Roisinitermesebogoensis gen. & sp. n., an outstanding drywood termite with snapping soldiers from Cameroon (Isoptera, Kalotermitidae)

Abstract Termites have developed a wide array of defensive mechanisms. One of them is the mandibulate soldier caste that crushes or pierces their enemies. However, in several lineages of Termitinae, soldiers have long and slender mandibles that cannot bite but, instead, snap and deliver powerful strikes to their opponents. Here, we use morphological and molecular evidence to describe Roisinitermesebogoensis Scheffrahn, gen. & sp. n. from near Mbalmayo, Cameroon. Soldiers of R.ebogoensis are unique among all other kalotermitid soldiers in that they possess snapping mandibles. The imago of R.ebogoensis is also easily distinguished from all other Kalotermitidae by the lack of ocelli. Our study reveals a new case of parallel evolution of snapping mandibles in termites, a complex apparatus responsible of one of the fastest biological acceleration rates measured to date.


Introduction
Termites are extremely abundant (Martius 1994, Eggleton et al. 1996 and colonies may contain millions of individuals attracting a wide variety of predators (Deligne et al. 1981). Additionally, termites experience strong intra-and inter-specific competition (Levings andAdams 1984, Thorne andHaverty 1991). To combat against the plethora of agonistic opponents, termites have developed a rich array of defensive strategies. The most important defenses are expressed in the soldier caste that is ancestral to all extent termites (Roisin 2000).
Soldiers are specialized sterile colony defenders possessing exaggerated morphology of the head and mandibles (Prestwich 1984). One of their most intriguing defenses is exemplified by long and slender snapping mandibles (Deligne et al. 1981). The snapping mandibles are paired with muscles to store potential energy which, when released, delivers a powerful strike producing one of the fastest accelerations known among animals (Seid et al. 2008). All termite species with snapping soldiers described so far belong to the Termitinae (Bourguignon et al. 2017), suggesting that snapping soldiers evolved several times independently within this subfamily. Alternatively, soldiers with snapping mandibles might have evolved once, and independently reverted to a biting strategy in several lineages.
The monumental revision of the Kalotermitidae by Krishna (1961) provided the morphological diagnoses for all extant genera with the exception of the recently described Longicaputermes (Ghesini et al. 2014). Aside from Longicaputermes, all new kalotermitid species described after Krishna's 1961 revision, ca. 115, have been assigned to one of the 21 genera he recognized. The soldier caste of several genera has unmistakable characters: e.g., the scooped out frons of Eucryptotermes, the massive third antennal article of Marginitermes, the large ovoid head of Pterotermes, or the spur on the fore tibia of Calcaritermes. We herein describe a new genus and species of Kalotermitidae, Roisinitermes ebogoensis, which possesses equally unmistakable soldiers. The soldier of R. ebogoensis is the first outside the Termitinae to have snapping mandibles.

Illustrations and measurements
Images of individuals were taken as multi-layer montages using a Leica M205C stereomicroscope with a Leica DFC 425 module run with Leica Application Suite software version 3. Preserved specimens, stored in 85% ethanol, were positioned in a transparent petri dish filled with Purell hand sanitizer (70% EtOH). Measurements (Tables 1-2) were obtained using an Olympus SZH stereomicroscope fitted with an ocular micrometer. A field photograph of live specimens placed in a small paper-lined Petri dish was taken with a Canon EOS 5DS R combined with a Canon EF 100mm f/2.8L Macro IS USM lens. Morphological terminology follows that of Krishna (1961).

Phylogenetic analyses
DNA was extracted from five individuals of R. ebogoensis, after removal of the digestive tract. The full mitochondrial genome was amplified with TaKaRa LA Taq in two long PCR reactions using primers specifically designed for termites (Bourguignon et al. 2016). Long PCR fragments were pooled in equimolar concentration, and 75bp paired-end reads were obtained using Illumina MiSeq. We subsampled a total of 10,000 reads and assembled the full mitochondrial genome with SPAdes, under default parameters (Bankevich et al. 2012). The total coverage of the assembly was 82-fold.
We used the mitochondrial genomes of ten species of Kalotermitidae, including one sample of Roisinitermes ebogoensis sequenced in this study. We used four non-Kalotermitidae termite species as outgroups to root the tree: Zootermopsis angusticollis, Hodotermopsis sjostedti, Coptotermes sjostedti, and Termitogeton planus. All mitochondrial genomes, except that of R. ebogoensis, have been published recently (Suppl. material 1: Table S1). Each gene of the mitochondrial genome was aligned separately using MAFFT v7.300b with the option "--maxiterate 1000 --globalpair" for higher accuracy. For protein-coding genes, we first aligned genes as protein, then converted protein sequence alignments into the corresponding codon alignments using PAL2NAL (Suyama et al. 2006). The 22 tRNAs and the two ribosomal RNAs were aligned as DNA. The resulting alignments were concatenated with FASTconCAT v1 (Kück and Meusemann 2010). Alignments were separated in five partitions: one for each codon position of the protein-coding genes, one for the combined ribosomal RNA genes, and one for the combined tRNA genes.
We reconstructed phylogenetic trees using Maximum Likelihood and Bayesian approaches. We ran the analyses twice, once with the third codon position included, and once without third codon position. The Bayesian phylogenies were implemented in MrBayes 3.2 (Ronquist et al. 2012) with unlinked partitions, each of four chains (three hot and one cold). The chain length was of two million generations with sampling every 2000 generations. 800,000 generations were discarded as burnin, to ensure that the chain reached convergence, as determined by Tracer 1.5 (Rambaut and Drummond 2007). We ran two replicates of each analysis to ensure consistency of the results. For each partition of the data, we assigned an independent Generalized Time Reversible model with gamma-distributed rate variation across sites and a proportion of invariable sites (GTR + G +I). The reconstruction of Maximum Likelihood phylogenies was carried out with RAxML (Stamatakis et al. 2008). We used the GTR-GAMMA model of rate heterogeneity across sites. Node support was estimated using 1000 bootstrap replicates.

Phylogenetic analysis
Our phylogenetic analyses supported the monophyly of Kalotermitidae ( Figure 1). The four analyses yielded identical tree topologies, with one exception: in the Bayesian analysis without third codon position Rugitermes was the sister group of Neotermes + Cryptotermes + Incisitermes + Roisinitermes, while in the other three analyses Rugitermes + Neotermes sp. A formed the sister group of Neotermes insularis + Cryptotermes + Incisitermes + Roisinitermes. Roisinitermes was consistently placed next to N. insularis.
Winged Imago. Ocelli not visible either by pigmentation or cuticular protrusion (Figure 2A-C). Fore wing with unsclerotized media and cubitus arising from a common vein distal from scale suture; radial sector with 5-6 anterior branches; subcosta very close and difficult to discern from costal margin ( Figure 2D). Hind wing with radial sector and cubitus arising from a common vein distal to suture. Tibial spurs 3:3:3; tarsi without arolia. The left imago/nymph mandible with anterior margin of their marginal tooth ca. 1.5 times longer than length of the posterior margin of the first plus second marginal tooth; right mandible with posterior margin of second marginal tooth 1.4 times as long as molar plate (Figure 3).
Diagnosis. The lack of visible ocelli is unique among all other Kalotermitidae. In Krishna's 1961 generic key, Roisinitermes would lead to couplet 2 (Epicalotermes). Diagnosis. Stick-like mandibles unique among all other kalotermitid soldiers. In Krishna's 1961 key, Roisinitermes leads to couplet 17 (Allotermes). In dorsal view, the mandibular blades of Allotermes, especially A. denticulatus Krishna 1962, somewhat resemble those of Roisinitermes as those of the former are long, rather narrow and with rudimentary dentition. In lateral view, however, the Roisinititermes mandibles differ from all other kalotermitids with projecting mandibles in that the Roisinitermes mandibles arch downward. Although the Roisinitermes imago venation and dentition is very similar to those of Epicalotermes, the soldier of Roisinitermes shares no major characters with the Epicalotermes soldier.
Etymology. The genus is named in honor of Dr. Yves Roisin for his many contributions to the study of termites.  Table 1) Head and pronotum light brownish orange; eye ovoid, anterior margin truncate abdominal tergites lighter, concolorous with legs and labrum; postclypeus nearly hyaline. Compound eyes black, of medium size and protrusion; ellipsoid but truncated near antennal socket, composed of approximately 85 facets. Ocelli not visible either by pigmentation or cuticular protrusion. Antennae with more than nine articles; formula 1>2=3=4<5. Pronotum width twice that of median length; several long and shorter setae project from lateral margins. Fore wing scale with basal origins of all major veins; wing membrane covered with papillae. Tibial spurs 3:3:3; tarsi without arolia.
Soldier (Figs 4-6; Table 2) Monomorphic. In dorsal view, head capsule yellowish orange in posterior grading to orange in middle and reddish brown from frons to ante-clypeus. Three proximal antennal articles sepia brown; distal articles light brown. Post clypeus and labrum yellowish with brown highlights. Eye spots prominent; large, dark brown, elliptical; formed from a mass of discrete ommatidia. Pronotum concolorous with posterior head capsule. Head capsule in dorsal view, subrectangular; lateral margins nearly parallel, length 1.5 times width. Posterior corners of head evenly rounded; posterior margin rectate. In lateral and oblique view, head capsule almost cylindrical with only slight dorso-ventral compression; frons bilobed in dorsal view, crested with rugose longitudinal stripes, rugosity lateral below frons to mandibles. In lateral view, frons sloping from vertex ~45°; mandibles bow upward to form a 15° arch. Setae short and sparse on pronotum and head capsule. Periantennal carina rugose, in dorsal view partially eclipsing the first antennal article. Small horn-like projection at terminus of ventral genae. Mandibles stick-like; long, blade narrower in middle than distal third, dentition very weak; left mandible with faint equilateral tooth approx. three fifths from base, serrations along blade from tooth to tip. Right mandible with single tooth approx. one third distance from base; blade narrowest before tooth; after tooth blade widens slightly and then gradually narrows at tip. In dorsal view, basal humps project sharply as rugose hemispheres. In lateral view, humps are columnar and equal in height to that of the mandibles. Anteclypeus shallowly incised in middle; labrum linguiform with gradual point; 4-5 long terminal setae. Antennae with 12-13 articles, third antennal article subclavate, barely shorter than fourth and fifth combined. Pronotum collar-shaped; much wider than long. Anterior margin weakly concave; lateral margins weakly convex, posterior margin forming 25° angle with incised middle. Femora moderately inflated, tibial spurs 3:3:3. Habitus as in Figure 6.
Brachypterous nymph (Fig. 3, Table 3) Body hyaline. Head, thorax, and abdomen similar in shape and pilosity of imago. Compound eyes with approx. 85 dark    relatively thin (3 cm) and long (over 3 m) broad-leaf tree branch suspended from the canopy approximately 2 m above the ground. The colony contained roughly 2,000 members. A second colony of R. ebogoensis was collected in a nearly pristine rain forest near the village of Ebogo II. The colony was taken from a dead liana branch (ca. 15 mm diam.) hanging from the canopy at a height of approx. 1 m above the ground. Liana stems have been generally overlooked as a colonization site for Kalotermitidae (Scheffrahn et al. 2018). In light of Emerson's 1925 description of Cryptotermes cubioceps from a single soldier collected from a dead liana, this host should be probed routinely as a colonization site for kalotermitids. Etymology. The species is named for the village of Ebogo II, the type locality for this termite.

Discussion
Kalotermitids inhabit a single woody item and are largely unable to move to a new food source once the original is exhausted. The lone exception is Paraneotermes simplicicornis that builds underground galleries connecting several wood pieces (Light 1937). The ability to feed on sound wood represents a defensive adaptation in itself as the hard food source acts as an efficient physical barrier against intruders. Kalotermitids thus show low soldier-to-worker ratios (see Haverty 1977) and soldiers reach a high level of polymorphism, reflected especially in the development of the headcap- sule and mandibles. Some genera such as Bicornitermes, Cryptotermes, Eucryptotermes, Calcaritermes, or Glyptotermes, possess very short mandibles and a plug-like headcapsule to prevent intruder entry into a nest gallery (phragmosis). In C. cryptognathus from Jamaica, the mandibles are reduced to small stubs that do not project beyond the frontogenal boundaries of the head capsule, and therefore cannot be used to bite opponents (Scheffrahn et al. 1998). Some other genera (e.g., Bifiditermes, Epicalotermes, Incisitermes, Kalotermes, Neotermes) possess long mandibles with robust dentition (crushing mandibles sensu Prestwich 1984) used to injure an opponent mechanically. This is often combined with release of defensive secretions originating in the labial glands (Šobotník et al. 2010, Sillam-Dussès et al. 2012. Epicalotermes pakistanicus has particularly long and serrated mandibles (Akhtar 1974). The defensive strategy of Roisinitermes soldiers does not match any of these; instead, Roisinitermes employs a unique strategy of snapping, achieved by long and slender mandibles pressed against each other in a defensive encounter. When this potential energy is released, the left mandible springs over the right and the resultant snap is forced onto the opponent if it is in the path of the strike. This singular mandibular modification was previously known in several lineages of Termitinae (Deligne et al. 1981, Prestwich 1984, Seid et al. 2008, and was portrayed as a defensive strategy unique to this group. Roisinitermes represents the first undisputable evidence of parallel evolution of snapping soldiers.
Our phylogenetic analyses consistently placed Roisinitermes on a long branch, next to N. insularis. Neotermes insularis is a large termite species from Northern Australia with soldiers endowed with biting mandibles of crushing type. The smaller Roisinitermes shares no obvious similarity with N. insularis, supporting its generic status. Currently, the number of mitochondrial genomes available for Kalotermitidae is limited to a handful of genera, and there is a possibility that future phylogenetic analyses will support affinities between Roisinitermes and yet-to-be sampled taxa. In any case, the highly unusual morphology of Roisinitermes suggests that it shares no close relatives among modern Kalotermitidae. Future studies should focus on whether the mechanisms used by soldiers of Roisinitermes to snap are like those of the distantly related Termitinae.