Statistical analyses
All statistical analyses were done in R version 3.6.1 (R Core Team 2019).
The homeostatic coefficient 1/H was calculated to analyze the response of fungal stoichiometric ratios to varying nutrient supply in glucose and cellulose media (Persson et al. 2010). For example, 1/HCN represents the regression slope of fungal log10 (C:N) correlated with log10 (C:N) in media, and takes values between 0 (strictly homeostatic) and 1 (non-homeostatic) (for more details see also Sterner and Elser (2002)). The same was calculated for 1/HCP and 1/HNP values, for the latter using the whole gradient of N and P manipulation in glucose media. Differences in isolate-specific 1/H values among different element gradients were calculated by paired t-tests. Variations in fungal stoichiometric ratios in fertilized SEA media were analyzed by linear mixed-effects models, taking isolate as random effect into account (lme(); package nlme (Pinheiro et al. 2020)). Here, in case of non-normality, data were log-transformed.
Details of the results and analyses of fungal growth responses to varying nutrient supply in glucose, cellulose and SEA medium are shown in Fig. S2, S3 and S4, respectively.
Total fungal element masses [mg] were correlated with fungal biomass [mg] in each medium type, in order to understand the stability of different elements in fungal mycelia under varying resource conditions. In this case, data were normalized within isolates for each element and medium type, respectively, by standardizations between 0 and 1. Model outputs are based on linear random slope models, using isolate as random factor. Explanatory power of biomass versus nutrient manipulation on total fungal element masses were assessed by the comparison of sums of squares, evaluated by type III analyses of variances of model outputs (package lmerTest (Kuznetsova et al. 2017)).
To assess simultaneous decreases in fungal N and P concentrations [%] in both, N and P manipulations, linear correlations were done by linear-mixed effects models with isolate as random factor for each medium and nutrient manipulation, respectively. Relative deviations in element concentrations [% deviation from maximum values of each isolate] were used to compare effect sizes of element reductions among isolates. In SEA media, added carbon sources were included as fixed effect, since element concentrations were significantly affected by cellulose and glucose additions. Marginal R² values were calculated using function r.squaredGLMM() (package MuMIn (Barton 2019)).
Differences in fungal C:N values of inner versus outer mycelium in SEA medium, as well as of fungi growing 12 or 26 days, respectively, were determined by paired t-tests (one-sided). To evaluate the effects of treatment (low vs. high N), position (in vs. outer mycelium) and isolate on fungal C:N and C:P values in low and high N glucose media, three-way type III analyses of variances were applied.
The predictive power of stoichiometric ratios for fungal growth and activity was determined by linear correlations of growth responses (namely fungal biomass, density, enzymatic activity and the ratio of leucine-aminopeptidase and phosphatase activity) with stoichiometric ratios, applying linear mixed-effects models (lmer(); packagelme4 (Bates et al. 2012)). Marginal (variance explained by fixed effects only) and conditional (variance explained by fixed effects and random effects due to varying intercepts among isolates) R² values were calculated with function r.squaredGLMM() .