The formation of carpophoroids associated with species of
E.
abortivum has traditionally been thought to be the result of an
Armillaria species attacking and parasitizing
Entoloma sporocarps (
14), hypothesized on the basis that
Armillaria species are widespread generalist forest pathogens that have a broad range of host plants (
15). However, subsequent studies suggested the opposite: the production of carpophoroids is the result of
E. abortivum disrupting the development of
Armillaria sporocarps (
17). Here, we employed RNA sequencing and differential gene expression analysis on field-collected fungal tissue of each of the three components of this association to better understand the mechanistic basis of this interaction. We determined that
E. abortivum reads in the metatranscriptome of the carpophoroid tissue—which can be interpreted as a measure of living tissue—are almost 35 times more abundant than
A. mellea reads (
Fig. 6). This finding suggests that carpophoroids are structures that result from
E. abortivum parasitizing, and eventually killing, its
Armillaria host under natural conditions.
The genetics of the Entoloma-Armillaria mycoparasitic interaction.
A crucial step in a successful mycoparasite’s life history is the ability to sense its host. Genes involved in the recognition of the fungal prey include those that code for GPCRs (
18,
20). However, we did not find any of these genes that were differentially upregulated by
E. abortivum in the carpophoroid tissue (
Fig. 3). Given the significantly lower number of
Armillaria reads in the carpophoroid tissue compared to
E. abortivum, we presume that these carpophoroids are relatively advanced in age, and expression of the genes used for sensing the presence of the host is no longer necessary.
We also identified three
E. abortivum transcripts that code for β-trefoil-type lectins—proteins that bind to galactose units of sugar chains (
23)—that were differentially upregulated in the carpophoroid tissue (
Fig. 3 and
5). In well-studied mycoparasitic interactions, the recognition, attachment, and coiling around a fungal substrate are mediated by lectins expressed by at least one of the fungal partners (
24–26). More specifically, basidiomycete β-trefoil-type lectins have previously been shown to play a role in the recognition of nonself glycans (
27). Interestingly, the most abundant and differentially upregulated transcript produced in the
Armillaria sporocarps, the substrate to which
E. abortivum hyphae attach, codes for a cell wall galactomannoprotein (
Fig. 4 and
5). These proteins belong to a group of glycans which consist of a mannose backbone with galactose side chains and are known to make up a major part of the cell wall of some fungal species (
28). This particular galactomannoprotein appears to be specific to
Armillaria species. Watling (
14) commented on the highly specific nature of this interaction and that it has been documented only occurring between
E. abortivum and
Armillaria species. One possible mediator of the specificity of this interaction could be the galactose sugars on the mannose protein (known thus far only from
Armillaria species) that are the means by which
E. abortivum β-trefoil-type lectins recognize and attach to the
Armillaria host. However, more genome sequencing of other Agaricales species is needed to determine whether this protein is truly specific to species in the genus
Armillaria.
During mycoparasitic interactions, the fungal host responds by mounting its own defense, and a successful mycoparasite must be able to cope with this counterattack (
18,
20). Oxalic acid (OA) is a virulence factor employed by some plant pathogens, including species of
Armillaria, to compromise the defense responses of the host plant by creating an acidic environment (
29,
30). One differentially upregulated
Armillaria transcript in the sporocarps codes for isocitrate lyase (
Fig. 5), which is involved in OA biosynthesis in other fungal pathogens (
31,
32), and we suspect that
Armillaria sporocarps may utilize OA to defend against the initial infection by
E. abortivum. However, two of the most abundant and differentially upregulated
E. abortivum transcripts in the carpophoroid tissue code for oxalate decarboxylases—enzymes responsible for the degradation of OA (
Fig. 3,
4, and
5D;
Table 1). In at least one known well-studied mycoparasitic interaction, OA is secreted by the fungal host,
Sclerotinia sclerotiorum, in reaction to penetration by its mycoparasite,
Coniothyrium minitans. The acidic environment created by the secreted OA inhibits conidial germination and suppresses mycelial growth of
C. minitans (
33). However,
C. minitans nullifies the growth-suppressing effects of OA or OA-mediated low pH by degrading the OA (
34,
35), an enzymatic process largely mediated by oxalate decarboxylase. Because of this, oxalate decarboxylase plays an imperative role in mycoparasitism as OA degradation is vital for infection of the fungal host (
36). Given the abundance of these genes produced by
E. abortivum in the carpophoroid, we suspect a similar scenario in this mycoparasitic interaction. Subsequent downregulation of isocitrate lyase by
Armillaria in the carpophoroids suggests that its genetic defense responses to infection were likely over at the time of our sampling.
Other ways that mycoparasites cope with the counterattack launched by their host include actively excreting host-secreted toxins. Here, we hypothesize that active extrusion of toxins secreted by the host occurs in the
E. abortivum carpophoroid tissue via membrane transporters in the ABC superfamily (
37–39). Three ABC transporters were differentially upregulated in the carpophoroid tissue (
Fig. 3). Another group of genes that were differentially upregulated by
E. abortivum in the carpophoroid belong to the major facilitator superfamily (MFS) transporters (
Fig. 3). In
C. rosea, there was selection for genes in this family that were related to drug resistance and the transport of secondary metabolites, small organic compounds, and carbohydrates (
40). Their importance to mycoparasitism in
C. rosea is predicted to invoke efflux-mediated protection against exogenous or endogenous secondary metabolites and nutrient uptake (
40). MFS transporters have also been shown to be induced in other mycoparasitic species (
19,
41), but their exact biological roles have not been investigated.
In mycoparasitism, the final death of the host often results from the synergistic actions of cell wall-hydrolytic enzymes and antifungal secondary metabolites (
18,
20). No secondary metabolite gene clusters identified in the
E. abortivum transcriptome were differentially upregulated in the carpophoroid tissue (
Fig. 3). In some mycoparasitic relationships, the secretion of secondary metabolites occurs early in the interaction, including in
Escovopsis weberi, which secretes toxic compounds that kill the leafcutter ant garden before contact (
5). In culture experiments between
Armillaria isolates and
E. abortivum, the growth of
Armillaria was severely inhibited by the presence of
E. abortivum (
42). This suggests that
E. abortivum may potentially secrete a toxic compound early in the interaction that inhibits the growth of
Armillaria. Given that significantly more of the living tissue in the carpophoroids belonged to
E. abortivum (
Fig. 6), it is possible that much of the
Armillaria tissue was killed preceding the full development of the carpophoroid.
Additionally, we hypothesize that the upregulated β-trefoil-type lectin in
E.
abortivum that may be important in hyphal recognition may also be cytotoxic toward
Armillaria. This type of lectin has sequence homology, as well as putative structural similarity, to the B-subunit of ricin, a toxic protein from the castor bean
Ricinus communis (
43). An array of β-trefoil-type lectins have been characterized from the sporocarps of the mushroom-forming species
Clitocybe nebularis (
44),
Coprinus cinerea (
27),
Macrolepiota procera (
45), and
Boletus edulis (
46). Besides being important for nonself recognition, these same mushroom lectins also exhibit entomotoxic activity (
47) as well as nematoxic activity (
27,
45,
48). Taken together, it is possible that the
E. abortivum β-trefoil-type lectins may also function as toxins toward
Armillaria. While the
E. abortivum transcripts coding for these lectins are not in the highest abundance in the carpophoroid tissue (
Fig. 5), this could be because most of the
Armillaria sporocarp tissue is already dead and the potential lethal effects produced by them are no longer necessary.
Chitin is an essential polymer in fungal cell walls (
49) and is an important target during mycoparasitic attack (
20). Indicative of the importance of chitinases in mycoparasitic interactions, members of the genus
Trichoderma, as well as
Tolypocladium ophioglossoides and
Escovopsis weberi, have an increased number of genes coding for them (
19,
50–53). Nine fungal chitinases were detected in the transcriptome of
E. abortivum, which is fewer than the 13, 19, 20, and 29 detected in the closely related species
Tricholoma matsutake (
54),
Clitocybe gibba (
55),
Lyophyllum atratum (
56), and
Lepista nuda (
55), respectively (
Fig. 2). Only two of those
E. abortivum chitinases were differentially upregulated in the carpophoroid tissue (
Fig. 3) and were not abundant in comparison to other genes, suggesting minimal significance at this stage in carpophoroid development. One possibility for this difference in abundance could be the result of the putatively acidic pH in the carpophoroid that we infer based on the high gene expression of oxalate decarboxylases. In
C. minitans, chitinase activity is positively correlated with ambient pH ranging from 3 to 8 (
57), so it is possible that chitinase activity in
E. abortivum will increase after a neutral pH is restored. Another possibility for the low abundance is that most of the
Armillaria host tissue was already broken down, reducing the need for chitinase activity.
Some of the putatively mycoparasitism-related genes outlined above were also differentially upregulated by
Armillaria in the carpophoroid tissue. These include genes that code for MFS, ABC transporters, chitinases, and secondary metabolite gene clusters (
Fig. 3). This suggests that
Armillaria may be using many of the same genetic mechanisms to defend itself against parasitism by
E. abortivum. Additionally, the degree of expression changes—in both the number of differentially upregulated transcripts and the log fold change (logFC)—between the sporocarp and carpophoroid is much greater in
Armillaria than in
E. abortivum (
Fig. 3 and
4), which could reflect an increase in the level of defense from
Armillaria. However, this defense is apparently not enough to overcome the parasitic adaptations of
E. abortivum.
Gene and CAZyme content of E. abortivum.
The number of predicted gene models in the transcriptome of
E. abortivum was 9,728, which is markedly fewer than the number of gene models in the genomes of its closest sequenced relatives (
Fig. 2, middle panel). Additionally, relative to other closely related mushroom species,
E. abortivum also exhibits a strong reduction in several gene families encoding CAZymes (
Fig. 2, right panel) and contains no cellobiohydrolases, xylanases, or polysaccharide monooxygenases. This finding is consistent with what has been observed in obligate mycoparasites and animal pathogens, which also have reduced CAZyme repertoires compared to fungi that utilize other nutritional strategies, hypothesized to be the result of their highly specialized interactions with their hosts (
50,
58,
59). Therefore, it is possible that
E. abortivum retained only the CAZymes and accessory genes necessary to interact with
Armillaria species. While a genome sequence of
E. abortivum will be necessary to confirm this reduction, the BUSCO analysis verified that the
E. abortivum transcriptome contains nearly 95% of the core set of eukaryotic genes, which suggests that our transcriptome is relatively complete, and it is unlikely that entire gene families are underrepresented.
Another possibility for the reduction in CAZymes could be explained by a broader nutritional strategy employed by
Entoloma species, some of which form ectomycorrhiza-like structures on host plant species (
60–62). Ectomycorrhizal species have a marked reduction in CAZymes in comparison to their saprotrophic ancestors (
63), which we also observe with
Tricholoma matsutake (
Fig. 2, right panel). However, microscopic analyses of
Entoloma ectomycorrhiza-like structures suggest that some species destroy root meristems and young root cells, suggestive of a more parasitic relationship (
60,
61). One explanation is that
Entoloma species, in general, are parasites of true ectomycorrhizae. More broadly, other species within the Entolomataceae are suspected mycoparasites, as they have been reported growing in close association with the sporocarps of other species (reviewed in reference
17), including
Entoloma parasiticum (=
Claudopus parasiticus) and
Clitopilus fasciculatus (
64,
65). This explanation would also add credence to the evidence that Entolomataceae species are difficult to culture and are slow growing (
62). Additional research utilizing genomes and laboratory studies to understand the nutritional strategy employed by this lineage will inform us as to whether fungal parasitism in this group is more common than it is currently understood to be. Lastly, it should be noted that
Entoloma species, and
E.
abortivum in particular, form sclerotia in culture and presumably in soil (
14,
17). These resting structures are perhaps a dormancy mechanism in the soil to survive seasonality when host tissue is unavailable. Because the timing and presence of
E.
abortivum fruitings have been hard to predict, we were not able to observe or acquire a transcriptome for the sclerotia in the field.
Gene content and identity of Armillaria species in this interaction.
The number of predicted gene models in the transcriptome of this
Armillaria species was 38,215 (
Fig. 2, middle panel). This is substantially greater than all other sequenced
Armillaria species, which range from 14,473 to 25,704 gene models (
22,
66). However, when we looked for gene models in our transcriptome that belong to known
Armillaria species, this reduced the total number of gene models to 29,936 (
Fig. 2, middle panel). The excess gene models in the
Armillaria transcriptome, compared to reference genomes, likely represent duplicated gene models with splice variants, a common artifact of transcriptome sequencing (
67). Additionally, nearly one-quarter of the gene models in the
Armillaria sporocarps were from organisms other than
Armillaria, including the yeast
Kodamaea, highlighting the fact that field-collected sporocarps are not composed of tissue from a single organism. However, none of these contaminating organisms had genes that were both abundant and differentially upregulated that we predicted to play a role in this interaction.
Phylogenomic analysis of the
Armillaria transcripts generated in this study suggests that the specific
Armillaria species parasitized in this relationship is sister to an
A. mellea specimen collected from western Europe (
Fig. 7A). An ITS-based phylogenetic analysis shows the
Armillaria specimen collected in this study is conspecific with other
A. mellea collections from eastern North America (
Fig. 7B). Before now, observations of
A. mellea sensu stricto fruiting in proximity to carpophoroids (
17) hinted that it may be a host to
E. abortivum. Here, though, we show for the first time using genomic data that
A. mellea sensu stricto can definitively serve as a host for
E. abortivum. However, our hypothesis is that this interaction does not appear specific to just
A. mellea sensu stricto as
Armillaria gallica,
Armillaria ostoyae,
Armillaria jezoensis,
Armillaria sp. Nag. E, and
Desarmillaria tabescens have been previously confirmed as hosts in this interaction as well (
17,
42,
68). However, a broad geographic sampling of carpophoroids using molecular markers or genomic information could address this question more thoroughly. Interestingly,
Armillaria species parasitized by
E. abortivum appear to be only those present in eastern North America and eastern Asia (
17,
42,
68).