The subfamily Ethopolyinae Chamberlin, 1915 is known to comprise some of the largest lithobiomorphs in the world, with several species reaching 45-50 mm in length. At present, the subfamily includes four more or less well defined genera: Bothropolys Wood, 1862 with around 40 species from North America and East Asia; Archethopolys Chamberlin, 1925 with three species from the southwestern USA, Zygethopolys Chamberlin, 1925 with four species from western Canada and the USA, and Eupolybothrus Verhoeff, 1907 with 23 valid and 15 doubtful species and subspecies assigned to seven subgenera ranging from Southern Europe and North Africa to the Near and Middle East, including the largest Mediterranean islands Corsica, Sardinia, Sicily, Crete and Cyprus (Zapparoli and Edgecombe 2006, Zapparoli and Edgecombe 2011). The genus Eupolybothrus exhibits the highest species diversity in the Italian and Balkan peninsulas (Zapparoli 2003), where a number of cave-dwelling species have restricted distribution ranges. A further 66 species-level taxa proposed in Eupolybothrus are currently considered to be junior synonyms, although their taxonomic status might change in the light of future taxonomic and molecular studies. The exact placement of genus Ethopolys Chamberlin, 1912, with twelve species in two subgenera from western Canada and the USA is uncertain, being treated in contemporary literature as either a synonym of Bothropolys (Zapparoli and Edgecombe 2006, Zapparoli and Edgecombe 2011) or a valid genus (Mercurio 2010).

While some species of Eupolybothrus and the genus itself have been treated recently in several publications (see e.g., Eason 1970, Zapparoli 1984, Zapparoli 1995, Zapparoli 1998, Zapparoli and Edgecombe 2006, Iorio 2008, Stoev et al. 2010), the other three genera, with few exceptions (e.g., Matic 1974, Ma et al. 2008, Ma et al. 2009, Ma 2012) have remained out of the scope of contemporary studies. Nevertheless, it is also far from being fully revised, as a number of problems are still in need of modern scrutiny. These mainly concern: 1) a high number of vaguely described or/and poorly known species and subspecies, mostly from the Balkans and Anatolia, known only from their original description; 2) an outdated subgeneric classification that lacks any phylogenetic framework; and 3) a high number of cryptic taxa in the Eupolybothrus nudicornis (Gervais, 1837), Eupolybothrus litoralis (L. Koch, 1867) and Eupolybothrus tridentinus (Fanzago, 1874) species-groups, as recently revealed by application of DNA barcoding (Porco et al. 2011, Komerički et al. 2012). Further, Stoev et al. (2010) found high interspecific divergence values (20.8% mean value) between two closely related Eupolybothrus species in another barcoding study with mitochondrial Cytochrome C Oxidase subunit I (COI). Two other studies (Edgecombe and Giribet 2003, Spelda et al. 2011) contributed genomic data by analysing DNA barcodes for Eupolybothrus fasciatus and Eupolybothrus tridentinus from Italy and Germany, respectively. The present study is part of an ongoing revision of the subfamily Ethopolyinae (Stoev et al. 2010, Porco et al. 2011, Komerički et al. 2012).

The present study is based on eight specimens of Eupolybothrus cavernicolus Komerički & Stoev sp. n. belonging to the Croatian Biospeleological Society (CBSS), the National Museum of Natural History, Sofia (NMNHS) and the Natural History Museum of Denmark (ZMUC). The specimens were preserved in ethanol (70 or 96%) or RNAlater (Qiagen, USA). The morphological study of the new species was performed at NMNHS and CBSSS with a Zeiss microscope. For scanning electron microscopy (performed at ZMUC), parts of the specimens were cleaned by ultrasonification, transferred to 96% ethanol and then to acetone, air-dried, mounted on adhesive electrical tape attached to aluminium stubs, coated with platinum/palladium and studied in a JEOL JSM-6335F scanning electron microscope. Images were edited in Adobe Lightroom 4.3 and Adobe Photoshop CS 5. All morphological images have been deposited in Morphbank. Terminology for external anatomy follows Bonato et al. (2010).

DNA barcode sequencing

DNA extraction was conducted in the the Canadian Centre for DNA Barcoding, Guelph on complete animals or part of the leg of the specimens preserved in 96% ethanol. Standard protocols of the Canadian Centre for DNA Barcoding were used for both DNA extraction and amplification. All specimen data are stored in the Barcode of Life Data System (BOLD) online database and are available also in the dataset DS-EUPCAV (http://dx.doi.org/10.5883/DS-EUPCAV), where they are linked to the respective Barcode Index Numbers clusters. This dataset contains sequences from ten species: Eupolybothrus cavernicolus Komerički & Stoev sp. n., Eupolybothrus leostygis (Verhoeff, 1899), Eupolybothrus obrovensis (Verhoeff, 1930), Eupolybothrus grossipes (CL Koch, 1847), Eupolybothrus gloriastygis (Absolon, 1916), Eupolybothrus nudicornis, Eupolybothrus litoralis, Eupolybothrus kahfi, Eupolybothrus transsylvanicus (Latzel, 1882) and Eupolybothrus tridentinus. In addition, all sequences were registered in GenBank (accession numbers KF715038-KF715064, HM065042-HM065044, HQ941581-HQ941585, JN269950, JN269951, JQ350447, JQ350449), one sequence of Eupolybothrus fasciatus (Newport, 1845) was recovered from GenBank (accession number AY214420). Two sequences from two Lithobius species were included as outgroups: Lithobius austriacus (Verhoeff, 1937) (MYFAB442-11) and Lithobius crassipes L. Koch, 1962 (MYFAB443-11). The final dataset comprises 39 sequences. Molecular delimitation of species was achieved by the implementation of the Automatic Barcoding Gap Discovery (ABGD) procedure as described in Puillandre et al. (2012) and by the reversed Statistical Parsimony (SP) approach as suggested by Hart and Sunday (2007). A Neighbor-Joining (NJ) tree was built for visualization.

For the ABGD method, we tested various model combinations to cross-check the obtained results: relative gap with (X) ranging from 0.05 to 1.5, minimal intraspecific distance (Pmin) of 0.001 and maximal intraspecific distance (Pmax) ranging from 0.02 to 0.11. Pmin and Pmax refer to the genetic distance area where the barcoding gap should be detected, whereas X defines the width of the gap. Distance calculation was based on the Kimura-2-parameter model and a transition/transversion ratio of 2.0. The method was performed in 100 steps. Statistical Parsimony networks for the delineation of species were reconstructed on the basis of 95% statistical confidence (i.e. connection probability) using the program TCS 1.21 (Clement et al. 2000). The NJ-topology was calculated in MEGA 5.0 (Tamura et al. 2011) using the K2P-model under the pairwise-deletion option and 1000 bootstrap replicates. Intra- and interspecific genetic K2P-distances were calculated in MEGA 5.0 as well.

Transcriptome sequencing

One entire adult male specimen of Eupolybothrus cavernicolus Komerički & Stoev sp. n. was crushed and preserved in liquid RNAlater (Qiagen, USA) immediately after being captured. To extract total RNA, TRIzol reagent (Invitrogen, USA) was used according to the manufacturer’s instructions. Messenger RNA (mRNA) was isolated from total RNA using a Dynabeads mRNA Purification Kit (Invitrogen, USA). The mRNA was fragmented and transcribed into first-strand cDNA using SuperScript™II Reverse Transcriptase (Invitrogen, USA) and N6 primer (IDT). RNase H (Invitrogen, USA) and DNA polymerase I (Invitrogen, Shanghai China; New England BioLabs) were subsequently applied to synthesize the second-strand of the cDNA. The double-stranded cDNA then underwent end-repair, a single ‘A’ base addition, adapter ligation, and size selection, indexed and PCR amplified to construct a library. The extracted cDNA was utilised for library construction with an insert size of 250 bp. Finally, the library was sequenced on the Illumina HiSeq2000 sequencing platform (Illumina, Inc., San Diego, California, USA) at BGI-Shenzhen using a 150bp pair-end strategy to generate a total of 2.5 Gb raw reads. Illumina HCS1.5.15.1 + RTA1.13.48.0 were applied to generate a “bcl” file which was then downloaded to local computers. Secondly, the “bcl” file was converted to qseq format using BclConverter-1.9.0-11-03-08. Finally, we separated individual sample data from multiplexed machine runs based on the specific barcode primer sequences, and converted the file format to fastq.

The micro-CT scanning of one adult female specimen was performed at Bruker microCT, Kontich, Belgium, using a SkyScan 1172 system with the following settings: 40kV, 0.43° rotation step, acquiring 839 projection images from 360° with a pixel size of 8µm. Prior to scanning, the sample was dehydrated in graded ethanol: 50%, 70%, 90%, 100%, for 2 hours in total, and then transferred to HMDS (hexamethyldisilasane) for 2 hours, and air dried. Reconstruction was done with the SkyScan software NRecon, using a modified Feldkamp algorithm, and adjusting for beam hardening and applying ring artefact correction resulting in 3865 cross sections in. bmp format, with image size 2000x2000 pixels. The video of 3D volume renderings was created with CTVox, using the flight recorder function, and saved as an AVI (Audio Video Interface) file. The obtained data were processed through a transfer function where the different voxels with different grey value were (or weren't) made opaque and where the color was assigned to a certain grey value. The image stack is stored in GigaDB (Stoev et al. 2013) under a Creative Commons CC0 public domain waiver. The only software used was CTVox, a viewing software, not analysis software (although you could argue that viewing the images is also a way of analyzing them).

T

– Tergite

The ABGD approach clustered the 37 Eupolybothrus specimens into 12 groups (Fig. 19). Ten of them agreed with morphospecies designations: Eupolybothrus grossipes, Eupolybothrus leostygis, Eupolybothrus kahfi, Eupolybothrus litoralis, Eupolybothrus nudicornis, Eupolybothrus fasciatus, Eupolybothrus transsylvanicus, Eupolybothrus obrovensis, Eupolybothrus gloriastygis and Eupolybothrus cavernicolus Komerički & Stoev sp. n. The remaining two genetic clusters are each formed by a specimen of the morphospecies Eupolybothrus tridentinus from Germany. The reversed SP networks support most of the ABGD results and morphospecies assignments, but split both Eupolybothrus leostygis and Eupolybothrus tridentinus in two and Eupolybothrus nudicornis in three clusters, respectively. We follow a conservative approach here and refer to the ABGD results, which largely correspond to morphology, Eupolybothrus tridentinus being the only exception which could suggest a case of cryptic diversity and will require further investigation. All delineated groups have a bootstrap support of 100 in the NJ-tree topology. The interspecific K2P genetic distance ranges from 10.7 % for the species pair Eupolybothrus tridentinus (GER1) – Eupolybothrus tridentinus (GER2) to 24.5 % for Eupolybothrus grossipes – Eupolybothrus cavernicolus Komerički & Stoev sp. n. (Table 3). Intraspecific K2P genetic distance is maximal for Eupolybothrus nudicornis (7.2 %) and minimal for individuals of Eupolybothrus obrovensis and Eupolybothrus gloriastygis and the species Eupolybothrus cavernicolus Komerički & Stoev sp. n. (0.0 %) (Fig. 19).

The raw data was first filtered by removing inadequate reads with: 1) adapter contamination; 2) ≥10 Ns; 3) ≥50 base pairs of low quality (quality value <65). The resulting 2 Gb of clean data were processed into subsequent assemblies using SOAPdenovo_trans (Xie et al. 2013) under default parameters. The abundance information was provided directly by SOAPdenovo_trans, and played no roles in the subsequent analysis steps. A total of 67,785 scaffolds were produced with an average length of 812 bp and N50 of 1,448 bp [see GigaDB (Stoev et al. 2013)]. Subsequent annotation was conducted by tracing homologs against currently available databases, including Nr, SwissProt and COG. Using this method, 22,866 scaffolds were functionally annotated (Fig. 20a, b, c). Annotated genes were then translated to peptide sequences via CDS prediction according to their blast results using GeneWise (Birney et al. 2004) (see GigaDB (Stoev et al. 2013)). Using orthoDB (http://cegg.unige.ch/orthodb6) (Waterhouse et al. 2012), 2,188 one to one orthologs were filtered out from four selected arthropod genomes: Drosophila melanogaster Meigen, 1830, Daphnia pulex (Linnaeus, 1758), Ixodes scapularis Say, 1821 and Strigamia maritima (Leach, 1817). HaMstR (Ebersberger et al. 2009) was applied to search corresponding orthologous genes in our transcriptome data, delivering 1,668 predicted orthologs of both nucleotide and protein sequences (see GigaDB (Stoev et al. 2013)).

According to the division of the subgenera of Eupolybothrus of Jeekel (1967), Eupolybothrus cavernicolus Komerički & Stoev sp. n. falls into subgenus Schizopolybothrus Verhoeff, 1934, characterized by the presence of triangular projections on tergites 9, 11, 13, a VCm spine on leg 15, one or more VCa spines and a single claw on the pretarsus of leg 15. The same author further distinguishes three species groups in the subgenus based on the morphology of male gonopods and presence/absence of modifications on leg 15:

Jeekel’s division of the genus (Jeekel 1967) is quite artificial and does not reflect real evolutionary relationships as it is merely based on a few morphological traits. Some species were certainly misplaced in these groupings, as for example Eupolybothrus excellens, of which, males show noticeable modifications on leg 15 (see Fig. 17b). Two other species, Eupolybothrus zeus and Eupolybothrus sissii, were even excluded from Schizopolybothrus (cf. Zapparoli 1994, Zapparoli 2002). Showing a prominent prefemoral knob on male leg 15 and having relatively short gonopods, Eupolybothrus cavernicolus Komerički & Stoev sp. n. unquestionably belongs in Group I, as defined by Jeekel (1967). The new species can be readily distinguished from other members of Eupolybothrus (Schizopolybothrus) by the presence of a large proximal knob surmounted by a characteristic cluster of setae, and distal setose protuberance of male prefemur 15. In addition, the species presents a different arrangement of spiniform setae on the intermediate tergite.

The new generation imaging technologies, such as magnetic resonance imaging (MRI) and micro-computed tomography (micro-CT) are opening new horizons in biology (Mietchen et al. 2008, Ziegler et al. 2008). Micro-CT is becoming widely used in comparative, developmental and functional biology (see e.g., Metscher 2009a, Metscher 2009b, Wojcieszek et al. 2012), paleontology (Błażejowski et al. 2011, Edgecombe et al. 2012), molecular biology (Metscher and Müller 2011) and taxonomy (Faulwetter et al. 2013, Michalik et al. 2013). By employing micro-CT scans in taxonomy, important morphological and anatomical characters can be examined in their natural position without damage to the original specimen. This allows researchers to re-assess character shape and functionality or even discover new diagnostic characters (Ziegler et al. 2010, Zimmermann et al. 2011, Faulwetter et al. 2013). To make type material continuously and simultaneously available to taxonomists and to improve access to high-quality morphological data, Faulwetter et al. (2013) suggested the creation of high-resolution virtual morphological and anatomical data libraries allowing reconstruction and interactive manipulation of type specimens, the so-called ‘cybertypes’.

The ‘cybertype’ notion is herewith tested for the first time with the newly described taxon (Fig. 21), for which a rich image library has been created to allow its subsequent recognition, virtual manipulation and reuse. This image library, from which the 3D model is created, has been deposited in the GigaScience database, GigaDB, as a zip and a gzipped tar archive containing BMP images (Stoev et al. 2013). The 3D model was converted into an AVI file, using the flight recorder of CTVox, and disseminated, along with the video of the living specimen (Fig. 22) through YouTube. According to Faulwetter et al. (2013), a ‘cybertype’ should be linked to the original type material and be retrievable and freely accessible. We comply with these requirements by a) including a set of Darwin Core files along with the deposited volumetric data which describe the attributes and deposition of the physical type material and b) using a CCZero license and rich metadata to make the "cybertype" retrievable and reusable. Furthermore, through the same set of Darwin Core files, the morphological data are also linked to the transcriptomic data at GigaDB, effectively extending the ‘cybertype’ concept and providing direct links to other data describing type material of the same species.

Whereas a lack of reference genomes in non-model organisms has hampered genetic and phylogenomic studies, transcriptomes may present a time and cost-effective substitute to whole genome sequencing for these types of studies and an efficient way to produce massive amounts of gene sequence data. While transcriptomic studies of centipede species, e.g. Alipes grandidieri Lucas, 1864 (Chilopoda; Scolopendromorpha), exist in the literature (Riesgo et al. 2012), centipede genome data in public and accessible repositories are still scarce and difficult to find. To address this deficiency, and to produce a model of an accessible resource for the community, all of the transcriptomic data have been made available under the highest metadata standards, both in relevant community specific databases (raw data in the SRA [SRA project accession: ERP003841] and transcriptomic in ArrayExpress [accession E-MTAB-1859]), as well as GigaDB (Stoev et al. 2013). GigaDB collects together all of the genomic and morphological data, and utilises the large computing infrastructure of the BGI and Aspera data transfer capabilities, able to host and deliver much larger and heterogeneous datasets than other repositories (Sneddon et al. 2012). Datasets are also issued with DOIs, which are discoverable through the DataCite metadata search engine and Thomson Reuters Data Citation Index, and can be integrated into a publication or independently cited.

In addition to making data publicly available, it is crucial to provide rich metadata to enable data interoperability and reuse. As there is only one transcriptome available, it is not possible to include additional ‘factor’ information. However, by including sequence reads, experimental design, protocols and processed data we were able to produce the maximum amount of (4*) MINSEQE compliant metadata (Brazma 2009). To maximise its interoperability, the metadata are also available from GigaDB in ISA-TAB format (Sansone et al. 2012).

For volumetric data created by techniques such as micro-CT and micro-MRI, no community standards exist yet. The DICOM standard (Digital Imaging and Communications in Medicine, http://dicom.nema.org/) used by the medical community is not tailored for taxonomic purposes, thus its usefulness for this research field still has to be investigated (Faulwetter et al. 2013). However, even in the absence of widely accepted standards, we provide rich metadata for the micro-CT data, based on the metadata descriptors at Morphosource (http://morphosource.org). The same set of descriptors has been used by GigaDB, where we also applied the ISA-TAB format in order to ensure re-usability and interoperability of the data (Sansone et al. 2012), describing all parameters and settings used to create the data. The data package of micro-CT deposited at GigaDB thus contain:

In combination with the Darwin Core files describing the specimen data, we thus fully annotate and document the ‘cybertype’ of Eupolybothrus cavernicolus Komerički & Stoev sp. n. The generation of large molecular and morphological data pools that originate from type specimens increases the applicability of taxonomic data in other scientific disciplines such as comparative morphology, evolutionary biology, medicine, ecology. The new holistic approach raises important questions and shows up new directions for developments of biodiversity data management about the lack of mechanisms for cross-linking molecular and morphological data and global metadata standards for micro-CT and transcriptomic data, as well as absence of reliable data repositories for micro-CT image libraries.

Also, as a pilot project, we annotate all currently valid Eupolybothrus (Schizopolybothrus) species with their original descriptions that were extracted from the original publications through applying optical character recognition (OCR) and additionally tagged by using Golden Gate software (Sautter et al. 2007). All species treatments are deposited at Plazi. This represents part of a more ambitious project aiming at digitization of all species descriptions and important taxonomic treatments of Eupolybothrus species that is currently being carried out in the framework of the pro-iBiosphere project (http://www.pro-ibiosphere.eu).

To create reliable links between the published sequence IDs and BOLD, an online dataset DS-EUPCAV was generated in the BOLD system, through which the respective Barcode Index Numbers (BINs) of the specimens barcoded for this study may be tracked (for the BIN concept see Ratnasingham and Hebert 2013). All COI sequences were registered in GenBank, following a newly launched metadata standard in the GenBank taxonomy database that flags sequences of type specimens.

This study demonstrates a holistic approach to the description of a new taxon, extending the conventional written description and two-dimensional illustrations with an array of different information types. While this novel approach contributes to the different aspects of the species' identity, its main aim is to provide an integrated approach to handling and publishing large data sets associated with a taxon. The generation of large molecular and morphological data pools that originate from type specimens increases the applicability of taxonomic data in other scientific disciplines such as comparative morphology, evolutionary biology, medicine, ecology, and others.

The concept of a “cybertype” is discussed in the study, but at the same time new questions arise, pertaining to the definition of such a “cybertype” and they will have to be addressed by the taxonomy community. Several different kinds of data belonging to the “cybertype” concept are treated in this study, from free text to sequence data, and from images to volumetric data. Questions have to be addressed such as whether a cybertype should only be restricted to morphological data, what data can be used to constitute a cybertype and whether a cybertype can be composite (i.e. consisting of several data types) or even distributed (different parts of the data residing on different physical servers). Further problems to be addressed are the lack of appropriate mechanisms for cross-linking molecular and morphological data, as well as the absence of global metadata standards and reliable data repositories for micro-CT image libraries. The metadata descriptors for micro-CT files used by the Morphosource and GigaDB repositories are a good starting point for that, as is the use of ISA-TAB to integrate everything together. Whatever the answers to these questions, there is one mandatory requirement for data that we can already identify: discoverability and accessibility.

With complex taxon descriptions such as the present one, we are entering new dimensions of data volumes that have to be managed properly to realise their true value. The deposition of large data pools in appropriate repositories is not yet straightforward, and such initiatives have started to emerge only recently. It is our task to ensure from the beginning that they do not develop into isolated data worlds but that they support community standards, describing the datasets in a way that they can be retrieved and cross-linked. Currently, even in modern taxon descriptions, different pieces of data are only linked through a central locus: the published article. In a future, data-centric world of taxonomy, articles published through next generation journal workflows will become an even more important node in a linked network of data elements describing the taxon. These data elements have to be defined and made accessible through persistent identifiers – not unlike the traditional practice for physical specimens that are accessible through their museum accession number. In combination with rich metadata standards, taxonomy will thus open itself up to the semantic Web with its possibilities for intelligent, complex queries.

In this study, we have taken a first step towards that direction. All data have been deposited in publicly accessible repositories, such as GigaDB, NCBI, BOLD, Morphbank and Morphosource, and the respective open licenses used ensure their accessibility and re-usability. GigaDB in this example provides direct links between the genomic and micro-CT data, through a Darwin Core CSV dataset describing the type specimens, as well as capturing all of the metadata in the interoperable ISA-TAB format. Molecular data and images are annotated with rich metadata to ensure discoverability and reuse. Techniques such as micro-CT are, however, still in their infancy, and no standardised metadata schemas exist yet – a gap that needs urgently to be addressed by the community if we are to avoid a proliferation of isolated datasets.

Taxonomy is at a turning point in its history. New technologies allow for creation of new types of information at high speed and in gigantic volumes, but without clear rules for communication standards, we will not be able to exploit their full potential. We need to focus our efforts on linking these bits and pieces together, by documenting them, by standardising them and by making them retrievable. If such an infrastructure is in place, unforeseen analytical powers can be unleashed upon these data, creating a revolution in our abilities to understand and model the biosphere.