Taxonomy is arguably man’s oldest profession and an important fundamental discipline for many other biological fields (Hedgpeth 1961, Wilson 2004). It helps us to inventory the biodiversity on earth by naming each classified group of organisms that shares certain attributes based on, for example, genetic and morphological evidence. However, the taxa names are not necessarily maintained in perpetuity, as classifications can be changed when more specimens and novel attributes are examined and compared. In fact, the continuous improvement of the existing classification scheme of any taxon is a fundamental characteristic in taxonomic science. As a result, a taxon name is a summary of hundreds of years’ of taxonomists’ attempts to classify biodiversity.

Despite this long history, only a small fraction of biodiversity has been named (Mora et al. 2011). Certainly, more resources (money, time and taxonomists) are needed to describe the remaining unknown biodiversity. For example, Carbayo and Marques (2011) estimated that over two hundred billion US dollars are needed to describe the remaining ca. 5.5 million undescribed species. They indicate that half of the cost is to be invested on training taxonomists and project budgets. Although it is not being discussed in Carbayo and Marques (2011), it is not hard to imaging that a significant proportion of the amount were spent on travelling to various museums for studying type specimens and other samples, compiling literature and maintaining all this information. Even if there were no financial constraint, given the current pace in describing species (ca. 18000 species per year, see Zhang 2008), about 400 years are still needed to describe all the species, which translates into 10 generations of taxonomists (Padial et al. 2010).

All organisms could eventually be scientifically described, but there are several pressing contemporary issues that could affect the advancement of taxonomy as a science. Firstly, the population of taxonomists is declining (Wägele et al. 2011), which will directly reduce the pace of species discovery and capacity building for future taxonomists. Secondly, the majority of taxonomic resources (e.g. type materials, literature, expertise) are centralised in developed countries, whereas the remaining undescribed species are concentrated in developing countries (Rodrigues et al. 2010). Neither taxonomists from developed countries nor taxonomists from developing countries can work effectively without having both biodiversity and taxonomic resources readily available to be integrated. Thirdly, many unnamed species are likely to go extinct under the current rates of habitat destruction in biodiversity hotspots (Giam et al. 2011). Thus, taxonomists have to find ways to be able to work more effectively and accelerate the pace of species discovery.

Several proposals to improve the practice of taxonomy have been published (e.g. Webb et al. 2010, Wägele et al. 2011, Miller et al. 2012). One of these suggests that taxonomists embrace the information and internet technologies in their routine workflow. This is particularly relevant because taxonomy is a data-rich science and taxonomists have to deal with all past taxonomic works, in terms of literature, collection data, genetic data, and ecology, that have been accumulated over hundreds of years. Thus, computer and internet technologies have become important tools to manage all these taxonomic data, especially regarding data storage, dissemination, and retrieval.

Over the past few years, many web-based databases have been developed to accommodate almost all kinds of the data that taxonomists deal with. For example, museum specimen data are held in the Global Biodiversity Information Facility (http://www.gbif.org, Edwards 2004), morphology data (images) in Morphbank (http://www.morphbank.net), genetic data in the Barcoding of Life Database (http://www.barcodinglife.com, Ratnasingham and Hebert 2007) and GenBank (http://www.ncbi.nlm.nih.gov/genbank, Benson et al. 1997), nomenclature data in ZooBank (http://zoobank.org, Pyle and Michel 2008), and literature in the Biodiversity Heritage Library (http://www.biodiversitylibrary.org, Gwinn and Rinaldo 2009). Although the need to use these facilities to improve taxonomy has been emphasised (e.g. Maddison et al. 2012, Wheeler et al. 2012, Miller et al. 2013), the majority of taxonomists has yet to join this movement. This hesitance may be caused by taxonomists’ fear for investing extra work to supply the data, and their doubt of the usefulness of these poorly integrated databases.

Recently, tools have become available for managing and integrating various types of taxonomic data by using either top-down approaches (e.g. Encyclopedia of Life http://eol.org, Wilson 2003) or bottom-up approaches (e.g. Lifedesks http://www.lifedesks.org, offered by EOL; Scratchpads http://scratchpads.eu, Smith et al. 2009). The top-down mechanism acts as an automatic data aggregator that harvests and pools semantic-tagged information from different databases. The bottom-up mechanism, on the other hand, acts as quality controller that checks, links and tags the different types of taxonomic data. Obviously, the key factor that determines the success of cybertaxonomy and the main challenge of these integration processes is the quality of the underlying taxonomic data (Parr et al. 2012).

Many data in the databases are outdated (e.g. nomenclature change), incorrect due to the limitation of technology (e.g. automation text extraction, Page 2011) or human error (e.g. misidentification, Yesson et al. 2007), and not linked or semantic-tagged. In fact, taxonomists spend much of the time in their careers to validate, link and tag all the existing and new taxonomic information. Thus, their contribution to the quality control of the data is essential. However, each database has its unique needs, standards, and format for the data and thus it might require taxonomists to do redundant work, such as uploading and key in the same data in different formats as required by the database.

In addition to the possible redundant efforts that need to be spent, it is not very clear how taxonomists as data suppliers would benefit most from such databases. In view of this, we demonstrate a working example of how these existing databases and platforms can be incorporated into revisionary taxonomic study processes, which begin with managing new specimen information, then establishing identity (literature study and examination of specimens in museum collections), and then writing taxonomic treatises (Winston 1999). Our working example consists of a revision of the taxonomy of the genus Plectostoma of Peninsular Malaysia, Indochina and Sumatra (i.e. non-Bornean) by using cybertaxonomy tools during the entire revisionary taxonomy process. With these cybertaxonomy tools, we show that various kinds of taxonomic information can be semantically tagged, linked and integrated in a more user-friendly interface (Penev et al. 2011). In addition, we demonstrate that using this technology and these databases could facilitate the hypothesis-testing nature of taxonomic research that deals with vast amounts of different kinds of information (e.g. Sluys 2013). The validated, linked and tagged data that are generated during the taxonomic revision in turn facilitates forthcoming taxonomy studies that will be conducted by future taxonomists. Finally, all this species information is made readily available and accessible for other users such as ecologists and conservationists, especially for those who have limited budget and are from developing countries.

The study of Opisthostoma sensu stricto and Plectostoma started during the English colonisation in India and North Borneo in the 1860s (Blanford and Blanford 1860; Adam 1865b). All early collections were made by English officers, who later described the species themselves or sent their collections to other malacologists for species identification and publication. Until 1880, there was only one Plectostoma species known–Plectostoma decrespignyi from Labuan, North Borneo. Between the 1880s and 1900s, A. H. Everett, C. Hose and S. Beddome collected more Plectostoma specimens from North Borneo and Sarawak, where they were working as either British colonial administrators or naturalists. Based on these specimens, a total of 20 Plectostoma species were eventually described from Borneo (Godwin-Austen 1889, 1890, Boettger 1893, Smith 1893b, 1894, 1904, 1905a, Fulton 1901). Many of these were originally described under Plectostoma, which was considered a subgenus of Opisthostoma.

After the intensive malacological documentation in Borneo in the late 19th and early 20^th century, more material of Plectostoma was collected in Sarawak and Sabah during explorations which principally had a different purpose, by the geologist G.E. Wilford and his associates in 1960s, by soil scientists K. Auffenberg and D.K. Dorman in the late 1980s, and by the botanist J.J. Vermeulen in the 1980s and early 1990s. All of this material was revised by Vermeulen (1994), under the genus Opisthostoma, which resulted in the description of 27 new Plectostoma species from Borneo. Since the 2000s, Plectostoma has attracted more interest, also regarding its phylogeny and evolution (Schilthuizen 2003, Schilthuizen et al. 2006, Webster et al. 2012) and ecology (Schilthuizen et al. 2003a, 2005).

In addition to the Bornean Plectostoma, two other Plectostoma species were described from Southern Thailand and Southern Vietnam in the 1900s (Sykes 1903, Dautzenberg and Fischer 1905). Between the 1930s and 1960s, M.W.F. Tweedie (then director of the Raffles Museum of Singapore) collected many Plectostoma specimens during explorations of Peninsular Malaysia. These specimens were later described as 13 new Plectostoma species (Tomlin 1938, 1948, van Benthem Jutting 1952, 1961). The research on the taxonomy of non-Bornean Plectostoma ceased in the 1960s with the retirement of Tweedie and the death of van Benthem Jutting. However, ecological investigations of one Plectostoma species–Plectostoma retrovertens–had started in 1960s by A.J. Berry, a professor at University Malaya (Berry 1961, 1962, 1963, 1964, 1966).

No publications on Plectostoma from Peninsula Malaysia appeared between 1966 and 1996, but many Plectostoma specimens were collected by Geoffrey W. H. Davison, mainly from the state of Kelantan, and other localities in Malaysia (e.g. Davison and Kiew 1990). In 1996, Plectostoma klongsangensis was described from Southern Thailand (Panha 1996). After that, four new species were described from Southern Thailand and Peninsula Malaysia (Maassen 2001). One year later, the first Plectostoma from Sumatra–Plectostoma kitteli, was decribed by Maassen (2002).

To sum up, a total of 69 Plectostoma are currently known. The hotspots of Plectostoma diversity are the Malay Peninsula (including the southern part of Thailand) and Borneo, harbouring 19 and 48 species, respectively. In addition, one species is known from Sumatra and another from Southern Vietnam.

The last revision of the non-Bornean Plectostoma species was done by van Benthem Jutting (1952, 1961). It is important to note that she described 10 new species on the basis of shell characters from only 21 samples from 16 locations. However, it is problematic to use conventional shell descriptions in this genus. The striking shell form of Plectostoma has attracted the attention of malacologists, but it also poses a challenge to describe the shell accurately. As mentioned by Benthem-Jutting (1952), “… it is evident that in such irregular shells … the measurements can only be given approximately, and never indicate the real proportion of the shell…” and “…after comparing over and over again did I succeed in checking the points of difference [between species], but even then it remained difficult to bring the true nature of these minor details into adequate words”. To date, Vermeulen’s (1994) approach in describing the Plectostoma shell is the most comprehensive, but it is still difficult to recognise the species from the written description alone.

As mentioned in the taxonomic history section above, additional Plectostoma specimens were collected by Geoffrey W. H. Davison in Peninsular Malaysia in the 1990s. Furthermore, we collected more specimens, including living ones for genetic study, during field trips to Peninsular Malaysian limestone hills between 2010 and 2011. These Plectostoma specimens are valuable to improve the taxonomy of Plectostoma in Peninsular Malaysia. Thus, it is timely to revise the non-Bornean Plectostoma based on these recently collected specimens by re-examining the species hypotheses formed on the basis of shell morphology. In addition, we also update the knowledge of Plectostoma regarding conservation status, distribution, and genetics.

Literature. In addition to traditional searching of, for example, the Zoological Record, we also searched for the terms “Geothauma, Plectostoma and Opisthostoma” in Google Scholar (on 21^st November 2012) and the Biodiversity Heritage Library (on 19^th November 2012). A URL link to the full-text article was provided as listed in bibliography whenever possible (e.g. from http://www.biodiversitylibrary.org). Each of the articles was tagged with the relevant Plectostoma taxon names. All the relevant references were catalogued as individual contents in bibliography of the Opisthostoma Lifedesks pages (http://opisthostoma.lifedesks.org/biblio). All the relevant bibliography metadata of each article were entered and stored according to the standard data entries of BibTeX (http://www.bibtex.org/) as implemented in Lifedesks. Furthermore, the bibliography metadata can be exported as BibTeX formatted file (.bib), which all the tagged metadata can be retrieved and reused. For video tutorials, see Appendix 1(1 and 2).

Nomenclature information and classification. After the relevant literature was identified, the relevant taxon names were extracted and organised by using the classification tool (tree editor) of Lifedesks (for detailed descriptions of the methodology, see: http://help.lifedesks.org/classification/edit). The classification and nomenclature information can be downloaded and saved according to the DarwinCore standard as xml file. For video tutorials, see Appendix 1(3).

Species information. In addition, the original species descriptions and important notes from the literature were imported to individual species pages as quotations (for detailed descriptions of the methodology, see: http://help.lifedesks.org/quickstart). New and unpublished data, together with the information extracted from literature were managed and stored in the relevant chapters under the headings of ‘Overview’, ‘Conservation’, ‘Description’, ‘Ecology and Distribution’, ‘Evolution and Systematics’, and ‘Relevance’ of each species webpage. As all the information was tagged accordingly in the form of xhtml format, the data of each species can be retrieved and reused. For video tutorials, see Appendix 1(4).

Managing museum collections data. As common curation practice, each collection (i.e. one museum lot) consists of a specimen or multiple specimens, which are kept either dry (empty shells) or wet (shells with animals preserved in ethanol) that were collected at a single sampling occasion, for example from a particular location at a particular day/time. In this study, each collection is regarded as a sample. For each sample, there are two categories of information that can be extracted, namely the physical properties and metadata of the samples.

For the physical properties of the samples, we recorded the exact number of shells or (for samples >10 shells) categorised the samples into four categories of sample size: 1) > 10 (10–24 shells); 2) > 25 (25–49 shells); 3) > 50 (50–100 shells); and 4) > 100 (>100 shells). We also obtained shell morphology data as follows. Whenever possible, each registered collection was photographed. The images for each unique collection were imported into Lifedesks as an individual content. Each of the images was then linked with the species name. For video tutorials, see Appendix 1(5).

Metadata consisted of collection reference number, collector information, collecting date, and locality of each sample. This information was published in Lifedesks as image description, except for the collection reference number, which was published as image caption. In addition to presenting collection data in a tabular format (Appendix 2), we also published the collection data in a more interactive manner, which can be used in Google Earth.

Whenever possible, location data of the collections were georeferenced. When the location description in the specimen label or in the publication was not clear, the itinerary of the collectors or expedition which had been published in other types of publication (e.g. maps and reports) was consulted. After the exact limestone hill where the collection had been made was identified, it was verified in Google Earth, after which the latitude and longitude were obtained. Although the coordinates as obtained from Google Earth are of high accuracy, they might be too accurate and lack an uncertainty estimate (Mesibov 2012). Thus, the coordinates that we report here should be interpreted as the location of the limestone hill and not the exact spot where the specimens were collected.

A Keyhole Markup Language (KML) file, which comprises the location, images, and collection reference number of each of the museum collections, was created by using Google Earth Spreadsheet Mapper v3.0 (http://www.google.com/earth/outreach/tutorials/spreadsheet3.html). For the data input for the spreadsheet (Template4), the species name was used for “Folder name”, and collection reference number for “Placemark Name”; the concatenation of species name and collection reference number for “Title”; URL of each collection’s original image in Lifedesks was named “Image URL”; detailed collection data as “Paragraph Text”. After that, the data in Spreadsheet Mapper were converted into a KML file, which allows semantic-tagged collection information to be retrieved and reused. When the KML file is opened in Google Earth, each of the museum collection (specimens) is shown as a single landmark on the virtual earth. For video tutorials, see Appendix 1(6).

BMNH Natural History Museum (previously known as British Museum (Natural History), London, United Kingdom.

BOR BORNEENSIS collection, Institute for Tropical Biology and Conservation, Universiti Malaysia Sabah, Kota Kinabalu, Malaysia.

RMNH Naturalis Biodiversity Center (previously known as Rijksmuseum van Natuurlijke Historie), Leiden, the Netherlands

V Jaap Vermeulen’s private collection, Leiden, the Netherlands.

YSC Chen Yansen’s private collection, Medan, Sumatra, Indonesia.

ZMA Naturalis Biodiversity Center (previously known as Zoological Museum of Amsterdam), Leiden, the Netherlands.

In the routine of conventional taxonomic revision practice, samples are sorted into groups (e.g. morphospecies) based on their morphology and distribution. Then, each group is assigned to an existing species name–when the morphology, distribution and/or other important characteristics fit with the species’ morphological description and distributional range as mentioned in the literature; or a newly designed species name–when the characteristics do not fit to any of the named species. In some cases, different species identities may have been assigned to the same specimen by multiple taxonomists. Thus, for each specimen, the collection data (morphology, distribution, genetics and others) are immutable, but the species name is mutable.

The key of this process is the taxonomists, who gather, integrate, sort and analyse, not only the biological specimens, but diverse and vast amounts of information from hundreds of specimens (Sluys 2013). This task has become more challenging for taxonomists and their successors because of the accumulative nature of taxonomic information. Thus, taxonomists have been using information technology to assist their routine work since the 1980s (Heywood 1974, Maxted 1992). However, the potential of information technology to be used by taxonomists remains underexploited, except for data storage. In fact, in additional to data storage, this technology can improve the efficiency and effectiveness of taxonomists in integrating, sorting, analysing and disseminating the information from the specimens.

As mentioned above, all key information from specimens and literature was digitised and tagged. Here, we integrated different types of information for different processes in taxonomy revision. For the specimen sorting process, we used a KML file to link and present the unique collection number, images and location data for each specimen. Each museum collection was shown on Google Earth as a landmark and these landmarks were sorted into respective species folders. When each landmark was selected (by clicking it), the information of the morphology (as shown in images), location (text description and map), and other relevant information was visible to the user (taxonomist). Species identification can be done in a single platform (i.e. Google Earth), where the morphological variation within a species or between the specimens across the genus’ geographical range can be examined. Based on this information, the species identity of the landmark (specimen) was determined by either keeping the landmark in the same species folder or moving to the other species folder. Likewise, whenever the coordinate of a specimen location was wrong, the landmark was edited by moving it to the correct location. Whenever necessary, the specimen itself was examined. For video tutorials, see Appendix 1(7).

After all the collection species identifications and location data in the KML file were verified and corrected, the data in the KML file were extracted with a customised Python script to update the data in Spreadsheet mapper and image species link in Lifedesks for all specimens in an automatic manner. This saved much time compared to the traditional laborious method of manually updating the collection database specimen by specimen after sorting and identification. Lastly, the taxonomic content was written in Lifedesks and then exported to the appropriate format and layout for publication. For video tutorials, see Appendix 1(8).

Since the information was stored and tagged digitally, a simple program can be customised to retrieve, integrate and process the data from many different online/offline databases and files. In our case, we used scripting language Python 2.73 (http://www.python.org). Its “urllib” and “re” modules were used to retrieve information from Internet resources and searching for patterns in text. In the same way, additional specimen information from Lifedesks (i.e. image pages) can be integrated into the KML file. Simultaneously, the literature and species information compiled in Lifedesks can be retrieved easily when necessary.

The application of a species concept in Plectostoma has been particular problematic. Nowadays, the most widely accepted species concepts (e.g., the biological species concept; Mayr 1942) include some aspect of genetic and/or reproductive cohesion. Such species concepts are, however, difficult to apply in taxa like Plectostoma, where all species and populations are restricted to isolated limestone outcrops. This island-like (allopatric) distribution pattern suggests very limited gene flow between populations. Furthermore, it would be impractical to verify experimentally “potential interbreeding” for each population, as Plectostoma populations occur in hundreds of different limestone outcrops.

Previous species circumscriptions in this genus have been mainly based on a morphological species concept. In gastropods, this is common practice. However, shell shape or even some shell structures, such as rib intensity, can be genetically variable and/or phenotypically plastic under different environmental conditions (e.g. Berry 1962, Kemp and Bertness 1984). Nevertheless, when the intra- and inter-specific variations in shells are understood (see below “Morphological analysis”), a morphological species concept may be used as one of the guidelines in species delimitation.

Finally, the phylogenetic species concept (sensu Cracraft 1983–“A species is the smallest diagnosable cluster of individual organism within which there is a parental pattern of ancestry and descent”) also cannot be used in the case of Plectostoma species, again because of their populations’ allopatric distribution on isolated limestone outcrops. The deposition age of Peninsular Malaysian limestone ranges from the Ordovician to the late Triassic (ca. 480 mya–200 mya). Though the exact time at which the limestone outcrop was exposed is unknown, van Benthem Jutting (1960) believed this began to happen after the Cretaceous - ca. 140 mya. Given the fact that Plectostoma species have been found on relatively young, Miocene age, limestone (ca. 24 mya–5 mya) in Borneo, it is likely that Plectostoma species will have colonised limestone hills in Peninsular Malaysia soon after the limestone hills were exposed. Hence, populations of the same Plectostoma species on each isolated limestone hill could have been separated for a long time and thus these isolated populations would appear as a several deeply diverged lineages in the phylogenetic tree (Liew TS, unpublished data). In view of this, the blind application of a phylogenetic species concept could inflate the number of species in Plectostoma.

Because of these problems, in this study, we have used elements of different species concepts for the delimitation of Plectostoma species. First we used a set of shell characters for initial morphological species delimitation. These groups were then checked for their distribution ranges. Whenever more than one of these morphological species was found at the same hill, we examined the genetic dissimilarity (in DNA barcode, COI) between these sympatric morphological species. In addition, the intra-specific genetic divergence was examined among several geographically separated populations of the same species (see below “COI Barcoding”). By reciprocal illumination from the morphological, genetic and distribution data, we determined for each species a set of characters that is stable within a species and diagnosable between Plectostoma species. We applied the same approach to the Bornean species (Liew et al. in prep).

In conventional conchology, shell descriptions and measurements are mainly made based on the standard apertural view of the shell. In this view, the shell is positioned so that the columella is vertical and the shell is rotated around this columella axis until the aperture faces the user. After the apertural view of the shell is set, the shell linear measurements are taken and descriptions of other shell characters are made (see, e.g., Vermeulen 1994). However, the irregularity in the orientation of the aperture of Plectostoma shells, caused by the distortion in the shell coiling hinder this traditional conchological approach.

We feel that in Plectostoma the usefulness of this traditional approach is limited because of the presence of the tuba that deviate from the coiling axis of the spire. The varying length and coiling mode of the tuba prevent any standardisation of the ‘frontal’ view of the spire (Figure 1; see also Sasaki 2010, who refers to “different direction in shell-mouth opening”). Nevertheless, all previous authors use this method to describe and illustrate Plectostoma shells. Here, we proposed a better approach to describe the shells of this, conchologically unusual, genus.

In Plectostoma, the only part of the shell that can serve as a landmark to determine a frontal view, while at the same time fixing the position of both the spire as well as the tuba, is not the aperture, but the constriction, the point where the spire ends and the tuba begins, and operculum rests. Therefore, we determined the frontal view of the shell as follows: the shell is held with the coiling axis of the spire vertical and with the operculum perpendicular to the line of view, on the right side (because all Plectostoma shells are dextral). Left lateral view, back view and right lateral view are obtained by turning the shell 90°, 180°, and 270° from this starting point (Figure 1). Accordingly, all the shell characters were described based on this positioning scheme.

Specimen images and 3D models. To examine the shell morphology, at least one digital 3D model of each species was obtained. We used microcomputed X-ray tomography (CT) to obtain 3D models of Plectostoma shells. For each species, several empty shells and ethanol-preserved shells with the soft body inside, spaning the morphological variation breadth, were selected. Microcomputed tomography was carried in a high-resolution micro-CT scanner (SkyScan, model 1172, Aartselaar, Belgium). 3D models were created from the reconstructed images with the manufacturer’s software CT Analyser v1.12.0.0 (Skyscan©) and saved as digital polygon mesh objects (*.ply). The 3D models were then simplified by quadric edge collapse decimation (to ca. 200, 000 faces) as implemented in the program MeshLab v1.3.2 (Cignoni et al. 2008).

The position and orientation of the 3D digital shell was manipulated so that the shell columella was in parallel with the z-axis and the operculum outer side was visible from a user perspective (Figure 1, Appendix 1(9)). Then, the outer operculum view of the shell was regarded as frontal view. After that, the field of view of the 3D model was set to orthographic, and an image was taken for each of the six perspectives: frontal view (A), left lateral view (B), back view (C), right lateral view (D), top view (E), and bottom view (F). In addition, two images were made of the constriction teeth of the parietal (G) and basal (H) inner shell whorl after clipping of the 3D model. All manipulation and imaging was done with MeshLab v1.3.2 (Cignoni et al. 2008). Thus, a total of eight images were made for each species (A–H). For video tutorials, see Appendix 1(10).

Shell characters and descriptions. The shell is an accretionary exoskeleton of the snail. The overall shape of the shell, which resembles a 3D spiral, results from changes in the curvature and torsion during the shell accretionary process, and form changes in aperture form during shell growth (Okamoto 1988). However, the exact quantifications of these changes might exceed the requirements of the practical purpose of this taxonomic paper. Therefore, we used traditional linear measurements to quantify the shell form, in a way that these measurements abstract the shell ontogeny and its 3D spiral properties.

After the six views of a shell were determined as described (Figure 1), the shell whorls were described for each of six major parts, according to the shell ontogeny order: (1) apex–protoconch and the first teleoconch (Figure 2A, and 3); (2) apical spire–the whole teleoconch except the last 1 1/2 whorls before the constriction (Figure 2B, and 4); (3) basal spire–the last 1 1/2 whorls before the constriction (Figure 2B, and 5); (4) constriction–the narrowest transitional part of the whorl between spire and tuba (Figure 2A, 6, and 7); (5) tuba whorl–teleoconch after the constriction (Figure 2A, and 8); and (6) aperture and peristome (Figure 2A, 9, and 10). The first three parts constitute the shell spire, for which size and shape were quantified from the left lateral view. The height, width and number of whorls of the shell were measured and counted from the spire (Figure 11A, and B). In addition to the description of the general shell form, we recorded the shell surface ornamentations, namely, (7) fine spiral striation (Figure 12), and (8) distinct radial ribs (Figures 2A and 13).

(1) Apex – Ranges from “distinctly convex”, via “moderately convex”, to “slightly convex” (Figure 3). The slightly convex apex has a teleoconch that grows with less torsion and greater curvature than the protoconch, and vice versa in the distinctly convex apex.

(2) Apical spire – Similar to the apex form, less torsion and greater curvature in shell growth produce a “depressed conical” apical spire, and the reverse produces an “oblong conical” apical spire. The oblong and depressed conical shape of the spire can be estimated by measuring the ratio between the apical spire height and width (Figure 4).

(3) Basal spire – The curvature of the basal spire determines the final form of the spire. The width of the spire base is related to how tightly the basal spire whorl coils towards the shell columella, and to the whorl width. The basal spire shape categories, namely, conical, ovoid, and ellipsoid, can be estimated by comparing the difference between two whorl width measurements from three consecutive whorl peripheries (Figure 5). This measurement is made at both the left and the right side of the shell, in left lateral view. The spire umbilicus may be open, partially closed or totally closed by the tuba.

(4) Constriction – The constriction is a short transitional and narrow part of the whorl between the spire and tuba (Figure 2A). This is the furthest point to where the snail can retract into the shell and where the operculum rests (Figure 1, and 2A). Inside the constriction of some species, there are calcareous structures that protrude from the inner shell wall–constriction teeth (Figure 6 and 7). The number, shape, and location of the constriction teeth are taxonomically informative characters.

(5) Tuba – It is difficult to describe the variable forms of tuba in words. For the sake of convenience, we categorised the tuba into three coiling regimes, which represent how many times torsion changes drastically. These are: type 1–no drastic changes in torsion at the beginning of the tuba as compared to the spire; type 2–drastic changes in torsion at the beginning and midway of the tuba; and type 3–drastic changes in torsion at the beginning of tuba (Figure 8). In addition to the coiling regime, the overall shell form is also determined by the final tuba length. However, estimation of tuba length is difficult. Hence, we quantified the tuba length by estimating the ratio of the tuba periphery length on the one hand, and the spire last whorl periphery length on the other. In addition, we estimated the proportion of the tuba that attaches to the spire. Finally, the difference between tuba forms can also be determined by comparing the shell perspective where the aperture opening is best visible.

(6) Aperture and peristome – Before we characterise the aperture form, we define the anatomical position of the aperture according to the orientation of the animal inside (Figure 9). Our definitions for the four areas of aperture side correspond to the conventional terminology used for the aperture of a regularly coiling shell: Anterior = palatal side, posterior = columellar side, right lateral = parietal side and left lateral = basal side. The convenient way to recognise these aperture areas is by identifying the posterior area where it has the densest ribs (Figure 9B). After that we describe and compare the shape of the aperture and the outer peristome (Figure 10). The shell has either a simple or a double peristome, and this character is species specific in Plectostoma. The prominence and shape of the outer peristome is described by how much the outer peristome is projected at the anterior, posterior, and both lateral sides, as compared to the inner peristome (Figure 10A).

(7) Spiral lines – Spiral line sculpture on the shell is composed of a row of granulated micro-structures. The intensity of the spiral lines depends on the size of these micro-structures. In general, Thick spiral lines should be visible under the dissecting microscope at 50× magnification (Figure 12A), whereas thin spiral lines are hardly visible at 50× magnification, but are visible at 100× magnification (Figures 12A and B, Appendix 3). Thick spiral lines are more widely spaced (< 7 lines per 100 μm) than thin lines (> 10 lines per 100 μm). Thin lines may be present in between sparse thick lines (Figure 12A), but on living snails these may fade away when the snail ages and may be hardly visible in old empty shells.

(8) Radial ribs – Radial ribs are produced by a change of shell ontogeny in both shell accretion direction and aperture dimensions. During rib formation, the shell material accretion direction around the aperture changes from longitudinal to orthogonal; meanwhile, aperture size increases and probably aperture shape changes. Thus, the formation of each rib represents a discontinuity in shell growth in the longitudinal direction and a change of the aperture shape and size. In view of these, three characters can be observed in the radial ribs.

The first character is the total number of ribs and the spacing between them on the shell. The number of ribs for a species can be highly variable between different individuals. The ribs are not evenly distributed on the shell surface; for example, the spacing between ribs consistently increases from the apex to the last whorl of the spire (Appendix 4). In view of this, we describe the rib density as the number of ribs within 1 mm on the whorl above the shell constriction in left lateral view (Figure 13A).

The second character is the intensity of the ribs in terms of length and thickness. Generally, rib length is related to the spacing between the ribs; for example, the greater the spacing (whorl length) before the rib, the longer the rib projects from the whorl periphery (Liew T.S. unpublished data). Thus, the lengths of the ribs, from the apex to the last whorl of the shell, change in a trend similar to the rib spacing. Besides, the thickness of ribs can vary between species, but less so within species. The thickness of the ribs depends on the number of shell layers and the thicker the ribs, the more likely it is that ribs persist in old shells (Figure 13B).

The third character is the form of the ribs. As mentioned above, each radial rib on a Plectostoma shell was actually a deformed aperture during shell ontogeny. Thus, a more biologically meaningful way to describe the radial ribs is to compare the rib edge form (deformed aperture) to the whorl before the rib (regular aperture). The rib form can be either the same or different in the spire and tuba parts of the shell. As is the case with the aperture and peristome, the rib edge is either slightly or distinctly projected at its anterior side (i.e. at whorl periphery) as compared to the lateral sides (hereafter “rib plate”). In addition, the shape of the rib plate can be straight, slightly curved, or single-humped (Figure 13C). Although thinner ribs are easily abraded in old shells, at least the rib plate form can be inferred from the abraded scar at the whorl periphery (Figure 13B).

Finally, we noticed variation among species in the rib inclination, which can be estimated with respect to the coiling axis. However, it is difficult to quantify this inclination accurately. Nevertheless, we add this character into the description of the shell in a qualitative manner in the terms of orthoclin (i.e. ribs are almost straight with respect to the collumella, as in Plectostoma christae and Plectostoma siphonostomum), prosoclin (i.e. ribs are distinctly tilted with respect to the collumella, as in Plectostoma tohchinyawi and Plectostoma salpidomon).

In brief, shells of all species were described following a template consisting of three major elements (Table 1). First, the shell was described in several parts which represent the chronology of the shell ontogeny. Second, we defined a set of characters that can be determined in each shell part. Lastly, we described each character in either a quantitative or quantitative manner. We regard this template as a morphological model for the taxonomy of the genus Plectostoma, and each element in this template may be updated by future taxonomists when necessary. We present the final description of each species in a uniform telegraphic format (e.g. semantic-tagged by bold text, and colon “:”) so that these morphological data can be mined effectively. By doing this, we hope to reduce the redundant process where species descriptions are done de novo each time a taxonomist revises the same taxa (Deans et al. 2012).

Digital model. Pictures are more effective than verbal descriptions for shell morphology. However, it is not feasible to have hundreds of pictures taken for each perspective of a shell. Many non-linear characteristics of a shell cannot effectively be representedby 2D images. Thus, an interactive 3D model shell improves the dissemination of morphological information. Presenting 3D models in digital publication has started five year ago (Ruthensteiner and Heß 2008), and since then more taxonomists have taken the initiative to embed 3D models in e-papers. However, in this paper we have refrained from embed 3D models in the paper itself, since this limits further analysis by readers. Instead, we provide sets of 3D data in *.blend files, which consists of all 3D models and which can be opened in Blender v2.63 (www.blender.org). The 3D models in the blend file can be exported to *.ply format, which can be opened in MeshLab v1.3.2 (Cignoni et al. 2008). Both Blender and MeshLab are freeware and can be used to analyse the 3D model further (e.g. measurements, modification, etc). For video tutorials, see Appendix 1(11 and 12).

We propose the conservation status for each species by following IUCN Red List Criteria and guidelines (IUCN Standards and Petitions Subcommittee 2013). We assess the conservation status based on our fieldwork in Malaysia between the year 2010 and year 2013 and the information obtained from the museum collections.

Taxon sampling. A total of 27 ingroup taxa of the genus Opisthostoma (n=11) and Plectostoma (n=16) were included in this study: six Opisthostoma species from Borneo, five Opisthostoma from Peninsular Malaysia, nine Plectostoma species from Borneo, and seven Plectostoma from Peninsular Malaysia. All of these ingroup taxa were selected on the basis of their distribution and shell forms which are representative for about 150 species in both genera. In addition to the ingroup taxa, eight outgroup taxa were included in the phylogenetic analysis. Sequence data for these outgroup taxa, which include three genera of the Diplommatinidae and a species of the Cochlostomatidae, were obtained from Webster et al. (2012). The details of these specimens and the Genbank accession numbers are listed in Table 2.

DNA extraction, PCR and sequencing. DNA extraction was done for each specimen (entire animal and its shell), with the E.Z.N.A. Mollusc DNA kit (OMEGA bio-tek). We followed the manufacturer’s extraction protocol. After extraction, PCR was carried out to amplify four regions, namely, 16S (mitochondrial, Palumbi 1996), COI (mitochondrial, Folmer et al. 1994), 28S (nuclear, Park and Foighil 2000), and 18S (nuclear, Stothard et al. 2000). We used the PCR reactions and programs of Webster et al. (2000). Positive PCR products were sequenced by Macrogen sequencing service (Macrogen Inc., Europe).

Phylogenetic inferences. Alignment of sequences was done with Bioedit v7.1.3 (Hall 1999) and adjusted manually. The final aligned data matrix consists of 2241 positions, of which 2092 can be aligned unambiguously (Appendix 17). The remaining 149 characters (91 from 16S and 58 from 28S) were excluded from further analysis. Mr.Modeltest v2.3 (Nylander 2004) was used to select the most appropriate model, based on the Akaike Information Criterion (AIC) for 16S, 18S and 28S, and as well as each of the three codon positions for COI. The best fits were: the GTR+I+Γ model for 16S, 28S, COI(2nd codon) and COI (3rd codon); the GTR+Γ for COI(1st codon); and the SYM+I+Γ for 18S.

Three phylogenetic analyses were done, namely, Bayesian inference (BI), Maximum Likelihood analysis (ML), and Parsimony analyses (PA). Bayesian inference was run in MrBayes v3.2.1 (Huelsenbeck and Ronquist 2001) with the following setting: mcmc ngen=5000000; nchains=4; samplefreq=100; average deviation of split frequencies < 0.01; and a burn-in value of 25%. A Maximum Likelihood analysis was run using RaxML v7.2.6 (Stamatakis 2006) as implemented on CIPRES portal v2.2 (Miller et al. 2010) on all the genes together, with 1000 rapid bootstraps using GTR + U. The data was divided into six partitions, all analyzed with a GTR substitution model. The Parsimony analyses were run using PAUP v4.0b (Swofford 1998). Gaps were set as fifth character state. A bootstrapped heuristic search with 1000 bootstrap replicates, and 10 random addition sequence heuristic search replicates, with no rearrangement limit per replicate was carried out, with 50% as the minimum bootstrap support included.

COI Barcoding. As mentioned above, the shell morphology alone may not be sufficient to delimit many species that have greater morphological variation within species. Thus, in addition to shell morphology, we sought another way to make the decision about the species delimitation. DNA barcoding is used to provide more insight into the species delimitation for gastropods (Boeters and Knebelsberger 2012, Puillandre et al. 2012). Therefore, we sequenced COI (mitochondrial, Folmer et al. 1994) for standard DNA barcoding analysis for a greater number of individuals than we used for the phylogenetic analysis. In total, we obtained COI sequences from 51 individuals, which comprise 19 species (including 8 new species) (Table 2). Then, the pairwise genetic distances for all the 51 sequences were computed by using Kimura 2-parameters in MEGA5 (Tamura et al. 2011).

Genetic data repository. The collection information for the specimens used were uploaded, stored and managed in Barcoding of Life Database (BOLD, http://www.boldsystems.org, Ratnasingham and Hebert 2007). After that, all DNA sequences were uploaded to NCBI GenBank (http://www.ncbi.nlm.nih.gov/genbank, Benson et al. 1997) via BOLD. Genetic data deposited in GenBank can be retrieved easily with the Python tool Biopython (e.g. Module “Entrez”) (http://biopython.org, Cock et al. 2009).

While uploading the data of 31 species, 62 references, 214 collections, 290 pictures, and 31 species pages to Lifedesks, they were simultaneously tagged and linked to each other (Figure 14). The textual information can be downloaded and accessed offline in the xhtml format (source code) (Figures 15). Similarly, the taxonomic information that was used in creating the KML file can be retrieved in KML format (source code) (Figures 15, Appendix 5). In addition to the taxonomic data, a total of 155 3D models of 29 Plectostoma species were constructed, which belong to 86 samples. For the ease of specimen and species comparison, all 3D models were saved in ten separate. blend files, each containing 20 layers (Appendix 6–15). The 3D models (i.e. specimens) that belong to the same species were saved in the same layer.

Most taxonomists will have established their own workflow when working with taxonomic data. However, often only fractions of such data are published in a taxonomic revision paper. Thus, extra work would be needed to upload the data to online data platforms whenever a taxonomic revision is done in a traditional manner (i.e. not using cybertaxonomic tools). Here, we demonstrated that, in fact, no extra works is required if a cybertaxonomic workflow is adopted. Taxonomists themselves, who maintain, tag, link and create the data will benefit most by using these cybertaxonomy tools. For example, with some existing tools, such as Google Earth, the information can be integrated, explored and analysed in an interactive way, which will increase the efficiency of the taxonomic methodology (sensu Figure 1 in Sluys 2013; Appendix 5).

Furthermore, all textual information is preserved, and can be simply accessed in raw format. In other words, this information is not locked and can be retrieved and integrated with very basic programming skills (such as Python) even when the desirable platform (internet) and software are not available. More importantly, adopting this workflow in taxonomy practice will help realise the vision of Wheeler et al. (2012), that “The resultant cyber-enabled taxonomy, or cybertaxonomy, would open access to biodiversity data to developing nations, assure access to reliable data about species, and change how scientists and citizens alike access, use and think about biological diversity information”.

Table 3 shows the morphological data matrix of 31 Plectostoma species and 11 general qualitative shell characters. Twenty-seven out of 31 Plectostoma species have a unique set of shell character states. The remaining four species share two unique sets of general characters states, namely, 1) Plectostoma salpidomon (van Benthem Jutting, 1952), and Plectostoma laemodes (van Benthem Jutting, 1961); and 2) Plectostoma dindingensis sp. n., and Plectostoma mengaburensis sp. n. However, each of these four species is distinguishable by other shell characters (see diagnosis and species description in the taxonomy section).

Our shell shape characterisation approach that views the shell as a petrified ontogeny provides a set of distinguishable general characters for species delimitation (Table 3). Most of the species in this study can be identified just by using these general shape characters. Furthermore, those species that are not distinguishable with these general shape characters, are distinguishable by using more specific shell shape and size characters (see Taxonomy Key).

However, it is important to note that the intra- and inter-specific variation in shell shape is more difficult to characterise than the variation in shell size. This is reflected in our species description, where the variation in shell size and countable shell characters, such as ribs, but not the variation in general shell shape are explicitly given. Nevertheless, the general qualitative shell shape characters do implicitly reflect the variation because many of these, such as spire shape, tuba, and spiral lines, are obtained by categorising the quantitative variation of these characters (e.g. Figures 4 and 5).

Overall, we suggest that 10 of the non-Bornean Plectostoma species are threatened and Plectostoma sciaphilum is extinct. Specifically, Plectostoma umbilicatum, Plectostoma senex, Plectostoma turriforme, Plectostoma retrovertens, Plectostoma charasense, and Plectostoma tenggekensis are in the Critically Endangered category; Plectostoma kubuensis is in the Endangered category; and Plectostoma dindingensis, Plectostoma palinhelix, and Plectostoma laidlawi are in the Vulnerable category. All of these species, except Plectostoma laidlawi, occur in limestone hills in the State of Pahang, Malaysia, where many of these hills are being quarried or are at risk of being quarried. Our assessments would eventually be submitted to IUCN.

As revealed by the Bayesian posterior probability (PP) and maximum likelihood analysis bootstrap (BS) values of the phylogenetic tree in Figure 16, Plectostoma is monophyletic (BI/ML/PA; 1.0/99/100) and is the sister taxon of the less well-supported clade of Opisthostoma + Arinia (0.5/<70/<70). Within the Opisthostoma + Arinia clade, all Opisthostoma except Opisthostoma vermiculum form a well-supported clade (0.98/88/71). The divergence between these clades is similar to the divergence between other genera in Diplommatinidae. Our phylogenetic analysis suggested that Opisthostoma vermiculum Clements & Vermeulen, 2008 in Clements et al. (2008) has been incorrectly assigned to the genus Opisthostoma.

Table 4 shows the Kimura 2-parameter distances for all sequence pairs within each group and the net average between groups of sequences (for sequence alignment see Appendix 16). This reveals that all species pairs exceed a divergence of 10% (n = 169), with the exception of Plectostoma crassipupa vs. Plectostoma christae and Plectostoma crassipupa vs. Plectostoma laidlawi. Genetic divergence within each species is below 9% (n = 14, mean = 2.6%, SD = 0.1%), with the exception of Plectostoma crassipupa.

One key determinant for the success of DNA barcoding is prior knowledge of intra- and interspecific genetic distances for the barcoding marker in question. The optimum intra- and interspecific threshold in gastropods is higher than the conventional value (Hebert et al. 2003) of 3% (e.g. 4% in Davison et al. 2009, 6% in Köhler and Johnson 2012, 9.8%–25% in Parmakelis et al. 2013). In our study, we also found higher values of intraspecific variation and interspecific divergence’ of COI for three well-defined species, namely, Plectostoma salpidomon, Plectostoma christae, and Plectostoma siphonostomum are larger and smaller than 10%, respectively.

A study on a pulmonate limestone-dwelling micro-landsnail in the same region also suggests intraspecific COI divergence not exceeding 10% based on the Kimura 2-parameter model (Hoekstra and Schilthuizen 2011), and similar values were obtained for Everettia, a Malaysian pulmonate not restricted to limestone (Liew et al. 2009). Hence, based on our results, together with the only other two studies on the COI variation of land snails in Sundaland, we advise caution in using a conventional threshold value in COI genetic variability for species delimitation, when the background genetic variability of COI is unknown for a particular taxon in a particular geographic region.