (Table 2). effects and “position effects” (i.e., the exact spatial layout of Community metrics explained 51% of the variability in species in the community, sensu Buss and Jackson [1979]). evenness, with frequency rank being the dominant explana- Our results suggest that these different community-assembly tory variable (Table 2). Biotic associations were significant processes may account for substantial variation in community predictors of evenness in univariate analysis (Fig. 3B), but structure and functioning, and that wood decomposition rates this relationship was no longer significant after accounting may thus be relatively insensitive to changes in temperature for differences in frequency rank (Table 2). Evenness and soil N availability, particularly in relation to other sapro- showed opposite patterns to richness, with a unit increase in trophic microbial communities (Frey et al. 2008, Conant et al.

FIG. 3. The relationships between richness, evenness, and the proportion of positive residual associations. The center lines denote the mean response and the colored bands denote the 95% confidence interval for the mean. In contrast to Fig. 2, these regression trends were calculated using simple linear regression, with the proportion of positive biotic associations as the sole explanatory variable. Even without accounting for additional variables, this metric is a strong predictor of richness and a moderate predictor of evenness. Circles denote control plots; squares denote +W plots; diamonds denote +N plots; and triangles denote W+N. April 2018 BIOTIC VS. ABIOTIC IN WOOD DECAY FUNGI 809

2011, Yergeau et al. 2012, Morrison et al. 2016). Nevertheless, vs. common taxa (Appendix S1: Fig. S5). These abiotic spe- approximately 10% of the fungi exhibited significant responses cies associations were substantially less informative of struc- to warming or N addition (Appendix S1: Table S6). Abiotic ture and function than the biotic species associations conditions can also indirectly affect community composition (Table 2). Supplemental analysis revealed that traditional and functioning by altering community-assembly dynamics, co-occurrence metrics (i.e., that do not separate abiotic from for example, by regulating the relative importance of priority residual correlations) reduced the explanatory power of the effects as a structuring force (Leopold et al. 2017). Thus, our models by 5–15% (Appendix S1: Table S8). Thus, these results should in no way be interpreted as implying that residual correlations capture aspects of community structure wood-decay fungi are unaffected by abiotic conditions. not explained by shared environmental preferences, suggest- Rather, these findings only suggest that aggregate community- ing that direct species interactions at least partially explain level properties may be relatively resistant to modest changes the observed relationships. in abiotic conditions. Positive co-occurrence patterns are often taken to indicate Two important considerations are that we only used a sin- mutualistic interactions such as facilitation or complementarity gle wood species (Acer rubrum), and that we only tracked (Ovaskainen et al. 2010, Ottosson et al. 2014, Morales-Castilla wood decomposition over a two-year period. There are thus et al. 2015). Indeed, we found, for example, that the sapro- likely to be important substrate-by-environment interactions trophic fungus Trichoderma koningiopsis exhibited a strong, that we cannot assess with this study design (e.g., Coates significant interaction with a Chloridium spp. (Appendix S1: and Rayner 1985, Frankland 1998, Weedon et al. 2009, Table S7), supporting previous inference of facilitative interac- Kubartova et al. 2012, Rajala et al. 2012). Indeed, only 7% tions between species in these two genera (Tiunov and Scheu (17 of 257) of the fungal OTUs were present in abundances 2005). Furthermore, approximately one-third of the significant >1% of total sequence reads, and nearly 40% of the OTUs positive biotic correlations involved a yeast and a non-yeast were found in fewer than 10 wood blocks. This preponder- fungus (Appendix S1: Table S7), which are likewise known to ance of rare taxa suggests that the wood traits and/or land- exhibit strong synergistic relationships due to differential sub- scape-level environmental conditions may serve as primary strate use (Blanchette 1978). Nevertheless, it remains unclear filters on species’ relative abundances, with biotic processes why communities with a high proportion of positive biotic emerging only when factors are held constant. The establish- associations exhibited lower richness and reduced functioning ment of multi-site, long-term decomposition experiments (Table 2, Figs. 2 and 3), as these results that are typically not using wood substrates from many tree species will be critical expected under facilitative interactions (Blanchette 1978, Tiunov for disentangling the importance of wood traits, decomposi- and Scheu 2005, Fukami et al. 2010). An alternate explanation tion stage, and abiotic conditions on landscape-level decom- is that these positive biotic co-occurrence patterns arise not from position dynamics (Cornelissen et al. 2012). facilitative interactions, but rather from combative deadlock or A challenge when using snapshot molecular sequence data intransitive competitive relationships among dominant individ- is disentangling cause from effect (Ottosson et al. 2014), uals (Boddy 2000, Fukami et al. 2010, Maynard et al. 2017a). which is why we measured fungal activity via SIR (indicative These combative interactions can reduce community function- of the resident fungal community) in addition to decomposi- ing in wood-decay fungi (Maynard et al. 2017b,c), and they can tion (indicative of the previous community). Although the res- reduce diversity due to displacement of the weakest competi- piration of fungi when fed a labile C source need not correlate tors by those near the top of the hierarchy (Maynard et al. with wood decomposition ability, community-level SIR has 2017a). Most likely, both facilitative and combative interac- been shown to correlate strongly with direct measures of fun- tions explain the observed results to varying degrees (Boddy gal biomass (Beare et al. 1990). Furthermore, fungal activity 2001, Ovaskainen et al. 2010), and isolating the relative and decomposition were moderately correlated (Appendix S1: contribution of each mechanism is an important next step. Fig. S1), and the statistical relationships were nearly identical Understanding how abiotic factors and biotic processes for decomposition and fungal activity (Fig. 2). These results alter the structure and function of wood-decay fungal com- thus suggest a close link between the functioning of the resi- munities is challenging given the opaque nature of the system. dent fungal community and the functioning of the historical By using a long-term global-change field experiment, we were fungal community. Nevertheless, it cannot be discounted that able to capture marked variation in abiotic conditions, while decomposition stage—and corresponding nutrient availability also ensuring that ancillary environmental factors were held and wood chemistry—caused deterministic changes in com- constant across communities. Our results demonstrate that munity composition and activity, and not vice-versa. However, abiotic factors—primarily temperature and nitrogen addi- if indeed this is the case, it is not clear what processes led to tion—are relatively uninformative of wood-decay fungal different decomposition rates among communities, given the community structure and function, whereas community char- large amount of variation that remained unexplained by the acteristics and species associations were strongly indicative of abiotic variables (Fig. 1, Table 1). wood decay rates and corresponding fungal activity. Pairwise The latent variable approach used here helps to isolate co-occurrence patterns were clearly linked to both diversity environmental-driven co-occurrence patterns from species and decomposition, suggesting that species interactions associations, but it does not permit identification of the among dominant fungi are important drivers of structure specific mechanisms underlying these relationships (Warton and function. Our results thus contribute to the growing evi- et al. 2015). The absolute magnitude of biotic associations dence that climate effects are secondary to competitive inter- (both positive and negative) was largest between common actions for wood-decay fungi, even over relatively large taxa (Appendix S1: Fig. S4), whereas the magnitude of abi- regional scales (Bradford et al. 2014). These findings empha- otic correlations was largest among rare vs. rare or common size the need to better understand how abiotic and biotic 810 DANIEL S. MAYNARD ET AL. Ecology, Vol. 99, No. 4 processes interactively alter fungal communities if we are Buss, A. L. W., and J. B. C. Jackson. 1979. Competitive networks: to successfully predict how decomposition processes will nontransitive competitive relationships in cryptic coral reef envi- – respond to novel environmental conditions. ronments. American Naturalist 113:223 234. Caporaso, J. G., et al. 2010. QIIME allows analysis of high-throughput – ACKNOWLEDGMENTS community sequencing data. Nature Methods 7:335 336. Chapela, I. H., L. Boddy, and A. D. M. Rayner. 1988. Structure and We thank Jason Paschke, Jonathan Schilling, and Mark Anthony development of fungal communities in beech logs four and a half for supplemental analyses, and we thank Connor Reed, Elena years after felling. FEMS Microbiology Letters 53:59–69. Karlsen-Ayala, and Charles Faires for laboratory assistance. We Coates, D., and A. D. M. Rayner. 1985. Fungal population and also thank two anonymous reviewers for their thoughtful comments community development in cut beech logs. III. Spatial dynamics, and suggestions. Funding for this experiment was provided in part interactions and strategies. New Phytologist 101:183–198. by the Yale Institute for Biospheric Studies (to D. S. Maynard) and Conant, R. T., et al. 2011. Temperature and soil organic matter the National Science Foundation DEB 1601036 (to D. S. Maynard decomposition rates—synthesis of current knowledge and a way and M. A. Bradford). Chemical analyses were performed by the forward. Global Change Biology 17:3392–3404. Yale Stable Isotope Center. Support for the initiation and mainte- Contosta, A. R., S. D. Frey, and A. B. Cooper. 2011. Seasonal nance of the Soil Warming and Nitrogen Addition Study at Harvard dynamics of soil respiration and N mineralization in chronically Forest has been provided by the National Science Foundation (to S. warmed and fertilized soils. Ecosphere 2:1–21. D. Frey) DEB 0447967, DEB 1237491, and DEB 1456528. Cornelissen, J. H. C., et al. 2012. Controls on coarse wood decay in temperate tree species: birth of the LOGLIFE experiment. Ambio 41(Suppl 3):231–245. LITERATURE CITED Crowther, T. W., S. M. Thomas, D. S. Maynard, P. Baldrian, K. A’Bear, A., T. Jones, E. Kandeler, and L. Boddy. 2014. Interactive Covey, S. D. Frey, L. T. A. van Diepen, and M. A. Bradford. effects of temperature and soil moisture on fungal-mediated wood 2015. Biotic interactions mediate soil microbial feedbacks to cli- decomposition and extracellular enzyme activity. Soil Biology mate change. Proceedings of the National Academy of Sciences and Biochemistry 70:151–158. USA 112:7033–7038. Altschul, S. F., W. Gish, W. Miller, E. W. Myers, and D. J. Lipman. Dickie, I., T. Fukami, J. P. Wilkie, R. B. Allen, and P. K. Buchanan. 1990. Basic local alignment search tool. Journal of Molecular 2012. Do assembly history effects attenuate from species to Biology 215:403–410. ecosystem properties? A field test with wood-inhabiting fungi. Baldrian, P. 2004. Increase of laccase activity during interspecific Ecology Letters 15:133–141. interactions of white-rot fungi. FEMS Microbiology Ecology Edgar, R. C., and H. Flyvbjerg. 2015. Error filtering, pair assembly 50:245–253. and error correction for next-generation sequencing reads. Bioin- Beare, M. H., C. L. Neely, D. C. Coleman, and W. L. Hargrove. formatics 31:3476–3482. 1990. A substrate-induced respiration (SIR) method for measure- El Ariebi, N., J. Hiscox, S. A. Scriven, C. T. Muller,€ and L. Boddy. ment of fungal and bacterial biomass on plant residues. Soil 2016. Production and effects of volatile organic compounds Biology and Biochemistry 22:585–594. during interspecific interactions. Fungal Ecology 20:144–154. Bengtsson-Palme, J., et al. 2013. Improved software detection and Fierer, N., J. P. Schimel, and P. A. Holden. 2003. Variations in extraction of ITS1 and ITS2 from ribosomal ITS sequences of microbial community composition through two soil depth pro- fungi and other eukaryotes for analysis of environmental sequenc- files. Soil Biology and Biochemistry 35:167–176. ing data. Methods in Ecology and Evolution 4:914–919. Frankland, J. C. 1998. Fungal succession – unravelling the unpre- Blanchette, R. A. 1978. Associations among bacteria, yeasts, and dictable. Mycological Research 102:1–15. basidiomycetes during wood decay. Phytopathology 68:631–637. Frey, S. D., M. Knorr, J. L. Parrent, and R. T. Simpson. 2004. Boddy, L. 1983. Carbon dioxide release from decomposing wood: Chronic nitrogen enrichment affects the structure and function of Effect of water content and temperature. Soil Biology and Bio- the soil microbial community in temperate hardwood and pine chemistry 15:501–510. forests. Forest Ecology and Management 196:159–171. Boddy, L. 2000. Interspecific combative interactions between Frey, S. D., R. Drijber, H. Smith, and J. Melillo. 2008. Microbial bio- wood-decaying basidiomycetes. FEMS Microbiology Ecology mass, functional capacity, and community structure after 12 years 31:185–194. of soil warming. Soil Biology and Biochemistry 40:2904–2907. Boddy, L. 2001. Fungal community ecology and wood decomposi- Frey, S. D., et al. 2014. Chronic nitrogen additions suppress decom- tion processes in angiosperms: from standing tree to complete position and sequester soil carbon in temperate forests. Biogeo- decay of coarse woody debris. Ecological Bulletins 49:43–56. chemistry 121:305–316. Boddy, L., and M. J. Swift. 1983. Wood decomposition in an Fukami, T., I. Dickie, J. Paula Wilkie, B. C. Paulus, D. Park, A. abandoned beech and oak coppiced woodland in SE England: Roberts, P. K. Buchanan, and R. B. Allen. 2010. Assembly history I. Patterns of wood-litter fall. Ecography 6:320–332. dictates ecosystem functioning: evidence from wood decomposer Boddy, L., B. Sciences, and B. Ba. 1983. Effect of temperature and communities. Ecology Letters 13:675–684. water potential on growth rate of wood-rotting basidiomycetes. Griffith, G., and L. Boddy. 1990. Fungal decomposition of attached Transactions of the British Mycological Society 80:141–149. angiosperm twigs I. Decay community development in ash, beech Bokulich, N. A., S. Subramanian, J. J. Faith, D. Gevers, J. I. and oak. New Phytologist 116:407–415. Gordon, R. Knight, D. A. Mills, and J. G. Caporaso. 2013. Harmon, M. E., et al. 1986. Ecology of coarse woody debris in tem- Quality-filtering vastly improves diversity estimates from Illumina perate ecosystems. Advances in Ecological Research 15:133–302. amplicon sequencing. Nature Methods 10:57–59. Hiscox, J., P. Baldrian, H. J. Rogers, and L. Boddy. 2010. Changes Bradford, M. A., R. J. Warren, P. Baldrian, T. W. Crowther, D. S. in oxidative enzyme activity during interspecific mycelial interac- Maynard, E. E. Oldfield, W. R. Wieder, S. A. Wood, and J. R. tions involving the white-rot fungus Trametes versicolor. Fungal King. 2014. Climate fails to predict wood decomposition at Genetics and Biology 47:562–571. regional scales. Nature Climate Change 4:625–630. Hiscox, J., M. Savoury, I. P. Vaughan, C. T. Muller,€ and L. Boddy. Burnham, K. P., and D. Anderson. 2002. Model selection and multi- 2015. Antagonistic fungal interactions influence carbon dioxide model inference: a practical information-theoretic approach. evolution from decomposing wood. Fungal Ecology 14:24–32. Springer, New York, New York, USA. Hiscox, J., M. Savoury, S. R. Johnston, D. Parfitt, C. T. Muller,€ Burnham, K. P., and R. P. Anderson. 2004. Multimodel inference: H. J. Rogers, and L. Boddy. 2016. Location, location, location: understanding AIC and BIC in model selection. Sociological priority effects in wood decay communities may vary between Methods and Research 33:261–304. sites. Environmental Microbiology 18:1954–1969. April 2018 BIOTIC VS. ABIOTIC IN WOOD DECAY FUNGI 811

Hui, F. K. C. 2016. boral – Bayesian ordination and regression anal- 2014. Species associations during the succession of wood-inhabiting ysis of multivariate abundance data in r. Methods in Ecology and fungal communities. Fungal Ecology 11:17–28. Evolution 7:744–750. Ovaskainen, O., J. Hottola, and J. Shtonen. 2010. Modeling species Ihrmark, K., et al. 2012. New primers to amplify the fungal ITS2 co-occurrence by multivariate logistic regression generates new region—evaluation by 454-sequencing of artificial and natural hypotheses on fungal interactions. Ecology 91:2514–2521. communities. FEMS Microbiology Ecology 82:666–677. Paulson, J. N., O. C. Stine, H. C. Bravo, and M. Pop. 2013. Differen- Keiser, A. D., M. S. Strickland, N. Fierer, and M. A. Bradford. 2011. tial abundance analysis for microbial marker-gene surveys. Nature The effect of resource history on the functioning of soil microbial Methods 10:1200–1202. communities is maintained across time. Biogeosciences 8:1477–1486. Prasad, A. M., L. R. Iverson, S. Matthews, and M. Peters. 2012. A Knorr, M., S. D. Frey, and P. S. Curtis. 2005. Nitrogen additions and climate change atlas for 134 forest tree species of the eastern litter decomposition: A meta-analysis. Ecology 86:3252–3257. United States. [database]. http://www.nrs.fs.fed.us/atlas/tree Koljalg,U.,R.H.Nilsson,K.Abarenkov,L.Tedersoo,A.F.S.Taylor, R Core Team. 2017. R: a language and environment for statistical and M. Bahram. 2014. Towards a unified paradigm for sequence- computing. R Foundation for Statistical Computing, Vienna, based identification of fungi. Molecular Ecology 22:5271–5277. Austria. www.r-project.org Kozich, J. J., S. L. Westcott, N. T. Baxter, S. K. Highlander, and Rajala, T., M. Peltoniemi, T. Pennanen, and R. Makip€ a€a.€ 2012. P. D. Schloss. 2013. Development of a dual-index sequencing Fungal community dynamics in relation to substrate quality of strategy and curation pipeline for analyzing amplicon sequence decaying Norway spruce (Picea abies [L.] Karst.) logs in boreal data on the miseq Illumina sequencing platform. Applied and forests. FEMS Microbiology Ecology 81:494–505. Environmental Microbiology 79:5112–5120. Tiunov, A. V., and S. Scheu. 2005. Facilitative interactions rather Kubartova, A., E. Ottosson, A. Dahlberg, and J. Stenlid. 2012. than resource partitioning drive diversity-functioning relationships Patterns of fungal communities among and within decaying logs, in laboratory fungal communities. Ecology Letters 8:618–625. revealed by 454 sequencing. Molecular Ecology 21:4514–4532. Toljander, Y. K., B. D. Lindahl, L. Holmer, and N. O. S. Hogberg.€ Leopold, D. R., J. P. Wilkie, I. Dickie, R. B. Allen, P. K. Buchanan, 2006. Environmental fluctuations facilitate species co-existence and T. Fukami. 2017. Priority effects are interactively regulated and increase decomposition in communities of wood decay fungi. by top-down and bottom-up forces: evidence from wood decom- Oecologia 148:625–631. poser communities. Ecology Letters 20:1054–1063. van der Wal, A., and W. De Boer. 2014. Neglected role of fungal Maynard, D. S., M. A. Bradford, D. L. Lindner, L. T. A. van Diepen, community composition in explaining variation in wood decay S. D. Frey, T. W. Crowther, J. Glaeser, and T. W. Crowther. 2017a. rates. Ecology 96:124–133. Diversity begets diversity in competition for space. Nature Ecology van Diepen, L. T. A., S. D. Frey, C. M. Sthultz, E. W. Morrison, R. and Evolution 1:0156. Minocha, and A. Pringle. 2015. Changes in litter quality caused Maynard, D. S., T. W. Crowther, and M. A. Bradford. 2017b. by simulated nitrogen deposition reinforce the N-induced sup- Fungal interactions reduce carbon use efficiency. Ecology Letters pression of litter decay. Ecosphere 6:1–16. 20:1034–1042. Veech, J. A. 2014. The pairwise approach to analysing species Maynard, D. S., T. W. Crowther, and M. A. Bradford. 2017c. co-occurrence. Journal of Biogeography 41:1029–1035. Competitive network determines the direction of the diversity– Vorıskova, J., and P. Baldrian. 2013. Fungal community on decom- function relationship. Proceedings of the National Academy of posing leaf litter undergoes rapid successional changes. ISME Sciences USA 114:11464–11469. Journal 7:477–486. Morales-Castilla, I., M. G. Matias, D. Gravel, and M. B. Araujo. Warton, D. I., et al. 2015. So many variables: joint modeling 2015. Inferring biotic interactions from proxies. Trends in Ecology in community ecology. Trends in Ecology and Evolution 30: and Evolution 30:347–356. 766–779. Morrison, E. W., S. D. Frey, J. J. Sadowsky, L. T. A. Van Diepen, Weedon, J. T., W. K. Cornwell, J. H. C. Cornelissen, A. E. Zanne, C. W. K. Thomas, and A. Pringle. 2016. Chronic nitrogen additions Wirth, and D. A. Coomes. 2009. Global meta-analysis of wood fundamentally restructure the soil fungal community in a temper- decomposition rates: A role for trait variation among tree species? ate forest. Fungal Ecology 23:48–57. Ecology Letters 12:45–56. Nguyen, N. H., D. Smith, K. Peay, and P. Kennedy. 2015. Parsing White, T., T. Bruns, S. Lee, and J. Taylor. 1990. Amplification and ecological signal from noise in next generation amplicon sequenc- direct sequencing of fungal ribosomal RNA genes for phylogenetics. ing. New Phytologist 205:1389–1393. Pages 315–322 in M. Innis, D. Gelfand, J. Sninsky, and T. White, Nguyen, N. H., Z. Song, S. T. Bates, S. Branco, L. Tedersoo, J. editors. PCR protocols: a guide to methods and applications 18. Menke, J. S. Schilling, and P. G. Kennedy. 2016. FUNGuild: An Academic Press, New York, New York, USA. open annotation tool for parsing fungal community datasets by Yatskov, M., M. E. Harmon, and O. N. Krankina. 2003. A ecological guild. Fungal Ecology 20:241–248. chronosequence of wood decomposition in the boreal forests of Niemela,€ T., P. Renvall, and R. Penttila.€ 1995. Interactions of fungi Russia. Canadian Journal of Forest Research 33:1211–1226. at late stages of wood decomposition. Annales Botanici Fennici Yergeau, E., S. Bokhorst, S. Kang, J. Zhou, C. W. Greer, R. Aerts, 32:141–152. and G. A. Kowalchuk. 2012. Shifts in soil microorganisms in Ottosson, E., J. Norden, A. Dahlberg, M. Edman, M. Jonsson,€ K.-H. response to warming are consistent across a range of Antarctic H. Larsson, J. Olsson, R. Penttila,€ J. Stenlid, and O. Ovaskainen. environments. ISME Journal 6:692–702.

SUPPORTING INFORMATION Additional supporting information may be found in the online version of this article at http://onlinelibrary.wiley.com/doi/10.1002/ecy. 2165/suppinfo

DATA AVAILABILITY Data and code associated with this study are available from Zenodo: https://doi.org/10.5281/zenodo.1146732.