Trichoderma fungi support sustainable agriculture by suppressing plant diseases and improving crop performance. However, emerging pathogenicity of Trichoderma warrants further ecological and genetic characterization. Here we used machine learning to correlate genomic data from 37 Trichoderma strains with over 140 phenotypic traits, spanning metabolic versatility, biotic interactions, stress tolerance and reproductive strategies. We determined Trichoderma to be an ancient, genetically cohesive and physiologically diverse genus with spores capable of germination in water and dispersal via air and water droplets. Metabolic preferences indicate universal adaptation to mycoparasitism and to niches like arboreal microbial mats, alongside broader saprotrophic versatility. Our analyses are consistent with character displacement among close relatives and convergent evolution in distant lineages, with both processes shaping ecological plasticity and traits including dispersal modes, terrestrialization or endophytism. Our findings reveal that while some Trichoderma species show traits of biosafety concern, its vast ecophysiological diversity enables the development of safe, targeted bioeffectors.
The use of microbes in sustainable agriculture requires careful assessment of their ecophysiological potential and adaptation mechanisms to minimize risks of persistence, non-target effects and pathogenicity, ensuring safe application. Some fungi from the genus Trichoderma (Hypocreales, Ascomycota) are recognized as versatile, plant-beneficial microorganisms 1,2 . When introduced to agricultural soil, Trichoderma can establish in the rhizosphere of most crops, enhancing their immune responses, stress tolerance, growth and yield. As innate mycoparasites, Trichoderma strains can directly antagonize plant-pathogenic fungi and nematodes, thereby influencing crop microbiomes 3 . Consequently, large quantities of Trichoderma spores are produced globally and incorporated into agricultural lands as biostimulants and plant protection products, with use projected to increase under policies supporting sustainable agriculture and reduced chemical pesticide application 4 .
Trichoderma comprises over 500 species 5 (www.Species Fungorum.org, accessed on 23 September 2025), most occurring on decaying wood or as mycoparasites 2 . Among many others, this lifestyle is represented by phytosaprotrophic and mycoparasitic Trichoderma reesei, also exploited for cellulolytic enzyme production 6 . However, agronomic uses rely on a small set of cosmopolitan, putatively edaphic species commonly found in soils and rhizospheres, including Trichoderma afroharzianum, Trichoderma asperellum, Trichoderma asperelloides, Trichoderma atroviride, Trichoderma harzianum, Trichoderma longibrachiatum and Trichoderma virens [1][2][3] . Although not closely related 7 , these species share ecological traits of environmental opportunism such as potent mycoparasitism, rapid growth in vitro and prolific sporulation. Some can infect immunocompromised humans 8 and/or cause green mould outbreaks in mushroom farms 9 , while others colonize woody stems, roots and leaves as endophytes [1][2][3] . Reports of crop pathogenicity are also increasing [10][11][12][13] , including recent T. afroharzianum ear rot on maize 10 , which led to the inclusion of this species on the European and Mediterranean Plant Protection Organization (EPPO) Alert List (www.eppo.int, accessed October 2025).
Because conditions that trigger conidial germination largely determine where Trichoderma can initiate its life cycle, we assessed spore germination across 95 carbon sources using BIOLOG FF microplates. l-Phenylalanine substantially improved early development in most Trichoderma strains, with glycogen also effectively enhancing germination. Other compounds such as adenosine, d-arabitol, i-erythritol and l-alanine supported germination to varying degrees, indicating strain-specific responses (Extended Data Fig. 1). Species used in agronomy like T. atroviride and Trichoderma guizhouense preferred glycogen and showed no response to l-phenylalanine, suggesting adaptations to nutrient-rich environments. Most strains, particularly T. atroviride, T. guizhouense and T. harzianum, germinated as effectively in pure water as in many carbon sources tested, reflecting adaptations to oligotrophic and moist habitats. Such carbon sources as fumaric acid, succinic acid and n-acetyl sugars were unfavourable for life-cycle initiation. This evaluation also revealed species-specific ecological adaptations: slower development in T. harzianum suggests parsimonious resource utilization beneficial in low-nutrient conditions, whereas Trichoderma velutinum Triveli and Trichoderma zeloharzianum showed rapid germination, indicating a competitive advantage on quickly colonizable substrates (Extended Data Fig. 1).
The genus-wide ability to germinate in water, together with stimulation by l-phenylalanine, glycogen and related compounds, suggests that Trichoderma life-cycle initiation may be particularly enhanced in niches where amino acids, storage polysaccharides and sugar alcohols are locally enriched. Such conditions occur, for example, in phyllosphere microbial mats and other plant-associated microhabitats, where these compounds are reported as common microbial osmoprotectants 17 .
The next vegetative stage, characterized by exponential growth, revealed consistently broad nutritional versatility across Trichoderma when tested on 95 carbon sources (Extended Data Figs. 2 and 3). Variability between clades was minor, but related species and strains showed marked differences, indicating diverse idioadaptations (fine-scale, lineage-specific ecological adjustments). Overall for the genus, at least 41 substrates and water supported growth, 36 were suboptimal, and 18 yielded biomass no greater than the water control.
The optimal substrate for all Trichoderma strains was N-acetyl-dglucosamine, the chitin monomer of fungal cell walls and arthropod exoskeletons, underscoring its mycoparasitic capacity, competitive role in fungal communities and occasional insect parasitism 2 . Sugar alcohols such as glycerol, d-sorbitol, i-erythritol, d-arabitol and d-mannitol also promoted robust growth, reflecting adaptation to osmoprotectants abundant in arboreal microbial mats (phylo-/xylosphere) 17 . These complex polysaccharide-and stress-protectant-rich habitats contrast with the simpler, root-exudate-driven systems of rhizosphere and bulk soil 18 .
Sugars from plant biomass decomposition-such as maltotriose, cellobiose, trehalose, glucose, fructose, mannose, ribose, galactose, melibiose and xylose-also supported robust but strain-specific growth, suggesting effective utilization in soil and rhizosphere environments where complex organic matter is abundant (Extended Data Fig. 2, and Supplementary Data 1 and 2).
Moderate growth on substrates such as other osmoprotectants (raffinose, lactulose), phenolics (arbutin), sugar alcohols (xylitol), unusual metabolites (ʟ-phenylalanine, ʟ-fucose) and polysaccharides (glycogen, dextrin) suggests broad ecological adaptability, although these rarely supported peak growth and may act as secondary energy sources or germination triggers (vide supra). By contrast, limited growth on specialized substrates-particularly organic acids (fumaric, succinic) Table 1. Icons summarize major ecological roles and reproductive modes. Genome-size bars show protein-coding gene (PCG)
counts normalized to T. atroviride (Supplementary Table 3). Pie charts summarize taxon-specific HOG composition (PFAM-annotated genes, SSPs and genes of unknown function); numbers above pies indicate the number of taxon-specific HOGs, and percentages below indicate the fraction of co-evolving genes. The inset (upper left) shows general average temperature (GAT, °C) across geological epochs 15 , with the Cretaceous-Paleogene (K-Pg) boundary and the warm, humid period that favoured tropical rainforests (relevant for Trichoderma evolution) highlighted. Time periods are abbreviated as follows: N, Neogene; Q, Quaternary.
https://doi.org/10.1038/s41564-026-02260-3 and amino acids (l-glutamic, l-asparagine)-indicates metabolic stress and points to environments (for example, acidic, fermentative) unfavourable for Trichoderma colonization (Extended Data Fig. 2).
Similar to spore germination, growth preferences showed no strict concordance with phylogeny, implying adaptive convergence. Extended Data Fig. 3 illustrates highly strain-specific 'nutritional barcodes'. Three Harzianum clade strains (T. zeloharzianum, Trichoderma lixii, Trichoderma breve), recently isolated from tropical leaves, grew exceptionally fast on d-sorbitol and other sugar alcohols, suggesting phyllosphere adaptation. Several Harzianum strains also used N-acetyl-d-mannosamine, a sialic-acid precursor typical of animal tissues and some prokaryotes 19 , pointing to possible links with insect gut microbiomes or vertebrate mucosa. In addition, selected Viride clade strains related to T. atroviride metabolized sucrose and maltose, implying potential for endophytic development (Extended Data Fig. 2).
In summary, the nutritional profile of Trichoderma aligns with its common roles as mycoparasite and saprotroph on microbially colonized wood, as endo-and epiphyte, and in the rhizosphere. It also suggests involvement in microbial interactions within arboreal and rhizospheric biofilms and confirms that these fungi can exploit plant biomass.
To further assess fitness-related traits of Trichoderma, we quantified asexual reproduction using REPAINT (Reproduction Potential Artificial Intelligence 20 ), which analyses time series images from BIOLOG microplates to extract numerical data on hyphal growth and conidial coverage.
N-Acetyl-d-glucosamine, sugar alcohols and simple sugars generally supported extensive aerial mycelium and abundant conidiation (Supplementary Data 3). By contrast, m-inositol, glycogen and dextrin promoted conidiation with little aerial mycelium. Certain amino acids induced aerial hyphae without conidiation in nearly all strains, suggesting either exploratory growth under metabolic stress or initiation of sexual reproduction. Many suboptimal carbon sources supported vegetative growth without conidiation.
We identified four distinct reproductive strategies with marked interspecific variability, suggesting underlying speciation processes (Fig. 2). T. asperellum consistently produced abundant conidia with little aerial mycelium, indicative of a copiotrophic, r-selected strategy favouring rapid reproduction in resource-rich environments; this pattern also characterized the mushroom pathogens Trichoderma pleuroticola and Trichoderma amazonicum and the endophyte Trichoderma endophyticum. By contrast, T. reesei and Trichoderma minutisporum, typically known in nature from sexual stages (Fig. 1), formed dense aerial mats with few conidia, reflecting a K-selected strategy emphasizing competitive ability and resource efficiency. Similar behaviour occurred in phylogenetically unrelated species such as Trichoderma spirale and Trichoderma brevicompactum. A third strategy-rapid vegetative growth with moderate conidiationwas characteristic of agronomically used species like T. atroviride and T. guizhouense, possibly compatible with adaptations to endophytic development. Finally, T. harzianum, T. afroharzianum and T. virens showed high reproductive plasticity, producing both dense aerial mycelium and conidia across diverse substrates, exemplifying ecological opportunism.
Variation was especially pronounced among close relatives, notably between T. velutinum strains and the formerly cryptic sympatric species T. asperellum and T. asperelloides 21 (Extended Data Fig. 4), pointing to divergent speciation within these clades. Such sharp differences in reproductive strategy among closely related and co-occurring taxa are suggestive of character displacement (vide infra) and idioadaptive processes likely driving speciation in Trichoderma.
We quantified pluviophilous (rain-mediated) and anemophilous (wind-mediated) spore dispersal, along with survivability under temperature extremes and desiccation (Fig. 3 and Supplementary Fig. 1). About one-third of strains were anemophilous, another third pluviophilous, and the remainder produced conidia not aligned with these mechanisms, likely dispersed through other media, for example, zoochory.
Trichoderma spores were more vulnerable to desiccation than to temperature extremes, with many strains failing to survive (Fig. 3). Air-dispersed conidia resisted extreme temperatures, whereas water-dispersed ones were highly susceptible to desiccation (Supplementary Fig. 1). Cold-resistant spores also tolerated heat, but desiccation tolerance was uncoupled from thermal resistance, indicating distinct adaptive mechanisms.
Spores of T. afroharzianum, widely used in agriculture and recently reported as a maize pathogen 10 , showed strong stress resistance and anemophilous dispersal. By contrast, T. asperellum, despite prolific conidiation (vide supra), dispersed poorly by air or water, suggesting reliance on alternative strategies. These contrasting profiles illustrate adaptive radiation within the genus and its occupation of diverse ecological niches.
We quantified fitness-related parameters across the Trichoderma life cycle, including growth rates on various carbon sources, reproductive potential, spore dispersal and stress resilience. These data were integrated into a relative fitness map, showing growth on universally optimal N-acetyl-d-glucosamine and differentiating substrates such as N-acetyl-d-mannosamine, saccharides and glycogen (Fig. 3). Metrics, combined with averaged REPAINT profiles and spore properties, were normalized to T. atroviride-closely related to the type species Trichoderma viride 5 -and plotted against the phylogenomic tree as in Fig. 1. Whereas Fig. 3 highlights variability in fitness traits, Extended Data Fig. 5 shows that genome inventories (proteins, transcription factors, transporters, enzymes, biosynthetic clusters) remain relatively compact. This asymmetry suggests that phenotypic differences among strains are greater than variation in genome content (Supplementary Table 3).
The relative fitness map revealed substantial ecophysiological diversity, with greater differences among closely related species than between clades and similarities between unrelated strains, consistent with processes such as character displacement 22 speciation and convergent adaptation within the genus (Fig. 3). Sympatric strain groups (for example, Tri5505-Tri5757, Trisp-Trien in Harzianum; Tristr and Triev in Viride) showed marked disparities in fitness despite co-isolation. Strains of T. amazonicum (Triam, Tripleu), which diverged ~0.4 Ma (Fig. 1), also contrasted strongly, with Triam outperforming Tripleu in this set-up. At deeper scales, contrasts appeared between Trivel and Triveli (T. velutinum) and between T. asperellum and T. asperelloides (~4-6 Ma divergence), the latter known to co-occur in nature 21 . Similarly, T. guizhouense (Trigui) and T. harzianum (Triha) strains, although not sympatric in this dataset, are reported to co-occur 20 at the species level. These examples highlight multiple instances where closely related taxa diverge in trait profiles, reflecting evolutionary dynamic consistent with possible character displacement.
In this approach, T. atroviride-widely marketed for mycoparasitism, plant immune induction and stress resilience 1,2 -was outperformed by T. afroharzianum and T. guizhouense, both showing greater conidial survivability (Fig. 3). Conversely, T. asperellum, despite prolific conidiation, showed poor dispersal and strong temperature sensitivity, consistent with adaptation to stable habitats. Such variability in fitness traits (Fig. 3) contrasted with the compact distribution of genome features (Extended Data Fig. 5), underscoring that no single strain dominates across metrics; each performs best in conditions matching its ecological niche.
https://doi.org/10.1038/s41564-026-02260-3 grew poorly in bulk soil (three soil types, sterile and non-sterile), with only T. virens and T. asperellum developing in soil if sterilized. Nevertheless, all strains promoted plant growth (Supplementary Fig. 2), indicating rhizosphere competence independent of bulk-soil growth. In general, Trichoderma also showed a clear preference for protein-rich substrates (fungal biomass, soybeans) over nitrogen-poor linear growth rate (bubble size, measured as optical density (OD 750 ) per 24 h; strain-and carbonspecific values are listed in Supplementary Data 1-3) on 95 carbon sources and water, and the corresponding development of aerial hyphae after 96 h (x axis) and conidiation after 120 h (y axis), expressed as percentage of well surface coverage. b, Principal component analysis of developmental profiles of 37 Trichoderma strains based on the complete REPAINT and BIOLOG dataset (Supplementary Data 1). Axes indicate PC1 (41.5% variance) and PC2 (32.5% variance). Bubble size represents the mean growth rate across all carbon sources, used here as a synthetic summary measure. c, Schematic diagram of the Trichoderma life cycle, illustrating asexual reproduction and the formation of aerial mycelium putatively associated with sexual development. Colours correspond to four putative reproductive strategies, consistent across panels: green, asexual/clonal r-strategy (abundant conidia, minimal aerial hyphae); yellow, putative sexual K-strategy (dense aerial mats, few conidia); blue, putative endophytic strategy (rapid vegetative growth, moderate conidiation); purple, holomorphic/opportunistic strategy (plastic responses across substrates).
To further probe ecophysiology of Trichoderma, we assessed multiple qualitative traits, finding universally strong proteolysis, robust mycoparasitism and negligible antibacterial activity (Supplementary Data 4). Despite frequent rhizosphere applications 1 , all species Enzyme activities such as amylase, cellulase and lipase were common but highly variable among strains, with no correlation to each other or to phylogeny (Supplementary Data 4). By contrast, cutinase activity on polycaprolactone (PCL) was characteristic of the Harzianum clade, whereas Viride strains excelled in siderophore-mediated iron sequestration (Supplementary Data 4).
Species varied in their ability to form buoyant colonies and produce conidia while floating, underscoring the importance of this trait in Trichoderma biology (Supplementary Data 4). Overall, the genus shows distinctive traits with frequent divergence among relatives traits (development, stress tolerance and dispersal). Stars mark strains whose spores did not survive cold or drought conditions tested in this study, resulting in a product value of 0. Values are displayed on a logarithmic scale, highlighting the drastic differences in performance across strains. Strains are ordered according to the phylogeny in Fig. 1. cellulosic materials (for example, rice straw, magnolia leaves) (Supplementary Data 4).
Light generally enhanced spore production, although without genus-wide circadian rhythms; some strains also conidiated in darkness (Supplementary Data 4). Conidial rings and conidiation-associated guttation occurred across clades but often contrasted sharply between close relatives (for example, two T. velutinum strains), suggestive of character displacement. Most strains thrived under oligotrophic conditions and showed sensitivity to mechanical injury, which variably influenced conidiation.
https://doi.org/10.1038/s41564-026-02260-3 including those of sympatric origin (for example, suggestive of character displacement) and convergence among unrelated taxa, while few adaptations align consistently with phylogeny.
Because agronomic use entails intentional mass release 4 , the observed phenotypic diversity has direct bearing on biosecurity. We therefore contextualize some of these readouts in a first-tier biosecurity framework for agronomically relevant strains that integrates taxonomy, fitness profiling and pathogenicity records in humans and plants (Fig. 4).
We demonstrated that Trichoderma spp. maintain genomic cohesion while undergoing adaptive radiation and developing considerable ecophysiological variability that often leads to phenotypic convergence among genetically distinct taxa, including several species important in agriculture. Using Support Vector Machine (SVM) analysis with Python scikit-learn v0.17, we evaluated each strain genome against more than c, Species-level evidence. Blue flag symbols denote literature reports for clinical occurrence 8 and for plant pathogenicity in T. lixii 11 , T. afroharzianum 10 , T. virens 12 and T. longibrachiatum 13 . Species in bold are those with documented plant and human pathogenicity. Pathogenicity in mushrooms is not mapped here, because although the number of reported species is relatively limited 9 , mycoparasitism is an innate trait of the genus 28 , making it reasonable to assume that all strains carry potential risk for mushroom growers. Similarly, the production of mycotoxins-such as gliotoxin in T. virens 31 , phytotoxins including harzianic acid in the Harzianum species complex 58 , sorbicillins in section Longibrachiatum 59 and trichothecenes in T. brevicompactum 60 -is not indicated here, as a comprehensive genus-wide survey of secondary metabolites has not yet been performed. d, Examples of biosecurity-relevant phenotypes assessed in this study. Yellow flags summarize traits relevant to environmental persistence and spread: dispersal (air), stress tolerance and high fitness (fast growth, abundant conidiation, versatile nutrition). These properties and species-level reports serve as screening cues, not hazard designations; their presence highlights strains that warrant additional, context-specific risk evaluation before deployment.
https://doi.org/10.1038/s41564-026-02260-3 140 phenotypic parameters, among which 72 showed statistical significance. These parameters were catalogued in a binary matrix, highlighting the presence, absence or uncertainty (in cases of heterogeneous or intermediate responses across strains) of each trait (Supplementary Data 1). Our approach identified phenotype-associated orthogroups (PAOGs) that were statistically significantly correlated with distinct phenotypic traits, facilitating predictions in cases with ambiguous phenotypic profiles. Unlike phenologues, which are cross-species phenotype-phenotype relationships inferred from shared orthologous gene sets 23 , the PAOG framework identifies genotype-phenotype associations within a clade of closely related species. Statistical analysis revealed substantial gene segregation across quantitative 45 growth profiles on specific carbon sources and 27 qualitative or summative phenotypes. Among others, these included traits such as mycoparasitic vigour and mechanical injury responses ranging from stimulation to non-response (Fig. 5 and Supplementary Table 4). For example, phenotypes such as the degradation of PCL correlated with the enhanced presence of certain oxidoreductase enzymes (PF00106), aldolases (PF03328), protective proteins (PF04479, PF07976, PF01494) and SSPs of unknown function, which collectively may enhance polymer breakdown and mitigate potential toxic effects. Conversely, subtilases (PF00082), NWD (NACHT-WD repeat) NACHT-NTPase-related domains (nucleotide-binding oligomerization domain proteins) (PF17100) and several intracellular proteins potentially weaken the response to mechanical injury by reducing cellular damage.
The comprehensive analysis of PAOGs across 72 phenotypes identified 775 occurrences of 600 unique HOGs, with the majority transcribed and attributed to shell genomes (the portion of a genome that is neither unique to a particular strain (strain-specific) nor common to all species within a genus (core genome)) (Supplementary Table 4 and Supplementary Data 1). Among these, 145 occurrences (20%) are intracellular proteins of unknown function (Fig. 6a). The second most numerous group comprised SSPs, emphasizing their role in fitness-related adaptations. Enzymes and transporter proteins showed more association with nutrition-related phenotypes, while genes involved in stress mitigation were detected in association with all kinds of phenotypes.
The most profound group of PAOGs with known functionality encoded such regulatory proteins as DNA-binding transcription factors and 'fungal-specific transcription factors' (PF11951, PF04082) that do not bind DNA but are involved in transcriptional regulation in fungi 24 . Transcription factors, especially zinc clusters (PF00172), dominated the regulatory PAOGs, with a significantly higher prevalence compared to the whole genome, suggesting a focused functional role in regulatory processes. However, homeodomain and CCCH-type Zn-finger (ZnF) families, although common in genomes, were absent from PAOGs, indicating selective regulatory deployment (Fig. 6b). This trend points to a bias towards Zn clusters over other major families such as C2H2 ZnF, which are underrepresented in PAOGs. Further phylogenetic analysis revealed clade-specific usage of transcription factors, underscoring the complex regulatory landscape that facilitates adaptation of Trichoderma to diverse environmental conditions (Extended Data Fig. 6). Genes encoding ankyrin repeats (PF12796 and PF00023) were also frequently detected among PAOGs being mainly associated with nutrition (Supplementary Table 4).
The gene distribution analysis across Trichoderma infrageneric clades, normalized to T. atroviride, revealed a marked reduction of PAOGs in the Longibrachiatum clade (Fig. 6c), suggesting either a true biological absence-also evident for secondary metabolite clusters-or that the phenotypes assayed in this study did not adequately capture the functional specializations of this clade.
The aim of our study was to apply a genus-wide phenogenomic approach, integrating comparative genomics with ecophysiological profiling, to clarify Trichoderma ecology and evolution and to support risk assessment in agronomy. Combined with previous insights 1,2,7 and in situ observations, our data portray Trichoderma as a conservatively evolved, sylvan genus with predominantly arboreal habitats, adapted for mycoparasitism within phyllosphere-, xylosphere-and root-associated microbial biofilms in humid environments. Its paradoxical universal association with woody plants, despite the inability to degrade lignin 7,25 , likely represents a plesiomorphic trait retained from the Late Cretaceous period, when early angiosperms had less lignin than modern trees 14 . This inference is supported by nutritional profiles consistent across the genus and by ecophysiological traits such as green pigmentation of spores and strong, often non-circadian, response to light and reaction to rain impact, consistent with canopy adaptation 26,27 .
Furthermore, our findings indicate that Trichoderma spores can be dispersed not only by air but also by water droplets and potentially via other media (for example, zoochory), well suited to arboreal habitats. This is consistent with observations that some Trichoderma species, such as T. virens or T. harzianum, produce spores embedded in a gelatinous matrix 20 and are therefore not primarily airborne. A distinctive genomic hallmark of Trichoderma is the expansion of surface-active cysteine-rich SSPs such as hydrophobins and cerato-platanins 2,7 , which reduce water surface tension and modulate surface properties of fungal bodies and their substrates 20 . These traits likely also support adaptation to rainforest canopies. We further detected considerable enrichment of Trichoderma genomes in genes encoding expansins, the accessory proteins for the degradation of plant biomass, most active at moderately low pH and high activity of water 16 . Consistently, our assays showed that desiccation, rather than extreme temperatures, poses a greater threat to Trichoderma spores, supporting the hypothesis of the adaptation to high-humidity environments. Finally, many species show the ability to form buoyant colonies, potentially adapting to the water layers enveloping the tropical phyllo-or xylosphere. This trait, as shown by ref. 20, was under selective pressure in one of two sister Trichoderma species, and we demonstrate its relevance across the genus.
Our analysis confirms that Trichoderma spp. are universally mycoparasitic 28 , with strong proteolytic and lipolytic activities likely inherited from entomopathogenic ancestors. This plesiomorphic association with arthropods and fungi is also reflected in their germination and nutritional profiles. By contrast, the cellulolytic machinery of Trichoderma represents a more recent acquisition via lateral gene transfer from plant-pathogenic ascomycetes 25 , likely explaining its antagonism of cellulose-walled oomycetes 29 . As these DNA exchanges occurred gradually over ~30 Ma during the Late Cretaceous 25 , Trichoderma expanded rather than shifted its nutritional capacities, resulting in ecological generalism. Phenotype Prediction ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? D-Xylose Annotation of PAOGs Occurrence in genomes Phenotype association Response to mechanical injury Yes No Unclear Copy number ? Yes No 0 1 2 3 Degradation of polycaprolactone Protein kinase domain PF00069 FAD binding domain PF00890 Fructosamine kinase PF03881 FSTF PF04082 Ureidoglycolate lyase PF04115 Ankyrin PF12796 Methyltransferase domain PF13649 Unknown PAOG Unknown PAOGs Unknown PAOGs Unknown PAOGs SSP Cytochrome P450 PF00067 Protein kinase PF00069 Subtilase family PF00082 PF00550 PF07993 CorA-like Mg 2+ transporter PF01544 FSTF PF04082 : Fungal Zn(2)-Cys(6) TF PF00172 DUF1752 PF08550 DUF3328 PF11807 NWD NACHT-NTPase PF17100 Subtilase PF00082 Berberine PF08031, FAD binding PF01565 SnoaL-like PF12680 NWD NACHT-NTPase PF17100, NACHT PF05729, WD40 PF00400 SSP SSP SSP SSP Short chain dehydrogenase PF00106 Fungal Zn(2)-Cys(6) TF PF00172 HpcH/HpaI aldolase/citrate lyase PF03328 RTA1 like protein PF04479 Phenol hydroxylase PF07976, FAD binding domain PF01494 Probability of 0 Harzianum Clade Longibrachiatum Viride Probability of 1 Probability of 1 Probability of 1 Probability of 1 Tri5640 Triag Trialn Triam Tripleu Tripleur Trivel Triveli Trice Trispi Trist Trivi Trihel Trinov Trior Trieff Trire Trifla Trisin Tri5649 Tritai Trikon Triat Tristr Trias Triaspe Triev Trimi Tribrev Tri5757 Trisp Trien Tri5505 Trigui Triha Triinh T22 Although nutritional profiles suggest capacity for soil feeding, growth of Trichoderma in bulk soil was limited, likely by microbiome interactions. Only opportunists like T. asperellum and T. virens grew in sterilized soils; none thrived in non-sterile soil. Thus, few Trichoderma species are genuinely edaphic, although several persist in disturbed soils, litter and rhizosphere. This is consistent with generally low antibacterial activity, despite some strains producing gliotoxin (for example, T. virens 31 ), peptaibols or other mycotoxins 2 . Secondary metabolism, although relevant for biosafety, remains poorly understood; most compounds are taxonomically restricted (for example, gliotoxin in T. virens, sorbicillins in T. reesei 6 ) and preclude genus-wide comparisons.
Regulatory genes comprised the most abundant annotated PAOGs, with Zn cluster transcription factors significantly enriched in HOGs, suggesting a central role in phenotypic differentiation. By contrast, homeodomain and C2H2 ZnF transcription factors, although common
The branching of the Trichoderma phylogenetic tree loosely coincides with periods of climatic cooling in the Eocene 15 and subsequent rainforest declines, suggesting that speciation events may have triggered occasional terrestrialization (transition from arboreal to soil-associated niches) in unrelated species, leading to niche partitioning and character displacement among sympatric taxa and convergent evolution among allospecies. Our phenogenomic analyses provide a mechanistic view of this process, showing that even closely related species differ markedly in ecophysiology, consistent with character displacement. Although rigorous sympatry and phenetic mapping will be required to formally demonstrate this phenomenon in Trichoderma, our findings indicate that species boundaries are often defined by ecolophysiological rather than morphological traits. Integrating gene function with ecophysiological profiling will therefore be critical for distinguishing species and may ultimately render the concept of 'cryptic speciation' 30 in fungi obsolete. b Zn(2)-C6 Zn cl + FSTF Zn(2)-C6 Zn cl bZIP Zn(2)-C6 Zn cl + ZF, C2H2 + FSTF Zn(2)-C6 Zn cl + ZF, C2H2 ZF, C2H2 + FSTF ZF, C2H2 Homeodomain ZF, CCCH-type FSTF Genome (%) TFs and FSTFs in PAOGs compared to the genome * * * * * PAOGs (%) Proteins of unknown function Small secreted proteins (75) Intracellular (145) DUF (14) Enzymes PAOGs Pept (11) Ankyrin repeats (26) AAA (6) NACHT (4) MFS (20) ABC Tran (9) MCP(5) Protein kinase (10) GH (17) SDH (16) FAD binding (16) PKS (16) P450 (15) Aldo/ Keto Red (14) Alpha/ Beta Hyd (13) Signalling and structural proteins Transporters Others (225) Regulatory proteins (118) a Tribrev Tri5505 Trigui Tri5757 Trisp Trien Triha T22 Triinh Tri5640 Triag Trialn Triam Tripleu Tripleur Trivel Triveli Trice Trivi Trispi Trist Trihel Trior Trire Trieff Trifla Trisin Trinov Tri5649 Tritai Trikon Triat Tristr Triev 1 1 1 1 1 Trias Triaspe Trimi PAOGs Total proteins Transcription factors Secondary metabolites Normalized gene counts (relative to T. atroviride) Glycoside hydrolases c T. harzianum s.l. Longibrachiatum Viride Harzianum Fig. 6 | Trichoderma phenotype-associated genes detected in this study. a, Annotation of 600 unique PAOGs associated with 72 phenotypes where statistically significant segregation was observed. Numbers indicate occurrences (total 775) within the dataset (Supplementary Table 4 and Supplementary Data 1). b, Composition of transcription factor families among PAOGs compared with their distribution in the full Trichoderma genomes. Asterisks (*) indicate transcription factor families detected among PAOGs but not recovered from the remainder of the genome in this comparative analysis (that is, present only in the PAOG subset). c, Comparative counts of PAOGs, secondary metabolite clusters, glycoside hydrolases, transcription factors and total proteins across Trichoderma clades, normalized to the 369 HOGs in T. atroviride. The Longibrachiatum clade shows a marked reduction in PAOGs and secondary metabolite clusters, while glycoside hydrolases (are only slightly reduced in proportion to the smaller total protein count. GH, glycoside hydrolase; PKS, polyketide synthase; TF, transcription factor; P450, cytochrome P450 monooxygenases; MFS, major facilitator superfamily transporters; SDH, short-chain dehydrogenase/ reductase; FAD, flavin adenine dinucleotide (FAD-binding proteins/domains); Hyd, hydrolases (α/β-hydrolase fold proteins); Pept, peptidases/proteases; AAA, ATPases associated with diverse cellular activities; NACHT, NACHT NTPase domain (nucleotide-binding oligomerization domain); ABC Tran, ATP-binding cassette transporters; MCP, mitochondrial carrier proteins; DUF, domain of unknown function; FSTF, fungal-specific transcription factor domain; ZF, zinc finger; bZIP, basic leucine zipper transcription factors. Resource https://doi.org/10.1038/s41564-026-02260-3 in genomes of Trichoderma and other Ascomycota 24 , were absent from PAOGs, indicating selective specialization. Fungal-specific transcription factor domains, although not DNA-binding, often co-occurred with Zn clusters and were particularly enriched in Harzianum HOGs, indicating roles in regulation. Rare dual-domain transcription factors (Zn(2)-C6 Zn cluster + C2H2) also appeared in PAOGs, implying roles in sugar metabolism.
The prevalence of uncharacterized genes in PAOGs (Conserved Gene Families of Unknown Function, 2024 (https://mycocosm.jgi. doe.gov/conserved-clusters/run/run-2024;daFFLV); DOE JGI Myco-Cosm), many encoding SSPs, highlights a key limitation of fungal postgenomic research. These genes are likely critical for fitness and warrant further investigation.
Our findings have direct implications for biosafety assessments of Trichoderma-based products, particularly because agronomic deployment relies on intentional mass release 4 . In our dataset, agronomic species generally showed greater vigour than arboreal taxa, with T. afroharzianum producing the most resilient and environmentally persistent spores. The inclusion of T. afroharzianum on the EPPO Alert List for maize pathogenicity 32 underscores emerging phytosanitary concerns and the need for heightened attention to the lineages of this species and the entire T. harzianum complex. T. longibrachiatum likewise raises concern because it is frequently isolated from clinical cases 8 , occasionally linked to plant disease 13 , and shows high opportunistic potential. More broadly, section Longibrachiatum, which includes the most clinically relevant taxa in the genus 8 , warrants particularly cautious consideration when selecting strains for agronomic use. By contrast, the Viride clade appears more promising, with relatively few reports of toxicity or clinical relevance, although species such as T. asperellum and T. asperelloides, which combine prolific conidiation with bulk-soil growth, still require careful strain-resolved evaluation. On a genus-wide scale, the universal mycotrophy and mesophily of Trichoderma 2,28 indicate inherent risks for mushroom farms and immunocompromised humans, while its affinity for the phyllosphere points to potential facultative biotrophic interactions with plants. Consequently, foliar applications of Trichoderma demand careful scrutiny by agencies such as European Food Safety Authority and United States Environmental Protection Agency, with our trait-based evaluation serving as a first-tier screening framework to prioritize candidate strains for detailed regulatory assessment.
In summary, this phenogenomic study resolves several previously unclear aspects of Trichoderma biology-from hydrophobin expansions and stress responses to nutritional profiles-yielding a holistic picture of the genus as predominantly arboreal fungi adapted to tropical rainforest canopies. It highlights the need to view Trichoderma in its sylvan context rather than only through soil or agronomy. While highly opportunistic species pose biosafety concerns, many unexplored taxa may provide beneficial traits, and strains with limited dispersal capacity or vulnerabilities to abiotic stress could represent safer candidates for agronomic use. Integrating genomic insights with ecophysiological data into biosafety assessments will be key to ensuring sustainable applications that safeguard ecosystem health, human wellbeing and biodiversity.
Strains were sourced from the Westerdijk Fungal Biodiversity Institute Culture Collection, Utrecht, The Netherlands; the TU Wien Collection of Industrial Microorganisms (TU CIM), Vienna, Austria; and Szeged Microbiological Collection, Department of Microbiology, Faculty of Science and Informatics, University of Szeged, Hungary (Supplementary Table 1). DNA Barcoding of the internal transcribed spacer of the rRNA gene cluster, tef1 and rpb2 loci were used for species identification following a standardized protocol for Trichoderma 5 . Genus confirmation required internal transcribed spacer similarity of ≥76%, while species-level identification was based on rpb2 similarity of ≥99% and tef1 similarity of ≥97% to reference sequences of the respective type strain. The final molecular species assignment required monophyletic phylogenetic clustering with recognized Trichoderma species based on tef1 and rpb2 sequences. If the identification remained ambiguous due to insufficient reference material, the designation 'cf.' (con forma) was applied. Phylogenetic analysis was performed using the maximum likelihood method in IQ-TREE 2.4.0 (ref. 33), incorporating sequences of ex-type strains for all related species at each locus. Molecular identifications were further validated against morphological and ecological data. If species identification was not possible, the strain was considered a putative new species. The DNA barcoding sequences for the Trichoderma strains analysed in this study have been submitted to the National Center for Biotechnology Information (NCBI) GenBank database, with accession numbers provided in Supplementary Tables 1 and 2.
Fungal growth was monitored using BIOLOG FF Microplates (BIOLOG), which contain 95 wells, each with a different carbon source and one control well with water. Spore suspensions (10 7 spores per ml) were prepared in sterile distilled water and adjusted to uniform turbidity using a BIOLOG turbidity meter at OD 590 (optical density at 590 nm). Each well was inoculated with 90 μl of the suspension, and plates were incubated in darkness at 25 °C. Growth was measured spectrophotometrically at OD 750 at 12, 18, 24, 36, 48, 60, 72, 96, 120, 144 and 168 h after inoculation.
Spore production was quantified using the REPAINT method 20 , which combines high-resolution imaging with artificial intelligence-based image analysis. Images of the BIOLOG FF MicroPlates were captured at 72, 96, 120, 144 and 168 h after inoculation using a Canon EOS 70D camera equipped with a 100 mm macro lens. Wells were segmented using a machine learning algorithm (U-Net), which classified pixels based on coverage by aerial hyphae or conidia. The percentage of conidial coverage per well was calculated to quantify sporulation dynamics and reproductive potential. It is maintained and distributed by KOLAIDO GmbH and is available from the company's website (https://kmlvision. atlassian.net/wiki/spaces/KB/pages/3450011692/Fungi+REPAINT+A pp+Versions).
Spore desiccation tolerance was assessed by drying spore suspensions (10 8 spores per ml) at 40 °C for 4 days. After treatment, 100 μl of re-suspended spores was spread on 9 cm potato dextrose agar (PDA) plates supplemented with 0.5% Triton-X100, and colony-forming units (c.f.u.) were counted after 48 h of incubation at 25 °C. The results were accepted if supported by at least two dilution series; otherwise, a third replicate was performed.
Spore freeze resistance was tested by freezing spore suspensions (10 8 spores per ml) at -80 °C for 12 h, followed by lyophilization. Dried spores were resuspended in sterile water and plated on PDA supplemented with 0.5% Triton-X100. The c.f.u. were counted after 48 h of incubation at 25 °C.
Spore UV tolerance was evaluated by spreading 100 μl of a suspension with 10 3 spores per ml on PDA plates supplemented with 0.5% Triton-X100. Plates were exposed to UV radiation (95 μW cm -2 ) for 7 min, and c.f.u. were counted after 36 h of incubation at 25 °C.
To assess aerial dispersal efficiency, fungal cultures were grown on PDA for 21 days. Conidial plugs (1 × 1 cm) were placed under constant airflow (0.3 m s -1 ) inside a 30-cm-long stainless steel pipe (diameter, 9 cm). Spores were collected on a PDA plate at the opposite end of the pipe, and c.f.u. were quantified after 30 h of incubation. The initial spore density of each plate was normalized based on OD 590 measurements.
Water dispersal was assessed following a modified protocol reported earlier 20 . A 200 μl droplet of water containing 0.05% Tween-80 was released onto a 1 × 2 cm conidiating culture plug inclined at 60°. The runoff was collected, and spores were counted using a haemocytometer. Spore abundance was normalized to initial conidial density.
Plant growth promotion test was performed for 10 selected Trichoderma spp. (Supplementary Fig. 2). Tomato seedlings (Solanum lycopersicum cv. HEZUO903) were grown in peat soil (200 g per pot) under controlled greenhouse conditions (25 °C, light/darkness at 8 h:16 h). Each seedling was inoculated with 1 ml of suspension with 10 6 spores per ml from a Trichoderma strain in soil, with the control left uninoculated. This experiment was conducted with 12 replicates per each group. Ten parameters regarding plant biomass and root development, including Soil-Plant Analysis Development (characterizing the chlorophyll content for leaves), shoot height, shoot fresh weight, shoot dry weight, root fresh weight, root dry weight, root total length, root average diameter, root volume and root tips, were recorded for each seedling 3 weeks after the fungal inoculation.
Trichoderma proliferation in soil was quantified using DNA metabarcoding. Three types of soil-forest soil (Entisols), saline soil (Oxisols) and red soil (Alfisols)-were collected from different locations (32° 3′ 26″ N, 118° 51′ 15″ E; 33° 13′ 39″ N, 120° 45′ 07″ E; 27° 56′ 1″ N, 115° 10′ 19″ E) with pH values of 4.59, 8.35 and 4.62, respectively. Soil was either sterilized by gamma radiation or left non-sterile, and each microcosm was inoculated with a spore suspension containing equally represented spores of 26 Trichoderma strains (=species) at a final concentration of 10 7 spores per ml. Inoculation was performed in custom-designed deep-well plates (127.76 mm × 85.48 mm) with six parallel trenches, each 20 mm deep, comparable in depth to standard deep-well plates. Each plate contained three soil-filled trenches, separated by empty trenches. Inoculation was conducted at one end of each soil-filled trench by pipetting 100 μl of spore suspension. Plates were incubated for 10 days at 25 °C in darkness under controlled conditions. Soil samples were collected at the end of the trench. Genomic DNA of each soil sample was extracted using the DNeasy PowerSoil Pro Kit (Qiagen). A fragment of the rpb2 gene (~380 bp) encoding RNA polymerase subunit B II was specifically designed for DNA barcoding distinguishing Trichoderma spp. at the species level. The primer pair ComRPB2-F: 5′-TGCGNAGRRTGAAYACNGA-3′ and ComRPB2-R: 5′-CCVACRCTGACYTARCACATN 3′. PCR reactions were performed in 20 μl volumes containing 4 μM of each primer, 50 ng DNA template, and 10 μl of 2× Phanta Max Master Mix (Vazyme Biotech). The thermal cycling conditions were as follows: initial denaturation at 95 °C for 3 min, followed by 32 cycles of 95 °C for 15 s, 58 °C for 15 s and 72 °C for 30 s, with a final extension at 72 °C for 5 min. Amplicons were purified using AMPure XP beads (Beckman Coulter), quantified with a Qubit 3.0 Fluorometer (Thermo Fisher), pooled equimolarly and sequenced on an Illumina MiSeq platform. Sequence data were processed using QIIME2 (ref. 34) and the taxonomic assignment against the local Trichoderma rpb2 database including sequences from >400 strains. Alphaand beta-diversity metrics were calculated, and differential abundance analyses were performed using edgeR 4.2.2. Statistical significance was determined using the Wilcoxon test for alpha-diversity and permutational multivariate analysis of variance for beta-diversity.
To assess growth on complex organic substrates, finely ground dried fungal biomass (basidiomas of Pleurotus sp.), dried soybean, dried rice straw and dried Magnolia grandiflora leaves were used as growth media. Each Trichoderma strain (5 μl of a 10 7 spores per ml suspension) was inoculated into 24-well plates containing sterile 0.5 g of each substrate with 50% water content. Plates were incubated at 25 °C in darkness for 7 days.
Each Trichoderma strain was inoculated onto 9 cm carboxymethyl cellulose plates (composition: 1.
4 g (NH 4 ) 2 SO 4 , 2 g KH 2 PO 4 , 0.3 g MgSO 4 •7H 2 O, 0.15 g KCl, 0.3 g CaCl 2 •2H 2 O, 0.3 g urea, 1.2 g casein, 2 g Tween 20, 5 mg FeSO 4 •7H 2 O, 1.7 mg MnSO 4 •H 2 O, 1.4 mg ZnSO 4 •7H 2 O, 2 mg CoCl 2 •2H 2 O, 10 g carboxymethyl cellulose and 20 g agar per litre) and incubated at 28 °C in darkness for 48 h. Plates were subsequently incubated at 50 °C for 12 h, followed by staining with 0.1% Congo red solution for 10 min and destaining with 1 M NaCl until a clear halo developed, indicating cellulase activity.
Strains were inoculated onto glyceryl tributyrate plates (2.5 g casein, 2.5 g peptone, 3 g yeast extract, 10 g glyceryl tributyrate and 20 g agar per litre, pH 7.5) and incubated at 28 °C in darkness for 7 days. Enzymatic activity was assessed based on halo size.
Trichoderma strains were inoculated on 9 cm plates with PCL amended (composition: 2 g (NH 4 ) 2 SO 4 , 4 g KH 2 PO 4 , 6 g Na 2 HPO 4 , 200 mg MgSO 4 •7H 2 O, 50 mg yeast extract, 1 ml Triton X-100, 20 g agar and 0.5 g of PCL powder dissolved in acetone prior, per litre) and incubated in darkness at 25 °C for 7 days. Enzymatic activity was assessed based on halo size.
Each strain was inoculated onto casein plates (6 g casein, 3 g Tris, 8.7 g NaCl and 10 g agarose per litre, pH 7.5) and incubated at 28 °C in darkness for 7 days. Enzymatic activity was assessed based on halo size.
For the qualitative measurement of amylase activity, each Trichoderma strain was inoculated on a 9 cm starch agar plate (composition: 5 g peptone, 3 g yeast extract, 10 g soluble starch, 5 g NaCl and 20 g agar per litre) and incubated at 28 °C in darkness for 48 h. Plates were then covered with Gram iodine solution (1 g iodine, 2 g KI, 3 g NaHCO 3 per 300 ml) and incubated at room temperature until a colourless halo appeared around the colony, indicating starch hydrolysis due to amylase activity.
Siderophore production was assessed using a modified chrome azurol S (CAS) agar plate assay to detect and quantify iron-chelating compounds secreted by Trichoderma strains. The CAS medium was prepared by mixing 900 ml of acetate-yeast (AY) medium (containing 0.27 g Na-acetate, 0.15 g yeast extract, 15 g agar per litre) with 100 ml of a separately autoclaved 10× CAS assay solution (comprising 20 ml of 10 mM Fe(NO 3 ) 3 , 40 ml of 10 mM CAS and 100 ml of 10 mM N-dodecyl-N ,N-dimethyl-3-ammonio-1-propanesulfonate).
To reduce toxicity at the growth surface, layered CAS plates were prepared by pouring acetate-yeast + CAS medium into Petri dishes, forming a dark blue bottom layer (20-25 ml). After cooling, an agar-solidified acetate-yeast top layer was added, allowing CAS Resource https://doi.org/10.1038/s41564-026-02260-3 diffusion overnight at 4 °C to create a diminishing blue gradient. Trichoderma strains were inoculated by transferring mycelium onto the assay plates and incubating at 25 °C in darkness for 12 days. Siderophore production was indicated by a colour change from blue to yellow or green, resulting from Fe(III) chelation 35 .
Interactions between Trichoderma strains and bacteria were tested using Escherichia coli DH5α, Ralstonia solanacearum RS1115, and Bacillus velezensis SQR9 (ref. 36). A 100 μl bacterial suspension (OD 600 = 1.0) was spread on 9 cm PDA plates, followed by inoculation of two fresh Trichoderma culture plugs (6 mm diameter) at opposite ends. Co-cultures were incubated in darkness at 25 °C for 72 h before imaging. Antibacterial effects were assessed based on inhibition zones.
To evaluate Trichoderma antagonism toward other fungi, Fusarium odoratissimum TUCIM 4848, Rhizoctonia solani TUCIM 3753 and Pestalotiopsis fici TUCIM 5788 (ref. 37) were used as target fungi. Pre-inoculated PDA plates were incubated at 25 °C for 48 h, after which Trichoderma was inoculated on the opposite side and the plates incubated for an additional 10 days. Interactions were monitored for inhibition zones or overgrowth.
To assess oligotrophic tolerance, each strain was inoculated on 9 cm water agar plates (20 g agar per litre) and incubated at 25 °C in darkness for 5 days. PDA-grown cultures served as controls. Oligotrophic growth was estimated based on colony expansion.
The ability of Trichoderma strains to form buoyant colonies and their photoresponse were assessed using 24-well plates containing six different media: potato dextrose broth, 1/3 Murashige and Skoog medium + 1% glucose, corn meal broth, synthetic nutrient broth, Czapek-Dox and glucose-salt medium. Strains were incubated under two conditions: constant darkness and light/darkness at 8 h:16 h at 25 °C for 5, 10 and 15 days. Colony buoyancy was evaluated based on floating mycelial mat formation, while the photoresponse was determined by changes in colony morphology and pigmentation under light exposure. Imaging was performed at each time point to document phenotypic variations.
To assess light-dependent developmental changes, strains were grown on PDA plates at 25 °C under two conditions: (1) constant darkness and (2) light/darkness at 8 h:16 h cycle. Morphological traits were recorded after 5 days.
Colonies were grown on 6 cm PDA plates at 25 °C in darkness for 48 h. A sterile scalpel blade was used to introduce standardized injuries in the shape of a number sign (#), followed by incubation under light/darkness at 8 h:16 h cycle for 48 h. Control plates were maintained uninjured.
The ability of Trichoderma strains to grow in soil and migrate through the substrate was assessed using two experimental set-ups: soil column assays and deep-well plate assays.
Three soil types-forest soil (Entisols), saline soil (Oxisols) and red soil (Alfisols)-were collected from different locations (32° 3′ 26″ N, 118° 51′ 15″ E; 33° 13′ 39″ N, 120° 45′ 07″ E; 27° 56′ 1″ N, 115° 10′ 19″ E) with pH values of 4.59, 8.35 and 4.62, respectively. Strains were first grown on PDA plates at 25 °C in darkness for a few days before introduction into soil. Fungal growth in soil was monitored for 14 days, and the ability of strains to penetrate a 10 cm soil column was recorded.
To qualitatively evaluate fungal growth and soil colonization under controlled conditions, the assay was repeated using deep-well plates filled with the same soil types. Trichoderma strains were inoculated into the wells and incubated at 25 °C in darkness. Growth progression was assessed by visualizing fungal expansion within the wells and quantifying colonization efficiency over time. Plates were imaged 10 days after inoculation.
Fungal strains were cultivated on PDA plates at 25 °C in darkness for 7 days. Colony morphology was recorded using a Canon EOS 70D camera equipped with a Canon 100 mm macro lens under white light.
Fungal surface structures were analysed using cryo-SEM with a Quorum PP3010T preparation system integrated into a Hitachi SU8010 FE-SEM (Hitachi). Fungal cultures were rapidly frozen in nitrogen slush. Samples were fractured at -140 °C and coated with a 5 nm platinum layer before imaging. All plate-based assays were imaged using a Canon EOS 70D with a Canon 100 mm macro lens (EF 100 mm f/2.8L IS USM, Canon) under white light.
All experiments, if not elsewhere specified, were conducted for at least three biological replicates. Data were analysed using one-way analysis of variance or multivariate analysis of variance in STATISTICA 6 (StatSoft) or in R (version 3.2.2). Heat maps, hierarchical clustering (average linkage) and principal component analysis plots were generated in R. The significance level was set at P < 0.05, unless otherwise stated. All tests were performed as two-sided; assumptions of normality and homoscedasticity were assessed where appropriate, and multiple comparisons were corrected using the Benjamini-Hochberg method. Data are presented as means with standard deviations unless otherwise noted; variation and associated uncertainty are indicated in the figure legends and supplementary tables.
Trichoderma genomes were sequenced on the Illumina platform using the Kapa Biosystems library preparation kit. DNA was sheared using a Covaris LE220 focused ultrasonicator. Sequencing was performed on an Illumina NovaSeq sequencer using NovaSeq XP v1.5 reagent kits (S4 flowcell) with a 2 × 150 bp indexed run. All raw Illumina sequence data were filtered for artefact and process contamination using the JGI QC pipeline. Genome assemblies were generated from the resulting non-organelle reads using SPAdes v3.8. 1 (ref. 38). Similar methodology using the UNITE rDNA database for molecular identification of fungi 39 was used to reassemble ribosomal DNA from the filtered reads.
Trichoderma transcriptomes were sequenced on the Illumina platform using the TruSeq Stranded mRNA Library Prep kit. Sequencing was performed on an Illumina NovaSeq sequencer using NovaSeq XP v1 reagent kits (S4 flowcell) with a 2 × 150 bp indexed run. Raw reads were filtered and trimmed using the JGI QC pipeline, including removal of artefacts, spike-in sequences, PhiX contamination, low-quality bases and short reads. Filtered reads were assembled into transcriptomes using Trinity (v2.11.0) 40 .
All genomes were annotated using the JGI Annotation Pipeline 41 , which combines several gene predictions and annotation methods with transcriptomics data and integrates the annotated genomes