Folding Wings like a Cockroach: A Review of Transverse Wing Folding Ensign Wasps (Hymenoptera: Evaniidae: Afrevania and Trissevania)

We revise two relatively rare ensign wasp genera, whose species are restricted to Sub-Saharan Africa: Afrevania and Trissevania. Afrevania longipetiolata sp. nov., Trissevania heatherae sp. nov., T. hugoi sp. nov., T. mrimaensis sp. nov. and T. slideri sp. nov. are described, males and females of T. anemotis and Afrevania leroyi are redescribed, and an identification key for Trissevaniini is provided. We argue that Trissevania mrimaensis sp. nov. and T. heatherae sp. nov. populations are vulnerable, given their limited distributions and threats from mining activities in Kenya. We hypothesize that these taxa together comprise a monophyletic lineage, Trissevaniini, tr. nov., the members of which share the ability to fold their fore wings along two intersecting fold lines. Although wing folding of this type has been described for the hind wing of some insects four-plane wing folding of the fore wing has never been documented. The wing folding mechanism and the pattern of wing folds of Trissevaniini is shared only with some cockroach species (Blattodea). It is an interesting coincidence that all evaniids are predators of cockroach eggs. The major wing fold lines of Trissevaniini likely are not homologous to any known longitudinal anatomical structures on the wings of other Evaniidae. Members of the new tribe share the presence of a coupling mechanism between the fore wing and the mesosoma that is composed of a setal patch on the mesosoma and the retinaculum of the fore wing. While the setal patch is an evolutionary novelty, the retinaculum, which originally evolved to facilitate fore and hind wing coupling in Hymenoptera, exemplifies morphological exaptation. We also refine and clarify the Semantic Phenotype approach used in previous taxonomic revisions and explore the consequences of merging new with existing data. The way that semantic statements are formulated can evolve in parallel, alongside improvements to the ontologies themselves.


Introduction
Elongate anatomical structures, along which the wing blade is creased, occur in many different forms and serve different functions in Pterygota [1,2]. Besides the wing veins, which are exclusively convex relative to the dorsal surface of the fore wing in Hymenoptera, two types of resilin-rich longitudinal anatomical structures are present on the fore wing. While the usually concave flexion lines are involved in the control of passive wing deformation during different stages of beat cycle, concave or convex fold lines play important roles in folding the wing to its resting position. The majority of fold lines are axillary folds that are located at the wing base. These folds act as hinges between axillary sclerites and play crucial roles in flexing the neopteran wing over the body. Occasionally other fold lines are present on the wing blade, which are involved in folding mechanisms that decrease the area of the wing blade. While flexion and axillary fold lines are arguably conservative enough to serve as landmarks for characterizing major wing regions across Pterygota, [1,3,4], fold lines involved in reducing wing blade area show greater diversity [5,6]. Four plane wing folding, where four flexible fold lines intersect in one point and hinge the four stiff plates between them, occurs in Coleoptera, Blattodea and Dermaptera, [6].
Hymenoptera exhibit enormous morphological diversity in their wing venation patterns, but the system of flexion and fold lines remains relatively conserved [5,7]. While wing venation contributes important and informative morphological characters to both higher-and lower-level classification, axillary folds are seemingly uniform throughout the order. So far, only one surface-reducing longitudinal fold has been reported for Hymenoptera [5].
Evaniidae (Hymenoptera) is a monophyletic lineage of cockroach predators, with striking morphological synapomorphies in both the meso-and metasoma that putatively reflect the shape and size of the cockroach ootheca in which they develop [8][9][10]). By contrast, evaniid wings are superficially similar to that of related taxa taxa (i.e., Gasteruptiidae and Aulacidae [11]) and are seemingly unaffected by constraints of host geometry.
Deans and Huben [8] developed numerous fore wing characters for their diagnostic key to evaniid genera. One of these character states was described as ''wings often folded apically'', and the illustration of the fore wing of Afrevania suggest that the fore wing of some evaniid taxa might be equipped with four-plane wing folding.
We characterize this trait here, including its underlying mechanisms, and discuss the morphological traits of evaniid wing vein geometry and wing base sclerites that possibly lead to the evolution of four-plane wing fold. We examine the possibility that wing folding Evaniidae form a monophyletic group by documenting the morphological and species diversity of a new tribe Trissevaniini. Finally, we compare the wing folding mechanism of Evaniidae with that of other four-plane folding insects.
Balhoff et al. [12] published a revision of the ensign wasp (Hymenoptera: Evaniidae) fauna of New Caledonia, in which they applied a recently developed semantic phenotype (SPs [13]) model to taxonomic descriptions. The application of this model allows a broader array of researchers to search and reuse these data increasing the value of taxonomic work. Balhoff et al. [12] proposed an annotation workflow for SPs generation and discussed limitations and potential benefits such as automated data integration and reasoner-driven queries. Issues related to the longevity and compatibility of SPs from different taxonomic treatments, however, were not challenged, and solutions for this conundrum were not proposed by Balhoff et al. [12]. We refine and clarify the Semantic Phenotype approach here and explore the consequences of merging new with existing data.

Materials and Methods
Specimens (Table S1) were studied from the following institutions (abbreviations after [14]; curators in parentheses after Glycerin-stored, critical point-dried, and air-dried specimens were used for morphological observations. For the species revision, 93 morphological characters are scored based on observations of critical point-dried and air-dried specimens, using Olympus SZX16 stereomicroscope with Olympus SDF PLAPO 2XPFC objective (230X magnification). Observations on wing base sclerites, wing veins, wing creases and thoracic musculature were carried out on glycerin-stored and critical point-dried specimens, using the same Olympus stereomicroscope. Blue-Tac (Bostik, Inc., Wauwatosa, Wisconsin, USA) was used as a base medium for stabilizing the specimens during dissections made with razor blades, #2 insect pins and #5 forceps.
Anatomical structures were visualized with bright field, confocal laser scanning microscopy (CLSM) and scanning electron microscopy (SEM) techniques. Bright field images were taken with an Olympus DP71 digital camera attached to an Olympus ZX41 compound microscope and subsequently processed with Combi-neZM [15] executing the ''do stack'' command. CLSM images were taken on glycerin-stored specimens with Zeiss LSM 710 Confocal Microscope. We used an excitation wavelength of 488 and an emission wavelength of 510-680 nm, detected using two channels and visualized separately with two pseudocolors (510-580 nm = green; 580-680 nm = red; Figs 1A, B). To visualizing resilin we used an excitation wavelength of 405 nm and an emission wavelength of 510-680 nm, detected on one channel and visualized with a blue pseudocolor ( Fig. 2A). The single SEM image was taken with FEI Nova 400 NanoSEM (Florida State University) on Au-Pd coated specimens. To facilitate understanding of the wing folding mechanism in Trissevaniini, and to demonstrate the rich flexion line system of the tribe relative to that of other Evaniidae, we have created a print-cut-fold model of the fore wing of Trissevania anemotis and Evania albofascialis. (Figure S1). Images and volume rendered media files were deposited at Figshare (http://dx.doi.org/10.6084/m9.figshare.961802) and at Morphbank (http://www.morphbank.net/?id = 835698). Abbreviations of anatomical structures used in Figures are listed in Table  S2.
The electronic edition of this article conforms to the requirements of the amended International Code of Zoological Nomenclature, and hence the new names contained herein are available under that Code from the electronic edition of this article. This published work and the nomenclatural acts it contains have been registered in ZooBank, the online registration system for the ICZN. The ZooBank LSIDs (Life Science Identifiers) can be resolved and the associated information viewed through any standard web browser by appending the LSID to the prefix ''http://zoobank.org/''. The LSID for this publication is: urn:lsid:zoobank.org:pub:ECF5A640-93F5-4DA6-B210-E404A7BCAAC7. The electronic edition of this work was published in a journal with an ISSN, and has been archived. It is available from the following digital repositories: PubMed Central and LOCKSS.

Semantic representations
In the following addenda we propose an expansion and elaboration of the semantic statement types described by Balhoff et al. [12] and Mullins et al. [16] and briefly discuss NL statements that we were not able express logically.
Sculpture. Sculpture is one of the most ambiguous character systems in Hymenoptera morphology descriptions [17]. Not only is the sculpture terminology flooded with synonymy [18], but the semantic representations of sculptural terms also are under almost constant debate and are highly context dependent. Sculptural terms are often used as quality descriptors, ''Axillae reticulate…'' but also as entity terms: ''Mesoscutum with raised and strong reticulation…'' (both statements from [19]).
The homonymic nature of surface sculpture terminology is perhaps the largest obstacle to developing a broadly accepted standard and to placing the concepts in an ontological framework. Here we propose a solution, in which we consider sculpture types as synonyms of the subclasses of the quality descriptor texture: 'has part' some (axilla and 'bearer of' some reticulate) and express sculpture-related entity terms as anatomical regions defined by the texture itself: 'has part' some (mesoscutum and 'has part' (some 'anatomical region' and 'bearer of' some reticulate')) Following our reasoning, Hymenoptera-specific sculpture terms can be easily incorporated into the general texture class system of PATO, making Hymenoptera sculptural phenotypes broadly accessible. On the other hand, further quality descriptors can be attached to the post-composed ''sculpture defined'' anatomical regions making possible the logical representation of statements using ''sculpture'' as entity terms. This approach makes it possible to handle widely controversial and mostly subjective complex sculptures like ''punctate-foveate'' (note that ''fovea'' is a synonym for depression). For the NL statement ''Upper face sculpture: punctuate and foveate'' the logical representation is 'has part' some ('upper face' and ('has part' some (region and ('bearer of' some foveate) and ('has part' some ((not (depression)) and ('bearer of' some punctate)))))).
Views. Views are critical when certain qualities are used (e.g., subclasses of distances between anatomical structures and shapes). We have expressed views using BSPO anatomical sides. For the NL statement ''Anterolateral mesopectal projection 2D shape lateral view: square shaped.'' the logical representation is 'has part' some ('anterolateral mesopectal projection' and 'has part' (some 'lateral side' and 'bearer of' some square)).
Length. Distances between anatomical structures play a central role in species diagnoses [20] and embody perhaps the most standardized portion of taxonomic descriptions [21]. Balhoff et al. [12] and Mullins et al. [16] used the PATO class length (http://purl.obolibrary.org/obo/PATO_0000122) in the formal representation of NL phenotypes describing anatomical lines (immaterial anatomical entities) that extends between anatomical structures (material anatomical entities) and can be measured. Although these phenotypes provide information about the nature of morphological characters (i.e. the given phenotypes provide some metric data of an anatomical structure), using the PATO class ''length'' in these instances result in ambiguous metric data that are neither comparable nor searchable. For instance, in phenotypes concerning ''scape length'' or ''mesoscutum length'', following the logic of PATO ''length'', we refer and compare any distance that is defined by any anatomical structures of the ''scape'' or ''anteromesoscutum''.
Ontologies provide the opportunity to define anatomical lines that can be used for refining the formal representation of length characters. Eye height can be explicitly defined as the anatomical line that is the longest of those lines that connect the dorsalmost and ventralmost points of the compound eye (http://purl. obolibrary.org/obo/HAO_0002254). Moreover, we can create a subclass hierarchy of these anatomical lines and post-compose individual lines using these general classes. In the present treatment we use two general anatomical lines, the ''median anatomical line'' (http://purl.obolibrary.org/obo/ HAO_0002272) and the ''proximodistal anatomical line'' (http://purl.obolibrary.org/obo/HAO_0002273), that can be applied to anatomical structures of the body and the appendages respectively. In the descriptions we refer these concepts as anatomical entities and provide their PATO lengths as numerical entities in the semantic statements. For the NL statement ''Mesoscutellum median length vs. anteromesoscutum median length: mesoscutellum longer than or equals anteromesoscutum'' the logical representation is 'has part' some (mesoscutellum and ('has part' some ('dorsal side' and (('has part' some 'median anatomical line') and ('bearer of' some (length and ('is quality measured as' some (('has measurement unit label' some (length and ('inheres in' some ('median anatomical line' and ((part_of some 'dorsal side') and (part_of some anteromesoscutum)))))) and ('has measurement value' some float[. = 1.0f]))))))))) and for the NL statement ''Female scape length vs. compound eye height: eye height is at least 26 as long as scape length'' the logical representation is 'has_part some ('eye height' and ('bearer of' some (length and ('is quality measured as' some (('has measurement unit label' some (length and (inheres_in some ('proximodistal anatomical line' and (part_of some scape))))) and ('has measurement value' some float[. = 2.0]))))))).
Comparing parts of an anatomical structure with remaining parts of an anatomical structure. Quality differences between a part of an anatomical structure and the remaining parts are often provided in taxonomic descriptions especially when color phenotypes are given. In the statement ''Mandible color: black with dark brown mandibular teeth.'' The coloration of the mandibular teeth is compared with the rest of the mandible. We expressed this statement logically in Balhoff et al. [12] as (has_part some (mandible and (has_part some ('proximal region' and ('bearer of' some black))))) and (has_part some (tooth and ('bearer of' some 'dark brown'))) where the proximal region of the mandible was vaguely defined and certainly did not refer to the mandible exclusive of the teeth. In the present treatment we have corrected this ambiguity and expressed the phenotype as ('has part' some ((not (tooth) and (part_of some (mandible and (bearer_of some black))))) and (has_part some (tooth and (bearer_of some 'dark brown'))).  Comments. Due to the limited number of specimens available for Afrevania leroyi and Trissevania mrimaensis we were not able to perform dissections and repositioning of these species. Only one female represents Trissevania mrimaensis. Given this situation we were not able to score some otherwise invariable characters, and thus they were not involved in the description of the new tribe. The phenotypes shared by all scored Trissevaniini taxa are: Median clypeal projection count: present; Median clypeal projection sharpness: blunt; Median conjunctiva of abdominal tergum 9 presence: present.
Description.  Comments. Although the repository for type specimen(s) of Trissevania anemotis Kieffer, 1913 is currently unknown, we were able to recognize specimens of this species based on the original description. The only Trissevaniini species with malar striation is T. anemotis (in Kieffer's description [22]): ''…joues fortement striées en long, convexes comme le front, sans sillon, égalant presque la longueur des yeux…'', which we translate as: ''cheeks strongly striated all along, convex as the forehead, without ridges, almost as long as the eye's length'').
Trissevania heatherae sp. nov. urn:lsid:zoobank.org:act:ACBF3813-5794-4E7F-9B5E-25A48CCBFB75 Diagnosis. Trissevania heatherae sp. nov. shares the presence of the mesopleural carina and the elongate body (head+mesosoma median length 1.86as long as mesosoma height) with T. mrimaensis sp. nov. and differs from it in the convex dorsal side of the mesosoma (T. mrimaensis is the only Trissevaniini with the dorsal side of the mesosoma flat).
Description. Body length: 2.6-2.7 mm. Mandible color: black with dark brown mandibular teeth. Malar distance vs. eye height: eye 26 as high as malar distance. Facial striae count: absent. Carinae laterally on frons count: absent. Antenna pilosity:      Etymology. The name of the new species refers to the locality of the holotype: Mrima Hill Forest in Kenya.

Morphological observations on the Trissevaniini fore wing
Although we were unable to observe live specimens, we were able to move and manipulate the fore wings of glycerin-stored specimens, transforming them between the ''fully folded'' (Fig. 2C) and ''fully unfolded'' (Fig. 2B) positions.
The wing blade is ''collapsed'' (Fig. 4C) along the dorsally convex anal-marginal fold line in most available specimens. We were able to force the wings from this ''collapsed'' position into a stable, ''fully unfolded'' position by gently pressing the area of the anal-marginal fold line (afl), delimited proximally by the intersection with the prestigmal fold line (pfl) and distally by the intersection with the poststigmal fold line (sfl) and the discal fold line (dfl) (stiffening region: Fig. 2A). This region is concave when the wing is in the ''fully unfolded'' (Fig. 2B) but convex if the wing is in the ''collapsed'' (Fig. 4C) or ''fully folded'' (Fig. 2C) positions. We could not force the wing blade into a stable ''fully folded'' position ( Fig. 2C) because it always restored itself to the ''collapsed'' position. In order to achieve a complete wing folding the wing must be folded first along the longitudinal fold line (afl: Figs 5, S1) and then along the transverse fold line (sfl+dfl: Figs 5, S1). The surface of the back-folded distal wing region (composed of the wing regions distal to sfl and dfl on Figs 5, S1) is curved in the fully folded wing. This curved region fits well in-between the anteroventral portion of the gaster and the posterior concavity on the mesosoma that receives the gaster (gastral scrobe). In two dried specimens with fully folded wings, they were folded in the same way, with the back-folded distal wing region tightly packed between the gaster and the mesosoma (Figs 10A, B).
While maneuvering the fore wing of glycerin-stored specimens, we observed that the posterior fore wing margin was often hitched to the dorsolateral part of the mesosoma. In these cases, setae of the dorsolateral setal patch (dsp: 3C, 4A, C) were inserted into the retinaculum (ventrally curved, gutter-like region on the posterior fore wing margin) (ret: Fig. 4B, 5A, B). Based on dissection of the mesosoma of critical point dried specimens, Trissevaniini shares the characteristics of the direct wing muscles (including the mesopleuro-basalare muscle) of other Evaniidae ( fig. 10D in [10]), and we were unable to define any differences in the trissevaniine wing base relative to other Evaniidae (Figs 1A, B). Our CLSM studies with the 405 nm laser show that fold lines on the fore wing of Afrevania longipetiolata are resilin rich (Fig. 2A).

Unresolved phenotypes and the compatibility of semantic statements
Fully representing the meaning of NL descriptions is impossible with current ontologies and OWL syntaxes. Textualizing our phenotype concepts with any syntax, however, limits expressivity, as even NL descriptions need incremental improvements. Although we have refined the semantics of numerous statements that were proposed in earlier taxonomic treatments [16,12], there remain many statements in need of further refinement (e.g., the HAO lacks many wing part classes).
This situation highlights an important issue related to the longevity of semantic statements: How can we trace refinements in Manchester syntax expressions of the same phenotype statement during the evolution of ontologies? Is it possible to connect specimens from earlier treatments to the new, refined statements? We propose here a simple mechanism to reattach refined statements to already published specimen data (see Materials and Methods). The resulting equivalence OWL file could be considered a long-term repository for further taxonomic treatments, and it provides the backbone for a semantic model system that meets the requirements of emerging eScience applications [23].

Trissevaniini and the evaniid phylogeny
Despite recent advances in evaniid systematics the phylogeny of Evaniidae is still unstable. Trissevaniini appears to be sister to Evania, based on analyses that included only one exemplar of the tribe, Trissevania anemotis [24]. Morphology seemingly contradicts these results. Evania spp. are relatively large (body length = 4-5 mm), elongate species with perhaps the most complex evaniid wing venation pattern (more than 10 cells on the fore wing). By contrast, genera of Trissevaniini are small (body length = 1-2 mm), stocky wasps, with less than 3 cells on the fore wing. The fore wing itself is relatively wider than those of the larger evaniids [8]. Apart from these general habitus (Gestalt) characters, Trissevaniini share numerous distinct phenotypes that appear to place them close to the genus Brachygaster: presence of posterior tooth of mandible (ptn: Fig. 3B Fig. 3A). Most of these traits, however, may be driven by spatial environmental constraints like the host geometry (scrobes on the mesosoma accommodate appressed appendages and the head) and might be evolutionary adaptations [25].
Evolutionary adaptation of the above listed morphological traits in Trissevaniini and Brachygaster is also supported by the differences in fore wing venation pattern between these taxa. The point-like or short and anterodistally-oriented fore wing 1CUa vein, and the absence of fore wing 1RS vein and the 1 st submarginal cell are shared by many short bodied evaniid taxa, such as Rothevania (Fig. 5C), Brachygaster (Fig. 5D) and Semaeomyia. Trissevaniini share the posteriorly oriented fore wing 1CUa vein and the presence of 1RS and RS+M and 2RS veins (Fig. 5B) with large bodied Evaniidae such as Evania (Fig. 5A) and Szepligetella.
Besides Trissevaniini, there is only one evaniid genus with the apparent ability to fold its wings, Brachevania. However, Brachevania has two parallel wing folds ( fig. 10 in [8]), and most importantly, does not share any of the diagnostic characters of the new tribe (e.g., the presence of dorsolateral setal patch of the metapectalpropodeal complex). Brachevania is monotypic and known only from one specimen. This situation limits further morphological examination.

Homology of wing creases across Evaniidae
The presence and pattern of wing folding/flexion lines are relatively conservative, and thus these structures render themselves to homologization across Hymenoptera [7,5,26]. According to Mason [27], only concave flexion lines are represented in Hymenoptera, so flexion lines are easy to differentiate from the always convex wing veins even if they are not tubular or less melanized. Wootton [1], however, reported a convex flexion line from Hymenoptera ( [27]) = medial folds sensu Richards [7]), posterior to the median vein (not illustrated; the anterior and posterior flexion lines are absent from Evaniidae). Unlike flexion lines, the single fold line of some Hymenoptera is always convex above (similarly to wing veins and Wootton's median flexion line), and thus the ventral wing surfaces are closed together during the wing folding process [5]. These straight wing folds are apparently not homologous with wing veins and flexion lines.
Homologizing the fold lines of Trissevaniini is seemingly difficult, due to the distally reduced wing venation of the new tribe (Figs 5A, B, S1). Based on their position relative to the pterostigma and medial vein, however, it is possible that the prestigmal flexion line is derived from the stigmal break [28] of other Hymenoptera, the poststigmal crease from the anterior radial flexion line, and the discal crease from the posterior radial flexion line of other Evaniidae (arf, prf: Fig. 5A). We were unable to find any crease on the fore wing of other Hymenoptera from which we could derive the anal-marginal fold line. The evaniid fore wing 4RS vein and the anterior radial flexion line [26] are bent and oriented anteriorly (4RS, arf: Fig. 5A). This phenotype is unique to Evaniidae and might favor the development of the transverse wing fold that extends parallel to the anteriorly curved portion of 4RS (dfl, sfl: Fig. 5B).
Possible wing folding/unfolding mechanism of Trissevaniini Some Blattodea, Dermaptera, and Coleoptera have four plane wing folding on their hind wings, primarily for reducing the exposed wing blade area so that it can be tucked under the modified fore wing (elytron or tegmin; see also [6]). These orders share a basic wing fold pattern, in which four flexible fold lines intersect in one point and delimit four stiff plates, but their wing folding/unfolding strategies are different. For example, hind wing folding/unfolding is facilitated by modified direct wing muscles and wing base sclerites in beetles allowing them to promote/ remote and unfold/fold their wings independently [29] while the wings are unfolded and folded solely by wing promotion and remotion in Blattodea [6] and in some Dermaptera [30]. The simplest wing folding mechanism and fold pattern are arguably that of the Blattodea. In this taxon the hind wing blade is folded longitudinally and then, in some species (e.g. Blaberidae), the apical region is folded over resulting in a small part of the wing tip lying on top of the folded wing [6]. This fold pattern consists of three convex and one concave folds and is named ''EXT+'' by Haas and Wootton ( fig. 2 in [31]). Until now, this simple pattern of wing folds was only known from some blaberid species [31].
Since we were not able to detect any differences between the wing base and direct wing muscles of Trissevaniini and that of other Evaniidae, it is the most likely that wings are folded/ unfolded during promotion and remotion, similar to Blattodea and Dermaptera. The basic pattern of Trissevaniini wing folds is most similar to that of Blattodea (Figs 2, 5, S1), representing the second known case of ''EXT+'' arrangement of wing folds [31].
Locking mechanisms play important roles in securing an unfolded wing position, a necessary condition for flight [5]. Wing regions that are involved in these mechanisms are usually at the intersection of concave and convex wing creases [5]. The stiffening region for the locking mechanism of trissevaniine fore wing is most likely on the anal-marginal fold line (stiffening region: Fig. 2A). The prestigmal fold line and the posterior radial fold line are not part of the main four plane folding line system, though they seem to play important roles in the locking mechanism. Without these creases the stiffening region would be restricted to one point at the intersection of the longitudinal fold and the transverse fold (intersection of sfl, dfl and afl on Fig. 5), which would result in a less stable unfolded position (Fig. S1).
High resilin content is necessary but not sufficient for the complete wing folding mechanism in Trissevaniini. External forces are required to keep the wing in the fully folded position (we were unable to find any wing blade based stiffening mechanism that stabilized the wing in a fully folded position), as well as for transforming it between the ''fully folded'' and ''fully unfolded'' positions. Based on our observations of specimens with fully folded wings, it is likely that the wing is secured in the fully folded position by being packed between the gaster and the mesosoma (Figs 10A,  B). Dislodging the wing from its folded position does not require much force, and thus the hind legs can probably initiate this activity. Reaching a stable, unfolded position, however, appears to be a more involved process that most likely requires the blade to be anchored at some points. The dorsolateral setal patch and the retinaculum of the fore wing serves most like as the anchoring device in the wing folding system of Trissevaniini (dsp: Figs 3C, 4A, C).

The cooption of the retinaculum
The presence of a dorsolateral setal patch (dsp : Figs 3, 4) is unique for the tribe. We hypothesize that this evolutionary novelty evolved in concert with the four plane wing folding mechanism in Trissevaniini. The retinaculum (ret : Figs 3, 4) is present in all winged Hymenoptera and is involved in wing coupling [32]. In Hymenoptera, a row of hook-like setae on the anterior hind wing margin, the hamuli, inserts into the retinaculum, locking the fore and hind wings together during the flight [33]. The involvement of the retinaculum in wing unfolding is an example of a morphological exaptation [34]. That is, the structure was produced by natural selection for wing coupling function and was then co-opted for wing folding function.

Conservation status of Kenyan Trissevania species
The specimens of Trissevania treated herein were collected as part of a widespread Malaise trap survey of Kenyan insects conducted between 2004 and 2012. They were found across the breadth of Kenya from coastal forest in the east to highland forests in the central and far western parts of the country. Wasps occurred at altitudes ranging from near sea level on the coast (Mrima Hill, 212 m above sea level) to 2507 m in the central highlands (Itieni Forest, Nyambene Hills) and 2474 m on Mt Elgon in western Kenya. They were also collected in Burundi in Malaise traps set in degraded shrub/grassland (Rusizi National Park, alongside Lake Tanganyika, 774 m), in mid-altitude gallery forest (Ruvubu National Park, 1382 m), and in Bururi National Forest at 1955 m.
The conservation status of the five Kenyan sites differs greatly. Itieni forest is managed by the Kenya Forest Service and is largely intact and under little threat from about 2100 m to over 2500 m. Mt. Elgon Forest lies within the national park and is also well protected. Gatamayu Forest is in the southernmost part of the Aberdare range, which has recently been completely fenced, and Castle Forest is above the altitude at which farming is permitted on the southern slope of Mt. Kenya. In contrast, the status of forest on Mrima Hill is precarious. Mrima Hill Forest (ca. 390 ha.) is part of the Coastal Forests of Eastern Africa biodiversity hotspot [35] comprising mostly small, isolated relict forests stretching from southern Somalia into Mozambique, and known for its high level of endemicity, with about 768 described endemic plants and animal species [36]. Many of these forests, including that of Mrima, are kaya, sites associated with religious and cultural rites of the coastal Mijikenda peoples. Recent gazettement of several kayas as National Monuments has brought them under the joint care and supervision of the National Museums of Kenya and the Kenya Forest Service. This and the past cultural importance of kayas have been largely responsible for their conservation in an otherwise vast expanse of farmland.
However, these small forests are under increasing pressure due to the declining influence of traditional customs in the face of westernization and the ever-present demands of a growing human population. Additionally, significant damage of a type specific to Mrima Hill Forest has accrued in the past few years, a result of invasive mining techniques (forest clearing for road building, hole drilling) by a foreign company hoping to extract Niobium, a rare earth mineral [37]. The operating license for this company has now been revoked. Nonetheless, we consider Trissevania mrimaensis and T. heatherae to be vulnerable (they meet criterion ''B'' of Vulnerable (VU) IUCN red list category [38]). Most coastal forest endemics are relict species, and except for a few plant and bird species, all have a very limited distribution even within the network of coastal forests [36].
Mrima Hill Forest is also home to several other interesting insect taxa. Recently two undescribed species of Megalyridae (Hymenoptera), a rarely collected wasp family, were sampled in Mrima (S. Shaw, unpublished data) as was a species of Rhipidioides (Coleoptera: Rhipiphoridae) (Z. Falin, unpublished data), a genus previously known only from Australia. Hopefully the discovery of these and other endemics such as Trissevania mrimaensis and T. heatherae will help build a case for increased vigilance in protecting Mrima and other coastal forest habitats. Figure S1 Print, cut and fold model of the evaniid fore wing (Trissevania anemotis on the left, Evania albofascialis on the right, anterior to the right). Convex wing creases are marked as mountain folds whereas concave ones as valley folds following the origami terminology. The wings should be cut along their margin and folded along the mountain and valley folds. The model demonstrates the wing folding mechanism of Trissevaniini and the complexity of fold and flexion line system of the tribe relative to other Evaniidae. The necessity and possible function of each crease can be demonstrated by comparing the cut outs: 1. wings without folded median flexion line (mfl) and claval flexion line (cfl) lack the ''Z-shape profile'' and bend at the basal region when the blade is moved; 2. Wing locking is restricted at the intersection of transverse and longitudinal wing folds and results an unstable folded position if the prestigmal fold line (pfl) and the posterior radial flexion line (prf) are not folded. Crease first the longitudinal fold line (afl) and after that the transverse fold (sfl+ pfl) for the accurate four plane wing folding! (TIF)