Laboulbeniales hyperparasites (Fungi, Ascomycota) of bat flies: Independent origins and host associations

Abstract The aim of this study was to explore the diversity of ectoparasitic fungi (Ascomycota, Laboulbeniales) that use bat flies (Diptera, Hippoboscoidea) as hosts. Bat flies themselves live as ectoparasites on the fur and wing membranes of bats (Mammalia, Chiroptera); hence this is a tripartite parasite system. Here, we collected bats, bat flies, and Laboulbeniales, and conducted phylogenetic analyses of Laboulbeniales to contrast morphology with ribosomal sequence data. Parasitism of bat flies by Laboulbeniales arose at least three times independently, once in the Eastern Hemisphere (Arthrorhynchus) and twice in the Western Hemisphere (Gloeandromyces, Nycteromyces). We hypothesize that the genera Arthrorhynchus and Nycteromyces evolved independently from lineages of ectoparasites of true bugs (Hemiptera). We assessed phylogenetic diversity of the genus Gloeandromyces by considering the LSU rDNA region. Phenotypic plasticity and position‐induced morphological adaptations go hand in hand. Different morphotypes belong to the same phylogenetic species. Two species, G. pageanus and G. streblae, show divergence by host utilization. In our assessment of coevolution, we only observe congruence between the Old World clades of bat flies and Laboulbeniales. The other associations are the result of the roosting ecology of the bat hosts. This study has considerably increased our knowledge about bats and their associated ectoparasites and shown the necessity of including molecular data in Laboulbeniales taxonomy.

Bats (Mammalia, Chiroptera) have received a great deal of attention due to their extraordinary morphological and ecological adaptations as well as their diversity in life history traits, qualities that make them ideal study organisms. Bats are parasitized by different groups of organisms, of which bat flies (Diptera, Hippoboscoidea, Nycteribiidae and Streblidae) are relatively well studied compared to other parasites. Published work has focused on host specificity, apparent male-domination and population structure of bat flies Dittmar, Porter, Murray, & Whiting, 2006;Olival et al., 2013) and on associations between functional traits of bats and parasitism by bat flies (Patterson, Dick, & Dittmar, 2007).
However, the addition of a second trophic level to the bat "microhabitat" is underexplored. Shockley and Murray (2006) reported two natural enemies of streblid bat flies (a hymenopteran parasitoid and a predaceous mirid bug). In addition, a handful of papers have discussed bacterial endosymbionts of bat flies in temperate and tropical regions (Duron et al., 2014;Hosokawa et al., 2012;Morse, Dick, Patterson, & Dittmar, 2012;Morse et al., 2013;Wilkinson et al., 2016).
In this study, we focus on the Laboulbeniales (Ascomycota, Laboulbeniomycetes), microscopic fungi that are obligate biotrophs on a wide range of arthropods, including bat flies. Prior to our current studies, the most recent papers dealing with Laboulbeniales on bat flies were published almost 40 years ago (Blackwell, 1980a(Blackwell, , 1980b.
Other papers on the same topic go back to the work of Harvard professor Roland Thaxter (1858Thaxter ( -1932. Some of his publications presented species descriptions and new records for Arthrorhynchus, a genus apparently restricted to Old World bat flies (Thaxter, 1896(Thaxter, , 1901(Thaxter, , 1908(Thaxter, , 1915(Thaxter, , 1931, and two genera that thus far have only been reported on neotropical bat flies, Gloeandromyces and Nycteromyces (Thaxter, 1917(Thaxter, , 1924(Thaxter, , 1931. Until we initiated our studies on bat flyassociated Laboulbeniales, five species were known from the type collections only (Haelewaters et al., 2017a(Haelewaters et al., , 2017bWalker et al., 2018). This illustrates how underexplored these hyperparasites are. Windsor (1990Windsor ( , 1995 made the claim "Equal Rights for Parasites!" arguing that whereas parasites are generally either ignored or seen as a threat to conservation of endangered organisms, they should be recognized as a legitimate part of the earth's biodiversity. This applies as well to hyperparasites. All organisms are almost sure to acquire a parasite during their lifetime, even parasites themselves. Laboulbeniales are one of three orders in the class Laboulbeniomycetes, the two others being Herpomycetales and Pyxidiophorales (Haelewaters et al., in review). All members of the class are obligately associated with arthropods for dispersal (Pyxidiophorales) or as biotrophs (Herpomycetales, Laboulbeniales).
What sets the Laboulbeniales apart is its diversity, with 2,200 described species and many more awaiting discovery, and its wide variety of arthropod hosts. Representatives of three subphyla serve as hosts to Laboulbeniales: Chelicerata, with harvestmen (Opiliones) and mites (Acari); Myriapoda, with millipedes (Diplopoda); and Hexapoda, with cockroaches and termites (Blattodea), beetles (Coleoptera), earwigs (Dermaptera), flies (Diptera), true bugs (Hemiptera), ants (Hymenoptera, Formicidae), crickets and allies (Orthoptera), lice (Psocodea), and thrips (Thysanoptera). As ectoparasites, Laboulbeniales are attached to the exoskeleton of the host where they form multicellular units of determinate growth, or thalli. They are developmentally unique among the fungi that usually have mycelia of unlimited growth. Laboulbeniales thalli are the result of subsequent divisions of a single two-celled ascospore. The ascospores are predominantly transmitted directly from infected to uninfected hosts (De Kesel, 1995).
Studying Laboulbeniales fungi has proven to be difficult for several reasons. The average size of Laboulbeniales thalli is around 200 μm, with extremes ranging from 35 μm (Rickia depauperata on mites of the genus Celaenopsis) to 4 mm (Laboulbenia kunkelii on Mormolyce phyllodes beetles). Because thalli are externally attached to a host, any study, morphological or molecular, requires micromanipulation with sterile techniques. Hosts may bear a large number of thalli, but often only few thalli are available for study. In some cases, thalli of a given species or morphotype may be restricted to a particular position on the host body (Goldmann & Weir, 2012;Goldmann, Weir, & Rossi, 2013). Unlike most fungi, Laboulbeniales have not been grown in culture to more than a few cells (never reaching maturity) (Whisler, 1968). The isolation of DNA has often been unsuccessful because of the often heavily pigmented cell walls (Weir & Blackwell, 2001b). This pigment, melanin, interferes during the PCR step by binding to the polymerase enzyme (Eckhart, Bach, Ban, & Tschachler, 2000). In addition, the cells are resilient in order to absorb impacts and friction on the host's integument. The combination of the melanized cell walls and resilient cells makes the thalli hard to break open.
Fungi of the order Laboulbeniales can display several types of specificity. Many species are host-specific; they are associated with a single host species or species in the same genus. Based on experimental work, De Kesel (1996) showed that this specificity is driven by characteristics of the integument and living conditions of the arthropod host, but also by the habitat chosen by that host. For a number of species, such as Euzodiomyces lathrobii, Hesperomyces virescens, Laboulbenia flagellata and Rhachomyces lasiophorus, many host species are known, often in more than one host family (Santamaria, Balazuc, & Tavares, 1991). Our work with H. virescens has demonstrated that it is impossible to make accurate species-level delimitations without molecular data (D. Haelewaters et al., unpublished data). It could be that more generalistic taxa are species complexes consisting of several species, whether cryptic or not, segregated by host. A different scenario is posed when hosts co-occur in a single micro-habitat. In this situation, opportunities exist for ascospores to be transmitted from a "typical" host to an "atypical" one. Such micro-habitats might be ant nests (Pfliegler, Báthori, Haelewaters, & Tartally, 2016), subterranean caves (Reboleira, Fresneda, & Salgado, 2017), or seaweed and plant debris on beaches (De Kesel & Haelewaters, 2014). Another type of specificity is displayed when a given fungus shows "a remarkable tendency to grow on very restricted portions of the host integument" (Benjamin & Shanor, 1952). This phenomenon is referred to as position specificity. For example, 13 species of Chitonomyces can be observed on restricted positions of the aquatic diving beetle Laccophilus maculosus. Based on the combination of molecular and ecological data, Goldmann and Weir (2012) confirmed that sexual transmission is the mechanism behind the observed position specificity patterns, as suggested by Benjamin and Shanor (1952 (2018), but only 24 Laboulbenia species are known from flies (Rossi & Kirk-Spriggs, 2011). Stigmatomyces is the secondlargest genus in the order, with 144 described species, all on flies (Rossi & Leonardi, 2013). The genera Arthrorhynchus, Gloeandromyces and Nycteromyces (Figure 1) are specific to bat flies, whereas none of the other genera have been recorded from bat flies.
Arthrorhynchus is restricted to Old World species of Nycteribiidae.
Kolenati (1857) (Blackwell, 1980b). Consequently, this taxon could be a complex of different species, each specialized to a single bat fly host or several hosts in a single genus-as is the situation in Hesperomyces virescens (D. Haelewaters et al., unpublished data).
The genera Gloeandromyces and Nycteromyces have hitherto only been found on streblid bat flies in the Americas (Haelewaters et al., 2017b;Thaxter, 1917Thaxter, , 1931Walker et al., 2018). The diversity of both genera is thus far limited, as is knowledge of their distribution and biology. After their original description (Thaxter, 1917), & Tortosa, 2017) and it has been shown that habitat disturbance affects parasitism of bats by bat flies (Pilosof, Dick, Korine, Patterson, & Krasnov, 2012 Figure 2). Bats were captured using three to four 6 m-wide 36 mm mesh ground-level mistnets with four shelves (Avinet, Portland, Maine, USA). Mistnets were set over trails that were presumably used by bats as flight pathways (Palmeirim & Etherdige, 1985). Nets were usually examined every 10-20 min between sunset and ~11 p.m. Bats were disentangled and processed immediately or kept in clean cotton bags until processing. Bats were released at the capture site immediately after processing. Bats were identified on site using dichotomous keys (Handley, 1981;Timm & LaVal, 1998). Bat taxonomy follows Simmons (2005). In this study, Artibeus intermedius was considered a junior synonym of A. lituratus (Barquez, Perez, Miller, & Diaz, 2015;Guerrero et al., 2008).

| Capture of bats and collection of bat flies
To remove bat flies from their bat hosts, 99% ethanol was applied using a paintbrush to reduce their activity. Subsequently, the bat flies were carefully removed using a rigid Swiss Style Forceps #5 with superfine tip (BioQuip #4535, Rancho Dominguez, California) or a Featherweight Forceps with narrow tip (BioQuip #4748). Some bat flies were collected using forceps alone or simply by hand.

| Sequence alignment and phylogenetic analyses
SSU and LSU rDNA datasets were constructed of newly generated sequences and sequences downloaded from GenBank, in order to assess (a) the position of bat fly-associated genera among Laboulbeniales from other hosts and (b) phylogenetic diversity in the genus Gloeandromyces. Alignments were done using Muscle v3.7 (Edgar, 2004) on the Cipres Science Gateway version 3.3 (Miller, Pfeiffer, & Schwartz, 2010) and manually edited in BioEdit v7.2.6 (Hall, 1999). The SSU and LSU aligned data matrices were concatenated in MEGA v7.0.21 (Kumar, Stecher, & Tamura, 2016). Maximum likelihood (ML) analysis of the SSU + LSU dataset was run using PAUP on XSEDE 4.0b (Swofford, 1991), which is available on Cipres. The appropriate nucleotide substitution model was selected by consider-

| Diversity in Gloeandromyces
To assess phylogenetic diversity within the genus Gloeandromyces, the LSU rDNA dataset was used. This region was put forward by previous studies to replace ITS as barcode for species delimitation in Laboulbeniomycetes (D. Haelewaters et al., unpublished data;Walker et al., 2018). Maximum likelihood (ML) analysis was run using the PAUP on XSEDE 4.0b tool (Swofford, 1991 (Gernhard, 2008;Yule, 1925) and the TPM2uf+G model of nucleotide substitution as selected by the Bayesian Information Criterion from jModelTest 2.1. The runs were performed from a random starting tree for 40 million generations, with sampling of parameters and trees every 4,000 generations.
The two resulting log files were combined in LogCombiner v1.8.4 with 10% burn-in. Consensus trees with 0% burn-in were generated and the MCC tree was constructed in TreeAnnotator v.1.8.4.

| Associations network
All presence/absence data of Laboulbeniales on bat flies and bat flies on bats were entered in a database. Data were partitioned to  (Dormann, Gruber, & Fründ, 2008).
Weighted data and the function plotweb were used to build a network showing host-dependencies and prevalence. Bats and bat flies that were not identified to genus level, bats without specimen label and infected bat flies with unidentified Laboulbeniales were excluded from the analysis. Bats and bat flies for which n < 10 were also excluded.

| Nucleotide alignment datasets
We generated 54 sequences of bat fly-associated Laboulbeniales during this study, of which 26 SSU and 28 LSU sequences. Our SSU + LSU concatenated dataset comprised 3,969 characters, of which 2,962 were constant and 789 were parsimony-informative. A total of 84 isolates were included (Table 1)
In the LSU dataset, Gloeandromyces forms six distinct clades ( Figure 4)

| Bats, bat flies and Laboulbeniales
Our complete dataset, prior to excluding specimens and par-  Notes. Also included are extraction protocols and numbers of thalli used per extraction for all isolates: 1% Triton 100-based protocol from Weir and Blackwell (2001a); 0.1 × TE buffer + dry ice protocol from Weir and Blackwell (2001b); heat extraction protocol, Extract-N-Amp Plant PCR Kit (ExNA) and QIAamp DNA Micro Kit (QIAamp Micro) from Haelewaters et al. (2015); REPLI-g Single Cell Kit (REPLI-g) from Haelewaters et al. (in review). GenBank accession numbers are provided (newly generated sequences in bold).

| Co-phylogenetic relationships between bat flies and Laboulbeniales
Our COI dataset of bat flies consisted of 15 taxa (one outgroup) and 677 characters, of which 410 were constant and 177 were parsimonyinformative. Our LSU dataset of Laboulbeniales consisted of 14 taxa (1 outgroup) and 998 characters, of which 610 were constant and 217 were parsimony-informative. The co-phylogeny plot is shown in Figure 7. There is congruence between the (basal-most) Old World clades, otherwise the evidence for coevolution is lacking.

| Prevalences
A comprehensive study of nycteribiid bat fly-associated Laboulbeniales was conducted by Blackwell (1980b). She screened 2,517 bat flies, of which 56 were infected with Arthrorhynchus eucampsipodae or A. nycteribiae, denoting a parasite prevalence of 2.2%.
In our larger study, we screened 7,949 bat flies of which 363 were infected by Laboulbeniales (4.6%). This includes both temperate and neotropical material. Taking only temperate flies into consideration (n = 2,001), parasite prevalence was again 4.6%. These low percentages can be explained by life history traits of the bat flies. Deposition of third instar larvae occurs on roosting substrates. Therein lies some risk, because flies need to return to their host within 25 hr. Since the flies are so closely tied to their bat host, we assume that transmission of ascospores of the fungi only occurs on the bat itself, most likely through direct contact (De Kesel, 1995). Host grooming is the main cause of death for bat flies (Marshall, 1981). Apparently, this TA B L E 2 Genera included in the concatenated SSU + LSU dataset, with classification up to ordinal level behavior is an important selective factor driving evolution of host specific and even position-specific parasites (ter Hofstede, Fenton, & Whitaker, 2004) and may to some extent be an explanatory factor in the observed patterns of Laboulbeniales.
Several studies confirm that bats are often infected by several bat fly species (Dick & Gettinger, 2005;Wenzel, 1976;Wenzel, Tipton, & Kiewlicz, 1966). At the same time, the average number of (nycteribiid) bat flies on their bat hosts is only 1.79 (Haelewaters et al.,

| Independent lineages of bat fly-associated Laboulbeniales
Parasitism of bat flies by Laboulbeniales arose at least three times independently, once in the Eastern Hemisphere and twice in the Western Hemisphere. The genus Gloeandromyces is placed sister to the speciose genus Stigmatomyces, species of which infect only flies.
The other two bat fly-associated genera form two separate clades, both sisters to a genus of Laboulbeniales that is associated with true bugs (Hemiptera). Arthrorhynchus and Prolixandromyces form a clade with moderate Bayesian support. The genus Prolixandromyces consists of eight species parasitizing taxa in the semi-aquatic family Veliidae (Weir, 2008). Nycteromyces forms a clade with Polyandromyces; the basal node of this clade received maximum support. Polyandromyces is a monotypic genus; its sole representative, P. coptosomalis, occurs on terrestrial species in the families Pentatomidae and Plataspidae. In other words, using the phylogenetic reconstruction of the SSU + LSU dataset, for the first time including molecular data from the rarely sampled bat fly-associated Laboulbeniales, we identified two interordinal host shifts (true bugs to bat flies). We hypothesize that two bat fly-associated lineages, Arthrorhynchus and Nycteromyces, have independently evolved from lineages of true bug ectoparasites. Tavares (1985) noted that bugs are secondary hosts to Laboulbeniales, and that their fungus parasites arose from taxa occurring on beetles (Coleoptera). We cannot F I G U R E 5 Host-parasite-parasite network of the final temperate dataset. Shown is the association of bat flies with their bat hosts (left) as well as the association of Laboulbeniales (right) and their bat fly hosts. Bar width represents the relative abundance of a species within each network level confirm this suggestion because our phylogenetic reconstruction is far from complete and does not encompass many taxa with beetle hosts. However, it is clear that Laboulbeniales on beetle hosts are evolutionary very successful; 80% of known species are reported from beetles (Weir & Hammond, 1997). In contrast, the numbers of known species from bugs is 4%, whereas the number from bat flies is less than 1%.
Is it possible bat fly-associated lineages have evolved from bugassociated lineages? Representatives of both host groups make use of the bat microhabitat and roost environment. Two families of terrestrial bugs are known as obligatory hematophagous ectoparasites: Cimicidae and Polyctenidae (Schuh & Stys, 1991). Both families belong to the superfamily Cimicoidea, along with Anthocoridae, Lasiochilidae, Lyctocoridae, and Plokiophilidae (Jung, Kim, Yamada, & Lee, 2010;Schuh & Stys, 1991). One lasiochilid, Lasiochilus pallidulus, has been found as a host to Cupulomyces lasiochili in Grenada, a member of the Stigmatomycetinae subtribe (Benjamin, 1992a). Benjamin (1992a) used the family name Anthocoridae for the host but he probably used this in the broad sense, whereas Schuh and Stys (1991) proposed to split up this non-monophyletic family into three, Anthocoridae sensu stricto, Lasiochilidae, and Lyctocoridae.
Lasiochilids live on the ground, under bark and in vegetation (Schuh & Slater, 1995). It is probable that transmission of ascospores occurs now and then between bugs and bat flies and that this at some point in time may have led to segregation of populations, microevolutionary changes and ultimately speciation.
Only C. lasiochili has been found on either cimicid or polyctenid bugs, but the limitation with Laboulbeniales reports is that the absence of reports on certain host groups is due to a lack of sampling and screening efforts. We recommend that future studies focus on screening bugs for Laboulbeniales parasites and on generating molecular data for taxa found on bugs. The cell II is positioned posterior and next to cell I, separated by an oblique septum, and cell II carries cells III obliquely and VI distally (Benjamin, 1981(Benjamin, , 1992a. In Cupulomyces, the perithecial wall cells are arranged in five tiers (Benjamin, 1992a). The situation has been described differently for Prolixandromyces, where in each vertical row of outer wall cells there are four tiers. However, Tavares (1985) mentioned that the fourth tier "may divide by maturity" even though the septa are extremely thin. Five tiers can be observed in drawings of mature thalli by Benjamin (1981: figure   13, reproduced here) and Weir (2008: figure 10). Consequently, also the perithecial outer wall structure is similar between both genera. Incorporating sequence data for Cupulomyces into our phylogenetic reconstruction will help elucidate whether contact between insects in the bat roost environment may have mediated host jumps to and subsequent speciation of Laboulbeniales on bat flies.

| Morphological diversity versus phylogenetic diversity
Based on morphological study, we identified seven species of Gloeandromyces. These are G. nycteribiidarum, G. pageanus, G. streblae and four undescribed, putative species (Figure 4a- Eberhard, 1989). In the case of G. streblae, this plasticity makes it hard to make morphologically based identifications. Some thalli are morphologically so similar to G. sp. nov. 4 that it is difficult to impossible to separate these taxa without sequence data. We have observed and included in our molecular work a range of G. streblae thalli, from short, stout, and curved to elongate, some with conspicuous bumps at the distal end of the perithecial venter. Even so, two clades were retrieved that are only segregated by host species. There is one exception: isolate D. Haelew. 1320b represents Gloeandromyces sp. nov. 2, which in reality is a morphotype.
This morphotype was removed from the last sternite/tergite. We

| CON CLUS IONS
This study has not only substantially increased our knowledge about bats and their ectoparasitic associates, but also shown the need to include molecular data in Laboulbeniales taxonomy.
Several phenomena come into play in the morphological and phylogenetic diversity of these parasites. Phenotypic plasticity and position-induced morphological adaptations go hand in hand. Position-induced morphotypes belong to the same phylogenetic species. In Chitonomyces, transmission of ascospores during mating between hosts seems to be the mechanism leading to position specific morphotypes (Goldmann & Weir, 2012).
For bat fly-associated Laboulbeniales, it is unclear what is driving morphological divergence within phylogenetic species. Another important contributor to diversity, whether or not ephemeral or incipient (Rosenblum et al., 2012), is host specialization.
Segregation by host species is observed for at least two bat flyassociated species. Concerning studies in diversity and taxonomy of Laboulbeniales, our main recommendation is to always include molecular data. The examples discussed in this study have made it clear that it has become impossible to assess diversity by morphology alone.

ACK N OWLED G M ENTS
All capturing and sampling procedures were licensed and ap-

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflicts of interest with regard to this article.

DATA ACCE SS I B I LIT Y
The complete dataset of bats, batflies and Laboulbeniales is available as Supporting Information (in Excel format