Masters of the Manipulator? A Comparative Genomics Study of Niveomyces gen. nov. and Two New Species of Mycoparasites Associated with Ophiocordyceps

Fungal hyperparasites can impact ecosystem composition and disease dynamics by modulating their parasite hosts’ population size and transmission rate. Despite their perceived ecosystem impacts and potential to be applied in disease control efforts, hyperparasites are vastly understudied. In this integrative study we formally describe a new genus and two new mycoparasite species that infect the fungus Ophiocordyceps camponoti-oridani, which manipulates the behavior of Florida carpenter ants. Phylogenetic analyses demonstrate that both fungal hyperparasites are distinct species within the families Cordycipitaceae and Ophiocordycipitaceae. The unique morphology, ecology and phylogenetic placement of the Cordycipitaceae species supports its placement in a new genus, Niveomyces. Our eld data alludes that both new species, Niveomyces ophiocordycipitis and Polycephalomyces oviedoensis, negatively impact O. camponoti-oridani survival and transmission. Moreover, their macromorphology, along with our eld data, suggest that these hyperparasites live an exclusively mycoparasitic lifestyle and do not infect, nor decompose O. camponoti-oridani’s ant host. To nd genome signatures that would further conrm their mycoparasitism, we sequenced and annotated both species’ genomes. We then compared the N. ophiocordycipitis and P. oviedoensis genomes to those of other mycoparasites, animal parasites, plant parasites, and saprophytes within the order Hypocreales and searched for previously reported mycoparasite genome signatures. However, our analyses indicate that these signatures are not greatly informative when expanding the number of genomes and lifestyles beyond those of previous studies, using the genome data that is currently available. This highlights the need for additional mycoparasite genomes within the order Hypocreales, as well as data on more closely related animal parasites. As such, this work contributes a starting point for more studies into fungal interactions between mycoparasites and entomopathogens, which have the potential to contribute important knowledge towards efforts to battle the global rise of plant and animal mycoses. We identied orthologous gene clusters between 2.4.0 (Emms and Kelly 2019) and visualized the summary output in a Venn diagram (http://bioinformatics.psb.ugent.be/webtools/Venn/). a Exact fungi occupying the basal portion of the Cordycipitaceae family (order Hypocreales). Mycelium white to light yellow, often covering the host entirely. Vegetative hyphae septate and hyaline. Synnemata multiple, erect, not branched, tapering towards the apex, covered by hymenium-like layer of conidiogenous cells. Conidiogenous cells polyblastic, elongated, irregular, hyaline, cylindrical, with characteristic denticles that are crowded on the apical part and less frequent towards the base. Conidia globose to ovoid formed singly on the denticles. Sexual morph unknown.


Introduction
Interspecies interactions are important drivers of evolution, biodiversity and ecosystem composition. Their signi cance is apparent throughout evolutionary history, where novel species interactions have led to major diversi cation events (Thompson 1999). Among the various forms of species-species interactions that exist, parasitism, where one species bene ts from another at a cost to the benefactor's tness, is highly common (Labaude et al. 2015). Considering the myriad of parasite species across all major taxa, including invertebrates, vertebrates, plants, and fungi, probably all free-living organisms are exploited by parasites. This highlights the importance of including parasites into studies that aim to understand the evolution, physiology, and behavior of organisms and their functioning within a larger ecological context (Windsor 1998). Indeed, previous studies have demonstrated both the direct and indirect effects of parasites on ecosystem composition through their ability to modulate their host's population size by increasing mortality or decreasing fecundity (Preston et al. 2016). In turn, these parasite-driven changes of host populations can signi cantly alter the entire community structure by impacting trophic cascades and other interspecies interactions (Frainer et al. 2018).
Since parasitism is such a successful lifestyle, which has evolved multiple times independently, most parasites are likely to also host parasites of their own. Such hyperparasites are expected to affect their host parasite's population size and disease dynamics, and exhibit additional indirect effects on ecosystem composition (Gleason et al. 2014).
Indeed, it has been hypothesized that hyperparasites could control infectious diseases in certain species by regulating their parasites (Parratt and Laine 2016). This has resulted in the successful employment of a few thoroughly studied hyperparasites as biological control agents. For example, the CHV1 mycovirus is used as a biopesticide to control Cryphonectria parasitica, a parasitic fungus that causes blight in chestnut trees across the world (Feau et al. 2014). This demonstrates the high potential for hyperparasites to be employed in conservation efforts and against The genus Ophiocordyceps is composed of approximately 281 species, strictly associated with insects across 11 orders (Crous et al. 2004;Araújo and Hughes 2016). Among those, one particular group stands out for its intriguing and fascinating biology, the O. unilateralis clade sensu (Araújo et al. 2018). Ophiocordyceps unilateralis s.l. are known for infecting and manipulating the behavior of ants, mostly Camponotini, across the globe (Andersen et al. 2009;Evans et al. 2011). Typically, Ophiocordyceps-infected ants, such as the Florida Carpenter ant Camponotus oridanus, are manipulated to leave their nest and ascend vegetation, where they exhibit a fungus-adaptive "death-grip" behavior (Andersen et al. 2009;Araújo and Hughes 2019;Will et al. 2020). Once dead, Ophiocordyceps utilizes the ant's body as a food source to grow spore-producing structures needed for infection (Evans et al. 2011;Hughes et al. 2011;de Bekker et al. 2015;Araújo et al. 2018). Species within this group are known to produce two types of spores, conidia (asexual) and ascospores (sexual). Ascospores are produced and actively shot from the ascoma onto the forest oor where they potentially attach to ant workers that pass by while foraging for food. In contrast, the mucilaginous conidia are produced along the stalk, not actively shot and presumably transmitted to other ants by direct contact with the fungal structure (Araújo and Hughes 2017).
So far, the fungal hyperparasites of four other entomopathogenic Ophiocordyceps species from China and Thailand have been described (Wang et al. 2015b, a;Zhong et al. 2016;Xiao et al. 2018). However, none of these species are associated with a behavior manipulating Ophiocordyceps. In addition, hyperparasites growing on Ophiocordycepsmanipulated ants, such as Ophiocordyceps camponoti-ru pedis in Brazil (Andersen and Hughes 2012) and Ophiocordyceps paltothyrei in Ghana (Araújo et al. 2020), have been reported. Yet, their formal species descriptions and phylogenetic placement are currently lacking, as well as their evolutionary history. These cases, along with several other unpublished observations of hyperparasites on Ophiocordyceps across North and South America (personal communications), suggest that the tri-trophic interactions we discovered in Florida are not unique, but an example of a world-wide phenomenon. In spite of this, formal species descriptions and reports have remained limited, no genomes have currently been sequenced, and little research has been done on their biology and the effects that these hyperparasites have on Ophiocordyceps disease dynamics and transmission.
The integrative study presented here formally describes one new genus and two new hyperparasite species that affect Ophiocordyceps-ant host interactions in Central Florida, through both morphology, ecology and phylogenetic analysis. The species reside within the Cordycipitaceae and Ophiocordycipitaceae families of the order Hypocreales and were named Niveomyces ophiocordycipitis and Polycephalomyces oviedoensis, respectively. We also set out to learn more about their natural history by determining the timing of emergence and potential effects on O. camponoti-oridani. To further analyze their presumed mycoparasitic capabilities, we sequenced and annotated both hyperparasites' genomes. We searched for genome signatures that could con rm their hyperparasitic (i.e., mycoparasitic) lifestyles by comparing the two new genomes to the available genomes of other Hypocrealean fungi that represented mycoparasites, animal parasites, plant parasites, and saprophytes. Based on previous comparative genomics studies of fungal mycoparasites, we expected a typical mycoparasite genome to stand out by containing high numbers of GH18 chitinases, protease genes, and secondary metabolite clusters (Karlsson et al. 2017).

Field Study
As part of a larger longitudinal eld study, we surveyed the occurrence of O. camponoti-oridani-infected ant cadavers from October 2018 to November 2019. Manipulated ant cadavers often cluster together in areas referred to as "graveyards" (Pontoppidan et al. 2009). As such, we chose nine of these graveyards to serve as our study sites. We set up three 9x9 meter graveyard plots in each of the following wilderness areas within During each plot visit, which occurred roughly every 1.5 months, we tagged new ant cadavers and recorded the growth and infection status of all previously tagged samples.

Sampling, DNA Extraction, Library Preparation & Whole-Genome Sequencing
We collected hyperparasitized ant cadavers from the Black Hammock Wilderness Area in Oviedo, Florida. These specimens are readily recognized in the eld by their characteristic cotton white hyphae and synnemata or dark perithecia that grow directly from the fungal hosts. Under a stereomicroscope, we removed hyperparasite tissues from the samples while taking careful consideration not to include O. camponoti-oridani tissue. We surface-sterilized the hyperparasite tissue in 70% ethanol and placed it into microcentrifuge tubes (Greiner, Monroe, NC, USA) along with two metal ball bearings (5/32" type 2B, grade 300, Wheels Manufacturing, Louisville, CO, USA). We snap-froze the fungal tissue in liquid nitrogen and used a 1600 MiniG tissue homogenizer (SPEX, Metuchen, NJ, USA) at 1300 RPM for 30 seconds to disrupt fungal cell walls. We then extracted DNA using a previously described phenol-chloroform extraction protocol ) and quanti ed it on a Qubit Fluorometer (Thermo Fisher, Waltham, MA, USA) with the Qubit dsDNA High Sensitivity Assay Kit (Thermo Fisher, Waltham, MA, USA). Subsequently, we prepared DNA libraries with the Nextera DNA Flex Library Preparation Kit (Illumina, San Diego, CA, USA) for sequencing on an Illumina MiSeq Sequencer to generate 2x250 bp paired-end reads with a 50x target coverage.

Genome Assembly & Gene Prediction
Prior to assembling the two hyperparasite genomes, we ltered and trimmed our raw sequence data using the BBDuk plugin in Geneious Prime (20.2.3) with default parameters. Subsequently, to con rm the quality of these trimmed reads, we used fastQC (Andrews 2010). We then assembled the genomes de novo using the SPAdes assembly algorithm (Bankevich et al. 2012) and con rmed the quality of the assemblies through QUAST (Gurevich et al. 2013).
As expected from samples taken directly from the eld, the QUAST outputs showed slight bacterial contamination in the genomic data indicated by the presence of two distinct mean G-C% peaks; one large peak comprising fungal reads (30-80%) and a second, much smaller peak comprising bacterial contaminants (0-30%). We removed the bacterial contaminants from both genomes by manually ltering out sequences with a mean G-C% that fell within the contaminant peak (i.e., < 30%). The effectiveness of this bacterial ltering was con rmed using MG-RAST (Keegan et al. 2016). We determined the completeness of the genomes after bacterial ltering with BUSCO (Seppey et al. 2019) by using the Hypocreales lineage (fungi_odb9), as well as CEGMA (Parra et al. 2007). Scaffolds shorter than 1000 bp were discarded. Gene prediction was performed using Augustus version 3.3 (Stanke and Morgenstern 2005). First, two separate sets of gene predictions were performed with the previously generated parameter set for Ophiocordyceps camponoti-oridani ) and the supplied parameter set of Fusarium graminearum. Next, these sets were combined into a nal set by adding genes from the second set into the rst set if those genes did not overlap with an existing gene in the rst set. The annotated genomes are available through GenBank (NCBI) under the accession numbers: JADHZA000000000 (Polycephalomyces oviedoensis) and JAFEME000000000 (Niveomyces ophiocordycipitis). Moreover, the genome assembly, gene predictions and functional annotations can be analyzed interactively at http://fungalgenomics.science.uu.nl.

Functional Annotation
We functionally annotated the predicted proteins using PFAM (El-Gebali et al. 2019) and mapped them to their corresponding gene ontology (GO) terms. We predicted transmembrane domains using TMHMM 2.0 (Krogh et al. 2001) and signal peptides using SignalP-5.0 (Nielsen et al. 1997). Proteins with a secretory signal were considered small secreted proteins if they were shorter than 300 amino acids and did not contain a transmembrane domain (except in the rst 40 amino acids). We used antiSMASH 5.0 (fungal version) to predict only well-de ned and complete secondary metabolite clusters (Blin et al. 2019). We used BlastP, with an E-value cutoff of 1e-10 to search the MEROPS database for proteases (Rawlings et al. 2018). Blast2Go was used to perform further blast searches on genes of interest (Gotz et al. 2008).

Phylogenetics
Using our de novo assembled and annotated hyperparasites as protein Blast databases, we obtained sequences for ribosomal 18S (SSU), ribosomal 28S (LSU), translation elongation factor 1-alpha (Tef), and RNA Polymerase II Subunits (RPB1 and RPB2) for phylogenetic placement. The sequences that we obtained from the genomes were aligned in Geneious to a database of 445 species (Additional File 1) within the order Hypocreales. Each locus was individually aligned with MAFFT (Katoh and Standley 2013) and concatenated into a single combined dataset using Geneious v 11.1.5. The concatenated les, along with a position (POS) le for each gene, were imported into CIPRES (Miller et al. 2012). The nal alignment length was 4,747 bp: 1,024 bp for SSU, 949 bp for LSU, 960 bp for TEF, 688 bp for RPB1 and 1,126 bp for RPB2. Maximum likelihood analysis was performed with RAxML v 8.2.4 (Stamatakis 2014) on a concatenated dataset containing all ve genes. The dataset consisted of 11 partitions, two for SSU and LSU and 9 for each codon position of the three protein coding genes: TEF, RPB1 and RPB2. The GTRGAMMA model of nucleotide substitution was employed during the generation of 1,000 bootstrap replicates. We visualized the tree using iTOL (Letunic and Bork 2019).

Morphology
To assess the macromorphology of the specimens, we took pictures using a Canon EOS 7D Mark II camera with a 35mm lens. To investigate their micromorphological features, we stained fungal tissue on microscope slides with a drop of Lacto-Fuchsin stain (0.1g Acid Fuchsin Powder and 100 mL 85% Lactic Acid). We visualized the slides using a Leica DMi8 inverted microscope, mounted with a Leica MC 170 HD camera (Leica Microsystems, Buffalo Grove, IL, USA). Type materials (holotypes and paratypes) are deposited at the New York Botanical Garden Fungarium (type numbers XXX and XXX for P. oviedoensis and N. ophiocordycipitis, respectively).

Phylogenetics
Our phylogenetics results recovered the overall topology presented in previous studies (Sung et al. 2007;Quandt et al. 2015;Kepler et al. 2017;Araújo et al. 2018). Both Ophiocordycipitaceae and Cordycipitaceae families are well supported (BS = 90 and 100, respectively) as well as many of the internal nodes ( Fig. 1). To determine the phylogenetic placement of the two hyperparasitic species that we discovered in Central Florida associated with O. camponoti-oridani, we generated a comprehensive phylogenetic tree of the order Hypocreales. Based on the phylogenetic results ( Fig. 1), one of the two fungal hyperparasite species that infects O. camponoti-oridani is a distinct new species within the genus Polycephalomyces (Family: Ophiocordycipitaceae), which we have named P. oviedoensis (BS = 97). We sampled the genus Polycephalomyces broadly for our phylogenetics analysis and included a variety of species representing distinct ecologies such as animal and fungal parasites (Wang et al. 2015b, a;Zhong et al. 2016;Xiao et al. 2018). Our phylogeny indicates Polycephalomyces as the most basal lineage within Ophiocordyceps and supports P. oviedoensis sp. nov. as a distinct species (BS = 97), closely related to a clade strictly associated with insects; P. cuboidea, P. paracuboidea and P. ryogamiensis (on Coleoptera) P. proli cus (on Hemiptera) (  Ophiocordycipitaceae, which have been highlighted. Additionally, the clades for the two new species P. oviedoensis and N. ophiocordycipitis, as well as their host O. camponoti-oridani, are shown. The colors indicate reported lifestyles for each of the fungal species that that we found to be closely related to the species that are the focus of this study. The second hyperparasite we found associated with O. camponoti-oridani is not only a new species, but a new genus that is sister of Blackwellomyces (Cordycipitaceae), a lineage that produces vivid orange to red stromata, strictly associated with larvae of Coleoptera and Lepidoptera (Mongkolsamrit et al. 2020). The BS support is not very strong (BS = 55), which could be explained by the limited available taxa for this part of the phylogenetic tree. However, the phylogenetic placement of this species (Fig. 1), in addition to its ecological and unique morphological features, strongly suggest that it represents a novel lineage within Cordycipitaceae. As such, a new genus is proposed herein, i.e., Niveomyces, to accommodate N. ophiocordycipitis sp nov. Furthermore, the fact that we found that one of the mycoparasites of O. camponoti-oridani to be an undescribed genus hints that hyperparasites of entomopathogens harbor a large reservoir of undocumented fungal diversity if we consider the diversity of its host worldwide (Araújo et al. 2018).

Taxonomy
Based on a combination of morphological, ecological and phylogenetic data, we introduce a new genus and two new species of mycoparasites. Polycephalomyces oviedoensis and Niveomyces (N. ophiocordycipitis) were both collected parasitizing Ophiocordyceps camponoti-oridani, a ubiquitous parasite of the ant Camponotus oridanus in Florida, USA. Etymology: Name re ects the "snowy" (Lat.: niveus) appearance of this fungus.
A genus of mycoparasitic fungi occupying the basal portion of the Cordycipitaceae family (order Hypocreales).
Mycelium white to light yellow, often covering the host entirely. Vegetative hyphae septate and hyaline. Synnemata multiple, erect, not branched, tapering towards the apex, covered by hymenium-like layer of conidiogenous cells.
Conidiogenous cells polyblastic, elongated, irregular, hyaline, cylindrical, with characteristic denticles that are crowded on the apical part and less frequent towards the base. Conidia globose to ovoid formed singly on the denticles. Sexual morph unknown.

Diagnosis
Fungus growing on Ophiocordyceps camponoti-oridani, producing black perithecia in multiple parts of the host, from synnemata and ascomata to fungal parts that emerge from the ant joints and sutures.
Mycelium cream to pale, growing sparsely; Perithecia 103.4 x 58.9 µm, black, ovoid to pyriform, formed in multiples parts of the host. Ascus and ascospores not observed. Asexual morph not observed.

Hyperparasite Infections in Nature
To determine the prevalence of P. oviedoensis and N. ophiocordycipitis, we included a survey of the presence of hyperparasites in a larger eld study that aimed to learn more about the natural history of O. camponoti-oridani. In this study, we tagged a total of 945 ant cadavers infected with O. camponoti-oridani in nine survey plots across three wilderness areas in Central Florida throughout the year. During this time, we found 62 ant cadavers that had their Ophiocordyceps visibly infected by a hyperparasite (Fig. 4A). Of these 62 hyperparasitized samples, 46 were infected by P. oviedoensis (i.e., had visible dark perithecia forming on top of O. camponoti-oridani mycelium) and only 16 were infected by N. ophiocordycipitis (i.e., had visible emergence of white cotton-like synnemata). Both hyperparasites were present in each of the three wilderness areas that we surveyed and often even co-occurred in the same plot.
However, we did not observe any Ophiocordyceps-infected cadavers that were infected by both hyperparasites at the same time. While new ant cadavers infected with O. camponoti-oridani seemed to appear year-round, we asked if the same was true for the two hyperparasite species P. oviedoensis and N. ophiocordycipitis or if their emergence was more seasonal. To answer this question, we noted the date on which we rst observed visible hyperparasite growth on previously tagged O. camponoti-oridani samples. We omitted the O. camponoti-oridani samples from this analysis that were rst discovered with a hyperparasite already present (n = 21). Our data indicates that new P. oviedoensis infections appeared consistently throughout the year, with a possible peak in March/April (Fig. 4B). In contrast, we almost exclusively found new N. ophiocordycipitis infections in November, with only one sample found in July (Fig. 4B). Our small sample size did not allow us to perform reliable statistics on our data to con rm if certain months of the year produce signi cantly more infections of either hyperparasite species compared to others. However, our data does suggest that, for both hyperparasite species, there is a potential seasonal trend to their infection biology, which is more prominent for N. ophiocordycipitis than for P. oviedoensis.
In addition to recording when both hyperparasite species emerged throughout the year, we monitored how much time it took for either hyperparasite to emerge on a new O. camponoti-oridani-infected ant. Such data provides insights into the lifestyles of P. oviedoensis and N. ophiocordycipitis and aids to con rm their assumed hyperparasitism, and, thus, mycoparasitism. Appearance on living, thriving mycelium and stalks of O. camponoti-oridani would suggest a mycoparasitic lifestyle while appearance on older, dying tissue would indicate a saprophytic one, and consistent emergence together with Ophiocordyceps would suggest insect co-infection and, thus, a more entomopathogenic lifestyle. For this analysis, we again removed the samples that were rst discovered with a hyperparasite already present, because of uncertainties with regards to timing of hyperparasite appearance. Among the samples that we monitored for hyperparasite emergence, both P. oviedoensis and N. ophiocordycipitis appeared as early as a week after a new O. camponoti-oridani-infected cadaver was found and as late as 9 months after (Fig. 4C). The later emergence of the hyperparasites, long after the ant's death, suggests that it is unlikely that either species would be an entomopathogen that coinfects the ant alongside O. camponoti-oridani. Additionally, the occurrence of both species early on, when O. camponoti-oridani is still actively growing its mycelium, stalk and ascoma, suggests that neither P. oviedoensis nor N. ophiocordycipitis would thrive on dead material, making a saprotrophic lifestyle very unlikely for both hyperparasitic species we studied. As such, these eld observations support the notion that both fungi are indeed to be considered mycoparasitic hyperparasites.
Parasites negatively affect their host's tness. Both hyperparasites we studied infect their O. camponoti-oridani host only after successful ant infection and manipulation, which are necessary for entomopathogen transmission.
However, before Ophiocordyceps can transmit, it also needs to successfully develop the stalk and ascoma that produce and release the infective spores (i.e., conidia and ascospores). As such, we documented the state of the O. camponoti-oridani stalk and ascoma after infection with P. oviedoensis and N. ophiocordycipitis to determine if these fungi would have a negative effect on transmission and, thus, their host's tness. We found that about half of all hyperparasitized samples never produced an ascoma (Fig. 4D). Additionally, of those that did, only very few P. oviedoensis and N. ophiocordycipitis-infected samples were able to develop an ascoma after hyperparasite infection (10.9% and 6.3% respectively) (Fig. 4D). Without ascoma formation, O. camponoti-oridani cannot produce ascospores, which hampers transmission to other ants. Moreover, macromorphological investigation of the samples showed that N. ophiocordycipitis tissue often covers O. camponoti-oridani entirely. This presents another possible reduction of O. camponoti-oridani transmission, by creating a physical barrier to the dispersal of both conidia and ascospores. In addition, both hyperparasite infections appeared to kill O. camponoti-oridani since our attempts to extract RNA from hyperparasitized stalks failed (while we were able to extract RNA from uninfected stalks) (Additional File 2). This suggests that the lifespan of Ophiocordyceps, and, thus, the amount of time over which it can transmit, is often critically reduced. Taken together, these eld observations support the hypothesis that both P. oviedoensis and N. ophiocordycipitis act as parasites and affect O. camponoti-oridani tness by decreasing transmission.

Comparative Genomics
As part of their species descriptions, we sequenced and assembled the genomes of the novel fungal hyperparasites P. oviedoensis and N. ophiocordycipitis. Detailed statistics on their assembly, gene predictions, annotations and completeness can be found in Additional File 3. To describe these genomes in more detail and determine if they contain features that would further con rm the mycoparasitic lifestyles of P. oviedoensis and N. ophiocordycipitis as hyperparasites of O. camponoti-oridani, we compared them to 15 genomes. These genomes represented other fungal mycoparasites, animal parasites, plant parasites, and saprophytes, largely within the order Hypocreales.
We rst used orthologous clustering to highlight the number of shared orthogroups between our two new hyperparasite species and the fungal genomes that represented the various fungal lifestyles in our comparison ( Fig. 5A and B, Additional File 4). When compared to mycoparasites, P. oviedoensis uniquely shared 27 orthogroups, comprising 34 genes, with fungi representing this lifestyle, (Fig. 5A, Additional File 4, sheet a), while N. ophiocordycipitis uniquely shared 41 orthogroups, comprising 44 genes (Fig. 5B, Additional File 4, sheet b). However, both species shared more unique orthogroups with animal parasites (55 for P. oviedoensis and 190 for N. ophiocordycipitis) than mycoparasites ( Fig. 5A and B, Additional File 4, sheet c and d). Both species uniquely shared the least amount of orthogroups with saprophytes (14 and 18 orthogroups, respectively). As such, our analysis indicates that the genomes of the two hyperparasites have more "parasite signatures" than "saprophytic signatures", which is in line with the expectations that we obtained from our eld observations. However, we cannot state from this orthologous clustering analysis that either species genome suggests a speci c mycoparasitic lifestyle. Analyzing the functional annotation terms of the shared orthologous genes between animal parasites and both hyperparasite species for potential enrichment showed that small secreted proteins (SSPs) were overrepresented. Similarly, we found an SSP enrichment when we analyzed the genes that the hyperparasites uniquely shared with both animal parasites and mycoparasites. However, no functional enrichments were found for the orthologous genes that the hyperparasites uniquely shared with only the mycoparasites in our comparison. In contrast, there was also an overrepresentation of SSPs among the 739 and 604 orthologs that were unique for each new hyperparasite species genomes ( Fig. 5A and B, Additional File 4, sheet e and f). Only nine of these orthologs overlapped between them (Additional File 4, sheet g). While seven of these overlapping orthologs were proteins without a predicted function, we obtained a chitinase precursor blast result for one of these genes and another was predicted to code for an acyl-CoAacyltransferase, which is an enzyme involved in the biosynthesis of ergosterol (Zhang et al. 2020). Figure 5: Orthologous clustering. Venn diagrams representing the shared orthogroups between the genomes of P. oviedoensis (A) and N. ophiocordycipitis (B) and the genomes of 15 other fungi that we grouped into different fungal lifestyles based on their ability to parasitize animal, fungal, or plant hosts, or decompose dead organic material.
We subsequently used our orthologous clustering analysis to look for genes of interest that could reveal something about the general biology of the two new hyperparasite species. Many of the genes that were unique to either of the two species or uniquely shared with the myco-or animal parasites were annotated to be related to generic cell processes or were classi ed as hypothetical proteins with an unknown function. However, there were a few notable genes that stood out to because of their annotated function (Additional File 4). For example, N. ophiocordycipitis uniquely shared several genes with animal parasites that were annotated to be involved in antagonistic fungal interactions such as a hirsutellin A toxin, the mycotoxin cyclochlorotine (Schafhauser et al. 2016), and an e ux pump protein linked to antifungal resistance (Prasad and Rawal 2014). Comparably, the uniquely shared genes between P. oviedoensis and animal parasites included the antimicrobial peptide defensin (Wu et al. 2014) and an enterotoxin. The function of enterotoxins in insect-infecting fungi has been implicated to be involved in infection and even behavioral manipulation of the host (de Bekker et al. 2015;Will et al. 2020). However, N. ophiocordycipitis only contains one enterotoxin gene and P. oviedoensis only two, while many more enterotoxin genes are generally found in insect and nematode infecting fungi, with the number of enterotoxins being particularly high in Ophiocordyceps species (de Bekker et al. 2017). Nevertheless, beyond animal parasites, enterotoxins are generally not found at all in fungal genomes (de Bekker et al. 2017), which our study con rmed. None of the other genomes beyond the animal parasites and new hyperparasites species included in our study contained any annotated enterotoxin encoding genes.
A notable gene uniquely shared between N. ophiocordycipitis and other mycoparasites encoded a putative Heterokaryon Incompatibility (HET) Protein. These proteins allow fungi to sense the difference between their own growth and the growth of other fungi (Uehling et al. 2017), which could be a trait that is of particular importance for fungal mycoparasites. Indeed, P. oviedoensis and N. ophiocordycipitis both contain many HET genes (i.e., 18 genes and 37 genes, respectively) while their insect parasite host O. camponoti-oridani contains far fewer (i.e., only 6). However, when counting the number of HET genes in the other fungal genomes included in this study, they all appeared to have many more and no particular pattern related to lifestyle seemed to exist. Another gene that peaked our interest, because it was uniquely orthologous in P. oviedoensis and all mycoparasites in our comparison, was an extracellular membrane protein with an annotated CFEM domain that is thought to be involved in fungal pathogenesis (Kulkarni et al. 2003). However, both hyperparasites and their Ophiocordyceps host contain the same number of genes with this annotation (i.e., 10 genes), which indicates that perhaps besides this one ortholog, having genes with CFEM domains is not necessarily a feature indicative of mycoparasitism. When looking at the genes that were unique to N. ophiocordycipitis and P. oviedoensis ( Fig. 5A and B, Additional File 3, sheet e and f), the majority (~ 80% and ~ 70%, respectively) did not have a known function. However, among the genes with a known function in N. ophiocordycipitis we found a ribotoxin alpha-sarcin, which has been found to exhibit strong antifungal properties (Citores et al. 2018) and might be involved in this species' mycoparasitic capabilities. Similarly, the unique P. oviedoensis genes included a putative gliotoxin gene and a Pathogenesis Related (PR-1-like) protein-encoding gene that has been linked to virulence in other fungal parasites (Prados-Rosales et al. 2012).
Our orthologous clustering analysis indeed suggested that the two new hyperparasites have a parasitic lifestyle. However, it did not speci cally highlight a mycoparasitic lifestyle since we found lots of overlap with fungal parasites of animals as well. As such we continued our investigation into the hyperparasite genomes by asking if they contained the typical mycoparasite signatures that have previously been reported (Karlsson et al. 2017). Fungi are unique in their ability to produce a wide range of secondary metabolites, many of which have important bioactive properties (Brakhage 2013). Because these metabolites are not directly necessary for growth and development, their importance often lies in their contribution during interspecies interactions such as parasitism (Steenwyk et al. 2020).
In addition, previous comparative genomic studies have suggested a link between high numbers of secondary metabolites and a mycoparasitic lifestyle (Karlsson et al. 2015). Although, these studies only compared mycoparasites to each other and did not include other fungal lifestyles. As such, the rst signature we compared across all the genomes included in our study was the number and type of secondary metabolites clusters. We found that the average number of secondary metabolite clusters was 35.2 for mycoparasites, 41.8 for animal parasites, 30.0 for plant parasites, and 25.0 for saprophytes (Table 1). Overall, these cluster averages suggested that fungal parasites generally produced more secondary metabolites compared to saprophytes, with animal and mycoparasites producing the most, and that the number of clusters found in the hyperparasite genomes t within this ballpark. However, when we further inspected the number of clusters for the individual species that represent the various fungal lifestyles, it became clear that there was a lot of variation between the species within each group ( Table 1). As such, the number of secondary metabolite clusters did not seem to be an especially indicative genome signature to con rm a speci cally mycoparasitic lifestyle as previously suggested. Nevertheless, investigating the potential secondary metabolites that these clusters produce, and their suggested function might still give us an indication about the lifestyles of the fungi that harbor them. The P. oviedoensis genome contained a total of 32 annotated secondary metabolite clusters that together comprised 200 genes (Table 1, Additional File 5, sheet 1). Of the 200 P. oviedoensis genes involved in secondary metabolism, 62.0% had orthologs in all other compared species while 7.0% were unique to this species. Out of the 32 clusters, only four of them matched to already known clusters within the antiSMASH database. One was 42.0% similar to a cluster that produces neosartorin, a compound known to have antibiotic properties in several Aspergillus species (Ola et al. 2014). The other three were all similar to clusters that are known to produce Cyclosporin C, a compound that has anti-fungal properties in several fungal species including Tolypocladium (Moussaïf et al. 1997). These bioactive secondary metabolites may play a role in infecting O. camponoti-oridani or protecting P. oviedoensis from attack by other microbes within their environment. We found that the N. ophiocordycipitis genome encoded a total of 38 secondary metabolite clusters that together included 347 genes (Table 1, Additional File 5, sheet 2). Of the 347 genes involved in secondary metabolism, 57.6% had orthologs in all other compared species while 4.6% were unique to N. ophiocordycepitis. Out of the 38 clusters, six matched to already known clusters. One of these matched 100% to the epichloenin A secondary metabolite cluster. Epichloenin A has been found to play a role in fungus-plant interactions (Jia et al. 2019) and could perhaps play a role in N. ophiocordycipitis interactions with its fungal host O. camponoti-oridani as well. We found another cluster that matched for 83.0% with the chrysogine secondary metabolite pathway. Chrysogene is a yellow pigment known to protect fungal tissue from UV radiation and to have antibacterial properties in Penicillium (Viggiano et al. 2017). The N. ophiocordycipitis genome also contained a cluster that putatively produces an aspirochlorine-like mycotoxin (13.0% match) (Wu et al. 2000). As such, it appears that both hyperparasites contain various secondary metabolite clusters that produce bioactive compounds, which are putatively involved in antagonistic microbe-microbe interactions. Table 1 Annotated secondary metabolite clusters. The total number of annotated secondary metabolite clusters in each genome compared in this study and the number of clusters per type of metabolite produced are presented in this  Parasitic fungi secrete many proteases during infection (Monod et al. 2002). Indeed, mycoparasitic Trichoderma species appear to upregulate their proteases during contact with their host and have a noticeable percentage of protease encoding genes in their genomes (Druzhinina et al. 2012). Therefore, it has been suggested that the presence of a high number of protease genes in the genome would be characteristic of mycoparasitic fungi (Karlsson et al. 2017). When counting the number of putative protease encoding genes for our new hyperparasite species, we found that P. oviedoensis had a total of 266 (i.e., 3.2% of all genes annotated) and N. ophiocordycipitis had a total of 319 (i.e., 3.6%) ( Table 2, Additional File 6). These numbers place both new species within the expected percentages for mycoparasites and animal parasites, which respectively contained 3.2% and 3.6% proteases on average. In contrast, with only 2.3% and 2.8% protease encoding genes on average, plant parasites and saprophytes contained less (Table 2, Additional File 6). However, these average percentages are arguably not that different across fungal parasites and, again, considerable variation within host groups exists. As such, merely counting the number of proteases also does not seem to be as informative as previously suggested. Instead, taking a closer look at the types of proteases that are encoded in the genome might be more revealing. Other studies have linked aspartyl, serine, and subtilisin-like proteases to mycoparasitism (Viterbo et al. 2004). Indeed, the P. oviedoensis genome contained each of these protease types and even had an overrepresentation of serine-type proteases among the genes that were found to be unique to this species. However, N. ophiocordycipitis did not have any aspartyl or subtilisin-like proteases. In addition, previous studies have shown an upregulation of trypsin-like proteases during mycoparasite interaction with host mycelia (Troian et al. 2014). However, the P. oviedoensis genome did not contain any trypsin-like proteases while the N. ophiocordycipitis genome did. As such, also by going into greater detail, protease annotations do not appear to provide genome signatures that are consistent enough across mycoparasites to aid in their distinguishing from other parasitic lifestyles. Chitin provides structural stability for fungal cell walls. As such, chitinases are important to fungi because of their involvement in several primary processes, including external chitin decomposition, cell wall degradation, and remodeling (Hartl et al. 2012). However, chitinases are thought to be especially important to mycoparasites and insect parasites that would need to break down the chitin in the cell walls or exoskeletons of their hosts to enter them (Hartl et al. 2012). The most prominent family of chitinases in fungi is the Glycoside Hydrolase Family 18 (Hartl et al. 2012).
Previous studies have proposed that a high number of these GH18 chitinases in the genome is a characteristic of mycoparasitism (Kubicek et al. 2011;Karlsson et al. 2015). As such, we asked if our new hyperparasite genomes indeed harbor a high number of GH18 chitinases and how this number compares to the numbers found in the other genomes in our study. For P. oviedoensis, we found 13 genes within the GH18 family and 85% of them had orthologs in all other compared species while N. ophiocordycipitis had 16 GH18 genes of which half had orthologs in all other compared species (Table 2, Supplementary File 7). For both hyperparasites, the number of chitinases was lower than the average numbers for mycoparasites (i.e., 22 GH18 genes) and animal parasites (i.e., 19.5 GH18 genes). Instead, their numbers were more in line with the average numbers found for plant parasites and saprophytes (i.e., 16 and 13.5 GH18 chitinases, respectively). Again, the numbers of chitinases within each lifestyle group were highly variable across the genomes included for each group. The previous study that highlighted the importance of GH18 chitinases to mycoparasitism did speci cally include other fungal lifestyles into their comparisons. However, their chosen representative species did not belong to the order Hypocreales (Karlsson et al. 2017). As such, a large number of GH18 chitinases is perhaps more characteristic of Hypocreales as a whole than mycoparasites speci cally.

Discussion
Although fungal hyperparasitism is a world-wide phenomenon that is often recognized by mycologists and disease ecologists, very few mycoparasitic species of entomopathogenic fungi have been formally described so far. Several hyperparasites of Ophiocordyceps species have been formally described so far (Wang et al. 2015b, a;Zhong et al. 2016;Xiao et al. 2018). However, hyperparasites associated with behavior-manipulating Ophiocordyceps have only been noted in the eld as an ecological aspect of those species (Andersen and Hughes 2012;Araújo et al. 2020). In this study, we present one new genus (Niveomyces) and two new fungal species, P. oviedoensis and N. ophiocordycipitis, that infect behavior manipulating Ophiocordyceps. We complemented our study with a natural history survey that aimed to determine the timing of the new species' emergence and potential effects on their fungal host in nature. This is in contrast with previously published reports that merely used the presence of fungal growth on Ophiocordyceps to classify a species as a hyperparasite. Additionally, we sequenced the genomes of the two new hyperparasite species proposed herein and compared them to the genomes of other species in the order Hypocreales.
The aim of our comparative genomics efforts was to determine if these genomes contain features that would further con rm a mycoparasitic lifestyle. However, the two new fungal hyperparasite species reside in parts of the fungal phylogenetic tree for which presently only very limited data is available, and the genomes of closely related species have not been sequenced yet. The fact that the genome for P. oviedoensis is the rst one sequenced among the 24 species of Polycephalomyces senso (Crous et al. 2004;Kepler et al. 2013) and that we had to erect a new genus to describe N. ophiocordycipitis highlights this lack of knowledge about these fungi. These data limitations complicated our comparative genomics efforts and emphasizes the need for the sequencing of additional genomes and conducting of future studies on hypocrealean hyperparasites.
Both of the hyperparasites that we discovered on O.camponoti-oridani in Central Florida are delineated by our comprehensive phylogeny as new species in the families Cordycipitaceae and Ophiocordycipitaceae of the order Hypocreales. One of the species, P. oviedoensis, placed con dently as a new species within the genus Polycephalomyces, as the sister to a clade that is strictly associated with insect parasites. However, many other species within Polycephalomyces exhibit mycoparasite lifestyles (Fig. 1). Moreover, reports on P. yunnanensis, on O. nutans and its stinkbug host (Wang et al. 2015b) and P. lianzhouensis coinfecting the larva host of O. crinalis (Wang et al. 2015a) suggest that some Polychephalomyces species might be able to act both as mycoparasites and entomopathogens. However, more detailed observations on a bigger range of specimens for both Polycephalomyces species would be needed to determine if this is actually the case. All previously described hyperparasites of Ophiocordyceps have been placed within Polycephalomyces (Wang et al. 2015b, a;Zhong et al. 2016;Xiao et al. 2018). This makes it di cult to determine the origin of P. oviedoensis mycoparasitism using the present data. However, the other new hyperparasite species resides within a part of the Hypocreales tree that, thus far, largely contains insect parasites. This suggests that its mycoparasitism might have evolved from a previous animal parasitic relationship (Fig. 1). Nevertheless, important species data is clearly missing for this part of the phylogeny as demonstrated by the new genus we had to invoke for our hyperparasite species (Fig. 1). Despite lower bootstrap value supporting Niveomyces we did feel con dent about erecting a new genus for the new hyperparasite species since the Blackwellomyces species in its sister clade are morphologically vastly different. Assessing the macromorphology, Niveomyces exhibits an asexual morph with snow white mycelium that produces multiple spiky synnemata on a subiculum, while Blackwellomyces in nature often produces only the sexual morph that is vivid orange to red. As for its micromorphology, Niveomyces exhibit a hymenium-like layer of conidiogenous cells with a unique crown of denticles on the apex (each producing a single conidia). In contrast, Blackwellomyces produces commonly two types of asexual morphs concomitantly; Marianneae, Evlachovaea and Acremonium-like, which usually produces clusters or chains of sticky conidia that remain attached to the conidiogenous cells (Mongkolsamrit et al. 2020). Both the morphological and lifestyle differences between Niveomyces and Blackwellomyces thus support our phylogenetic data to assign this species both a new species and a new genus name.
While little research has been done on the effects of fungal hyperparasites on their hosts, Polycephalomyces sp. GIMCC 3.570 has been found to shorten the lifespan and signi cantly reduce the amount of ascospores produced by O. sinensis (Zhong et al. 2016). Moreover, undescribed hyperparasites of O. camponoti-ru pedis have been found to 'castrate' their hosts and prevent transmission . Therefore, it is parsimonious to state that both N. ophiocordycipitis and P. oviedoensis negatively affect O. camponoti-oridani transmission. Indeed, the majority of observed P. oviedoensis-infected samples did not grow an ascoma after hyperparasite infection, suggesting that this hyperparasite interferes with O. camponoti-oridani transmission. In addition, N. ophiocordycipitis seems to negatively impact O. camponoti-oridani transmission by largely preventing the growth of a new ascoma or by completely covering an existing stalk and ascoma before consuming it, which prevents the release of spores. Moreover, O. camponoti-oridani stalks appeared to have died after infection by either hyperparasite species. This suggests that N. ophiocordycipitis and P. oviedoensis also shorten the lifespan of their host and prevent the production and transmission of spores beyond the timespan of our eld study. As such, both hyperparasites affect the disease ecology of Ophiocordyceps-ant infections. More studies are needed to con rm the extent of the impacts of P. oviedoensis and N. ophiocordycipitis on O. camponoti-oridani transmission.
The fungi that grow on Ophiocordyceps have always been considered hyperparasitic mycoparasites. However, one could perhaps argue that they could also be growing saprophytically on dead insect or fungal tissue or act as entomopathogens that co-infect the insect host of Ophiocordyceps. The latter is especially enticing considering the phylogenetic placement of both species among species of entomopathogenic fungi. However, based on their morphology, which indicates that both P. oviedoensis and N. ophiocordycipitis only grow on top of O. camponotioridani tissue, it is more likely that they are indeed mycoparasites. Our eld surveys support this conclusion. Although in some cases P. oviedoensis and N. ophiocordycipitis growth was observed less than a week after a new O. camponoti-oridani-infected ant cadaver was found, they also appeared up to nine months after initial Ophiocordyceps infection. Moreover, O. camponoti-oridani mycelium begins to emerge from a fresh cadaver within one to two days after its death, providing already su cient tissue for the mycoparasite to inhabit. This seemingly rules out the possibility of either species being an entomopathogen that coinfects the ant. Moreover, already within the rst hours after an ant's manipulation and following death, Ophiocordyceps completely colonizes the ant's body, consuming all ant tissue besides the cuticle to gain the energy needed to grow a stalk and fruiting body (de Bekker et al., 2015). This makes it unlikely that either of the two species that we describe here would be able to saprophytically consume the insect cadaver. Moreover, the appearance of both presumed hyperparasite species in the rst few weeks after death of the ant host, when O. camponoti-oridani is fresh and actively growing, seems to rule out the possibility of either species being an opportunistic saprophyte that merely feeds off of dead O. camponoti-oridani tissue. In fact, even months after ant manipulation and death have taken place, Ophiocordyceps is often found to be alive and is still able to produce stalks with ascoma. Taken together, our morphological and natural history data con rm previous assumptions that fungi found in association with Ophiocordyceps species, that emerged from their previously infected insect hosts, are indeed mycoparasitic hyperparasites.
In addition to performing a species description, phylogeny and natural history study, we sequenced the genomes of P. oviedoensis and N. ophiocordycipitis and compared them to available genomes of the most closely related fungi we could nd that spanned a diversity of lifestyles. In doing so, we aimed to detect signatures of their mycoparasitic lifestyle. Nevertheless, while our analyses all largely pointed towards a parasitic lifestyle for both species, no clear distinction between mycoparasitism or animal parasitism could be made. This might re ect an animal parasitic origin for these mycoparasites, as our phylogeny suggests. However, this result could also be an indication that a more meaningful and conclusive analysis would have been achieved if we would have been able to compare the genomes to those of species that are more closely related. We were especially limited by the number of available mycoparasite genomes within the order Hypocreales beyond the genus Trichoderma. This argues for the sequencing and investigation of more mycoparasites while better knowledge of their infection biology might hold keys to treating animal mycoses and the biological control of devastating crop diseases.
Indeed, our orthologous clustering analysis did not directly re ect a mycoparasitic lifestyle as both P. oviedoensis and N. ophiocordycipitis shared more unique orthogroups with the genomes of the fungal animal parasites that we included in our study than with the other mycoparasites. Yet, the unique orthogroups that were shared with plant parasites and saprophytes were clearly lower. When we investigated shared orthogroups for the presence of gene functions involved in infection biology and antagonistic interactions with other microbes, we found that genes with such functions were shared with both mycoparasites and animal parasites. In fact, we found that SSPs were overrepresented among the genes that were found to be unique to either new hyperparasite species, and those that were uniquely shared with animal parasites and both animal and mycoparasites. Small secreted proteins are often bioactive compounds that have a strong link to self-protection and host association (Kim et al. 2016). As such, both new hyperparasite species appear to have infection strategies that are unique to either species, as well as shared with other Hypocrealean parasites. Quite some of these SSPs were annotated to be toxins or antimicrobials, which further indicates that they likely play a role in the infection of O. camponoti-oridani. In addition, the high number of uncharacterized SSPs highlights the potential for new bioactive compound discovery in these species that could lead to the development of new applicable antifungals.
Our further comparative genomics analyses followed the lead of previous studies that have singled out high numbers of secondary metabolite clusters, proteases, and GH18 chitinases as genome signatures of a mycoparasitic lifestyle. However, these studies either did not include other lifestyles beyond mycoparasitism in their comparisons (Druzhinina et al. 2012;Karlsson et al. 2015), or chose representative species for other lifestyles outside of the order Hypocreales (Karlsson et al. 2017). We found differences in numbers of secondary metabolites, proteases, and chitinases that were not consistent or not large enough between lifestyle groups to be conclusive. As such, our results show that the functional gene categories discussed in earlier studies are not necessarily informative when conducting cross-lifestyle comparisons within the order Hypocreales. Instead, some of the suggested genome signatures may be merely characteristic of Hypocreales as a whole. Nevertheless, the numbers of functional annotations that we obtained for the secondary metabolite clusters, proteases and chitinases in the P. oviedoensis and N. ophiocordycipitis genomes, did generally place these species among the averages for mycoparasites and animal parasites, which is in line with our orthologous clustering results. As such, our genome comparisons could be indicating that fungal parasites that infect animal and fungal hosts largely evolved similar strategies, while plant parasites use other strategies.
Alternatively, this could be another suggestion that the mycoparasitic species that infect O. camponoti-oridani evolved from entomopathogens. Indeed, both hyperparasites have at least one gene that is annotated to encode for an enterotoxin. Beyond fungal animal parasites (which contain many enterotoxins) most fungal genomes do not contain a single PFAM annotation for this function (de Bekker et al. 2017). In line with this, none of the other mycoparasitic fungi in our study contain any putative enterotoxins in their genomes. As such, the remaining enterotoxin genes could indicate an earlier entomopathogenic ancestry for both P. oviedoensis and N. ophiocordycipitis.

Conclusion
To understand the complex dynamics of inter-species interactions more thoroughly, there is a need to incorporate tritrophic levels. Nevertheless, the natural enemies of parasites in parasite-host interactions are still vastly understudied. Our study contributes systematics, natural history, and genomic data to the current very limited knowledge of fungal hyperparasites of entomopathogenic fungi to shed a light on hyperparasitism and advocate for more in-depth investigations of fungal species that infect other fungi. We formally described and sequenced the genomes of two new hyperparasite species that infect the ant behavior manipulating parasite O. camponoti-oridani: Polycephalomyces oviedoensis and Niveomyces ophiocordycipitis. Our need to describe a new genus within the family Cordycipitaceae for one of the new species highlights the current limited data for this part of the Hypocreales phylogenetic tree. Moreover, our study adds considerable data to the very limited number of available mycoparasite genomes by providing two new genomes, of which one is the rst genome in the genus Polycephalomyces. Our natural history data suggests that both species described in this study impact the transmission of the behaviormanipulating ant parasite O. camponoti-oridani, warranting further detailed studies into this tri-trophic fungusfungus-ant system. More detailed integrative studies into the infection biology of these and other mycoparasitic fungi, from the molecular to the ecological level, will result in the leveraging of new strategies to battle mycoses. Such discoveries are especially timely with fungal disease being on the rise. The annotated genomes are deposited in GenBank: accession numbers JADHZA000000000 (Polycephalomyces oviedoensis) and JAFEME000000000 (Niveomyces ophiocordycipitis). The genome assembly, gene predictions and functional annotations can also be analyzed interactively at http://fungalgenomics.science.uu.nl. All other data has been included as additional les to this manuscript.

Competing Interests
The authors declare that they have no competing interests.

Funding
The work presented in this manuscript has been supported by startup funds from the University of Central Florida, made available to CdB.

Author's Contributions
This study was conceived by BML and CdB who both also collected the samples used in this study and have written the manuscript. All other authors contributed to editing the manuscript. BML isolated the DNA, assembled the genomes, performed the functional annotations and comparative genomics analyses, and analyzed the eld data.
JPMA performed the phylogenetic analysis and the taxonomy. SRL collected the eld data. RAO annotated the genomes and AB performed the library preparations and genome sequencing. for each of the fungal species that that we found to be closely related to the species that are the focus of this study.    Orthologous clustering. Venn diagrams representing the shared orthogroups between the genomes of P. oviedoensis (A) and N. ophiocordycipitis (B) and the genomes of 15 other fungi that we grouped into different fungal lifestyles based on their ability to parasitize animal, fungal, or plant hosts, or decompose dead organic material.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.