Valuable crop resources require maintenance and protection. In ant fungus gardens, Escovopsis fungi are well known as specialized parasites, especially in the tropics . Infection of the fungus garden by Escovopsis decreases colony fitness and can lead to colony collapse . However, Escovopsis does not regularly parasitize fungus gardens cultivated by Trachymyrmex septentrionalis, the northernmost fungus-growing ant, a finding based on a single culture-based study that detected Trichoderma and several other microfungi but not Escovopsis in T. septentrionalis fungus gardens and that did not test its infectiveness . Many other fungi have also been isolated from ant fungus gardens, although the culture-based methods used and the often nonsystematic sampling obscures their ecological distribution and symbiotic role . Beyond Escovopsis, only Trichoderma and Syncephalastrum have been shown to exhibit some degree of pathogenesis toward Atta ant fungus gardens [but not toward T. septentrionalis. More work is needed to fully understand the diversity and ecology of these and other potential fungus garden pathogens and the mechanisms by which ants respond to protect their fungus gardens. Ants have developed numerous chemical and behavioral mechanisms to avoid infection of their fungus gardens and prevent colony collapse . Chemical defenses include the application of antimicrobials from ant metapleural gland and fecal secretions , and from antibiotic-producing Pseudonocardia bacteria that the ants host on their cuticles . Ant behavioral defenses include grooming their own bodies and those of their nestmates , task partitioning between different members of the colony , grow racks and preprocessing of foraged material before its incorporation into the fungus garden . Fungus garden grooming and especially weeding represent important ant behavioral responses to fungal pathogens that have invaded the fungus garden .

Although these defensive responses have been described in detail, how fungus-growing ants detect threats to their fungus gardens remains poorly understood . Insects are well-known to communicate using chemical cues. For example, Mastotermes darwiniensis termite soldiers respond to the cuticular hydrocarbon p-benzoquinone with increased mandible openings, indicating excitement . In another example, diverse ant species destroy diseased pupae in response to cuticular hydrocarbons emitted during fungal infection . Although the mechanism by which fungus-growing ants detect pathogen infections remains unknown, they do detect and respond to chemical cues that promote other aspects of fungus-garden health. For example, Acromyrmex lundii ants use carbon dioxide as a spatial cue to position their fungus gardens at optimum soil depth . Based on unfavorable CO2 levels, ants will relocate their gardens, a remarkable demonstration of their sensitivity to small molecule cues. Pathogenic fungi also produce metabolites that can affect fungus-growing ant health and behavior such as shearanine D produced by Escovopsis fungus garden pathogens, which reduces ant movement, inhibits Pseudonocardia growth, and directly causes ant death . Thus, chemical communication mechanistically underpins diverse symbiotic interactions in ant fungus gardens. In this study, we sought to identify the chemical cues that induce hygienic weeding behavior during infections of ant fungus gardens. We established that Trichoderma fungi are common in T. septentrionalis ant fungus gardens and that T. septentrionalis ants weed their fungus gardens in response to treatments with live Trichoderma spores, Trichoderma chemical extracts and fractions, and Trichoderma-derived pure compounds. Our results suggest that peptaibol metabolites are produced by Trichoderma fungi during fungus garden infection and cue ant weeding behaviors that promote fungus garden hygiene.

This study fills the gap between the well-studied hygienic behavioral responses of fungus-growing ants and the hitherto unknown chemical cues that induce them. Such chemical cues are likely widespread in other agricultural systems where they are used by hosts to detect and prevent pathogen infections.We used internal transcribed spacer region 2 community amplicon sequencing to investigate microfungal communities in field-sampled and apparently healthy T. septentrionalis fungus gardens from across the Eastern USA . Reads classified as Trichoderma, a common genus of mycoparasites that also includes some saprophytes and plant mutualists , were both the most abundant and prevalent noncultivar reads in field-sampled T. septentrionalis fungus gardens, with a median relative abundance of 1.2% and a maximum of 68.6% in the most extreme case . Other noncultivar fungi in these fungus garden samples were only rarely abundant , and no reads in this dataset matched the common tropical fungus garden pathogen Escovopsis. Only four samples out of 83 were dominated by a noncultivar fungus; two of these were dominated by Meyerozyma, one by Trichoderma, and one by an unclassified member of the family Stephanosporaceae . Although the sampled fungus gardens did not visually appear to be diseased at the time of their collection, our ITS sequencing results suggest that Trichoderma spp. may be a low level but constant threat to T. septentrionalis fungus gardens in situ. In a parallel analysis, we also generated environmental metabolomes from 53 field-sampled T. septentrionalis fungus gardens, 18 of which were the same as those sampled for our ITS dataset. Using untargeted liquid chromatography-tandem mass spectrometry and Global Natural Products Social molecular networking , we identified chemical evidence for the presence of Trichoderma spp. in T. septentrionalis fungus gardens.

After searching our network of specialized metabolites from these fireshly collected fungus gardens for metabolites produced by potential pathogens, we identified one cluster that contained a feature whose molecular weight and fragmentation pattern were consistent with the peptaibol trichodermide D along with a suite of related peptaibols, most of which were originally isolated from Trichoderma virens CMB-TN16 . A related series of peptaibol features were detected in three fungus gardens collected from North Carolina, as shown by the nodes in the network that were closely related to known peptaibol features . Given that peptaibols are characteristic of mycoparasitic members of the Hypocreales and especially prevalent in Trichoderma , our detection of peptaibols from these samples further supports that Trichoderma is present and metabolically active in field-sampled T. septentrionalis fungus gardens.To identify the underlying mechanisms by which Trichoderma inoculation induced waste production by the ants, we exposed T. septentrionalis fungus gardens to extracts of Trichoderma sp. JKS001884. Colonies were tested multiple times, including both intracolony and intercolony replicates . In all tests, ant waste production was greater in extract-treated fungus gardens compared to the negative controls treated only with DMSO or left untreated . These results suggest that ant waste production was induced by metabolites from the Trichoderma extracts and that ant responses were not solely due to the physical presence of Trichoderma cells. To identify the metabolites responsible for this bioactivity, we evaluated semi-purified fractions of our Trichoderma extract for their ability to induce ant behavioral responses. Fractions B, D, and E induced the greatest amount of ant waste production , although further analyses determined that fraction B was chemically dissimilar to fractions D and E and was instead highly similar to fraction A , one of the least bio-active fractions . Given that such early fractions often contain pan-assay interference compounds , we prioritized fractions D and E for further analysis. Comparative metabolomics was used to prioritize and identify metabolites that were highly abundant in fractions D and E. These fractions grouped together along non-metric multidimensional scaling axis 1, distinct from all other fractions , demonstrating their high chemical similarity, as also determined using Spearman’s correlation . Comparisons of total ion chromatograms indicated considerable overlap of peaks in fractions D and E especially between 7 and 8 min, planting racks retention times at which numerous peptaibols elute. In addition, a large number of features that cooccurred in fractions D and E had molecular weights above 1,000 Da, consistent with peptaibol metabolites . This motivated further chemical analysis to identify features that may underpin the ant behaviors induced by fractions D and E.Based on our prioritization of fractions D and E , we generated a heat map to identify features shared between fractions D and E and dereplicated these features using NP Atlas . Fractions D and E exclusively shared 118 features , although an additional suite of features was highly enriched in fractions D and E but present in lower abundances in other fractions. Based on their molecular weights, retention times, and fragmentation patterns, many of these shared features were consistent with the structure of peptaibols. Three peptaibol-like features were exceptionally abundant in fractions D and E : [M + Na]+ peaks m/z 1197.7557 and 1183.7406, and [M + H]+ peak m/z 1452.8756. Together, the ion abundances of these three features represented a combined 35.5% and 75.5% of total metabolite abundance in fractions D and E, respectively. Further exploration of the accurate masses and fragmentation patterns of these features confirmed that all three likely represent peptaibols, two of which have masses consistent with the peptaibols trichodermides B/C and D/E, although several other peptaibols have similar masses. These results prompted further ant weeding assays using a small library of purified peptaibols to determine if specific, individual peptaibols induce ant behavior and if this behavior is a result of collective and/or nonspecific peptaibol metabolites. We tested six peptaibols isolated from Trichoderma arundinaceum and one purchased peptaibol to evaluate their ability to induce ant waste production. Although these purified peptaibols varied in their bioactivity, all induced ant waste production, withsome replicates having higher levels of bioactivity than the extract . Two of these peptaibol metabolites, 1 and 2 , are previously undescribed compounds given the trivial names trichokindins VIII and IX, respectively. Mass spectrometric analysis confirmed that both metabolites were present in our Trichoderma extract and in the bioactive fractions D and E .

Full characterization of 1 and 2 indicated that these previously undescribed compounds have a classical peptaibol structure, being composed entirely of amino acids, including characteristic α-aminoisobutyric acid moieties. Because each isolated peptaibol induced ant weeding, this behavior is likely characteristic of the peptaibol class, in general, and may not be specific to individual peptaibols. Our results strongly suggest that ant waste production is induced by Trichoderma-derived peptaibol metabolites; however, we cannot discount that other fungal secondary metabolites were present in our Trichoderma extracts that may also induce waste production. Using comparative metabolomics, we identified a feature suggestive of the fungal metabolite roselipin 1A , although this feature had similar abundances across fractions D, E, and F. Because fraction F did not induce substantial ant waste production, this feature did not likely cause the observed ant weeding behavior. Features with masses matching two other common fungal metabolites, gliovirin and heptelidic acid , exhibited similar patterns of abundance in both inactive and bioactive fractions, and thus were also unlikely to induce the observed ant weeding behavior. Thus, the unique correlation between peptaibol-enriched fractions and increased ant waste production underscores the relationship between Trichodermaderived peptaibols and ant waste production.Our data show that Trichoderma spp. are common in wild T. septentrionalis fungus gardens sampled from across a broad geographic range , which suggests that at least some Trichoderma spp. may naturally cause fungus garden disease. Supporting this hypothesis, our isolation of Trichoderma spp. from diseased gardens, experimental infection of healthy gardens, and detection of Trichoderma during experimental infections using ITS2 amplicon sequencing all provide experimental evidence that Trichoderma spp. can be opportunistic pathogens of T. septentrionalis fungus gardens and fully satisfy the experimental aspects of Koch’s postulates for disease causality, which include the isolation of a pathogen from a diseased host, experimental infection of a naive host using that isolate, and reisolation of that disease-causing isolate . The ecological aspects of Koch’s postulates are partially fulfilled in this study by our detection of Trichoderma in environmental fungus gardens, even though these were not visually diseased at the time of collection, a difficult to detect event due to acute disease leading to rapid colony collapse and our detecting colonies to collect based on the presence of active ants, likely indicative of colony health. We also note that fungus garden diseases can be present but not visually apparent , making it challenging to definitively link Trichoderma presence to disease in the field, and that a focus on pathogen presence/absence without consideration of pathogen load, microbiome composition, or environmental conditions is a noted weakness of Koch’s postulates . Although the species identity, ecological source, and relationship to disease of the low levels of Trichoderma present in T. septentrionalis fungus gardens remains unclear, our results indicate the consistent presence of these potential disease-causing agents that could at least in some cases necessitate defensive responses by the ants. Together, our ecological and experimental data support our conclusion that, at least under some conditions, Trichoderma spp. can infect T. septentrionalis fungus gardens. Metabolomics analyses of lab-reared fungus gardens inoculated with T. septentrionalis-isolated Trichoderma spp. revealed the presence of peptaibols in infected gardens .