Statistical Analyses

All statistical analyses were performed at a level of 100% (unique) and 97% ITS sequence identity. We pooled the colonies from the three plates of each medium, and the three types of medium were used as biological replicates for each sample because there were no significant differences in the community composition (Supplementary Figure S1), species diversity (Supplementary Figures S2A,B), and isolate abundance (Supplementary Figure S2C) among the three types of medium, with the exception that fungal community variation was significantly associated with the medium at 97% sequence identity (Supplementary Figure S1B; F = 2.2, p = 0.002) but only explained very little variation [0.8% adjusted explained variation (AEV)]. Communities were defined as the sequenced fungal strains obtained from a given plant tissue, site, or environmental source.

Non-metric multidimensional scaling (NMDS) was used to visualize the similarity of fungal communities among plant tissues or environmental samples, and principal coordinate analysis (PCoA) was used to assess the spatial variation of endophytic and withered leaf fungal communities in different plant tissues during the geographical expansion of A. adenophora. At the unique sequence identity level, two outliers [NE–RS–NS and NE–AS–MEA; NE represents the sampling site, RS (rhizosphere soil) and AS (ambient soil) represent the source, and NS and MEA represent the culture medium; detailed in Supplementary Table S2] were eliminated in the NMDS analysis of the environmental samples. The correlation between fungal communities and the classified factors was tested by redundancy analysis (RDA). All distance matrices for community composition analyses were based on the Bray–Curtis dissimilarity index.

The Shannon diversity index was calculated for fungi from both plant tissue and the environment. ANOVA was used to compare the diversity and abundance of fungi (containing endophytic and withered leaf fungi) across plant tissues and geographical sites. Assumptions of normality were tested by Kolmogorov–Smirnov test, and the assumptions were met. Post hoc comparisons were performed using Duncan’s tests for equal variance and Dunnett’s T3 tests for unequal variance. Linear regression was used to analyze the relationship between the Shannon diversity index or average abundance and latitude. Mann–Whitney U non-parametric tests were used to compare the diversity and average abundance of fungi from different environmental sources. Fungal relative abundance was calculated to analyze the relationship between plant tissue and environmental fungi.

Multivariate analyses were performed using CANOCO, version 5.0 (Ter Braak and Šmilauer, 2012), and all other analyses were executed using SPSS version 22.0 (SPSS Inc., Chicago, IL, United States). Visualization of diversity and abundance data were realized with GraphPad Prism 7 (GraphPad Software Inc., San Diego, CA, United States).