Stars in subtropical Japan: a new gregarious Meteorus species (Hymenoptera, Braconidae, Euphorinae) constructs enigmatic star-shaped pendulous communal cocoons

A new gregarious braconid parasitoid wasp of Euphorinae, Meteorus stellatus Fujie, Shimizu & Maeto sp. nov., is described from the Ryukyu Islands in Japan, based on an integrative taxonomic framework. The phylogenetic position of the new species within the Meteorini was analyzed based on DNA fragments of the mitochondrial cytochrome c oxidase 1 (CO1) and the nuclear 28S rDNA genes. The new species was recovered as a member of the versicolor complex of the versicolor + rubens subclade within the pulchricornis clade. The new species is a gregarious parasitoid of two Macroglossum species (Lepidoptera: Sphingidae) and constructs single or several unique star-shaped cocoon masses separately suspended by very long threads. The evolution of gregariousness and spherical cocoon masses is discussed.


Introduction
The pupae of parasitoid wasps cannot actively escape various risks, such as predation, parasitism, pathogenesis, and environmental stresses. Therefore, cocoons and mummies play important roles in protecting soft and exarate pupae from such risks (Gauld and Bolton 1988;Shaw and Huddleston 1991;Maeto 2018).
Members of Braconidae, one of the most diverse hymenopteran families, form various types of cocoons and mummies to adapt to various natural enemies and environmental threats. Many gregarious braconids produce communal cocoon masses, while discrete cocoons for each individual are also constructed (Shaw and Huddleston 1991). Final-instar braconid larvae emerge from host organisms and then spin a cocoon using silk from the labial glands (e.g., Shaw and Huddleston 1991;Quicke 2015). In some gregarious species, wasp larvae not only spin individual cocoons but also cooperate to construct massive communal cocoons (Zitani and Shaw 2002).
The cosmopolitan braconid genus Meteorus Haliday consists of more than 300 valid species (Yu et al. 2016). As the genus Zele Curtis is apparently nested within the Meteorus species tree, Meteorus is a paraphyletic group and its rearrangement into several monophyletic genera is pending (Maeto 1990b;Stigenberg and Ronquist 2011). Maeto (1990b) has divided the Meteorus species into seven morphologically and biologically defined species groups, while monophyly has been supported partially (Stigenberg and Ronquist 2011).
Meteorus species are solitary or gregarious koinobiont endoparasitoids of Lepidoptera or Coleoptera larvae (Huddleston 1980;Stigenberg and Ronquist 2011;Maeto 2018). Their cocoons are either suspended (pendulous) or not, and the cocoon suspension is one of diagnostic characters to identify the pulchricornis group (Maeto 1989a). The gregarious species construct either independent (e.g., M. acerbiavorus Belokobylskij, Stigenberg & Vikberg, M. heliophilus Fischer, and M. rubens Nees) or communal cocoons (e.g., M. congregatus Muesebeck, M. komensis Wilkinson, M. kurokoi Maeto, and M. townsendi Muesebeck) (Maeto 1989b;Zitani and Shaw 2002;Zitani 2003;. The latest molecular phylogeny has suggested that the gregariousness is a derived character in the pulchricornis + rubens group complex (Stibenberg and Ronquist 2011). So far, six types of cocoon architectures have been observed in the gregarious species of Meteorus, as is shown in Table 1. However, the evolution of gregariousness and cocoon architectures within the genus Meteorus has been poorly studied, and more bionomical and phylogenetic information is thus needed.
Although the cocoon structure of Meteorus is quite mysterious, larval behavior associated with cocoon formation has received little attention, with only a few reports examining it (Askari et al. 1977;Zitani and Shaw 2002;Maeto 2018). Recently, unique star-shaped cocoon masses of an undescribed gregarious species of Meteorus have been observed in subtropical Japan (Mitamura 2013). Therefore, this study aims to describe that gregarious species of Meteorus based on integrative morphological and molecular evidence and observe its cocoon-mass formation behavior. The phylogenetic position of the new species and evolution of gregariousness and cocoon masses are also discussed.

Study fields
The field collection of host moth larvae to observe the cocoon formation behavior of emerged larvae of wasps was conducted at Okinawa Municipal Museum, Okinawa City, Okinawa-hontô, Okinawa Prefecture, Japan. Some materials were also collected within Okinawa-hontô (Okinawa Prefecture) and Amami-ôshima (Kagoshima Prefecture), Japan. All materials were from the middle part of the Ryukyu Islands, the subtropical Oriental region in Japan.

Morphological observation and terms
Morphological observation was conducted with a stereoscopic microscope (SMZ800N, Nikon, Tokyo, Japan). Specimens and cocoons were photographed using a Digital Microscope (VHX-1000, Keyence, Osaka, Japan) with a 10-130× lens. Multi-focus photographs were stacked in the software associated with the Keyence System. Multifocus photographs of cocoon masses were taken using a single lens reflex camera (α7II, Sony, Tokyo, Japan) with a micro-lens (A FE 50 mm F2.8 Macro SEL50M28, Sony). The RAW format photographs were developed using Adobe Lightroom CC v.2.2.1 (Adobe Systems Inc., San Jose, CA, USA), and stacked using Zerene Stacker v.1.04 (Zerene Systems LLC., Richland, WA, USA). The holotype of M. komensis, deposited in the Natural History Museum, London, UK was also examined by the second author  Stigenberg and Ronquist (2011). The morphological terms and measurements follow those of Richards (1977) and van Achterberg (1988). The following abbreviations are used: OOL = ocelli-ocular line, OD = ocelli diameter of a posterior ocellus, and POL = posterior ocellar line.
The abbreviations for repositories are listed below:

Observation of cocoon formation behavior
The cocoon formation behavior of wasp larvae was observed at a laboratory of OMM, in June 2019 by the third author. It was recorded with video cameras (Sony Handycam, HDR-CX470 and HDR-XR150, Sony). The suspended larvae were blown with air currents created by breathing, as there was no wind, which would enhance the merging of each individually suspended larva in the laboratory as it would in natural conditions. A single silk thread spun by an individual larva is called a "thread", and intertwined threads are called a "cable" as in Barrantes et al. (2011). A short movie showing the cocoon formation behavior is available on YouTube (https://www.youtube.com/watch?v=AuHarLHolPM).

Gene selection
To delimit a species, fragments of a mitochondrial protein encoding gene, cytochrome c oxidase 1 (CO1), were selected, because its evolutionary rate is more or less rapid and it is one of the most common genes used for population to species level phylogenetic analysis (this is well-known as the DNA barcoding gene). To infer the phylogenetic relationships among species of Meteorini (Meteorus and Zele), CO1 and a nuclear noncoding gene, 28S rRNA (28S), were selected. 28S is a gene that has evolved more slowly than CO1 and is usually used for species-groups or higher-level phylogeny; therefore, the combined CO1 and 28S analysis can provide a higher resolution of species phylogeny.

Taxon sampling and outgroups
A total of 44 species of Meteorus including five morphospecies were sampled as ingroups. Five species of Zele were also sampled as ingroups because Zele is deeply nested within the Meteorus tree (Stigenberg and Ronquist 2011). Three species from different tribes of euphorine genera were sampled as outgroups (Syrrhizus Förster, Syntretus Förster, and Peristenus Förster). A total of 177 sequences of CO1 and 172 of 28S were compiled from GenBank. Sequences obtained from databases sometimes contain unreliable information (e.g., Klimov et al. 2019;Shimizu et al. 2020), and an evaluation of such sequences is always strongly recommended to ensure that the analysis is accurate. In the present study, the sequences used by Stigenberg and Ronquist (2011) and several additional sequences were considered as reliable sources. The complete information of the sampled taxa and sequences is available in the Suppl. material 2: Table S2.

Species delimitation
Partial fragments of CO1 were used for species delineation. A total of 189 sequences were used for analysis (Suppl. material 2: Table S2). To obtain an accurate multiple sequence alignment (MSA), MSA was conducted using MEGA v.10.0.5 (Kumar et al. 2018) based on amino acids. First, the codon positions of all nucleotide sequences were adjusted, the nucleotide sequences were translated to amino acids, the amino acid sequences were aligned by CLUSTAL W (Thompson et al. 1994) implemented in MEGA with default settings, the amino acid alignment was checked by eye, and finally the aligned amino acid sequences were retranslated to nucleotides. The final dataset was 657 bp without indels.

Topology-based method (GMYC)
The General Mixed Yule Coalescent (GMYC) analysis was employed. GMYC analysis requires an ultrametric tree (UTree) as an input. To construct the UTree, the model and parameters was selected on a web server of the smart model selection (SMS) (Lefort et al.  Schulmeister (2003) 2017) (available at: http://www.atgc-montpellier.fr/phyml-sms/): the GTR+G+I model was selected as the best fit model under the Bayesian information criterion (BIC). The UTree was generated using BEAST v.2.6.3 (Bouckaert et al. 2019), with a random starting tree, the uncorrelated lognormal relaxed clock model, and the coalescent tree prior. A Bayesian Markov chain Monte Carlo (MCMC) was run for 40,000,000 generations, with trees sampled every 5,000 generations, and a burnin of anterior 25%. The convergence of run was assessed using Tracer v.1.6 (Rambaut and Drummond 2007): run reached a stationary distribution and all effective sample sizes (ESS) were greater than 200. A majority-rule consensus ultrametric tree was finally generated using TreeAnnotator v.2.6.3 (Bouckaert et al. 2019). GMYC analysis was run using the GMYC function of the R-package Splits (Fujisawa and Barraclough 2013) (available at: http://r-forge.rproject.org/projects/splits/) using R v.3.6.3 (R Core Development Team 2020).

Phylogenetic analysis
The phylogeny of Meteorini species was inferred with both the Bayesian Inference (BI) and maximum likelihood (ML) approaches using a concatenated CO1 and 28S fragments.

MSA
Although the MSA of CO1 was already performed in the species delimitation, MSA for 28S was conducted in the MAFFT online service (Katoh et al. 2019), using the Q-INS-i algorithm, which is the structural alignment method for RNA (Katoh and Toh 2008;Katoh and Standley 2013). Ambiguously aligned regions were automatically removed from the dataset using trimAl v.1.2 (Capella-Gutíerrez et al. 2009), with default parameters. The final datasets were 657 (CO1) and 560 (28S) bp in length: the concatenated CO1 and 28S dataset was 1,217 bp.

Terminal species selection
In order to exclude the taxon sampling bias, a single sequence for each species was selected based on the conservative results of the species delimitation analysis by ABGD: sequences of 61 Meteorus species, six Zele species, and three outgroup species were finally selected (Table 3).

Model selection
Each codon position within the CO1 fragment was treated as a different data block, but not for noncoding 28S. The best-fit substitution model was determined using Par-titionFinder v.2.1.1 (Lanfear et al. 2017) with the greedy search algorithm under the corrected Akaike information criterion (AICc): the selected model was the GTR+I+Γ model for the first and second codon position of CO1 and 28S, and GTR+Γ for the third codon position of CO1.

Analysis
The BI analyses were conducted using MrBayes v.3.2.2 (Ronquist et al. 2012). A Bayesian MCMC analysis was ran with the following settings: four independent runs, 20 chains each, heating 0.05, random starting trees, and trees sampled every 1,000 th generation for 10,000,000 generations. The convergence of the MCMC runs was checked by the average standard deviation of split frequencies (ASDSF) in MrBayes (i.e., AS-DSF < 0.01) (Ronquist and Huelsenbeck 2003) and chain stationarity in Tracer v.1.6 (Rambaut and Drummond 2007). Then, we discarded the anterior 25% of the generations as burn-in, obtained estimates for the harmonic means of the likelihood scores from the remaining 75% of the generations using the sump command, and conducted a final check of the convergence of the runs by the value of a potential scale reduction factor (PSRF); if the runs were convergent enough, PSRF was less than 5% divergent from 1.0. Finally, a consensus tree with the Bayesian inference posterior probabilities was obtained using the sumt command in MrBayes.

Species identification
Although the molecular species delimitation was conducted using (1) whole datasets (i.e., Meteorus plus Zele plus outgroups), (2) Meteorus plus Zele datasets, and (3) Meteorus datasets, the results were congruent among all datasets in both the ABGD and GMYC methods. The number of recognized species was higher in the GMYC than in the ABGD method (Fig. 1). However, all results indicated that Meteorus stellatus sp. nov. was a single species. Baded on morphological data, M. stellatus sp. nov. ran to the versicolor subgroup of the pulchricornis group based on Maeto's (1989a, b) criteria, but it could not be identified as any described species of Meteorus (see also Differential diagnosis). Etymology. The specific name is a masculine Latin word, "stellatus", meaning "starry", which is derived from the unique shape of the cocoon masses.
Description. Female (holotype; Fig. 2). Body length 3.6 mm. Head (Fig. 2B, C, F, H, I). Width of head 1.7× median height. Length of eye 1.7× length of temple in dorsal view. Temple roundly narrowed posteriorly. Eyes large and moderately convergent ventrally. Face with width 1.6× height; distinctly and densely transversely striate with fine granulation. Clypeus as wide as face, distinctly separated from face, and punctate-rugose. OOL / OD = 1.2. POL / OD = 1.4. Frons widely smooth, anteriorly with a pair of obscure carinae. Vertex and temple almost smooth. Length of malar space 1.1× basal mandibular width. Antennae with 26 segments; 4 th segment 3.1× longer than wide; and penultimate one 1.9× longer than wide.
Habitats. Despite the multiple field collection sessions at primary forest areas in the Okinawa-hontô and Amami-ôshima Islands, only one specimen of M. stellatus sp. nov. was sampled from a secondary evergreen forest in the latter island. Most other specimens of M. stellatus sp. nov. were collected from a campus of the University of the Ryukyus, urban parks, and back yards in Okinawa-hontô Island, by finding suspended cocoon masses or rearing host larvae. As the host sphingids and their host plants are abundant in or around the edges of sparse forests, M. stellatus sp. nov. likely prefers rather open forests.
Phenology. The emergence of adult wasps occurred from April to June and from October to January, but not during the hottest season from July to September (Fig. 5).
Secondary sex ratio. The proportion of males (secondary sex ratio) ranged from 0.20 to 0.64, showing a gradual increase with the total number of wasps per host larva (Fig. 6). The positive effect of the total number of wasps on the sexual ratio was significant (B = 0.010, Wald Chi-Square = 18.129, df = 1, p < 0.001). The estimated mean of the proportion of males was 0.36 (95% confidence interval = 0.31-0.42, significantly less than 0.50) in the average number of total wasps (50.5), indicating an overall female-biased sex ratio.  (Fig. 7A). The larvae commenced hanging down from the host plant substance with their own suspensory threads as soon as they emerged (Fig. 7B). The larvae began to cluster by actively rotating, twisting, swaying, and horizontally stretching (Fig. 7C). When the larvae found the threads of other individuals, they actively went down, intertwined with said threads, and eventually merged together. Once they formed a large mass, the mass did not descend any more (Fig. 7D). In rare cases, several larvae moved from the upper mass to the lower mass as the cable of the former intertwined with that of the latter, owing to the blowing wind. No further larval transfer was observed after approximately 70 min of the emergence of the larvae. Initially, the shape of the larval masses was irregular (Fig. 7E), but gradually the larvae adopted a spherical shape (Fig. 7F, G). The larvae twisted their upper bodies and spun the thread at the posterior of their body, namely inside the cocoon mass. The silk walling action lasted approximately 40 min, along with the spinning of their own individual cocoons (Fig. 7H, I). Finally, three cocoon masses were completed approximately 2 h after larval emergence (Fig. 7J). A video of the entire process of cocoon mass formation is available at the following address: https://www.youtube.com/watch?v=AuHarLHolPM.
The host sphingid died on the following day after wondering. The color of the cocoons gradually darkened over a few days. After 8 days, 68 females and 23 males of M. stellatus sp. nov. emerged from these three cocoon masses. The wasps emerged simultaneously, cutting the tip of each cocoon.  Characteristics of the cocoon masses. The cocoon masses of M. stellatus sp. nov. (Fig. 8) were light brown to brown, 7-14 mm in width, 9-23 mm in length, and regularly spherical to ovoid with minimally 12 (Fig. 8D) to maximally over 100 cocoons (Fig. 8C). Exceptionally, approximately 200 cocoons formed a collapsed large mass in an artificial breeding case (Fig. 8B). Each cocoon mass was suspended by a single thick cable. The cable was 12-100 cm in length. Although most larvae constructed such cocoon masses, sometimes a few single larvae formed their own cocoons on the cable (Fig. 8F). The cable consisted of individual threads, which were tightly intertwined, like a rope (Fig. 8G). The anterior third to half of individual cocoons was exposed outward and fairly distributed on the spherical or ovoid surface. The posterior half of individual cocoons was invisible under the dense silk wall. Adults emerged by opening an anterior outside cocoon cap, which was circular shaped and tapering (Fig. 8F). Cocoons with such a regular cap are typical of the pulchricornis clade of Meteorus (Askari et al. 1977;Maeto 1989aMaeto , b, 1990a. Phylogeny of Meteorini and affinity of M. stellatus sp. nov. The Meteorini phylogeny is illustrated in Fig. 9. The Meteorini classification was also revised based on our phylogeny and those of Maeto (1990b) and   (Table 3). Although our topology was poorly resolved at the species level, it was mostly congruent with that of Stigenberg and Ronquist (2011). Meteorini was recovered as a monophyletic group. Meteorini species were divided into five clades ( Fig. 9; Table 3). Zele was recovered as a robustly supported monophyletic clade and nested within Meteorus species. Monophyly was robustly supported for the ictericus and pulchricornis clades but not for the unresolved clade. The pulchricornis clade was divided into four internal subclades (the colon, pendulus, pulchricornis, and rubens subclades).
Meteorus stellatus sp. nov. was recovered as an ingroup of the versicolor complex of the rubens subclade within the pulchricornis clade and sister to M. tarius.

Discussion
Our observation of M. stellatus sp. nov. shows that gregarious cocoon masses were constructed by the highly elaborated cooperation of larvae. The larvae never merged immediately after emergence from their host, but initially just descended. This seems to reinforce the idea that the suspended cocoon makes the pupating wasp inaccessible to some potential enemies (Shaw and Huddleston 1991;Quicke et al. 2006;Zitani 2003;Shirai and Maeto 2009;Maeto 2018). It seems that predators, like ants, seldom encounter suspended larvae because the threads of larvae are attached to the plant subtract only by a small area.
The cable of gregarious Meteorus is thought to be very resistant to breaking and highly tolerant to environmental stress (Barrantes et al. 2011). A cable of M. stellatus sp. nov. consists of a lot of individual threads and seems to be very strong, pretty much like that of M. restionis Shaw & Jones (Barrantes et al. 2011). Interestingly, cocoon masses of M. townsendi are suspended by a fairly long cable, the length of which is approximately 3.0 m (Zitani and Shaw 2002;Zitani 2003), while the longest cable of M. stellatus sp. nov. is approximately 1.0 m. According to our observation, the long individual threads and the adequate wind during the hanging period seem to make it easier for the larvae of M. stellatus sp. nov. to merge with each other. The larvae continue to spin thread until they are able to merge into a cocoon mass, and in most cases, they never go down unnecessarily after that. The star-shaped cocoon masses of M. stellatus sp. nov. can reduce the risk of hyperparasitism, because the exposed area of each individual cocoon is apparently smaller than the solitary cocoon or non-star-shaped cocoon masses, as suggested by the spherical cocoons of M. komensis (Zitani 2003). The outer cocoons of non-circular cocoon masses of gregarious Cotesia glomerata (Linnaeus) (Braconidae, Microgastrinae) are actually more easily parasitized than the inner ones (Tagawa and Fukushima 1993; Figure 9. Maximum likelihood tree of Meteorini generated using IQ-TREE (BIPP, Bayesian inference posterior probabilities; SH-aLRT, a Shimodaira-Hasegawa-like approximate likelihood ratio test; UF-Boot2, ultrafast likelihood bootstrap replicates). Tanaka and Ohsaki 2006). Therefore, the evolution of gregariousness and spherical cocoon masses seems reasonable.
The sex ratio has been studied in gregarious species of Meteorus, while a similar pattern of female-biased sex ratio has been shown in Macrostomion sumatranum (Enderlein) (Braconidae, Rogadinae), which is also a gregarious parasitoid of matured sphingid larvae (Maeto and Arakaki 2005). Both in the gregarious parasitoid category, the number of wasps emerged from each host varies widely (8-122 in M. stellatus sp. nov. and 26-160 in Ma. sumatranum) and the proportion of males increases with it. The female-biased sex ratio could be a result of the local mate competition, as expected in inbreeding gregarious parasitoids (Hamilton 1967;Godfray 1994;Smart and Mayhew 2009), in which the increase of the male proportion may be caused by the oviposition of multiple females on a single host larva (Werren 1983). This prediction will be tested by the examination of mating systems, oviposition behavior, and primary sex ratio.
The pulchricornis and rubens species-groups belong to the monophyletic lineage of pulchricornis clade (Fig. 10), in which both solitary and gregarious species are included and cocoons are usually suspended by a spun thread (Maeto 1990b;Stigenberg and Ronquist 2011). It is thus likely that communal cocoon masses of type F have evolved through individually or sparsely suspended gregarious cocoons of type B and C, and subsequently loosely clumped and suspended gregarious cocoons of type D, from suspended solitary cocoons within the versicolor complex. However, the evolutionary pathways are not yet clarified because only a few gre- garious species are placed in the present phylogram. Further and comprehensive analyses including more gregarious species are necessary to confirm and expand this evolutionary scenario.