Fungal strains were isolated from 18 sampling sites in Yunnan province, China, from 2012 to 2017 (Supplementary Table 1). Isolates were obtained from healthy leaves, leaf spots, the stem, and leaf litter of Ageratina adenophora. The canopy air and rhizosphere soil associated with A. adenophora were sampled, as were the leaf spots of native plants from invaded and noninvaded ranges. Native plants growing immediately next to A. adenophora were selected for leaf spot collection. If such plants were unavailable, native plants growing close (<5 m) to A. adenophora were selected instead.

To isolate leaf-based endophytes, the fourth pair of healthy leaves from the top of fully grown A. adenophora, which harbor the greatest fungal diversity (Jiang et al., 2011), were collected from July to August. Endophytes were obtained using the leaf fragment incubation technique (Arnold and Lutzoni, 2007). To obtain sufficient pathogen diversity, in any one sample site, diseased leaves with morphologically different symptoms were collected from at least three individuals. Healthy leaf tissues and the margins of diseased tissues of each leaf spot were cut into six sections (6 mm²) and the surface was sterilized. The sterile fragments were then plated into PDA and incubated at ambient temperature for 6–8 days or until mycelia were observed growing from the leaf fragments.

For collection of the rhizosphere soil, plants were dug out and lightly shaken, and the soil that remained attached to the root surface was carefully collected using a brush. Rhizosphere soils were packed into polyethylene bags and placed on ice packs in a cooler after collection and immediately transported to the laboratory where they were stored at 4°C until processing. To obtain soil suspensions, soil samples (10 g) were placed in triangular bottles containing 100 ml of sterile water and several glass beads and incubated at 28°C for 30 min with shaking (180 rpm). Then, 0.1 ml of soil suspension (10⁻² and 10⁻³ dilutions) was placed on a PDA dish and incubated at ambient temperature for 6–8 days. Fungi were collected from the air by exposing PDA Petri dishes 60 cm above the canopy for 15 min. The Petri dishes were then sealed using a sealing membrane and returned to the laboratory where they were incubated at 28°C for 3–4 days.

All fungal colonies were purified and used for the determination of their phylogenetic position based on the ITS locus sequence. Those belonging to the family Didymellaceae were selected and further subjected to multi-locus phylogenetic analyses and disease experiments. All fungi were maintained as pure cultures at Yunnan University (Kunming, China).

Total genomic DNA was extracted from fresh mycelia using the CTAB method (Murray and Thompson, 1980). The first ITS (ITS1) region was amplified with the ITS4 and ITS5 primers (White et al., 1990); LR5 and LR7 (Vilgalys and Hester, 1990) were used for large subunit (LSU) rRNA (LSU rRNA) amplification; Btub2Fd and Btub4Rd (Woudenberg et al., 2009) were used for the partial amplification of the β-tubulin (tub2) gene; and finally, RPB2-5F2 (Sung et al., 2007) and fRPB2-7cR (Liu et al., 1999) were used for the amplification of the second-largest subunit of RNA polymerase II (rpb2). Amplicons for each locus were generated by Sanger sequencing following the protocols listed in Chen et al. (2017). Before multi-locus phylogenetic analyses, the sequences of all the loci generated in this study were used as queries to search similar DNA sequences in GenBank using BLAST. The isolates displaying the highest sequence similarity to Didymellaceae species were selected for analysis. The sequences of isolates and reference strains based on Chen et al. (2017) and Valenzuela-Lopez et al. (2018) were aligned and combined using BioEdit version 7.0 (Hall, 1999).

Bayesian (BI) analysis was performed on MrBayes v.3.2.1 (Ronquist et al., 2012) based on the models selected by MrModeltest v.2.3 according to the protocol described by Chen et al. (2017). The best-fit models of evolution for the four loci tested (GTR + I + G for ITS, LSU, rpb2, and tub2) were estimated by MrModeltest. Leptosphaeria conoidea (CBS 616.75) and L. doliolum (CBS 505.75) were selected as outgroup species. To identify the phylogeny of the isolates, the fungi identified by Chen et al. (2017) were used as the reference strains in the phylogenetic analysis (Supplementary Table 2).

Based on the multi-locus phylogenetic analysis, 131 isolates belonging to 11 genera were selected to evaluate virulence against A. adenophora and 14 native plant species commonly found in Yunnan province (Institute of Botany, 2019), including three trees [Celtis tetrandra (Ulmaceae), Cyclobalanopsis glauca (Fagaceae), and Lindera communis (Lauraceae)], five vines [Ampelopsis delavayana (Vitaceae), Fallopia multiflora (Polygonaceae), Petunia hybrid (Solanaceae), Pueraria peduncularis (Fabaceae), and Zehneria maysorensis (Cucurbitaceae)], five shrubs [Urena lobata (Malvaceae), Abelmoschus moschatus (Malvaceae), Achyranthes bidentata (Amaranthaceae), Rubia cordifolia (Rubia cordifolia), and Reinwardtia indica (Linaceae)], and one grass [Arthraxon hispidus (Poaceae)]. A disease experiment was performed in the field as previously reported for the testing of the virulence of necrotrophic pathogens in tropical forests (Gilbert and Webb, 2007), with a minor modification. Briefly, fungi were grown on PDA for 7 days, after which agar disks (6 mm²) containing fungal mycelia were used to inoculate mature and healthy leaves. Small wounds were made by lightly touching the underside of the leaf with a sterilized toothpick, resulting in a wound area of ~0.2 cm². The inoculum agar was pressed against the wound on the underside of the leaf using Scotch tape and clipped in place with a bent hair clip. Control inoculation was performed with agar without fungal mycelia. One week after inoculation, the leaves were harvested and photographed, and symptomatic areas were measured. Because the control inoculation showed only a minimal wound reaction, strains inducing significant wound symptoms were considered pathogenic. The field site was located in Xishan Forest Park, Kunming, at an altitude of 2,214 m, a latitude of 24°58′24″ N, and a longitude of 102°37′17″ E. The experimental period ran from June to the end of October 2017–2018, the primary growing season for plants in Kunming.

After this analysis, 13 potential biocontrol agents were screened for virulence in an additional 15 hosts that included Alnus nepalensis (Fagales), Amphicarpaea edgeworthii (Fabaceae), Bambusa emeiensis (Poaceae), Cynanchum otophyllum (Apocynaceae), Hypoestes triflora (Acanthaceae), Michelia figo (Magnoliaceae), Myrica rubra (Myricaceae), Oreocnide frutescens (Urticaceae), Parthenocissus tricuspidata (Vitaceae), Quercus variabilis (Fagaceae), Rosa xanthina (Rosaceae), Rubus parvifolius (Rosaceae), Sechium edule (Cucurbitaceae), Smilax scobinicaulis (Smilacaceae), and Zanthoxylum bungeanum (Rutaceae). The disease experiment was performed as described above. Four Epicoccum isolates and one Didymella isolate served as positive and negative controls, respectively.

To screen candidates for their biological control potential against A. adenophora, three isolates with high virulence and a narrow host range were selected for the validation of virulence in vitro in leaves of A. adenophora. Meanwhile, to determine if the host, A. adenophora, evolves resistance to fungal virulence, the tested plants were grown from seeds collected from A. adenophora populations with 40 and 80 years of invasion history (Wang and Wang, 2006). Fresh hyphae (3 g) were crushed in a magnetic stirrer and diluted in 12 ml of a solution containing 10% sterilized glycerinum. A volume of 10 μl of glycerinum with hyphae was dripped into the back of injured and uninjured leaves that had been placed in a Petri dish moistened with 10% glycerin and sterilized filter paper. The size of the diseased spot was recorded daily. Three or more replicates were set for each strain. Because leaf spots covered whole leaves in some isolates by day 5 after inoculation, day 4 was selected for the evaluation of virulence against A. adenophora of differing invasion histories. Meanwhile, 10 ml of glycerinum with hyphae was used to spray 12 plants (6 with injured and 6 with uninjured leaves) and the disease spots were recorded. All leaf spots were reisolated, thus achieving Koch’s postulates.

To construct a plant–pathogen evolutionary association network, three DNA sequences—rbcL, matK, and ITS1—were obtained from GenBank for each of the 15 plant species (Supplementary Table 2). Two early-diverging gymnosperm species—Abies alba and Cycas rumphii—served as the outgroup (Chen et al., 2019). Also, isolates used for the evaluation of virulence were selected for further analysis. The phylogenetic tree of the hosts was then built using the RAxMLGUI v1.5 algorithm (Stamatakis, 2014) under a maximum likelihood model with a General Time Reversible (GTR) + Gamma nucleotide substitution model of evolution (1,000 bootstrap replicates). After removing the two outgroup species, a final molecular phylogeny was obtained for the 15 host species. The phylogenetic tree of all these pathogenic isolates was built using BI analysis, as mentioned above. Global and individual ParaFit tests (Legendre et al., 2002; Chen et al., 2019) were performed to examine whether there was a non-random plant–pathogen association network.

Heatmaps for leaf spot sizes were plotted using the R package pheatmap (Kolde, 2018). The Kruskal-Wallis test (for three and more groups) and the Mann–Whitney U test (for two groups) were used to test between-group differences in SPSS v.25.0 (IBM, Chicago, IL, USA). Diagrams were plotted in GraphPad Prism 7 (GraphPad Software, Inc., La Jolla, CA, USA).

A total of 213 Didymellaceae isolates were collected, including 189 (89%) from 12 sites with and 24 (11%) from 6 sites without Ageratina adenophora invasion. Of these, 195 (92%) were obtained from plant samples and 18 (8%) from environmental samples, including rhizosphere soil (2 isolates) and air (16 isolates). Most of the strains (145 isolates) were isolated from A. adenophora (101 from leaf spots, 29 from healthy leaves, 5 from stems, 7 from roots, and 3 from rotted leaves); 50 isolates were obtained from native plant leaf spots, including 26 from invaded ranges and 24 from noninvaded ranges (Supplementary Tables 3 and 4).

Phylogenetically, as defined by ITS sequencing, a total of 52 unique OTUs and 6 OTUs based on 97% similarity were obtained. Multi-locus analysis identified a total of 131 OTUs (Supplementary Table 5). These fungi were distributed in 10 genera, including Epicoccum (58 isolates, 27%), Didymella (28 isolates, 13%), Remotididymella (12 isolates, 6%), Boeremia (5 isolates, 2%), Stagonosporopsis (5 isolates, 2%), Nothophoma (2 isolates, 0.5%), Allophoma (1 isolate, 0.5%), Leptosphaerulina (1 isolate, 0.5%), Neoascochyta (1 isolate, 0.5%), and Paraboeremia (1 isolate, 0.5%), as well as two new groups (named Mesophoma speciosa, 80 isolates, 37%; M. ageratinae, 19 isolates, 9%) (Supplementary Figures 1, 2, the novel genus report in other article Yang et al., 2022 pers. comm.). Isolates belonging to Didymella, Epicoccum, Remotididymella, and Mesophoma accounted for 93% of the total isolates (Figure 1A and Supplementary Table 5. The distribution of the seven main groups is shown in Figure 1B. Epicoccum, Didymella, Remotididymella, and Boeremia were isolated from both invaded and noninvaded ranges, as well as from leaf spots of both A. adenophora and native plants. However, isolates belonging to Stagonosporopsis, M. speciosa, and M. ageratinae were only collected from sites with A. adenophora invasion. Epicoccum, Didymella, and Stagonosporopsis isolates were obtained from canopy air, while Remotididymella was found in rhizosphere soil and canopy air.

Based on the multi-locus phylogenetic analysis, 131 typical isolates from both known genera and unclassified groups were selected for the field disease experiment (Figure 2). These fungi varied greatly in virulence and host range, even those in the same genus (Figure 2A). In particular, 21 strains belonging to Epicoccum showed extremely strong virulence and broad host range, and clustered primarily in a small clade phylogenetically close to species E. latusicollum and E. sorghinum (Figure 3, red frame).

The virulence of fungi against A. adenophora and native plants differed significantly among genera (p = 0.000, Kruskal-Wallis test) (Figure 2B). Both Boeremia and Epicoccum were strongly virulent toward all tested native plants, including A. adenophora; in contrast, Remotididymella and Didymella were only weakly virulent toward all tested hosts. Interestingly, most members from M. speciosa and M. ageratinae were weakly virulent toward native plants but strongly virulent against A. adenophora. The genus Stagonosporopsis was associated with median spot size in A. adenophora and large spot size in native plants. Host range also differed significantly among genera (p = 0.000, Kruskal-Wallis test; Figure 2C). Genera exhibiting relatively high virulence, namely, Boeremia, Epicoccum, and Stagonosporopsis, showed a wider host range than those displaying weak virulence (Didymella and Remotididymella); as the host number increased, Epicoccum replaced Didymella as the dominant pathogenic fungus. M. speciosa and M. ageratinae had a median host range (Figure 2C). There was no strong phylogenetic correlation between pathogen and host (global ParaFit tests, p = 1), i.e., a wide range of phylogenetic strains could infect the same host, and the same phylogenetic strain could infect a wide range of hosts (Figure 4). Relatively, R. indica was the most susceptible host and 92% of the isolates were virulent against it; in contrast, C. glauca, with only 44% of virulent isolates, was the most resistant to Didymellaceae infection.

For 18 isolates with high virulence against A. adenophora, we expanded the virulence test to a further 15 native plants (Figure 5A). Again, compared with other genera, Epicoccum exerted the strongest virulence and had the widest host range. Three isolates—S188 (M. speciosa), G56 (M. ageratinae), and Y122 (M. ageratinae)—displayed strong virulence toward A. adenophora but were only weakly virulent toward native plants and had a narrow host range (Figure 5B).

Using detached leaves, three isolates—S188 (M. speciosa), G56 (M. ageratinae), and Y122 (M. ageratinae)—were further tested for their virulence toward A. adenophora with different invasive histories (Figure 6). Disease development after inoculation with either Y122 or S188 was slightly faster than that for inoculation with G56; however, leaf spot size did not differ for all three isolates (Mann–Whitney U test, p > 0.05) (Figure 6A). On average, the lesion size of uninjured leaves was even slightly greater than that of injured leaves; isolates Y122 and S188 exerted stronger virulence than G56 (Figure 6B). The spray experiments using seedlings showed that Y122 and S188, but not G56, could cause lesions in the leaves of A. adenophora (Supplementary Figure 3).