Disease management in two sympatric Apterostigma fungus‐growing ants for controlling the parasitic fungus Escovopsis

Abstract Antagonistic interactions between host and parasites are often embedded in networks of interacting species, in which hosts may be attacked by competing parasites species, and parasites may infect more than one host species. To better understand the evolution of host defenses and parasite counterdefenses in the context of a multihost, multiparasite system, we studied two sympatric species, of congeneric fungus‐growing ants (Attini) species and their symbiotic fungal cultivars, which are attacked by multiple morphotypes of parasitic fungi in the genus, Escovopsis. To assess whether closely related ant species and their cultured fungi are evolving defenses against the same or different parasitic strains, we characterized Escovopsis that were isolated from colonies of sympatric Apterostigma dentigerum and A. pilosum. We assessed in vitro and in vivo interactions of these parasites with their hosts. While the ant cultivars are parasitized by similar Escovopsis spp., the frequency of infection by these pathogens differs between the two ant species. The ability of the host fungi to suppress Escovopsis growth, as well as ant defensive responses toward the parasites, differs depending on the parasite strain and on the host ant species.


| INTRODUC TI ON
Individuals within insect societies receive benefits and share responsibilities for the maintenance and propagation of their group (Hölldobler & Wilson, 1990;Weber, 1972). Some challenges require individual responses, and others require collective action (Hart, 1990). Adaptations at individual and social levels enable the development of task allocation strategies, including those for the prevention and control of disease (Schmid-Hempel, 1995, 1998Wilson, 1971). Variation in responses of individuals from different colonies could result in the development of different disease management strategies among colonies or species (Franceschi et al., 2005). Moreover, at local scales, when two or more species are challenged by the same pathogens, complementary defensive responses may reduce the prevalence or overall impact of the parasites by decreasing the likelihood that a pathogen evolves counterstrategies to a single host species.
When alternative host species are present, these hosts may function as a general pool of acceptable hosts with which a parasite evolves. To the extent that each host has unique defense strategies, it is potentially harder for a parasite to evolve to overcome all defenses. Cocktails of antimicrobial chemicals may be used to treat pathogens in order to minimize the likelihood of the evolution of resistance (Raymond, 2019). The fungus-growing ant system provides an excellent opportunity to explore the degree to which different pathogens attack sympatric hosts, and how the hosts respond. This is in part because of the experimental tractability of the ants, the ants' cultivated fungi, their common fungal parasites (Escovopsis spp.), and other associated microbes Sen et al., 2009).
Previous work on Escovopsis infections in other sympatric, closely related attine species [i.e., Cyphomyrmex costatus, C. muelleri, and C. longiscapus (Birnbaum & Gerardo, 2016;Gerardo et al., 2004); Atta spp. and Acromyrmex spp. (Taerum et al., 2010)] indicates that colonies can be infected by the same parasites, though individual parasite strains may be more likely to infect colonies of one species than the others. In Cyphomyrmex, observed patterns of natural infections are consistent with differences in the parasites' abilities to infect the ants' gardens in experimental infections (Birnbaum & Gerardo, 2016;Gerardo et al., 2004). These differences, however, have not been clearly linked to ant behavioral defenses toward the parasites. Here, we test whether the colonies of two sympatric ant species, Apterostigma dentigerum and A. pilosum, are infected by the same parasite strains. We characterize the diversity of these parasites both genetically and phenotypically. We then assess whether the outcomes of parasite infections are similar in both colony types or whether these hosts present unique defensive challenges for these parasites to overcome.

| Diversity and prevalence of Escovopsis morphotypes in colonies of Apterostigma spp
To assess the prevalence and distribution of Escovopsis among host colonies of two ant species, we utilized two methods (M1 and M2) to maximize the likelihood of isolating parasite strains. First (M1), within five days of colony collection, from each of 32 A. dentigerum and 14 A. pilosum colonies, we isolated 50 pieces of fungal garden material (3-5 mm diameter) using sterilized forceps. Ten pieces were plated aseptically on a Petri dish containing potato dextrose agar (PDA, Scharlau culture media, 39 g/L) without antibiotics, and all five dishes per colony were incubated at room temperature for 10 days.
Each day, the dishes were scanned for Escovopsis growth, which was transferred to a new PDA dish to obtain pure cultures. Second (M2), within five days of colony collection, 5 to 20 larger fungus garden pieces (~0.5 g) per colony (N = 12 for each species) without workers and brood were placed in sterile Petri dishes (100 × 15 mm), containing sterile, water-soaked paper, sealed with parafilm. These garden fragments were incubated at room temperature and allowed to grow for 7 to 10 days. Each day, all dishes were scanned for Escovopsis growth, and fungi were transferred to a new PDA dish to obtain pure cultures.

| Phenotypic characterization of prevalent Escovopsis morphotypes
Characterization based on patterns of growth. Four Escovopsis morphotypes isolated from Apterostigma species were classified based on spore color and growth rate. Hereafter, we refer to these morphotypes of Escovopsis based on color as Y (yellow conidia and fluffy mycelium, from A. pilosum colony); B1 (brown conidia, pulverulent mycelium with a single concentric ring of white conidia formed away from the inoculum, from A. dentigerum colony); B2 (brown conidia, pulverulent mycelium with a single concentric ring of white to light brown conidia formed away from the inoculum, from A. dentigerum colony); and B3 (dark brown conidia, pulverulent mycelium with a single concentric ring of light to dark brown conidia formed away from the inoculum, from A. pilosum colony) (Figure 1c-f). As an outgroup, given the specificity reported for the Escovopsis-fungal cultivar symbiosis (Birnbaum & Gerardo, 2016;Taerum et al., 2007), we isolated a fifth morphotype of Escovopsis from Trachymyrmex zeteki (TB) (cottony hard mycelium growth and brown colored conidia) (Appendices S1-S2).
To obtain micrographs of Escovopsis conidia using scanning electron microscopy, we selected a single strain randomly from each of the five morphotypes. We grew the strains on PDA for seven days at room temperature and fixed their fungal structures with osmium tetroxide vapor. To determine the fungal growth rates of Escovopsis morphotypes, we collected individual fungal pieces of Escovopsis conidia from pure culture using a sterilized inoculating needle with a looped end (1 mm 2 ). We placed each piece in the center of a Petri dish (100 × 15 mm) containing PDA without antibiotics. We plated ten pieces per morphotype. The dishes were sealed with parafilm and incubated at 12:12 light:dark and at room temperature (~27°C) for 20 days. Each Petri dish then was photographed using a Nikon Coolpix camera. Total area of growth was analyzed using ImageJ v1.4 (Schneider et al., 2012).
Characterization based on chemical profiling. To determine the chemical profiles of the five Escovopsis morphotypes, we grew 200 PDA dishes/morphotype at room temperature for 20 days.
All the fungal growth on agar was cut into small pieces with a sterile spatula and transferred to Erlenmeyer flasks (500 ml). We

| In vitro Escovopsis-cultivar interactions
We conducted confrontation bioassays between the Escovopsis strains and cultivars of A. dentigerum (CAd.) and A. pilosum (CAp.) to assess whether the parasites vary in their ability to grow in the presence of alternative cultivars. We first plated small pieces of cultivar mycelium on the edges of PDA Petri dishes (100 × 15 mm) and allowed them to grow for 20 days. Then, we used a steri- and three replicates of controls with each Escovopsis strain alone (Y Ct, B1 Ct, B2 Ct, B3 Ct, and TB Ct). All plates were monitored and photographed every five days for 20 days, and the growth area of Escovopsis (in cm 2 ) was measured using ImageJ 1.4v (Schneider et al., 2012;Appendix S3). We performed a two-way ANOVA with interaction effect between treatment (fungal cultivar isolated from each ant species) and Escovopsis strains (five morphotypes). All analysis and plots were performed with the statistical software R (R Development Core, 2008). Two-way ANOVA was performed using the aov() function; normality and homoscedasticity of the data were analyzed using the shapiroTest() and leveneTest() functions.
Pairwise comparisons were adjusted using the Tukey HSD function.

| Behavior of ants in response to Escovopsis infections of their gardens
To evaluate antiparasitic behavioral defenses (e.g., fungal grooming; Currie & Stuart, 2001) in response to alternative Escovopsis, we scored behavior of ants from 22 ant colonies not previously inoculated with Escovopsis (10 colonies of A. dentigerum and 12 colonies of A. pilosum).
We conducted experiments using entire colonies. Prior to infection, we cultured a representative of each of the five Escovopsis morphotypes on PDA media at room temperature for 20 days. We then weighed each colony and recorded the number of ants. We cut 0.5 cm PDA plugs of Escovopsis conidia as described above. The Escovopsis conidia were inoculated randomly in five areas of the fungus garden (five morphotypes per dish), maintaining a distance of ∼1 cm between pieces. The Escovopsis plugs were placed gently on the fungus garden, applying conidia to the garden. We inoculated each of the fungus gardens with 4.4 ± 0.17 × 10 6 dry conidia (mean ± SD) of each of the morphotypes of Escovopsis and waited five minutes for acclimation of the workers. Using stereomicroscopy (Leica 10x, with a dim, not hot light), we recorded the number of workers fungal grooming the areas with each Escovopsis morphotype every 10 min for two hours; grooming behaviors involved scanning the fungal garden surface to remove contaminants (Currie & Stuart, 2001;Fernández-Marín et al., 2013). The Petri dishes were gently moved, without affecting the behavior of the ants. For each nest, we summed the total number of grooming events observed in each of the five infected garden regions. We assessed whether the ant species differed in their propensity to groom infections of each Escovopsis type using multinomial logistic regression via the multicomp() function in the nnet package of R. Significance was assessed using a likelihood ratio test.
To better understand the outcome of grooming in response to Escovopsis garden infection, we evaluated the accumulation of microbes in infrabuccal pellets after Escovopsis infection of their gardens. Infrabuccal pellet comprised of detritus and other materials that are packed into a specialized pocket on the head after ant grooming behavior. They are then discarded to trash piles (Fernández-Marín et al., 2006;Little et al., 2006). We formed five subcolonies, each consisting of 0.3 g of fungus garden and 10 workers, from each of five colonies of A. dentigerum and 10 colonies from A. pilosum, and placed them onto separate, sterile Petri dishes (100 × 15 mm).
We cut 0.65 cm PDA plugs covered with Escovopsis conidia as before. The fungus garden from each subcolony was inoculated with 5.3 ± 0.47 × 10 6 conidia (mean ± SD) by rubbing a sterilized swab of Escovopsis conidia onto the center of the garden, one morphotype per subcolony. Eight hours after inoculation, we removed the workers, and all infrabuccal pellets were collected and plated on PDA. The infrabuccal pellets were incubated at room temperature to score germination of Escovopsis, antibiotic-producing actinomycete bacteria, mutualistic cultivar, and other nonmutualistic fungi.
We performed a one-way MANOVA to evaluate the effect of host ant species, Escovopsis morphotype, and their interaction. Data were transformed (log + 1) when the assumption of normality was violated, and homoscedasticity of the data was checked with the levene-Test() function. Outliers were detected using box plots and removed from the analysis. To detect the global effect of each variable, we conducted a univariate one-way ANOVA using the aov() function.

| Diversity and prevalence of Escovopsis morphotypes in sympatric Apterostigma spp
We isolated 48 Escovopsis isolates from a total of 2,300 small pieces of fungal garden (M1)  (B1, B2, B3) grew significantly more than the Y morphotype Escovopsis in the same time period but did not differ in growth rate from one another; the TB morphotype grew significantly more than all four of the morphotypes from Apterostigma colonies (Figure 3).

| Characterization based on chemical profiling
We detected 171 ion signals across chemical profiles based on their molecular weights. Of these 171 signals, we identified 16 based on mass as belonging to a group of cyclic peptides, suggesting the presence of a series of diketopiperazine-type compounds in the extracts produced by these Escovopsis (Appendix S4). Phylogenetic analyses based on chemical profiles yielded a single most parsimonious tree (Figure 4a). Bootstrap analyses strongly supported monophyly among Escovopsis brown morphotypes, while the yellow Escovopsis morphotype formed a distinct clade.

| In vitro Escovopsis-cultivar interactions
Escovopsis morphotypes significantly varied in area of growth during in vitro bioassays ( Figure 5, there were no significant differences in Escovopsis mycelial growth based on the origin of the cultivar (i.e., from A. dentigerum or A. pilosum colonies).

| Behavior of ants in responses to Escovopsis infections of their gardens
In response to inoculation of colonies with the conidia of multiple

| Diversity and frequency of Escovopsis morphotypes between ant congeners
Our results highlight that the fungal cultivars of colonies of two closely related, sympatric Apterostigma ant species are infected by the same  Gerardo, 2016;Gerardo et al., 2004). Similarly, sympatric colonies of Atta spp. and Acromyrmex spp. leaf-cutting ants are attacked by the same Escovopsis parasites Taerum et al., 2007).
There are, however, differences in the prevalence of Escovopsis strains infecting colonies of Apterostigma dentigerum and A. pilosum, in the degree to which cultivars suppress growth of the pathogen species, and in the degree to which pathogen species stimulate ant defensive behaviors.
Colonies of both A. dentigerum and A. pilosum were frequently infected with multiple Escovopsis strains (11.4% and 23.07%, respectively), a phenomenon also observed in relation to colonies of other fungus-growing ant species (Taerum et al., 2010). Based on theory, coinfecting parasites are expected to compete with each other for host resources (competitive suppression) but incidentally may together overwhelm host defenses (Abdullah et al., 2017;Bashey, 2015). The impact of coinfection can be influenced by the degree of genetic similarity between coinfecting parasites (Read & Taylor, 2001) and by changes in parasite behavior in the presence of coinfecting parasites (Rayner, 1991

| The utilization of multifaceted, specific defense strategies
In response to Escovopsis infection, fungus-growing ants groom their gardens and apply antimicrobial compounds. This behavior facilitates removal of infectious parasite spores. The congeneric Apterostigma ants groomed garden areas infected by some parasite species more than others. Both species tended to groom areas infected with the yellow Escovopsis morphotype more than areas infected with the brown Escovopsis morphotype, but this was significant only for A. pilosum. Most strikingly, the ants tended the garden area infected with TB Escovopsis, which they may not encounter in nature much less the garden areas with other Escovopsis.
The ants' differential responses to different parasites suggest that the ants can discriminate between different Escovopsis strains as previously suggested (Goes et al., 2020). However, we do not know whether ants' attention to particular morphotypes can be modified by adding or removing other Escovopsis morphotypes or pathogens. Previous in vivo infection experiments studies demonstrated that leaf cutter ants in the genera Atta and Acromyrmex adjust their hygienic defensive strategies (e.g., metapleural glands grooming) to respond to distinct fungal conidia (Fernández-Marín et al., 2013), and adjust the quantity of their antifungal metapleural gland secretions in response to different pathogens (Fernández-Marín et al., 2015;Yek et al., 2012). These findings, coupled with our own, indicate that fungus-growing ants are able F I G U R E 7 Emergence of microbes from infrabuccal pellets collected from (a) A. dentigerum colonies and (b) A. pilosum colonies. Colonies were subdivided into five subcolonies, each receiving a different Escovopsis infection (Y, B1, B2, B3, or TB, as in previous figures). Emergence of microbes was recorded from recovered pellets (n = 0 to 29 pellets per subcolony, mean = 12, SD = 7); no pellets were recorded from colony 6 when infected with Y morphotype Escovopsis. Colonies vary significantly in the types of microbes that emerge from pellets, and the Escovopsis morphotype with which the subcolony was infected also influences microbial emergence from pellets to differentially respond to conidia of distinct fungal types, including Escovopsis spp. that threaten their gardens' health.
This specificity of the defensive response requires a mechanism for the ants to recognize and distinguish different Escovopsis strains (Currie & Stuart, 2001;Fernández-Marín et al., 2009), but nothing is known of the sensory mechanisms for detection and discrimination. We hypothesize that this recognition could be driven by differential parasite chemistry, as the ants exhibited similar frequencies of fungal grooming toward Escovopsis that were chemically similar (Figure 4a,b). However, aspects of parasite morphology, growth, and other phenotypes could also modulate the behavioral responses of the ant hosts.
In a separate experiment, we isolated microbes from infrabuccal pellets to assess the degree to which the ants may be utilizing two defensive strategies, grooming (Currie & Stuart, 2001) and actinomycete-based antibiotic inhibition (Currie, Scott, et al., 1999). Infrabuccal pellets reflect material that worker ants have removed from the garden or microbes being used to actively kill pathogens (Fernández-Marín et al., 2006). Presence of Escovopsis in the pellets likely reflects active grooming of infected areas. In terms of antibiotic suppression, previous research has shown that upon infection with Escovopsis, infrabuccal pellets are more likely to contain actinomycete bacteria that produce antibiotics that can suppress Escovopsis growth (Little et al., 2006).
Thus, increased proportions of pellets containing these bacteria in response to some Escovopsis and not others could suggest that some Escovopsis stimulate a stronger actinomycete-based antimicrobial response. The disease management implications of the presence of cultivar and nonmutualistic fungi in the pellets is less clear. Interestingly, Escovopsis morphotype, ant species, and the interaction between the two all had significant impacts on the microbial composition of isolated pellets, suggesting that the ants use defenses differently based on the type of parasite.
In addition to the ant-and bacteria-derived defenses, fungal cultivars produce compounds with antifungal properties that can inhibit Escovopsis fungal growth or spore germination (Oh et al., 2009;Wang et al., 1999). The result of our fungal cultivar versus Escovopsis interaction experiments showed no effect of the fungal cultivars on growth of the Escovopsis TB morphotype, but inhibition of both the yellow and brown Escovopsis morphotypes from Apterostigma spp. The ants also showed reduced behavioral response to the TB morphotype compared with the yellow and brown Escovopsis strains. This is consistent with the hypothesis that both the ants and their mutualistic fungi are under selection to maintain defenses that suppress the parasites that infect them in nature, and it is also consistent with previous work indicating that both ant genotype and cultivar lineage influence the outcome of Escovopsis infections (Kellner et al., 2018). However, previous work suggested that cultivars are more likely to inhibit the growth of non-native pathogens than the pathogens that attack them in nature (Birnbaum & Gerardo, 2016;Custodio & Rodrigues, 2018).
Overall, patterns of these bioassays may be highly dependent on the parasite and host strains utilized, which would be consistent with a model of ongoing coevolution. Thus, difference across studies, particularly when few host and parasite combinations are utilized (as here), is not surprising.

| CON CLUS IONS
Apterostigma ants utilize multiple defenses against Escovopsis infections, but the prevalence of Escovopsis infections suggests that the parasites possess mechanisms to overcome these defensive strategies, establish infections, and prevent clearance at least some colonies. The utilization of multiple defensive strategies, and the differential utilization of different strategies between congeneric ant species in sympatry, might help reduce the likelihood of the evolution of resistance of Escovopsis against colony-level host defenses. Studies integrating different approaches at distinct scales are needed to better understand how variation in colony defenses within populations of more than one host species shape ecology and evolution dynamics of these infections. To further elucidate population-level disease dynamics and defense, more detailed taxonomic and molecular investigations of Escovopsis are needed (see Augustin et al., 2013;Masiulionis et al., 2015;Meirelles, Montoya, et al., 2015;Montoya et al., 2019). In particular, future research should study geographic patterns of hostparasite association in this system and coupled this work with experimental studies of coinfection, infection outcomes, and defense utilization.
We thank Jorge Ceballos for the SEM images and Celestino Aguilar for helping with phylogenetic analysis. We thank the Ministerio de Ambiente for collecting permits and the Smithsonian Tropical