A revision of Evaniscus (Hymenoptera, Evaniidae) using ontology-based semantic phenotype annotation

Abstract The Neotropical evaniid genus Evaniscus Szépligeti currently includes six species. Two new species are described, Evaniscus lansdownei Mullins, sp. n. from Colombia and Brazil and Evaniscus rafaeli Kawada, sp. n. from Brazil. Evaniscus sulcigenis Roman, syn. n., is synonymized under Evaniscus rufithorax Enderlein. An identification key to species of Evaniscus is provided. Thirty-five parsimony informative morphological characters are analyzed for six ingroup and four outgroup taxa. A topology resulting in a monophyletic Evaniscus is presented with Evaniscus tibialis and Evaniscus rafaeli as sister to the remaining Evaniscus species. The Hymenoptera Anatomy Ontology and other relevant biomedical ontologies are employed to create semantic phenotype statements in Entity-Quality (EQ) format for species descriptions. This approach is an early effort to formalize species descriptions and to make descriptive data available to other domains.

introduction Deans et al. (2012) recently opined that phenotype data collected by taxonomists, i.e., the natural language character statements found in diagnoses and descriptions, could, if presented in a broadly accessible, searchable manner, be used to address big questions in biology. Other components of the taxonomic process -names, specimens, DNA sequences, images, etc. -are already digitized and therefore contribute to discoveries in other contexts (Patterson et al. 2010;Padial et al. 2010). Here we offer a real example of natural language descriptions that are annotated with semantic phenotype statements, modeled after the EQ representation referred to by Deans et al. (2012) (see also Mikó and Deans 2009;Mungall et al. 2010;Mabee et al. 2007;Patterson et al. 2010;Balhoff et al. in prep), expressed in Web Ontology Language (OWL) and therefore ready for the Semantic Web. A formal model has been developed (Balhoff et al. 2011), and its advantages and limitations are discussed by Balhoff et al. (in prep).
Our taxonomic subject is the ensign wasp genus, Evaniscus (Hymenoptera: Evaniidae). Ensign wasps develop as solitary predators within cockroach egg cases (Dictyoptera: Blattodea). The family is common across the world except in polar regions, and species diversity is highest in the Neotropics (Deans 2005). There are 21 extant genera and 580 described species of Evaniidae in the world (Deans 2005;Kawada 2012); ten genera of fossil evaniids are also known (Deans 2005;Peñalver et al. 2010). There is a paucity of prey records for Evaniidae in general, and none is known for Evaniscus (Deans 2005).
Evaniscus Szépligeti, 1903 is a relatively small genus of New World ensign wasps with four previously known, rarely collected species (Deans and Huben 2003). The genus belongs to a New World clade that exhibits reduced wing venation, along with Semaeomyia, Hyptia, Decevania and Rothevania (Deans et al. 2006). Originally described by Szépligeti in 1903 for an unusual species from Venezuela, Evaniscus tibialis, the genus has not been previously revised. Only two other New World evaniid lineages, Alobevania and Decevania, have undergone revision recently (Deans and Kawada 2008, Kawada and Azevedo 2007, Kawada 2011. Deans and Huben (2003) diagnosed Evaniscus by the following characters: "RS+M vein missing in the fore wing, coxae evenly-spaced, head hemispherical in lateral view, antennae 13-segmented and arising mid-height on the head, and metasoma ovoid". In addition to the type species, three other species are currently included within Evaniscus: E. marginatus (Cameron, 1887), E. rufithorax Enderlein, 1905, andE. sulcigenis Roman, 1917. Two hundred-fifty years of ensign wasp taxonomy has thus far yielded a corpus of species descriptions that lack utility beyond the realm of descriptive taxonomy (and even very little utility within this domain, as descriptions are usually short and lexically cryptic). For almost all Evaniidae, identification of species must be done by direct comparison with type specimens since there is a shortage of useable species descriptions or identification keys.
The three primary goals of this paper are to 1) provide diagnostic characters for the identification of Evaniscus species as well as a phylogeny, annotated images, and distribution records for species (i.e., a robust taxonomic revision), 2) apply new descrip-Phylogenetics. Outgroups from four different genera were chosen from the closest known relatives based on estimated evaniid relationships using 16S and 28S ribosomal RNA (rRNA) data in MrBayes (Deans et al. 2006) and morphological similarity to Evaniscus (Deans and Huben 2003), including a more distantly related species, Alobevania gattiae Kawada and Deans 2008). To discover new characters of phylogenetic importance, we examined as many individuals of each species as possible and extracted homologous characters across species. A total of 31 parsimony-informative morphological characters were analyzed in this study. A parsimony analysis was performed with an exhaustive search in PAUP* version 4, beta 10 (Swofford 2002). The root was placed at A. gattiae. Jackknife and bootstrap values were calculated using default settings with 1000 pseudoreplications.
Data management. Morphological characters, taxonomic concepts, descriptive language, electronic keys, and georeferenced collecting events were maintained in mx (Yoder et al. 2006). Over twenty researchers currently contribute to the development of the Hymenoptera Anatomy Ontology (Yoder et. al 2010). Phylogenetic datasets, trees and associated metadata, such as specimen information and matrices, were exported from mx as NeXML and are deposited into TreeBASE. Semantic, marked-up phenotype annotations expressed in OWL are deposited in the Dryad Data repository. Mx-generated species pages are provided to the Encyclopedia of Life via XML exports.
Distribution map. Google Maps ® is used to produce distribution maps for each species. Collecting locality data are available on species pages at the Evanioidea Online (http://evanioidea.info/) descriptive web pages and are also shared with EOL.
Images. Specimens were examined using an Olympus SZX16 Research Stereo Microscope (at NCSU) and Leica MZ12.5 (at MZSP). Images for figures were obtained using the Passport Storm Portable Imaging System by Visionary Digital and combined with Combine ZP © (Hadley 2009) or a Leica M205C magnifying glass attached to a Leica DFC 295 video camera with images combined using Leica LAS (Leica Application Suite V3.6.0) Microsystems by Leica (Switzerland) Limited. All images were cropped and brightness and contrast were adjusted in Adobe Photoshop ® CS4 when necessary. Images included within this study are available at Morphbank (http://morphbank.net).
Material examined. Specimens (Appendix C) were borrowed from museums (see Acknowledgments). Nine specimens of E. rufithorax and four specimens of E. marginatus (including the holotype for E. marginatus and three syntypes of E. rufithorax), and two additional specimens of E. tibialis were observed and imaged at the Natural History Museum in London, UK and Museum für Naturkunde, Berlin, Germany, but were not assigned NCSU barcode numbers.

Data resources
The data underpinning the analyses reported in this paper are deposited in the Dryad Data Repository at doi: 10.5061/dryad.2jd88 and at TreeBASE (http://purl.org/phylo/treebase/phylows/study/TB2:S13316). Etymology. The specific epithet honors José Albertino Rafael, a great collector in the Amazon forest and an entomologist at INPA.

Diagnosis. Evaniscus lansdownei
Diagnosis. This species shares the following character states with E. tibialis: posterior margin of the metapectal-propodeal complex ventrally of the propodeal foramen curvature in lateral view: straight; scape length vs compound eye height: scape longer than half compound eye height; mesoscutum shape: wider than long (length of mesoscutum < width of mesoscutum); dorsal region of petiole sculpture: wrinkled. The following character states are present in E. rafaeli but not in E. tibialis: ventral region of occipital carina curvature in lateral view: straight; median notch of transverse pronotal carina presence: present; transverse pronotal carina length: long, extending posterolaterally of epomia; parapsidal signum conspicuousness: inconspicuous.
Description. Head. Head color: orange. Mandible color vs clypeus color: mandible color same as clypeus color. Subantennal carina length: extending dorsally of medial margin of lower face. Preorbital carina length: extending dorsally to the ventral margin of the anterior ocellus. Upper face sculpture: foveate. Malar space length vs. half compound eye height (male): shorter than half compound eye height. Ocellar ocular line length vs. lateral ocellus diameter: as long or longer than lateral ocellus diameter. Posterior ocellar length vs. lateral ocellus diameter: 1.5× as long as the diameter of the lateral ocellus. Ventral region of occipital carina curvature in lateral view: straight. Ventral region of the postoccipital carina shape: raised. Radicle color: red. Scape color: red. Scape length vs compound eye height: scape longer than half compound eye height. Female flagellomere 1-8 shape: distinctly wider than long (length of flagellomere < width of flagellomere). Shape of occiput: as high as wide.
Description. Head. Head color: dorsal half of upper face and vertex color black; ventral half of upper face and lower face color red or yellow; orange. Mandible color vs clypeus color: mandible color same as clypeus color. Subantennal carina length: extending ventrally of medial margin of lower face. Preorbital carina length: extending dorsally to ventral margin of the antennal foramen. Upper face sculpture: foveate. Malar space length vs. half compound eye height (male): shorter than half compound eye height. Ocellar ocular line length vs. lateral ocellus diameter: shorter than lateral ocellus diameter. Posterior ocellar length vs. lateral ocellus diameter: 1.5x as long as the diameter of the lateral ocellus. Ventral region of occipital carina curvature in lateral view: straight. Ventral region of the postoccipital carina shape: not raised. Radicle color: yellow; orange. Scape color: yellow; orange. Scape length vs compound eye height: scape shorter than half compound eye height. Female flagellomere 1-8 shape: distinctly longer than wide (length of flagellomere > width of flagellomere). Shape of occiput: as high as wide. Mesosoma.

Comments.
A lectotype was designated because syntypes were from specimens with different localities and one syntype varies in color; the male specimen chosen as lectotype from Peru is in very good condition and fits the description well. The male paralectotype from Bolivia that is at ZMPA varies in color (only) from the other type specimens and all known other material; the antero-dorsal region of mesosoma color is red in all E. rufithorax specimens, but is black in this type specimen. Diagnosis. Evaniscus tibialis is the largest species of the genus, is the only species that ever has an entirely black body, and can be distinguished from other species by the combination of the following characters: ocellar ocular line length vs lateral ocellus diameter: as long or longer than lateral ocellus diameter; transverse pronotal carina length: short, not extending postero-laterally of epomia; parapsidal signum conspicuousness: conspicuous; posterior propodeal projection shape in lateral view: raised.

Evaniscus tibialis
Description. Head. Head color: black; dorsal half of upper face and vertex color black; ventral half of upper face and lower face color red. Mandible color vs clypeus color: mandible color same as clypeus color. Subantennal carina length: extending dorsally of medial margin of lower face. Preorbital carina length: extending dorsally to the ventral margin of the anterior ocellus. Upper face sculpture: foveate. Malar space length vs half compound eye height (male): as long as or longer than half compound eye height. Ocellar ocular line length vs. lateral ocellus diameter: as long or longer than lateral ocellus diameter. Posterior ocellar length vs. lateral ocellus diameter: 1.5× as long as the diameter of the lateral ocellus. Ventral region of occipital carina curvature in lateral view: curved. Ventral region of the postoccipital carina shape: raised. Radicle color: black; red. Scape color: black; red. Scape length vs compound eye height: scape longer than half compound eye height. Female flagellomere 1-8 shape: distinctly longer than wide (length of flagellomere > width of flagellomere). Shape of occiput: higher than wide.

Discussion
Semantic phenotypes. Through the descriptive process, taxonomists stand to contribute an immense body of knowledge that could be used to address a broad array of questions in many realms of biology (Deans et al. 2012). How might phenotypes be correlated to climate change? Or how might changes in phenotype correspond with the environment? Presently, queries of characters that reference a specific part of the anatomy are already possible (Deans et al. 2012). There are, however, some current limitations in the workflow of semantic phenotype construction, e.g., ontologies often do not have sufficient content, using the software to manually create the statements can be complex, and it may prove challenging for taxonomists to alter their workflow (Deans et al. 2012). Though some of the methods and tools used to build semantic phenotype annotations are in their infancy, semantic phenotypes hold the potential to unlock valuable data within taxonomic descriptions. For example, semantic phenotypes could become more meaningful when mapped across large-scale phylogenies; if we apply semantic phenotype annotations to specimens, phenotype data can be connected to evolutionary history through the organism's phylogenetic relationships. Also, semantic phenotypes help make available millions of unambiguous data to a broad array of scientists (Deans et al. 2012). The descriptive statements within this manuscript represent one of the first efforts in descriptive taxonomy to capture phenotypes using formalized, semantic methods. As a result of employing these new methods, a contemplative, calculated approach was taken in selecting terminology for characters and character states. The natural language descriptions in this manuscript were originally written with controlled vocabulary using terms present in ontologies that made the process of translation into semantic phenotypes relatively straightforward; for instance, a phenotype statement was written with the anatomical character followed by the descriptive character state, e.g., "dorsal area of metapectal-propodeal complex foveate". The semantic phenotype statements resulting from these descriptions are more objective and less ambiguous than those frequently found in traditional taxonomic descriptions.
In the original description of E. marginatus, the mesonotum was described as "shining, bearing some large scattered punctures". This description is vague. To make it more explicit, we changed the character name to "distance between depressions vs. diameter of depressions on internotaular area" and the states to "0: greater than (or 1: less than) the diameter of one depression". This is more specific than the presence of some large scattered punctures on the mesonotum, but making a semantic statement from this character was not particularly intuitive (for semantic statement, see Appendix B). For the majority of characters in the descriptions presented here, the process of translation to more meaningful semantic statements was not as complicated.
Semantic phenotypes in these taxonomic descriptions were created in a logical manner by means of extracting anatomical information from an organism-specific ontology, such as the HAO, and pairing this with a quality from a general trait ontology (PATO). The natural language description persists, but a machine-readable interpretation is constructed that can be stored on the Semantic Web, where the valuable phenotypic data can easily be mined by computers and captured for use by taxonomists, biologists, or, essentially, anyone who wants to query the database of descriptions. Taxonomy that includes ontology-based descriptions, such as those presented in this manuscript, avails phenotype data to experts in all domains through bioinformatics applications.
Geographic distribution. In addition to the discovery of two new species, our results expand upon the geographical range of the four previously described species. Subsequently to the original descriptions, E. tibialis has been collected in northeast Guyana, northern Brazil, and Trinidad. The range of E. rufithorax has been expanded into north-central western Brazil, northeastern and southern Ecuador, and southern Colombia. Evaniscus marginatus was described from Guatemala, and has now been collected in Costa Rica, Mexico, Brazil and Ecuador.
The majority of described evaniid species to date are from tropical regions (Deans and Huben 2003), which is consistent with the primarily tropical and subtropical distribution of cockroaches (Vélez 2008). In Colombia alone, there are 133 species of cockroaches present, 10 of which are found in Amazonas and five of which are nonblaberids (Vélez 2008). Since the holotype of E. lansdownei was collected in Amazonas, one of these cockroach species could potentially represent the host of E. lansdownei.
Interestingly, four of the E. tibialis specimens in this study were collected at the entrance to Tamana Caves, Trinidad. Blaberus posticus dominates the cockroach fauna in this area, but because this is a blaberid and retains the ootheca within the abdominal wall, as do all other species dwelling in the caves, it is highly unlikely that they could be the host of E. tibialis (Darlington 1970).
Color variation. Some insects exhibit varying levels of intraspecific polychromatism or heterochromatism. For example, Berniker and Weirauch (2012) identified species of the reduviid genus Apiomerus that exhibit intraspecific polychromatism; some species showed discrete color morphs while others showed only limited polychromatism in the pronotum and the corium. Some individuals of Apiomerus californicus collected along the Sierra Nevada mountain range in California showed an increased red pigmentation, which suggested a possible correlation between color variation and elevation. In the ichneumonid Aphidius smithi, adult wasps that were reared at different temperatures during the mid to late pupal stage displayed constant differences in integumental coloration, especially in the face, thorax and petiole; when wasps were reared at higher tempteratures, the face and mesothorax were orange, but when reared at lower temperatures, these parts in adults were black (Shu-Sheng and Carver 1982).
In Evaniscus, heterochromatism is limited to the head and mesosoma in E. rufithorax. The most distinct color morph of E. rufithorax has the dorsal half of the upper face and vertex black while the ventral half of the upper face and lower face is red or yellow. However, a few males and females have an entirely red head. A similar color pattern can be applied to the mesosoma in this species; the majority of specimens have the postero-ventral region of the mesosoma black and the antero-dorsal region red, but some males and females have an entirely red mesosoma. The same variation in head color pattern applies to some specimens of E. tibialis, including the holotype.
With the limited availability of specimens in this study, it is difficult to determine if there is any correlation with color variation and biogeography in E. rufithorax, or if rearing temperature plays a role in adult coloration. The females with entirely red heads were collected in Ecuador with the exception of one specimen collected in Brazil. This heterochromatism is not female-limited, however, since males of the species do exhibit the same red color morph. There is geographical overlap between the red morphs and the more common specimens with the dorsal half of the upper face and vertex black and the postero-ventral region of the mesosoma black and the antero-dorsal region red. In addition to Evaniscus, intraspecific polychromatism has also been observed in Hyptia thoracica specimens collected in the Sandhills Gamelands in North Carolina, with up to 8 different color morphs all present in the same area (personal observation, PLM).
Systematics. In the morphological analysis, Evansicus was well supported as a monophyletic lineage (bootstrap support=90, jackknife support=89). Evaniscus rafaeli was placed as sister to E. tibialis, and these two share several synapomorphic character states: posterior margin of the metapectal-propodeal complex ventrally of the propodeal foramen curvature in lateral view: straight; scape length vs compound eye height: scape longer than half compound eye height; mesoscutum shape: wider than long (length of mesoscutum < width of mesoscutum); distance between depressions vs. diameter of depressions on internotaular area: less than the diameter of one depression; dorsal region of petiole sculpture: wrinkled; metafemur length vs. metatibia length: metafemur longer than metatibia. While these two species share several characteristics, the support for their monophyly is moderate (bootstrap=72, jackknife=67). Nearly 25% of parsimony informative characters were apomorphies for E. tibialis. Despite the morphological analysis placing E. tibialis as sister to E. rafaeli, to the naked eye this species looks distinctly different from other Evaniscus species, and our understanding of the placement of E. tibialis would benefit from future molecular analyses.
The clade of Evaniscus comprising E. lansdownei, E. marginatus, E. rufithorax and E. sulcigenis was moderately supported (bootstrap support=76, jackknife support=72), but all species share several derived character states: posterior margin of the metapectal-propodeal complex ventrally of the propodeal foramen curvature in lateral view: curved; scape length vs compound eye height: scape shorter than half compound eye height; mesoscutum shape: as long as wide (length of mesoscutum > width of mesoscutum); distance between depressions vs. diameter of depressions on internotaular area: greater than the diameter of one depression; dorsal region of petiole sculpture: foveate.
In addition to those published by Deans and Huben 2003, 18 characters that all Evaniscus species share have been documented in this manuscript. Previous to this study, very few females of Evaniscus had been observed, and it was thought that the ovipositor is short and completely hidden within the metasoma (Deans and Huben 2003). However, we have seen the ovipositor in several female specimens, and it extends to the tip of the metasoma; if the ovipositor was concealed when the insect was preserved, the female metasoma looks identical to that of the male.
Based on the lack of morphological variation and the results of the present analysis, we consider Evaniscus sulcigenis to be a junior synonym of E. rufithorax. Both species share all external morphological character states examined in this study, except for some variable color patterns. Further comprehensive morphological studies in addition to molecular studies are required to fully confirm this hypothesis; however, since the only available specimen of E. sulcigenis is the holotype specimen, these future studies may not be practical.
Sexual dimorphism. Evaniidae are usually sexually dimorphic in their antennal morphology, body coloration, facial sculpturing, and/or metasomal morphology (Deans and Huben 2003;Deans and Kawada 2008). For example, females of Decevania have a distinctly sculptured head, flattened, small eyes, the antennal segments are enlarged progressively from the fourth flagellomere, and the posterior region of the metasoma is expanded dorso-ventrally with the ovipositor usually concealed; males usually have larger bulging eyes, antennal segments all the same diameter, and the posterior region of the metasoma constricted dorsoventrally with genitalia protracted (depending upon preservation) (Kawada and Azevedo 2007;Kawada 2011). In addition, color pattern variation in male and female specimens of many species have also been observed (personal observation). All Evaniscus species are identical in coloration, except for those of E. rufithorax (as discussed above).
Many hymenopterans have sexually dimorphic antennae (Deans and Kawada 2008;Gauld and Fitton 1987;Wharton 1980;Onagboloa et al. 2009). Antennae of most species of Evaniidae are also sexually dimorphic (Deans and Kawada 2008;Kawada and Azevedo 2007;Kawada 2011). In Evaniscus, females have a ventral sensillar patch on flagellomeres 6-12 or 8-12, whereas males do not, and many females also have flagellomeres that are distinctly wider than long, where male flagellomeres are as long as wide or longer than wide. In E. marginatus and E. tibialis, for example, flagellomeres 1-8 are distinctly wider than long in the female, but not in the male. Evaniscus rufithorax is likely unique among Evaniscus species in that the antennal flagellomeres do not exhibit the flagellomere shape sexual dimorphism; however, a ventral sensillar patch on flagellomeres 6-12 is present in females but not in males. We cannot be certain this is the only species that exhibits this character state, as the male is not yet known for E. rafaeli and females are still unknown for E. lansdownei and E. sulcigenis.
Another difference between male and female Evaniidae specimens is the connection between the petiole and the first abdominal segment. This difference can be observed in a longitudinal section through the junction between the two sclerites in the petiole. In males, the foramen of the petiole receives the connection to the junction of the first abdominal tergite and sternite. The first abdominal tergite has a folding anterior edge along with the first abdominal sternite. These are generally divided into two sclerites: a lower tubular sclerite and another larger sclerite, which covers a large area of the first abdominal sternum (Fig. 32). In the females, the first abdominal tergite and sternite are expanded to cover the distal region of the petiole. Internally, the anterior portion of the tergite and sternite are curved to the inner wall of the petiole and connected to it by a thin membrane (Fig. 33).

Appendix B
Character descriptions in Manchester syntax.
"Dorsal area of the