Phylogeny- and morphology-based recognition of new species in the spider-parasitic genus Gibellula (Hypocreales, Cordycipitaceae) from Thailand

Abstract Thailand is known to be a part of what is called the Indo-Burma biodiversity hotspot, hosting a vast array of organisms across its diverse ecosystems. This is reflected by the increasing number of new species described over time, especially fungi. However, a very few fungal species from the specialized spider-parasitic genus Gibellula have ever been reported from this region. A survey of invertebrate-pathogenic fungi in Thailand over several decades has led to the discovery of a number of fungal specimens with affinities to this genus. Integration of morphological traits into multi-locus phylogenetic analysis uncovered four new species: G. cebrennini, G. fusiformispora, G. pigmentosinum, and G. scorpioides. All these appear to be exclusively linked with torrubiella-like sexual morphs with the presence of granulomanus-like asexual morph in G. pigmentosinum and G. cebrennini. A remarkably high host specificity of these new species towards their spider hosts was revealed, and for the first time, evidence is presented for manipulation of host behavior in G. scorpioides.


Introduction
To arthropodologists or even arachnologists, it is surprising that fungal pathogens of spiders seem to be generally neglected when the host can be completely overgrown by the pathogens to be unrecognizable as a spider. Nonetheless, this group of fungi has been known and studied for more than two centuries (Evans 2013). Recently, over 80 fungal species, mostly distributed in Cordycipitaceae, have been reported as spider pathogens (Shrestha et al. 2019). Among them, only Gibellula and Hevansia are obligate parasites of spiders whereas others appear to be natural enemies of different insects and do not show apparent host specificity.
Gibellula is well-known to be a specialized spider-parasitic genus widely distributed worldwide, mostly found in tropical regions (Shrestha et al. 2019). Originally, the type species, G. pulchra (Sacc.) Cavara was known as Corethropsis pulchra Sacc. collected from Italy, recognized by producing primarily synnematous, aspergillus-like conidiophores with terminal vesicles, each gives rise to phialides produced on metulae (Saccardo 1877;Shrestha et al. 2019). After establishing a genus Gibellula Cavara in honor of Prof. Giuseppe Gibelli by Cavara (1894), a number of species in this genus were recorded across the world (Shrestha et al. 2019). Currently, nearly 40 species have been described and listed in the global fungal databases Index Fungorum (www. indexfungorum.org) and MycoBank (www.mycobank.org). According to the review of Shrestha et al. (2019), many of them including G. arachnophila (Ditmar) Vuill., G. arachnophila f. arachnophila (Ditmar) Vuill., G. arachnophila f. macropus Vuill., G. aranearum P. Syd., G. globosa Kobayasi & Shimizu, G. globosostipitata Kobayasi & Shimizu, G. haygarthii Van der Byl, G. suffulta Speare and G. tropicalis Sawada were synonymized with G. pulchra whereas G. arachnophila f. leiopus Vuill. ex. Maubl., G. araneae Sawada and G. perexigua (Kobayasi) Koval were synonymized with G. leiopus (Vuill. ex Maubl.) Mains. In addition to these species, the identities of several other species reported in this genus still remain doubtful. Petch (1932) expressed uncertainty about the identities of G. aspergilliformis (Rostr.) Vuill. and G. phialobosia Penz. & Sacc. by pointing out that the narrow metulae and spherical conidia in chains present in G. aspergilliformis and the flask-shaped phialides in the latter species were not common features of Gibellula. Moreover, description of G. eximia Höhn. did not point to the genus. Since Gibellula is well-known as an obligate parasite of spiders, Mains (1950) reported that the assignment of G. elegans Henn. to this genus might be erroneous, as this species is found occurring on locusts. According to Mains (1950), the description of G. capillaris Morgan did not fit the concept of Gibellula and re-examination of the type specimen is unfortunately infeasible since it is no longer in a good condition. Tzean et al. (1997) doubted the identity of G. araneicola Sawada that produces an isarioid morph instead of Gibellula. In the case of G. petchii Humber & Rombach, it is still unclear whether the species name should be retained or abandoned. Gibellula petchii Humber & Rombach was proposed to accommodate Cylindrophora aranearum Petch, which was originally described as the conidial state of Torrubiella albolanata Petch and later elevated to generic rank as a new genus, Granulomanus de Hoog & Samson (de Hoog 1978;Humber and Rombach 1987;Petch 1944). From the point of view of Humber and Rombach (1987), Granulomanus should be synonymized with Gibellula as it almost never occurs in the absence of Gibellula and/or its torrubiellalike sexual morph. Cylindrophora aranearum (≡Granulomanus aranearum (Petch) de Hoog & Samson) was henceforth synonymized with G. petchii. On the other hand, Samson and Evans (1992) argued that Granulomanus naturally occurs independently on spider hosts either with or without Gibellula. Thus, the genus should be retained as an independent asexually typified genus resulting in rejection of G. petchii. According to a recent taxonomic revision of the Cordycipitaceae, which was largely based on molecular data, several generic names including Granulomanus were suppressed (Kepler et al. 2017). Nevertheless, the taxonomic dilemma of G. petchii cannot yet be resolved owing to the lack of its sequence data. Based on these facts, only 17 species have been accepted in Gibellula (Shrestha et al. 2019).
Thailand is one of the most biodiverse countries in Southeast Asia and the BIO-TEC culture collection has more than 700 Gibellula strains. Despite this number, only very few Gibellula species with distinct features could be recognized morphologically (Luangsa-ard et al. 2008(Luangsa-ard et al. , 2010. Gibellula gamsii is the most recently described species reported from Thailand . Our continuous survey of invertebrate-pathogenic fungi in Thailand for over two decades has led to the BIOTEC Bangkok Herbarium (BBH) and the BIOTEC Culture Collection (BCC) owning a very large herbaria, and culture collections, which greatly facilitates the exploration of existing species including Gibellula. Here, phylogeny within Gibellula species from the ribosomal internal transcribed spacer (ITS) regions analyzed prior to this study enabled recognition of four distinct clades. The morphological and multi-gene phylogenetic data confirm their identities as well as taxonomic placements. Herein, new species are described that are illustrated morphologically and phylogenetically and compared with other species in the same genus.

Collection of fungal materials and isolation of pure cultures
Spiders parasitized by Gibellula spp. firmly attached on the underside of living leaves were collected from various locations throughout Thailand, mostly in the Northeastern region. The leaf bearing the parasitized spider was carefully detached from the tree, placed in a plastic box and transported to the laboratory for immediate isolation of a pure culture according to the protocols described by Kuephadungphan et al. (2014) and Mongkolsamrit et al. (2018). Briefly, the conidia located on the synnemata were gently swiped with small agar plugs of potato dextrose agar (PDA) which were then placed on a PDA plate. The conidia were allowed to germinate at 25 °C for a few days. Thereafter, each agar plug with actively growing mycelia was transferred to a fresh PDA plate where the fungus could readily grow for another 6-8 weeks. For specimens bear-ing sexual morphs, pure cultures were isolated by enabling ascospores from the perithecia to be discharged onto PDA plates and allowing them to grow at 25 °C for a certain amount of time depending on the growth rate of each individual strain. All cultures were required to be deposited in the BCC, Thailand while the fungal specimens were dried in an electric food dryer (50-55 °C) before being stored at the BBH, Thailand.

Morphological characterization
Microscopic characteristics were studied based on observation of synnemata and perithecia. Each of them was detached from the stroma and mounted on a microscope slide containing a drop of lactophenol cotton blue solution. Shapes and sizes of individual character were determined and measured according to Mongkolsamrit et al. (2018).

Identification of spider hosts
The mummified spiders were identified based on morphological characteristics. To better understand the host-pathogen relationship, posture of spider at attachment on leaf surface (touching or lifting), position of spider on the leaf (under or upper side), and leaf type (monocots or dicots) were herein recorded.
DNA sequences were checked manually for ambiguous base calls and all sequences were assembled using BioEdit v.7.2.5 (Hall 1999;Hall et al. 2011). Sequence alignment was conducted using MAFFT 7.017 with G-INS as the algorithm and default settings used for gap opening and gap extension penalties (Katoh and Toh 2008). Manual adjustments were subsequently made in BioEdit. Concatenation of multiple loci was performed in GENEIOUS® 7.1.19 (http://www.geneious.com, Kearse et al. 2012).
Phylogenetic relationships were inferred using maximum likelihood (ML) with GTRCAT as the substitution model in RAXML 7.2.8 (Stamatakis 2006) and the rapid bootstrap analysis algorithm. Relative support for the branches was obtained from bootstrap analysis with 1,000 replicates. Bayesian analysis was performed according to Mongkolsamrit et al. (2019) using MRBAYES v.3.2.7 (Ronquist andHuelsenbeck 2003) on XSEDE via the online CIPRES Science gateway using SYM+G selected by MRMODELTEST 2.2 (Nylander 2004) as the best nucleotide substitution model. Posterior probabilities were performed by Markov Chain Monte Carlo Sampling (MCMC) in which four chains were run for 5,000,000 generations with a tree sampling frequency of 100 and a burn-in of 10% of the total run.
Culture characteristics. Colonies derived from conidia, on PDA slow-growing, attaining a diam of 1.1±0.03 cm in 20 days at 25 °C, white, velvety; reverse cream, becoming light brown with age toward center (Fig. 3k). Sporulation not observed.
Culture characteristics. Colonies derived from ascospores, on PDA slow-growing, attaining a diam of 1.5±0.2 cm in 4 weeks at 25 °C, white, floccose; reverse light brown, darkening with age toward center (Fig. 4j). Sporulation not observed. Notes. Gibellula pigmentosinum shares similarity with G. pulchra (Mains 1950) in producing cylindric, yellowish white synnemata bearing aspergillate conidiophores with fusoid-ellipsoid conidia and superficial, reddish brown, ovoid perithecia containing bacilliform part-ascospores. The synnemata in G. pulchra are more copious and sometimes more violaceous than in G. pigmentosinum. Remarkably, G. pigmentosinum distinctly differs from G. pulchra in having a granulomanus-like conidial state. Etymology. Refers to the outer appearance of the fungus resembling the posture of a scorpion.
Additional specimens  Notes. The morphology of G. scorpioides appeared to be very close to G. clavulifera var. clavulifera (Samson and Evans 1977), G. clavulifera var. major (Tzean et al. 1997) and G. clavulifera var. alba (Humber and Rombach 1987). The penicillate conidiophores were largely absent from the whip-like stroma in G. clavulifera var. alba but distinctly present on a synnema of G. scorpioides. Based on a comparison of microscopic characteristics among G. scorpioides, varieties clavulifera, major and alba, the latter three were found to produce much longer conidiophores (up to 100 μm) than G. scorpioides (20-29(-30) × 4 μm) while the other characters such as metulae, phialides as well as conidia were considered to be not significantly different in both shape and size. Considering the presence of the torrubiella-like sexual morph, perithecia of G. clavulifera var. alba were produced sparingly and separately on the host abdomen while those of G. scorpioides distinctly appeared in groups, only on the spider's legs and basally on synnema. Nevertheless, an examination of additional specimens has led us to conclude that the sexual morph is not always present in G. scorpioides.

Discussion
A torrubiella-like sexual morph is well-known to be connected with Gibellula (Evans 2013;Kepler et al. 2017;Shrestha et al. 2019). Shrestha and colleagues (2019) recently reviewed spider-pathogenic fungi within Hypocreales including Gibellula where its sexual morph links are listed. Torrubiella globosa Kobayasi & Shimizu, Torrubiella globosostipitata Kobayasi & Shimizu, Torrubiella arachnophila var. pulchra Mains and Torrubiella gibellulae Petch were synonymized with G. pulchra, species where their conidial and torrubiella-like sexual morphs often concurrently occur on the same substrates. Gibellula pigmentosinum appeared to be remarkably close to G. pulchra in producing nearly identical microscopic characters in both shapes and sizes. Nonetheless, G. pigmentosinum distinctly differs from G. pulchra in having a granulomanus-like conidial state. Considering its phylogenetic placement, G. pigmentosinum was significantly placed far from the taxon representing G. pulchra which supported the morphological differences between them. Noticeably, G. pigmentosinum formed a very strongly supported clade together with Gibellula cf. alba. It is interesting that Gibellula cf. alba was not proposed as a species and it is unfortunate that the herbarium specimen of NHJ 11679 is no longer in a good condition. According to its placement in the RAxML/Bayesian tree inferred from multiple loci (Fig. 1), Gibellula cf. alba NHJ 11679 could unambiguously be assigned to G. pigmentosinum.
The morphological resemblance between G. cebrennini and G. fusiformispora as well as a multi-gene phylogenetic analysis indicate a very close relationship among these species. Moreover, they can be distinguished from each other by the length of conidiophores, the shape of ascospores as well as the presence of a granulomanus-like conidial state.
In nature, a torrubiella-like sexual morph may occur on spider hosts without the presence of Gibellula. It may be premature to assign the new species to Gibellula on the basis of sexual morph, when more than one asexually reproductive genus are known to be linked to a torrubiella-like sexual morph. Gibellula cebrennini and Akanthomyces thailandicus (Mongkolsamrit et al. 2018) are good examples of such a phenomenon. Based on an investigation of Thai specimens, G. cebrennini tended to produce torrubiella-like perithecia on the spider hosts in the absence of Gibellula and granulomanus-like asexual morphs, whereas A. thailandicus is an obligate parasite of spiders, of which only its torrubiella-like sexual morph has so far been recorded. These characteristics could lead to misidentification between these two species. The size and shape of partspores are considered as the only morphological characters that have potential in species discrimination according to the evidence that G. cebrennini mostly produces longer bacilliform part-spores than A. thailandicus. In G. cebrennini and G. fusiformispora, it may also be difficult to discriminate them at first glance as the only distinguishing feature is the shape of their part-spores. These similiarities, and the occasional overlap in shape and size of morphological characters, were also demonstrated by  in the Ophiocordyceps irangiensis and O. myrmecophila species complex, by Mongkolsamrit et al. (2019) in Paraisaria phuwiangensis and P. yodhathaii, by Luangsaard et al. (2018) in Ophiocordyceps spp. with superficial perithecia, and  among termite pathogens in Ophiocordyceps. To improve species delimitation among closely related species with such very low morphological differentiation, integrative taxonomy combining a variety of data such as molecular, chemical, biogeographical, ecological characters, etc. is suggested to be very useful (Pante et al. 2015).
It has been over a half century since host specialization was suggested as one of the taxonomic criteria for parasitic fungi (Johnson 1968). In most cases of fungi that have a narrow host range or are restricted to a single host species, host specificity is considered as an important feature that can be used for identification at the species level (Vialle et al. 2013). In the case of invertebrate-pathogenic fungi, host specificity is usually taken into account mostly to evaluate their virulence and potential in terms of using them as biocontrol agents. For taxonomic purposes, Kobmoo and co-workers (2012) proved that host specificity has great potential for reflecting the divergent evolution of the ant-parasitic Ophiocordyceps unilateralis. Their success has therefore driven us to put effort for the first time to define the host species of Gibellula. Bishop (1990), Hughes et al. (2016), and Savić et al. (2016) reported spider hosts of Gibellula distributed among 11 families consisting of Anyphaenidae, Agelinidae, Araneidae, Corinnidae, Linyphiidae, Pholcidae, Salticidae, Sparassidae, Theridiidae, Thomisidae, Zodariidae which represent approximately 10% of described families worldwide (World Spider Catalog 2020). Herein, three of those including Salticidae, Thomisidae and Zodariidae were found to be the hosts of G. scorpioides, G. cebrennini and G. pigmentosinum, respectively, whereas the family Deinopidae is reported here for the first time as a Gibellula host for G. fusiformispora.
According to the effort of putting toward species identification of spider hosts (Bris-towe1930;Tikader 1965;Jocqué and Bosmans 1989;Deeleman-Reinhold 2001, 2009Jocqué and Dippenaar-Schoeman 2007;Lehtinen et al. 2008;Benjamin 2011Benjamin , 2016Miller et al. 2012;Dhali et al. 2017;Jocqué et al. 2019), G. pigmentosinum, G. cebrennini as well as G. scorpioides were found to be exclusively specific to spider species with the exception of G. fusiformispora wherein only one specimen allowed us to identify the spider only to the family rank. Despite Gibellula being a well-known spider specialist, only Nentwig (1985), Hughes et al. (2016), and Savić et al. (2016) have ever indicated its hosts at genus or species ranks. In the current study, only the host of G. cebrennini could be identified to the species level as Cebrenninus cf. magnus, whereas hosts of G. pigmentosinum and G. scorpioides were assigned to Storenomorpha sp. (Zodariidae) and Portia sp. (Salticidae), respectively. One common problem is that the fungus tends to cover the host body completely, which can obscure the spider's morphological features making identification infeasible. Tarsal claws and scopulae are important morphological features used to identify spider species (Wolff and Gorb 2012;Wolff et al. 2013;Labarque et al. 2017), and the morphology of spider feet was herein targeted and carefully studied. Since their legs appeared to be the only part that were slightly covered by fungal mycelia, it was thus considered to be the most significant character for distinguishing spider species infected by Gibellula. Host identification can be especially challenging for old fungal herbarium specimens that are dry and damaged. We suggest delivering specimens to arachnological taxonomists immediately after field work to allow the identification of spider hosts to species rank. Furthermore, as new species of spider continue to be described (e.g. Miller et al. 2012;Dhali et al. 2017;Jocqué et al. 2019), accurate taxonomy of spider hosts could be important for taxonomy of fungal pathogens on them.
It is notable that all seven host individuals used for identification in this study were found under a leaf. Portia sp. has a unique life history. These spiders are not only web invaders or cursorial hunters but are also web builders. Thus, they exist both on and off their own webs. Their webs are used for various activities including trapping, baiting, resting, molting, mating, oviposition, brooding (Jackson and Blest 1982;Jackson and Hallas 1986;Nelson and Jackson 2011a). They are day-active hunters and stay on their own webs at night (Barth 2002;Herland and Jackson 2004;Nelson and Jackson 2011b). Moreover, moribund web-building spiders infected by pathogenic fungi are presumed to stay motionless on their webs (Anotaux et al. 2016). This behaviour may promote growth and reproduction of G. scorpioides since spider silks possibly have antifungal properties (Tahir et al. 2015;Phartale et al. 2019). Another distinct feature of the spiders infected by G. scorpioides is its firm attachment to the substrate by only the mycelia growing over the tips of their legs allowing the host to sprawl and elevate their bodies upward. Interestingly, the host of G. scorpioides, Portia sp. is an araneophagic jumping spider that usually assumes such a posture during hunting (Forster and Murphy 1986). Furthermore, spiders are generally found dead with a posture of their legs flexed beneath their body (Pollard et al. 1987;Pizzi 2006;Starr and Taggart 2006). Such dead posture may also support growth and reproduction of G. scorpioides by keeping host from predation (Deeleman-Reinhold 2001; Honma et al. 2006), at least at the beginning stage of a fungal growth. From these observations, we believe that the fungus may influence the behaviour of the spider host by forcing it to stay firmly in place and assume an active posture during infection.
Since G. scorpioides can be cultured, it is possible to apply fungal spores to Portia spiders and study the spider-fungus interaction. It will be particularly interesting to investigate death sites, on or off web and death posture, resting or hunting postures between uninfected control and fungal infected spiders, which may give insight into behavioural manipulation before death by the fungus. Additionally, cultured G. scorpioides could be used to test the antifungal properties of Portia spider silk.
It is remarkable that our finding has revealed the high possibility to incorporate the host specificity in molecular and morphological criteria for classification and identification of Gibellula.
The biggest challenge for molecular phylogeny-based classification of Gibellula is the lack of reliable sequence data from type specimens. From sequences available in public databases, identities often appear erroneous, e.g. G. clavispora , G. curvispora (Han et al. 2013) and G. shennongjiaensis  appeared to be closer to other ascomycetes than Gibellula based on ITS sequence data. In the past, no attempts were made to establish the described species as pure cultures, or the attempts failed, thus making molecular analysis impractical. The lack of sequence data from type strains of Gibellula makes it difficult to establish whether query sequences from new specimens represent new or rediscovered taxa.
Despite molecular phylogeny currently being the most powerful approach available in modern fungal classification and taxonomy (Ariyawansa et al. 2014), many attempts to incorporate alternative or polyphasic approaches, such as chemotaxonomy, have been made. This approach provides high-informative data to support morphological and molecular data for identifying fungal species, facilitates solving taxonomic problem as well as unraveling asexual morph-sexual morph links (Frisvad et al. 1990;Stadler et al. 2003;Helaly et al. 2018). As part of our ongoing research on taxonomy and secondary metabolite production of Thai invertebrate-pathogenic fungi, chemotaxonomy has been employed, resulting in the discovery of unprecedented secondary metabolites, including pigmentosin B from G. pigmentosinum and gibellamines from G. gamsii (Helaly et al. 2019;Kuephadungphan et al. 2019). These compounds were found to be species-specific and could be designated as chemotaxonomic markers for the species. However, it is premature to use such compounds as markers for Gibellula since the exploration of their secondary metabolite production is limited to only a few species. Chemotaxonomy must therefore be expanded to other taxa, in particular G. cebrennini, G. fusiformispora as well as G. scorpioides.